Preface to GEB's Twentieth-anniversary Edition

So what is this book, Gdel, Escher, Bach: an Eternal Golden Braidusually known by its acronym, "GEB" really all about? That question has hounded me ever since I was scribblin its first drafts in !en, way back in "#$%& 'riends would inquire, of course, what I was so ri!!ed by, but I was hard !ressed to e(!lain it concisely& ) few years later, in "#*+, when GEB found itself for a while on the bestseller list of The New York Times, the obli atory one-sentence summary !rinted underneath the title said the followin , for several weeks runnin , -) scientist ar ues that reality is a system of interconnected braids&- )fter I !rotested vehemently about this utter ho wash, they finally substituted somethin a little better, .ust barely accurate enou h to kee! me from howlin a ain& /any !eo!le think the title tells it all, a book about a mathematician, an artist, and a musician& 0ut the most casual look will show that these three individuals per se, au ust thou h they undeniably are, !lay but tiny roles in the book1s content& There1s no way the book is about those three !eo!le2 3ell, then, how about describin GEB as -a book that shows how math, art, and music are really all the same thin at their core-? ) ain, this is a million miles offand yet I1ve heard it over and over a ain, not only from nonreaders but also from readers, even very ardent readers, of the book& )nd in bookstores, I have run across GEB racin the shelves of many diverse sections, includin not only math, eneral science, !hiloso!hy, and co nitive science 4which are all fine5, but also reli ion, the occult, and 6od knows what else& 3hy is it so hard to fi ure out what this book is about? 7ertainly it1s not .ust its len th& 8o, it must be in !art that GEB delves, and not .ust su!erficially, into so many motley to!icsfu ues and canons, lo ic and truth, eometry, recursion, syntactic structures, the nature of meanin , 9en 0uddhism, !arado(es, brain and mind, reductionism and holism, ant colonies, conce!ts and mental re!resentations, translation, com!uters and their lan ua es, :8), !roteins, the enetic code, artificial intelli ence, creativity, consciousness and free willsometimes even art and music, of all thin s2that many !eo!le find it im!ossible to locate the core focus&

The Key Images and Ideas that Lie at the Core of GEB
8eedless to say, this wides!read confusion has been quite frustratin to over the years, since I felt sure I had s!elled out my aims over and over in the te(t itself& 7learly, however, I didn1t do it sufficiently often, or sufficiently clearly& 0ut since now I1ve ot the chance to do it once moreand in a !rominent s!ot in the book, to bootlet me try one last time to say why I wrote this book, what it is about, and what its !rinci!al thesis is& In a word, GEB is a very !ersonal attem!t to say how it is that animate bein s can come out of inanimate matter& 3hat is a self, and how can a self come out of stuff that is as selfless as a stone or a !uddle? 3hat is an -I-, and why are such thin s found 4at least so far5 only in association with, as !oet ;ussell Edson once wonderfully !hrased it, -teeterin bulbs of dread and dream-that is, only in association with certain kinds of

ooey lum!s encased in hard !rotective shells mounted ato! mobile !edestals that roam the world on !airs of sli htly fu<<y, .ointed stilts? GEB a!!roaches these questions by slowly buildin u! an analo y that likens inanimate molecules to meanin less symbols, and further likens selves 4or -I- 1s or -souls-, if you !referwhatever it is that distin uishes animate from inanimate matter5 to certain s!ecial swirly, twisty, vorte(-like, and meaningful !atterns that arise only in !articular ty!es of systems of meanin less symbols& It is these stran e, twisty !atterns that the book s!ends so much time on, because they are little known, little a!!reciated, counterintuitive, and quite filled with mystery& )nd for reasons that should not be too difficult to fathom, I call such stran e, loo!y !atterns -stran e loo!s- throu hout the book, althou h in later cha!ters, I also use the !hrase -tan led hierarchies- to describe basically the same idea& This is in many ways why /& 7& Escheror more !recisely, his artis !rominent in the - olden braid-, because Escher, in his own s!ecial way, was .ust as fascinated as I am by stran e loo!s, and in fact he drew them in a variety of conte(ts, all wonderfully disorientin and fascinatin & 3hen I was first workin on my book, however, Escher was totally out of the !icture 4or out of the loo!, as we now say5= my workin title was the rather mundane !hrase -6>del1s Theorem and the ?uman 0rain-, and I ave no thou ht to insertin !arado(ical !ictures, let alone !layful dialo ues& It1s .ust that time and a ain, while writin about my notion of stran e loo!s, I would catch fleetin lim!ses of this or that Escher !rint flashin almost subliminally before my mind1s eye, and finally one day I reali<ed that these ima es were so connected in my own mind with the ideas that I was writin about that for me to de!rive my readers of the connection that I myself felt so stron ly would be nothin less than !erverse& )nd so Escher1s art was welcomed on board& )s for 0ach, I1ll come back to his entry into my -meta!horical fu ue on minds and machines- a little later& 0ack to stran e loo!s, ri ht now& GEB was ins!ired by my lon -held conviction that the -stran e loo!- notion holds the key to unravelin the mystery that we conscious bein s call -bein - or -consciousness-& I was first hit by this idea when, as a teen-a er, I found myself obsessedly !onderin the quintessential stran e loo! that lies at the core of the !roof of @urt 6>del1s famous incom!leteness theorem in mathematical lo ica rather arcane !lace, one mi ht well think, to stumble across the secret behind the nature of selves and -I-s, and yet I !ractically heard it screamin u! at me from the !a es of 8a el and 8ewman that this was what it was all about& This !reface is not the time and !lace to o into detailsindeed, that1s why the tome you1re holdin was written, so it would be a bit !resum!tuous of me to think I could outdo its author in .ust these few !a es2but one thin has to be said strai ht off, the 6>delian stran e loo! that arises in formal systems in mathematics i!e!, collections of rules for churnin out an endless series of mathematical truths solely by mechanical symbol-shuntin without any re ard to meanin s or ideas hidden in the sha!es bein mani!ulated5 is a loo! that allows such a system to -!erceive itself-, to talk about itself, to become -self-aware-, and in a sense it would not be oin too far to say that by virtue of havin such a loo!, a formal system ac"uires a self!

Meaningless Symbols Acquire Meaning

es!ite Themsel"es

3hat is so weird in this is that the formal systems where these skeletal -selvescome to e(ist are built out of nothin but meanin less symbols& The self, such as it is, arises solely because of a s!ecial ty!e of swirly, tan led pattern amon the meanin less symbols& 0ut now a confession, I am bein a bit coy when I re!eatedly ty!e the !hrase -meanin less symbols- 4as at the ends of both of the !revious sentences5, because a crucial !art of my book1s ar ument rests on the idea that meanin cannot be ke!t out of formal systems when sufficiently com!le( isomor!hisms arise& /eanin comes in des!ite one1s best efforts to kee! symbols meanin less2 Aet me re!hrase these last cou!le of sentences without usin the sli htly technical term -isomor!hism-& 3hen a system of -meanin less- symbols has !atterns in it that accurately track, or mirror, various !henomena in the world, then that trackin or mirrorin imbues the symbols with some de ree of meanin indeed, such trackin or mirrorin is no less and no more than what meanin is& :e!endin on how com!le( and subtle and reliable the trackin is, different de rees of meanin fulness arise& I won1t o further into this here, for it1s a thesis that is taken u! quite often in the te(t, most of all in 7ha!ters B, C, D, #, and ""& 7om!ared to a ty!ical formal system, human lan ua e is unbelievably fluid and subtle in its !atterns of trackin reality, and for that reason the symbols in formal systems can seem quite arid= indeed, without too much trouble, one can look at them as totally devoid of meanin & 0ut then a ain, one can look at a news!a!er written in an unfamiliar writin system, and the stran e sha!es seem like nothin more than wondrously intricate but totally meanin less !atterns& Thus even human lan ua e, rich thou h it is, can be drained of its seemin si nificance& )s a matter of fact, there are still quite a few !hiloso!hers, scientists, and so forth who believe that !atterns of symbols per se 4such as books or movies or libraries or 7:;E/1s or com!uter !ro rams, no matter how com!le( or dynamic5 ne#er have meanin on their own, but that meanin instead, in some most mysterious manner, s!rin s only from the or anic chemistry, or !erha!s the quantum mechanics, of !rocesses that take !lace in carbon-based biolo ical brains& )lthou h I have no !atience with this !arochial, bio-chauvinistic view, I nonetheless have a !retty clear sense of its intuitive a!!eal& Tryin to don the hat of a believer in the !rimacy, indeed the uniqueness, of brains, I can see where such !eo!le are comin from& Such !eo!le feel that some kind of -semantic ma ic- takes !lace only inside our -teeterin bulbs-, somewhere behind !airs of eyeballs, even thou h they can never quite !ut their fin er on how or why this is so= moreover, they believe that this semantic ma ic is what is res!onsible for the e(istence of human selves, souls, consciousness, -I-s& )nd I, as a matter of fact, quite a ree with such thinkers that selves and semanticsin other words, that me1s and meanin sdo s!rin from one and the same source= where I take issue with these !eo!le is over their contention that such !henomena are due entirely to some s!ecial, thou h as yet undiscovered, !ro!erties of the microsco!ic hardware of brains& )s I see it, the only way of overcomin this ma ical view of what -I- and consciousness are is to kee! on remindin oneself, un!leasant thou h it may seem, that the -teeterin bulb of dread and dream- that nestles safely inside one1s own cranium is a !urely !hysical ob.ect made u! of com!letely sterile and inanimate com!onents, all of

which obey e(actly the same laws as those that overn all the rest of the universe, such as !ieces of te(t, or 7:-;E/1s, or com!uters& Enly if one kee!s on bashin u! a ainst this disturbin fact can one slowly be in to develo! a feel for the way out of the mystery of consciousness, that the key is not the stuff out of which brains are made, but the patterns that can come to e(ist inside the stuff of a brain& This is a liberatin shift, because it allows one to move to a different level of considerin what brains are, as media that su!!ort com!le( !atterns that mirror, albeit far from !erfectly, the world, of which, needless to say, those brains are themselves deni<ens and it is in the inevitable self-mirrorin that arises, however im!artial or im!erfect it may be, that the stran e loo!s of consciousness start to swirl&

Kurt G#del Smashes through Bertrand $ussell%s Maginot Line

I1ve .ust claimed that the shift of focus from material com!onents to abstract !atterns allows the quasi-ma ical lea! from inanimate to animate, from nonsemantic to semantic, from meanin less to meanin ful, to take !lace& 0ut how does this ha!!en? )fter all, not all .um!s from matter to !attern ive rise to consciousness or soul or self, quite obviously, in a word, not all !atterns are conscious& 3hat kind of !attern is it, then, that is the telltale mark of a self $ GEB's answer is, a stran e loo!& The irony is that the first stran e loo! ever foundand my model for the conce!t in eneralwas found in a system tailor%made to keep loopiness out! I s!eak of 0ertrand ;ussell and )lfred 8orth 3hitehead1s famous treatise &rincipia 'athematica, a i antic, forbiddin work laced with dense, !rickly symbolism fillin u! volume after volume, whose creation in the years "#"+-"#"% was s!arked !rimarily by its first author1s des!erate quest for a way to circumvent !arado(es of self-reference in mathematics& )t the heart of &rincipia 'athematica lay ;ussell1s so-called -theory of ty!es-, which, much like the rou hly contem!oraneous /a inot Aine, was desi ned to kee! -the enemy- out in a most staunch and waterti ht manner& 'or the 'rench, the enemy was 6ermany= for ;ussell, it was self-reference& ;ussell believed that for a mathematical system to be able to talk about itself in any way whatsoever was the kiss of death, for self-reference wouldso he thou htnecessarily o!en the door to self-contradiction, and thereby send all of mathematics crashin to the round& In order to forestall this dire fate, he invented an elaborate 4and infinite5 hierarchy of levels, all sealed off from each other in such a manner as to definitivelyso he thou htblock the dreaded virus of self-reference from infectin the fra ile system& It took a cou!le of decades, but eventually the youn )ustrian lo ician @urt 6>del reali<ed that ;ussell and 3hitehead1s mathematical /a inot Aine a ainst selfreference could be most deftly circumvented 4.ust as the 6ermans in 3orld 3ar II would soon wind u! deftly sideste!!in the real /a inot Aine5, and that self-reference not only had lurked from :ay Ene in &rincipia 'athematica, but in fact !la ued !oor &' in a totally unremovable manner& /oreover, as 6>del made brutally clear, this thorou h riddlin of the system by self-reference was not due to some weakness in &', but quite to the contrary, it was due to its strength! )ny similar system would have e(actly the same -defect-& The reason it had taken so lon for the world to reali<e this astonishin fact is that it de!ended on makin a lea! somewhat analo ous to that from a brain to a self, that famous lea! from inanimate constituents to animate !atterns&

'or 6>del, it all came into focus in "#%+ or so, thanks to a sim!le but wonderfully rich discovery that came to be known as -6>del numberin -a ma!!in whereby the lon linear arran ements of strin s of symbols in any formal system are mirrored !recisely by mathematical relationshi!s amon certain 4usually astronomically lar e5 whole numbers& Fsin his ma!!in between elaborate !atterns of meanin less symbols 4to use that dubious term once a ain5 and hu e numbers, 6>del showed how a statement a(out any mathematical formal system 4such as the assertion that &rincipia 'athematica is contradiction-free5 can be translated into a mathematical statement inside number theory 4the study of whole numbers5& In other words, any metamathematical statement can be im!orted into mathematics, and in its new uise the statement sim!ly asserts 4as do all statements of number theory5 that certain whole numbers have certain !ro!erties or relationshi!s to each other& 0ut on another level, it also has a vastly different meanin that, on its surface, seems as far removed from a statement of number theory as would be a sentence in a :ostoevsky novel& 0y means of 6>del1s ma!!in , any formal system desi ned to s!ew forth truths about -mere- numbers would also wind u! s!ewin forth truthsinadvertently but ine(orablyabout its own !ro!erties, and would thereby become -self-aware-, in a manner of s!eakin & )nd of all the clandestine instances of self-referentiality !la uin &' and brou ht to li ht by 6>del, the most concentrated doses lurked in those sentences that talked about their own 6>del numbers, and in !articular said some very odd thin s about themselves, such as -I am not !rovable inside &'"! )nd let me re!eat, such twistin -back, such loo!in -around, such sell enfoldin , far from bein an eliminable defect, was an inevitable by-!roduct of the system1s vast !ower& 8ot too sur!risin ly, revolutionary mathematical and !hiloso!hical consequences tumbled out of 6>del1s sudden revelation that self-reference abounded in the bosom of the bastion so carefully desi ned by ;ussell to kee! it out at all costs= the most famous such consequence was the so-called -essential incom!leteness- of formali<ed mathematics& That notion will be carefully covered in the cha!ters to come, and yet, fascinatin thou h it is, incom!leteness is not in itself central to GEB's thesis& 'or GEB, the most crucial as!ect of 6>del1s work is its demonstration that a statement1s meaning can have dee! consequences, even in a su!!osedly meanin less universe& Thus it is the meaning of 6>del1s sentence 6 4the one that asserts -6 is not !rovable inside &'-5 that uarantees that 6 is not !rovable inside &' 4which is !recisely what 6 itself claims5& It is as if the sentence1s hidden 6>delian meanin had some kind of !ower over the vacuous symbol-shuntin , meanin -im!ervious rules of the system, !reventin them from ever !uttin to ether a demonstration of 6, no matter what they do&

&!side-do'n Causality and the Emergence of an (I(

This kind of effect ives one a sense of cra<ily twisted, or u!side-down, causality& )fter all, shouldn1t meanin s that one chooses to read into strin s of meanin less symbols be totally without consequence? Even stran er is that the onl) reason sentence 6 is not !rovable inside &' is its self-referential meanin = indeed, it would seem that 6, bein a true statement about whole numbers, ought to be !rovable, butthanks to its e(tra level of meanin as a statement about itself, assertin its own non!rovabilityit is not&

Somethin very stran e thus emer es from the 6>delian loo!, the revelation of the causal !ower of meanin in a rule-bound but meanin -free universe& )nd this is where my analo y to brains and selves comes back in, su estin that the twisted loo! of selfhood tra!!ed inside an inanimate bulb called a -brain- also has causal !oweror, !ut another way, that a mere !attern called -I- can shove around inanimate !articles in the brain no less than inanimate !articles in the brain can shove around !atterns& In short, an -I- comes aboutin my view, at leastvia a kind of vorte( whereby !atterns in a brain mirror the brain1s mirrorin of the world, and eventually mirror themselves, whereu!on the vorte( of -I- becomes a real, causal entity& 'or an im!erfect but vivid concrete analo ue to this curious abstract !henomenon, think of what ha!!ens when a TG camera is !ointed at a TG screen so as to dis!lay the screen on itself 4and that screen on itself, etc&5what in GEB I called a -self-en ulfin television-, and in my later writin s I sometimes call a -level-crossin feedback loo!-& 3hen and only when such a loo! arises in a brain or in any other substrate, is a persona unique new -I-brou ht into bein & /oreover, the more self-referentially rich such a loo! is, the more conscious is the self to which it ives rise& Hes, shockin thou h this mi ht sound, consciousness is not an onIoff !henomenon, but admits of de rees, rades, shades& Er, to !ut it more bluntly, there are bi er souls and smaller souls&

Small-souled Men) Be'are*

I can1t hel! but recall, at this !oint, a horribly elitist but very droll remark by one of my favorite writers, the )merican -critic of the seven arts-, James ?uneker, in his scintillatin bio ra!hy of 'rKdKric 7ho!in, on the sub.ect of 7ho!in1s Ktude E!& BL, 8o& "" in ) minor, which for me, and for ?uneker, is one of the most stirrin and most sublime !ieces of music ever written, -Small-souled men, no matter how a ile their fin ers, should avoid it&-Small-souled men-?2 3hew2 :oes that !hrase ever run a ainst the rain of )merican democracy2 )nd yet, leavin aside its offensive, archaic se(ism 4a crime I, too, commit in GEB, to my reat re ret5, I would su est that it is only because we all tacitly do believe in somethin like ?uneker1s shockin distinction that most of us are willin to eat animals of one sort or another, to smash flies, swat mosquitoes, fi ht bacteria with antibiotics, and so forth& 3e enerally concur that -men- such as a cow, a turkey, a fro , and a fish all !ossess some s!ark of consciousness, some kind of !rimitive -soul-, but by 6od, it1s a ood deal smaller than ours isand that, no more and no less, is why we -men- feel that we have the !erfect ri ht to e(tin uish the dim li hts in the heads of these fractionally-souled beasts and to obble down their once warm and wi lin , now chilled and stilled !roto!lasm with limitless usto, and not to feel a trace of uilt while doin so& Enou h sermoni<in 2 The real !oint here is that not all stran e loo!s ive rise to souls as rand and lorious as yours and mine, dear reader& Thus, for e(am!le, I would not want you or anyone else to walk away from readin all or !art of GEB, shake their head and say with sadness, -That weird ?ofstadter uy has convinced himself that ;ussell and 3hitehead1s &rincipia 'athematica is a conscious !erson with a soul2?orsefeathers2 0alderdash2 Po!!ycock2 6>del1s stran e loo!, thou h it is my !ara on for the conce!t, is nonetheless only the most bare-bones stran e loo!, and it resides in a system whose com!le(ity is !athetic, relative to that of an or anic brain& /oreover, a

formal system is static= it doesn1t chan e or row over time& ) formal system does not live in a society of other formal systems, mirrorin them inside itself, and bein mirrored in turn inside its -friends-& 3ell, I retract that last remark, at least a bit, any formal system as !owerful as &' does in fact contain models not .ust of itself but of an infinite number of other formal systems, some like it, some very much unlike it& That is essentially what 6>del reali<ed& 0ut still, there is no counter!art to time, no counter!art to develo!ment, let alone to birth and death& )nd so whatever I say about -selves- comin to e(ist in mathematical formal systems has to be taken with the !ro!er rain of salt& Stran e loo!s are an abstract structure that cro!s u! in various media and in varyin de rees of richness& GEB is in essence a lon !ro!osal of stran e loo!s as a meta!hor for how selfhood ori inates, a meta!hor by which to be in to rab a hold of .ust what it is that makes an -I- seem, at one and the same time, so terribly real and tan ible to its own !ossessor, and yet also so va ue, so im!enetrable, so dee!ly elusive& I !ersonally cannot ima ine that consciousness will be fully understood without reference to 6>delian stran e loo!s or level-crossin feedback loo!s& 'or that reason, I must say, I have been sur!rised and !u<<led that the !ast few years1 flurry of books tryin to unravel the mysteries of consciousness almost never mention anythin alon these lines& /any of these books1 authors have even read and savored GEB, yet nowhere is its core thesis echoed& It sometimes feels as if I had shouted a dee!ly cherished messa e out into an em!ty chasm and nobody heard me&

The Earliest Seeds of GEB

3hy, one mi ht wonder, if the author1s aim was merely to !ro!ose a theory of stran e loo!s as the cru( of our consciousness and the source of our irre!ressible -I-feelin , did he wind u! writin such a vast book with so many seemin di ressions in it? 3hy on earth did he dra in fu ues and canons? 3hy recursion? )nd 9en? )nd molecular biolo y? Et cetera&&& The truth of the matter is, when I started out, I didn1t have the fo iest idea that I would wind u! talkin about these kinds of thin s& 8or did I dream that my future book would include dialo ues, let alone dialo ues based on musical forms& The com!le( and ambitious nature of my !ro.ect evolved only radually& In broad strokes, it came about this way& I earlier alluded to my readin , as a teen-a er, of Ernest 8a el and James ;& 8ewman1s little book Gdel's &roof! 3ell, that book .ust radiated e(citement and de!th to me, and it !ro!elled me like an arrow strai ht into the study of symbolic lo ic& Thus, as an under raduate math ma.or at Stanford and a few years later, in my short-lived career as a raduate student in math at 0erkeley, I took several advanced lo ic courses, but to my bitter disa!!ointment, all of them were arcane, technical, and utterly devoid of the ma ic I1d known in 8a el and 8ewman& The u!shot of my takin these hi hbrow courses was that my keen teen interest in 6>del1s wondrous !roof and its -stran e loo!iness- was nearly killed off& Indeed, I was left with such a feelin of sterility that in late "#D$, almost in des!eration, I dro!!ed out of math rad school in 0erkeley and took u! a new identity as !hysics rad student at the Fniversity of Ere on in Eu ene, where my once-ardent fascination with lo ic and metamathematics went into dee! dormancy&

Several years !assed, and then one day in /ay of "#$B, while browsin the math shelves in the Fniversity of Ere on bookstore, I stumbled across !hiloso!her ?oward :eAon 1s su!erb book * &rofile of 'athematical +ogic, took a chance on buyin it, and within weeks, my old love for the reat 6>delian mysteries and all they touch on was reawakened& Ideas started churnin around like mad inside my teeterin bulb of dread and dream& :es!ite this .oy, I was very discoura ed with the way my !hysics studies and my life in eneral were oin , so in July I !acked all my belon in s into a do<en or so cardboard bo(es and set out on an eastward trek across the vast )merican continent in Muicksilver, my faithful "#LD /ercury& 3here I was headed, I wasn1t sure& )ll I knew is that I was lookin for a new life& )fter crossin the beautiful 7ascades and eastern Ere on1s desert, I wound u! in /oscow, Idaho& Since Muicksilver had a little en ine trouble and needed some re!air, I took advanta e of the s!are time and went to the Fniversity of Idaho1s library to look u! some of the articles about 6>del1s !roof in :eAon 1s annotated biblio ra!hy& I !hotoco!ied several of them, and in a day or so headed off toward /ontana and )lberta& Each ni ht I would sto! and !itch my little tent, sometimes in a forest, sometimes by a lake, and then I would ea erly !lun e by flashli ht into these articles until I fell aslee! in my slee!in ba & I was startin to understand many 6>delian matters ever more clearly, and what I was learnin was truly enthrallin &

+rom Letter to ,am!hlet to Seminar

)fter a few days in the 7anadian ;ockies, I headed south a ain and eventually reached 0oulder, 7olorado& There, one afternoon, a host of fresh ideas started ushin out in a s!ontaneous letter to my old friend ;obert 0oenin er& )fter several hours of writin , I saw that althou h my letter was lon er than I1d e(!ectedthirty handwritten !a es or so I1d said only about half of what I1d wanted to say& This made me think that maybe I should write a !am!hlet, not a letter, and to this day, ;obert has never received my unfinished missive& 'rom 0oulder I headed further east, bouncin from one university town to another, and eventually, almost as if it had been beckonin me the whole time, 8ew Hork 7ity loomed as my ultimate oal& Indeed, I wound u! s!endin several months in /anhattan, takin raduate courses at 7ity 7olle e and teachin elementary !hysics to nurses at ?unter 7olle e, but as "#$% rolled around, I faced the fact that des!ite lovin 8ew Hork in many ways, I was even more a itated than I had been in Eu ene, and I decided it would be wiser to return to Ere on and to finish raduate school there& )lthou h my ho!ed-for -new life- had failed to materiali<e, in certain res!ects I was relieved to be back& 'or one thin , the F of E in those days had the enli htened !olicy that any community member could invent and teach a for-credit -SE);7?course, as lon as one or more de!artments a!!roved it& )nd so I !etitioned the !hiloso!hy and math de!artments to s!onsor a s!rin -quarter SE);7? course centered on 6>del1s theorem, and my request was ranted& Thin s were lookin u!& /y intuition told me that my !ersonal fascination with stran e loo!snot only with their !hiloso!hical im!ortance but also with their esthetic charmwas not .ust some unique little neurotic obsession of mine, but could well be infectious, if only I could et

across to my students that these notions were anythin but dull and dry, as in those fri id, sterile lo ic courses I1d taken, but ratheras 8a el and 8ewman had hintedwere intimately related to a slew of !rofound and beautiful ideas in mathematics, !hysics, com!uter science, !sycholo y, !hiloso!hy, lin uistics, and so on& I ave my course the half-di!!y, half-romantic title -The /ystery of the Fndecidable- in the ho!es that " mi ht attract students from wildly diverse areas, and the trick worked& Twenty-five souls were sna ed, and all were enthusiastic& I vividly remember the lovely blossoms I could see out the window each day as I lectured that s!rin , but even more vividly I remember :avid Justman, who was in art history, Scott 0uresh, who was in !olitical science, and )vril 6reenber , who was an art ma.or& These three sim!ly devoured the ideas, and we talked and talked endlessly about them& /y course thus turned out very well, both for the sna ees and for the sna er& Sometime durin the summer of "#$%," made a stab at sketchin out a table of contents for my -!am!hlet-, and at that !oint, the ambitiousness of my !ro.ect started dawnin on me, but it still felt more like a !am!hlet than a tome to me& It was only in the fall that I started writin in earnest& I had never written anythin more than a few !a es lon , but I fearlessly !lun ed ahead, fi urin it would take me .ust a few daysmaybe a week or two& I was sli htly off, for in fact, the very first draft 4done in !en, .ust like my letter to ;obert, but with more cross-outs5 took me about a montha month that overla!!ed in time with the -Hom @i!!ur war-, which made a very dee! im!ression on me& I reali<ed this first draft was not the final !roduct, but I felt I had done the ma.or work and now it was .ust a question of revision&

E-!eriments 'ith Literary +orm Start to Ta.e ,lace

)s I was writin that draft, I certainly wasn1t thinkin about Escher !ictures& 8or was I thinkin about 0ach1s music& 0ut one day I found myself on fire with ideas about mind, brain, and human identity, and so, shamelessly borrowin Aewis 7arroll1s odd cou!le of )chilles and the Tortoise, whose droll !ersonalities amused me no end, I sat down and in absolute white heat dashed off a lon , com!le( dialo ue, all about a fictitious, unima inably lar e book each of whose !a es, on a one-by-one basis, contained e(haustive information on one s!ecific neuron in Einstein1s brain& )s it ha!!ened, the dialo ue featured a short section where the two characters ima ined each other in another dialo ue, and each of them said, -Hou mi ht then say this&&& to which I mi ht well re!ly as follows&&& and then you would o on&&&- and so forth& 0ecause of this unusual structural feature, after I1d finally !ut the final !eriod on the final s!eech, I fli!!ed back to the to! of !a e one and there, on a whim, ty!ed out the sin le word -'F6FE-& /y Einstein-book dialo ue was not really a fu ue, of coursenot even close and yet it somehow reminded me of one& 'rom earliest childhood, I had been !rofoundly moved by the music of 0ach, and this off-the-wall idea of marryin 0ach-like contra!untal forms to lively dialo ues with intellectually rich content rabbed me with a !assion& Ever the ne(t few weeks, as I tossed the idea around in my head, I reali<ed how much room for !lay there was alon these lines, and I could ima ine how voraciously I as a teen-a er mi ht have consumed such dialo ues& Thus I was led to the idea of insertin contra!untal dialo ues every so often, !artly to break the tedium of the heavy ideas in my

cha!ters, and !artly to allow me to introduce li hter, more alle orical versions of all the abstruse conce!ts& The lon and the short of it is that I eventually decidedbut this took many monthsthat the o!timal structure would be a strict alternation between cha!ters and dialo ues& Ence that was clear, then I had the .oyous task of tryin to !in!oint the most crucial ideas that I wanted to et across to my readers and then somehow embodyin them in both the form and the content of fanciful, often !unnin dialo ues between )chilles and the Tortoise 4!lus a few new friends5&

GEB Is +irst Cooled off) Then $eheated

In early "#$C, I switched Ph&:& advisors for the fourth and final time, takin on a totally unfamiliar !roblem in solid-state !hysics that smelled very sweet thou h it threatened thorniness& /y new advisor, 6re ory 3annier, wanted me to !lun e in dee!ly, and I knew in my ut that this time it was sink or swim for me in the world of !hysics& If I wanted a Ph&:&a !recious but horribly elusive oal toward which I had been stru lin for almost a decade by thenit was now or never2 )nd so, with reat reluctance, I stowed my beloved manuscri!t in a desk drawer and told myself, -?ands off2 )nd no !eekin 2- I even instituted food-de!rivation !unishments if I so much as o!ened the drawer and riffled throu h my book-in-the-makin & Thinkin GEB thou hts or rather, GT*T,B thou htswas strictly #er(oten! S!eakin of 6erman, 3annier was scheduled to o to 6ermany for a si(-month !eriod in the fall of "#$C, and since I had always loved Euro!e, I asked if there was any way I could o alon & Gery kindly, he arran ed for me to be a wissenschaftlicher *ssistent essentially a teachin assistantin !hysics at the FniversitNt ;e ensbur , and so that1s what I did for one semester s!annin the end of "#$C and the start of "#$L& It was then that I ot most of the work done for my Ph&:& thesis& Since I had no close friends, my ;e ensbur days and ni hts were lon and lonely& In a !eculiar sense, my closest friend durin that tou h !eriod was 'rKdKric 7ho!in, since I tuned in to ;adio 3arsaw nearly every ni ht at midni ht and listened to various !ianists !layin many of his !ieces that I knew and loved, and others that were new to me and that I came to love& That whole stretch was GEB%#er(oten time, and thus it continued until the end of "#$L, when finally I closed the book on my thesis& )lthou h that work was all about an e(quisite visual structure 4see 7ha!ter L of this book5 and seemed to offer a ood launch!ad for a career, I had suffered too many blows to my e o in raduate school to believe I would make a ood !hysicist& En the other hand, the rekindlin of old intellectual flames and es!ecially the writin of GT*T,B had breathed a new kind of selfconfidence into me& Jobless but hi hly motivated, I moved to my home town of Stanford, and there, thanks to my !arents1 unquestionin and enerous financial su!!ort 4-a two-year ?ofstadter 'ellowshi!-, I .okin ly called it5, I set out to -retool myself- as an artificialintelli ence researcher& Even more im!ortant, thou h, was that I was able to resume my !assionate love affair with the ideas that had so rabbed me a cou!le of years earlier& )t Stanford, my erstwhile -!am!hlet- bloomed& It was rewritten from start to finish, because I felt that my earlier drafts, thou h focused on the !ro!er ideas, were immature and inconsistent in style& )nd I en.oyed the lu(ury of one of the world1s earliest and best word-!rocessin !ro rams, my new friend Pentti @anerva1s tremendously

fle(ible and user-friendly TG-Edit& Thanks to that !ro ram, the new version .ust flowed out, and ever so smoothly& I .ust can1t ima ine how GEB could have been written without it& Enly at this sta e did the book1s unusual stylistic hallmarks really emer ethe sometimes-silly !layin with words, the concoctin of novel verbal structures that imitate musical forms, the wallowin in analo ies of every sort, the s!innin of stories whose very structures e(em!lify the !oints they are talkin about, the mi(in of oddball !ersonalities in fantastic scenarios& )s I was writin , I certainly knew that my book would be quite different from other books on related to!ics, and that I was violatin quite a number of conventions& 8onetheless I blithely continued, because I felt confident that what I was doin sim!ly had to be done, and that it had an intrinsic ri htness to it& Ene of the key qualities that made me so believe in what I was doin is that this was a book in which form was bein iven equal billin with contentand that was no accident, since GEB is in lar e !art about how content is inse!arable from form, how semantics is of a !iece with synta(, how ine(tricable !attern and matter are from each other& )lthou h I had always known of myself that, in many as!ects of life, I was concerned as much with form as with content, I had never sus!ected how dee!ly I would et cau ht u!, in the writin of my first book, in matters of visual a!!earance on all levels& Thus, thanks to the ease of usin TG-Edit, whatever I wrote underwent !olishin to make it look better on the screen, and thou h such control would at one time have been considered a lu(ury for an author, I was very attached to it and loath to ive it u!& 0y the time I had a solid version of the manuscri!t ready to send out to !ublishers, visual desi n and conce!tual structure were intimately bound u! with each other&

The Clarion Call

I1ve oft been asked if I, an unknown author with an unorthodo( manuscri!t and an off-the-wall title, had to stru le for years a ainst the monolithic !ublishin industry1s fear of takin risks& 3ell, !erha!s I was .ust lucky, but my e(!erience was far more !leasant than that& In mid-"#$$, I sent out a little sam!le to about fifteen hi h-quality !ublishers, .ust as a feeler, to which most re!lied !olitely that this was -not the ty!e of thin - they dealt in& 'air enou h& 0ut three or four e(!ressed interest in seein more, and so, by turns, I let them take a look at the whole thin & 8eedless to say, I was disa!!ointed when the first two turned it down 4and in each case the vettin !rocess took a few months, so the loss of time was frustratin 5, but on the other hand, I wasn1t overly disheartened& Then near 7hristmastime, /artin @essler, head of 0asic 0ooks, a !ublishin outfit I had always admired, ave me some ho!eful thou h tentative si nals& The winter of "#$$-$* was so severe that Indiana Fniversity, where I was now a fled lin assistant !rofessor, ran out of coal for heatin , and in /arch the university was forced to close down for three weeks to wait for warmer weather& I decided to use this free time to drive to 8ew Hork and !oints south to see old friends& 7lear as a bell in my oft-blurry memory is my brief sto! in some din y little diner in the town of 7larion, Pennsylvania, where from a chilly !hone booth I made a quick call to /artin @essler in 8ew Hork to see if he had a verdict yet& It was a reat moment in my life when he said he would be -deli hted- to work with meand it1s almost eerie to think that this si nal event occurred in that well-named hamlet, of all !laces&&&

$e"enge of the /oley $ollers

8ow that I had found a !ublisher, there came the question of turnin the manuscri!t from crude com!uter !rintout to a finely ty!eset book& It was a !iece of true luck that Pentti, to enhance TG-Edit, had .ust develo!ed one of the world1s first com!uter ty!esettin systems, and he stron ly encoura ed me to use it& @essler, ever the adventurer, was also willin to ive it a try!artly, of course, because it would save 0asic 0ooks some money, but also because he was by nature a shrewd risk-taker& :o-it-yourself ty!esettin , thou h for me a reat break, was hardly a !iece of cake& 7om!utin then was a lot more !rimitive than it is today, and to use Pentti1s system, I had to insert into each cha!ter or dialo ue literally thousands of cry!tic ty!esettin commands, ne(t cho! each com!uter file into several small !iecesfive or si( !er file, usuallyeach of which had to be run throu h a series of two com!uter !ro rams, and then each of the resultin out!ut files had to be !unched out !hysically as a cry!tic !attern of myriad holes on a lon , thin roll of !a!er ta!e& I myself had to walk the B++ yards to the buildin where the hole-!uncher was located, load the !a!er ta!e, and sit there monitorin it carefully to make sure it didn1t .am& 8e(t, I would carry this batch of oily ta!es another quarter-mile to the buildin where The -tanford .ail) was !rinted, and if it was free, I would use their !hototy!esettin machine myself& :oin so was a lon , elaborate o!eration involvin cartrid es of !hotosensitive !a!er, darkrooms, chemical baths with rollers throu h which the !a!er had to be !assed to et all the develo!in chemicals off, and clotheslines on which all the five-foot lon alleys with my te(t on them would be hun out to dry for a day or two& The !rocess of actually seeing what my thousands of ty!esettin commands had wrou ht was thus enormously unwieldy and slow& Truth to tell, thou h, I didn1t mind that= in fact, it was arcane, s!ecial, and kind of e(citin & 0ut one day, when nearly all the alleys had been !rintedtwo to three hundred of themand I thou ht I was home free, I made a horrendous discovery& I1d seen each one emer e with .et-black !rint from the develo!in baths, and yet on some of the more recently-dried ones, the te(t looked brownish& 3hat2? )s I checked out others, sli htly older, I saw li ht-brown !rint, and on yet older ones, it was oran y, or even !ale yellow2 I couldn1t believe it& ?ow in the world had this ha!!ened? The sim!le answer left me feelin so an ry and hel!less, the a in rollers, havin worn unevenly, no lon er wi!ed the alleys clean, so acid was day by day eatin the black !rint away& 'or the .ail)'s !ur!oses, this didn1t matterthey chucked their alleys in a matter of hoursbut for a book, it s!elled disaster& 8o way could a book be !rinted from yellow alleys2 )nd the !hotoco!ies I1d made of them when they were newborn were shar!, but not shar! enou h& 3hat a ni htmare2 Fntold labor had .ust one u! in smoke& I was filled with the des!air of a football team that1s .ust made a ##-yard downfield march only to be sto!!ed dead on the o!!onent1s one-yard line& I1d s!ent almost all summer "#$* !roducin these alleys, but now summer was drawin to a close, and I had to o back to Indiana to teach courses& 3hat on earth to do? ?ow could I salva e GEB? The only solution I could see was, on my own money, to fly back to Stanford every weekend of the fall, and redo the whole thin from scratch& Auckily, I was teachin only on Tuesdays and Thursdays, and so each Thursday afternoon I would <oom from class, catch a !lane, arrive at Stanford, work like a maniac until /onday afternoon, and then dash off to the air!ort to return to Indiana& I will never for et

the worst of those weekends, when I somehow mana ed to work for forty hours strai ht without a wink of slee!& That1s love for you2 In this ordeal there was a savin race, thou h, and it was this, I ot to correct all the ty!esettin errors I1d made in the first batch of alleys& The ori inal !lan had been to use a bunch of correction alleys, which would have had to be sliced u! into little !ieces in 0asic1s 8ew Hork offices and !asted in wherever there were litchesand in that first batch I1d made litches alore, that1s for sure& Such a !rocess would !robably have resulted in hundreds of errors in the layout& 0ut thanks to my ##-yard drive havin been halted at the one-yard line, I now had the chance to undo all these litches, and !roduce a nearly !ristine set of alleys& )nd thus, althou h the chemical catastro!he delayed the actual !rintin of GEB for a cou!le of months, it turned out, in retros!ect, to have been a blessin in dis uise&

0o!s111
There were of course many ideas that vied with each other for entry into the book takin sha!e durin those years, and some made it in while others did not& Ene of the ironies is that the Einstein-book dialo ue, which in its -fu ality- was the ins!iration for all dialo ues to come, was cho!!ed& There was another lon and intricate dialo ue, too, that was cho!!ed, or more accurately, that wound u! ettin transmo rified nearly beyond reco nition, and its curious story is connected with an intense debate that was ra in inside my brain at that time& I had been made acutely aware, by some leaflets I1d read in the student union at Ere on in "#$+, of se(ist lan ua e and its insidious unconscious effects& /y mind was awakened to the subtle ways that eneric -he- and -man- 4and a host of similar words and !hrases5 contribute to the sha!in of one1s sense of what is a -normal- human bein and what is an -e(ce!tion-, and I welcomed this new !ers!ective& 0ut " was not a writer at that timeI was a !hysics rad studentand these issues didn1t seem all that close to my own life& 3hen I started writin dialo ues, thou h, thin s chan ed& There came a !oint when it dawned on me that the characters in my dialo ues)chilles, the Tortoise, the 7rab, the )nteater, and a cou!le of others with cameo roleswere without e(ce!tion males! I was shocked at my own havin fallen victim to the unconscious !ressures !ushin a ainst the introduction of female characters& )nd yet, when I toyed with the idea of oin back and !erformin a -se(-chan e o!eration- on one or more of these characters, that really rubbed me the wron way& ?ow come? 3ell, all I could tell myself was, -0rin in females and you wind u! im!ortin the whole confusin world of se(uality into what is essentially a !urely abstract discussion, and that would distract attention from my book1s main !ur!oses&- This nonsensical view of mine stemmed from and echoed many tacit assum!tions of western civili<ation at that time 4and still today5& )s I forced myself to ra!!le with my own u ly attitude, a real battle started u! in my mind, with one side of me ar uin for oin back and makin some characters female, and the other tryin to maintain the status "uo! Eut of this internal battle suddenly came a lon and rather amusin dialo ue in which my various characters, havin come to the reali<ation that they are all males, discuss why this mi ht be so, and decide that, des!ite their sense of havin free will, they must in fact be merely characters in the mind of some se(ist male author& Ene way or

another, they mana e to summon this )uthor character into their dialo ueand what does he do when accused of se(ism? ?e !leads innocent, claimin that what his brain does is out of his controlthe blame for his se(ism must instead fall on a se(ist 6od& )nd the ne(t thin you know, 6od !oofs into the dialo ueand uess what? She turns out to be female 4ho ho ho5& I don1t remember the details of how it went on from there, but the !oint is, I was dee!ly torn, and I was ra!!lin in my own way with these com!le( issues& To my re retthat is to say, to the re ret of the me of the years that followed the side that wound u! winnin this battle was the se(ist side, with .ust a few concessions to the other side e!g!, the tower of :.inns in the dialo ue -Aittle ?armonic Aabyrinth-, and )unt ?illary in -Prelude&&& )nt 'u ue-5& GEB remained a book with a dee! se(ist bias sewn into its fabric& Interestin ly, it is a bias that very few readers, females or males, have commented on 4which in turn su!!orts my belief that these kinds of thin s are very subtle and insidious, and esca!e nearly everyone1s !erce!tion5& )s for eneric -man- and -he-, I certainly disliked those usa es at that time, and I tried to avoid them whenever I could 4or rather, whenever it was eas)/, but on the other hand I wasn1t !articularly concerned about cleansin my !rose of every last one of them, and as a consequence the book1s !a es are also marred, here and there, by that more obvious, more e(!licit form of se(ism& Today, I crin e whenever I come across sentences in GEB that talk about the reader as -he-, or that casually s!eak of -mankind- as if humanity were some hu e abstract gu)! Ene lives and learns, I uess& )nd lastly, as for that soul-searchin dialo ue in which the )uthor and 6od are summoned u! by )chilles and com!any to face the accusation of se(ism, well, that was somehow transformed, in a series of many, many small chan es, into the dialo ue with which GEB concludes, -Si(-!art ;icercar-& If you read it with its enesis in mind, you may find an e(tra level of interest&

Mr1 Tortoise) Meet Madame Tortue

) few years later, a wholly une(!ected chance came alon to make amends, at least in !art, for my se(ist sin& That o!!ortunity was afforded me by the challen e of translatin GEB into various forei n lan ua es& 3hen I was writin the book, the idea that it mi ht someday a!!ear in other lan ua es never crossed my mind& I don1t know why, since I loved lan ua es and loved translation, but somehow it .ust never occurred to me& ?owever, as soon as the idea was !ro!osed to me by my !ublisher, I was very e(cited about seein my book in other lan ua es, es!ecially ones that I s!oke to some e(tentmost of all 'rench, since that was a lan ua e that I s!oke fluently and loved very dee!ly& There were a million issues to consider in any !otential translation, since the book is rife not only with e(!licit word!lay but also with what Scott @im dubbed -structural !uns-!assa es where form and content echo or reinforce each other in some une(!ected manner, and very often thanks to ha!!y coincidences involvin s!ecific En lish words& 0ecause of these intricate medium-messa e tan les, I !ainstakin ly went throu h every last sentence of GEB, annotatin a co!y for translators into any lan ua e that mi ht be tar eted& This took me about a year of on-a ain, off-a ain toil, but finally it was done, and .ust in the nick of time, because contracts with forei n !ublishers started flowin thick and fast around "#*B& I could write a short booka !am!hlet?on the

cra<y, deli htful, knotty !u<<les and dilemmas that arose in translatin GEB, but here I will mention .ust onehow to render the sim!le-seemin !hrase -/r& Tortoise- in 'rench& 3hen in the s!rin of "#*%, Jacqueline ?enry and 0ob 'rench, the book1s e(cellent translators into 'rench, be an to tackle the dialo ues, they instantly ran headlon into the conflict between the feminine ender of the 'rench noun tortue and the masculinity of my character, the Tortoise& 0y the way, I must ruefully mention that in the marvelous but little-known Aewis 7arroll dialo ue from which I borrowed these deli htful characters 4re!rinted in GEB as 1Two-!art Invention-5, the Tortoise turns out, if you look carefully, never to have been attributed either ender& 0ut when I first read it, the question never entered my mind& This was clearl) a he-tortoise& Etherwise, I would have known not only that it was female but also wh) it was female& )fter all, an author only introduces a female character for some s!ecial reason, ri ht? 3hereas a male character in a -neutral- conte(t 4e!g!, !hiloso!hy5 needs no raison d'0tre, a female does& )nd so, iven no clue as to the Tortoise1s se(, I unthinkin ly and uncritically envisa ed it as a male& Thus does se(ism silently !ervade well-meanin but susce!tible brains& 0ut let1s not for et Jacqueline and 0ob2 )lthou h they could sim!ly have blud eoned their way throu h the !roblem by inventin a -/onsieur Tortue- character, that route felt distinctly unnatural in 'rench, to their taste, and so, in one of our many e(chan es of letters, they rather in erly asked me if I would ever consider lettin them switch the Tortoise1s se( to female& To them, of course, it !robably seemed !retty farfetched to ima ine that the author would even ive such a !ro!osal the time of day, but as a matter of fact, the moment I read their idea, I sei<ed u!on it with reat enthusiasm& )nd as a result, the 'rench GEB's !a es are raced throu hout with the fresh, fantastic fi ure of /adame Tortue, who runs !erverse intellectual circles around her male com!anion )chilles, erstwhile 6reek warrior and amateur !hiloso!her& There was somethin so deli htful and ratifyin to me about this new vision of -the Tortoise- that I was ecstatic with her& 3hat !articularly amused me were a few bilin ual conversations that I had about the Tortoise, in which I would start out in En lish usin the !ronoun -he-, then switch to 'rench and to elle as well& Either !ronoun felt !erfectly natural, and I even felt I was referrin to the selfsame -!erson- in both lan ua es& In its own funny way, this seemed faithful to 7arroll1s tortoise1s se(ual neutrality& )nd then, redoublin my !leasure, the translators into Italian, another lan ua e that I adored and s!oke quite well, chose to follow suit and to convert my -/r& Tortoiseinto -si norina Tartaru a-& Ef course these radical switches in no way affect the !erce!tions of GEB1s !urely an lo!hone readers, but in some small way, I feel, they hel! to make u! for the lamentable outcome of my internal battle of a few years earlier&

2en Buddhism) 3ohn Cage) and My 4oguish Irrationality

The 'rench translation was reeted, overall, very favorably& Ene s!ecially ratifyin moment for 0ob, Jacqueline, and myself was when a truly lowin full-!a e review by Jacques )ttali a!!eared in the most !resti ious 'rench news!a!er, +e 'onde, not .ust !raisin the book for its ideas and style, but also makin a !articular !oint of !raisin its translation&

) few months later, I received a !air of reviews !ublished in successive issues of ,umanisme, an obscure .ournal !ut out by the Society of 'rench 'reemasons& 0oth had issued from the !en of one author, )lain ?oulou, and I tackled them with interest& The first one was quite len thy and, like that in +e 'onde, lowed with !raise= I was ratified and rateful& I then went on to the second review, which started out with the !oetic !hrase *pr1s les roses, les 2pines!!! 4-)fter roses, thorns&&&-5, and which then !roceeded for several !a es, to my ama<ement, to ri! GEB a!art as un pi1ge tr1s gra#e 4-a very dan erous tra!-5 in which the mindless bandwa on of 9en 0uddhism was ea erly .um!ed on, and in which a rabidly antiscientific, beatnik-influenced, hi!!ie-like irrationality ty!ical of )merican !hysicists was embraced as the su!reme !ath to enli htenment, with the iconoclastic 9en-influenced )merican com!oser John 7a e as the !atron saint of it all& )ll I could do was chuckle, and throw my hands u! in bewilderment at these Tatiesque #acarmes de monsieur ,oulou! Somehow, 4his reviewer saw me !raisin 7a e to the skies 4-6>del, Escher, 7a e-?5 and mana ed to read into my coy allusions to and minor borrowin s from 9en an uncritical acce!tance thereof, which in fact is not at all my stance& )s I declare at the start of 7ha!ter #, I find 9en not only confusin and silly, but on a very dee! level utterly inimical to my core beliefs& ?owever, I also find 9en1s sillinesses!ecially when it ets reall) sillyquite amusin , even refreshin , and it was sim!ly fun for me to s!rinkle a bit of Eastern s!ice into my basically very 3estern casserole& ?owever, my havin s!rinkled little traces of 9en here and there does not mean that I am a 9en monk in shee!1s clothin & )s for John 7a e, for some odd reason I had felt very sure, u! till readin ?oulou1s weird about-face, that in my -7anon by Intervallic )u mentation- and the cha!ter that follows it, I had unambi uously hea!ed scorn on 7a e1s music, albeit in a somewhat res!ectful manner& 0ut wait, wait, waitisn1t -hea!in res!ectful scorn- not a contradiction in terms, indeed a !atent im!ossibility? )nd doesn1t such coy flirtin with self-contradiction and !arado( demonstrate, e(actly as ?oulou claims, that I am, dee! down, both antiscientific and !ro-9en, after all? 3ell, so be it& Even if I feel my book is as often misunderstood as understood, I certainly can1t com!lain about the si<e or the enthusiasm of its readershi! around the world& The ori inal En lish-lan ua e GEB was and continues to be very !o!ular, and its translated selves hit the bestseller lists in 4at least5 'rance, ?olland, and Ja!an& The 6erman GEB, in fact, occu!ied the O" rank on the nonfiction list for somethin like fi#e months durin "#*L, the %++th birthyear of J& S& 0ach& It seems a bit absurd to me& 0ut who knowsthat anniversary, aided by the other 6ermanic names on the cover, may have crucially s!arked GEB's !o!ularity there& GEB has also been lovin ly translated into S!anish, Italian, ?un arian, Swedish, and Portu uese, and!erha!s une(!ectedlywith reat virtuosity into 7hinese& There is also a fine ;ussian version all ready, .ust waitin in the win s until it finds a !ublisher& )ll of this far transcends anythin I ever e(!ected, even thou h I can1t deny that as I was writin it, es!ecially in those heady Stanford days, I had a rowin inner feelin that GEB would make some sort of s!lash&

My Subsequent Intellectual ,ath5

ecade 6

Since sendin GEB off to the !rinters two decades a o, I1ve somehow mana ed to kee! myself !retty busy& )side from strivin , with a team of e(cellent raduate students, to develo! com!uter models of the mental mechanisms that underlie analo y and creativity, I1ve also written several further books, each of which I1ll comment on here, thou h only very briefly& The first of these, a!!earin in late "#*", was The 'ind's 3, an antholo y coedited with a new friend, !hiloso!her :aniel :ennett& Eur !ur!ose, closely related to that of GEB, was to force our readers to confront, in the most vivid and even .oltin manner, the fundamental conundrum of human e(istence, our dee! and almost ineradicable sense of !ossessin a unique -I--ness transcendin our !hysical bodies and mysteriously enablin us to e(ercise somethin we call -free will-, without ever quite knowin .ust what that is& :an and I used stories and dialo ues from a motley crew of e(cellent writers, and one of the !leasures for me was that I finally ot to see my Einstein-book dialo ue in !rint, after all& :urin the years "#*"-"#*%, I had the o!!ortunity to write a monthly column for -cientific *merican, which I called -/etama ical Themas- 4an ana ram of -/athematical 6ames-, the title of the wonderful column by /artin 6ardner that had occu!ied the same slot in the ma a<ine for the !recedin BL years5& )lthou h the to!ics I dealt with in my column were, on their surface, all over the ma!, in some sense they were unified by their incessant quest for -the essence of mind and !attern-& I covered such thin s as !attern and !oetry in the music of 7ho!in= the question of whether the enetic code is arbitrary or inevitable= strate ies in the never-endin battle a ainst !seudoscience= the boundary between sense and nonsense in literature= chaos and stran e attractors in mathematics= ame theory and the Prisoner1s :ilemma= creative analo ies involvin sim!le number !atterns= the insidious effects of se(ist lan ua e= and many other to!ics& In addition, stran e loo!s, self-reference, recursion, and a closely related !henomenon that I came to call -lockin -in- were occasional themes in my columns& In that sense, as well as in their wanderin throu h many disci!lines, my -/etama ical Themas- essays echoed the flavor of GEB! )lthou h I sto!!ed writin my column in "#*%, I s!ent the ne(t year !ullin to ether the essays I1d done and !rovidin each of them with a substantial -Post Scri!tum-= these BL cha!ters, alon with ei ht fresh ones, constituted my "#*L book 'etamagical Themas: 4uesting for the Essence of 'ind and &attern! Ene of the new !ieces was a rather <any )chilles-Tortoise dialo ue called -3ho Shoves 3hom )round Inside the 7areenium?-, which I feel ca!tures my !ersonal views on self, soul, and the infamous -I--wordnamely, -I-2!erha!s better than anythin else I1ve writtenmaybe even better than GEB does, thou h that mi ht be oin too far& 'or several years durin the "#*+1s, I was afflicted with a severe case of ambi rammitis-, which I cau ht from my friend Scott @im, and out of which came my "#*$ book *m(igrammi! )n ambi ram 4or an -inversion-, as Scott calls them in his own book, 3n#ersions/ is a calli ra!hic desi n that mana es to squee<e two different readin s into the selfsame set of curves& I found the idea charmin and intellectually fascinatin , and as I develo!ed my own skill at this odd but ele ant art form, I found that selfobservation ave me many new insi hts into the nature of creativity, and so *m(igrammi,

aside from showcasin some B++ of my ambi rams, also features a te(tin fact, a dialo uethat is a lon , wanderin meditation on the creative act, centered on the makin of ambi rams but branchin out to include musical com!osition, scientific discovery, creative writin , and so on& 'or reasons not worth oin into, *m(igrammi: 5n microcosmo ideale per lo studio della creati#it6 was !ublished only in Italian and by a tiny !ublisher called ?o!eful /onster, and I re ret to say that it is no lon er available&

My Subsequent Intellectual ,ath5

ecade II

)s I said above, writin , thou h crucial, was not my only intellectual focus= research into co nitive mechanisms was an equally im!ortant one& /y early hunches about how to model analo y and creativity are actually set forth quite clearly in GEB's 7ha!ter "#, in my discussion of 0on ard !roblems, and althou h those were .ust the erms of an actual architecture, I feel it is fair to say that des!ite many years of refinement, most of those ideas can be found in one form or another in the models develo!ed in my research rou! at Indiana Fniversity and the Fniversity of /ichi an 4where I s!ent the years "#*C-"#**, in the Psycholo y :e!artment5& )fter a decade and a half of develo!ment of com!uter models, the time seemed ri!e for a book that would !ull all the main threads to ether and describe the !ro rams1 !rinci!les and !erformance in clear and accessible lan ua e& Thus over several years, 7luid 8oncepts and 8reati#e *nalogies took sha!e, and finally a!!eared in !rint in "##L& In it are !resented a series of closely related com!uter !ro ramsSeek-3hence, Jumbo, 8umbo, 7o!ycat, Tableto!, and 4still in !ro ress5 /etacat and Aetter S!iritto ether with !hiloso!hical discussions that attem!t to set them in conte(t& Several of its cha!ters are co-authored by members of the 'luid )nalo ies ;esearch 6rou!, and indeed ');6 ets its !ro!er billin as my collective co-author& The book shares much with GEB, but !erha!s most im!ortant of all is the basic !hiloso!hical article of faith that bein an -I- in other words, !ossessin a sense of self so dee! and ineradicable that it blurs into causalityis an inevitable concomitant to, and in redient of, the fle(ibility and !ower that are synonymous with intelligence, and that the latter is but another term for conceptual fle9i(ilit), which in turn means meaningful s)m(ols! ) very different strand of my intellectual life was my dee! involvement in the translation of GEB into various lan ua es, and this led me, !erha!s inevitably, in retros!ect, to the territory of verse translation& It all started in "#*$ with my attem!t to mimic in En lish a beautiful 'rench miniature by si(teenth-century 'rench !oet 7lement /arot, but from there it s!un off in many directions at once& To make a lon story short, I wound u! writin a com!le( and dee!ly !ersonal book about translation in its most eneral and meta!horical sense, and while writin it, I e(!erienced much the same feelin of e(hilaration as I had twenty years earlier, when writin GEB! This book, +e Ton (eau de 'arot: 3n &raise of the 'usic of +anguage, winds throu h many diverse terrains, includin what it means to -think in- a iven lan ua e 4or a blur of lan ua es5= how constraints can enhance creativity= how meanin erminates, buds, and flowers in minds and mi ht someday do so in machines= how words, when !ut to ether into com!ounds, often melt to ether and lose some or all of their identity= how a lan ua e s!oken on a neutron star mi ht or mi ht not resemble human lan ua es= how !oetry written hundreds of years a o should be rendered today= how translation is

intimately related to analo y and to the fundamental human !rocess of understandin one another= what kinds of !assa es, if any, are intrinsically untranslatable= what it means to translate nonsense !assa es from one lan ua e to another= the absurdity of su!!osin that today1s mostly money-driven machine-translation immickry could handle even the sim!lest of !oetry= and on and on& The two middle cha!ters of +e Ton (eau de 'arot are devoted to a work of fiction that I had recently fallen in love with, )le(ander Pushkin1s novel in verse, Eugene :negin! I first came into contact with this work throu h a cou!le of En lish translations, and then read others, always fascinated by the translators1 different !hiloso!hies and styles& 'rom this first flame of e(citement, I slowly was drawn into tryin to read the ori inal te(t, and then somehow, des!ite havin a !oor command of ;ussian, I could not !revent myself from tryin to translate a stan<a or two& Thus started a sli!!ery slo!e that I soon slid down, eventually stunnin myself by devotin a whole year to recreatin the entire novelnearly C++ s!arklin sonnetsin En lish verse& Ef course, durin that time, my ;ussian im!roved by lea!s and bounds, thou h it still is far from conversationally fluent& )s I write, my :negin has not yet come out, but it will be a!!earin at .ust about the same time in "### as the book you are holdin the twentiethanniversary version of Gdel, Escher, Bach! )nd the year "### !lays an equally im!ortant role in my E:'s creation, bein the B++th birthyear of )le(ander Pushkin&

+or'ard-loo.ing and Bac.'ard-loo.ing Boo.s

+e Ton (eau de 'arot is a bit lon er than GEB, and on its first !a e, I o out on a limb and call it -!robably the best book I will ever write-& Some of my readers will maintain that GEB is su!erior, and I can see why they mi ht do so& 0ut it1s so lon since I wrote GEB that !erha!s the ma ical feelin I had when writin it has faded, while the ma ic of +eT(' is still vivid& Still, there1s no denyin that, at least in the short run, +eT(' has had far less im!act than GEB did, and I confess that that1s disa!!ointed me quite a bit& Permit me to s!eculate for a moment as to why this mi ht be the case& In some sense, GEB was a -forward-lookin - book, or at least on its surface it ave that a!!earance& /any hailed it as somethin like -the bible of artificial intelli ence-, which is of course ridiculous, but the fact is that many youn students read it and cau ht the bu of my own fascination with the modelin of mind in all of its elusive as!ects, includin the evanescent oals of -I- and free will and consciousness& )lthou h I am the furthest thin in the world from bein a futurist, a science-fiction addict, or a technolo y uru, I was often !i eonholed in .ust that way, sim!ly because I had written a lon treatise that dealt quite a bit with com!uters and their vast !otential 4in the most !hiloso!hical of senses5, and because my book was quite a hit amon youn !eo!le interested in com!uters& 3ell, by contrast, +e Ton (eau de 'arot mi ht be seen as a -backward-lookin book, not so much because it was ins!ired by a si(teenth-century !oem and deals with many other authors of the !ast, such as :ante and Pushkin, but because there sim!ly is nothin in the book1s !a es that could be confused with lib technolo ical lit< and surreal futuristic !romises& 8ot that GEB had those either, but many !eo!le seemed to see somethin va uely alon those lines in it, whereas there1s nothin of that sort to latch onto, in +eT('! In fact, some mi ht see it almost as technolo y-bashin , in that I take

many artificial-intelli ence researchers and machine-translation develo!ers to task for wildly e(a erated claims& I am not an enemy of these fields, but I am a ainst vast oversim!lifications and underestimations of the challen es that they re!resent, for in the end, that amounts to a vast underestimation of the human s!irit, for which I have the dee!est res!ect& )nyone who has read GEB with any care should have seen this same -backwardlookin - flavor !ermeatin the book, !erha!s most e(!licitly so in the key section -Ten Muestions and S!eculations- 4!!& D$D-D*+5, which is a very romantic way of lookin at the de!th of the human s!irit& )lthou h my !rediction about chess-!layin !ro rams !ut forth there turned out to be embarrassin ly wron 4as the world saw with :ee! 0lue versus @as!arov in "##$5, those few !a es nonetheless e(!ress a set of !hiloso!hical beliefs to which I am still committed in the stron est sense&

To Tam!er) or to Lea"e ,ristine7

6iven that I was quite wron in a !rediction made twenty years a o, why not rewrite the -Ten Muestions and S!eculations- section, u!datin it and talkin about how I feel in li ht of :ee! 0lue? 3ell, of course, this brin s u! a much lar er issue, that of revisin the "#$# book from to! to bottom, and comin out with a s!ankin new "### edition of GEB! 3hat mi ht militate for, and what mi ht militate a ainst, undertakin such a !ro.ect? I don1t deny that some deli htful, if small, im!rovements were made in the translated versions& 'or e(am!le, my ma istrally 0ach-savvy friend 0ernie 6reenber informed me that the -0)7? oblet- I had invented out of whole cloth in my dialo ue -7ontracrosti!unctus- actually e(ists2 The real oblet is not 4as in my dialo ue5 a !iece of lass blown by 0ach, but rather a ift from one of his !ri<e students= nonetheless, its key featurethat of havin the melody -0)7?- etched into the lass itselfis .ust as I said in the dialo ue2 This was such an ama<in coincidence that I rewrote the dialo ue for the 'rench version to reflect the real oblet1s e(istence, and insisted on havin a !hoto ra!h of the 0)7? oblet in the 'rench GEB! )nother delicious touch in the 'rench GEB was the re!lacement of the very formal, character-less !hoto of 6>del by a far more en a in sna!shot in which he1s in a s!iffy white suit and is strollin with some old cod er in a forest& The latter, decked out in a flo!!y hat and ba y !ants held u! by awky sus!enders, looks every inch the quintessential rube, so I rewrote the ca!tion as ;urt Gdel a#ec un pa)san non identifi2 4-@urt 6>del with unknown !easant-5& 0ut as anyone who has lived in the twentieth century can see in a s!lit second, the pa)san non identifi2 is none other than )& Einstein& 3hy not, then, incor!orate those amusin chan es into a revised edition in En lish? En a more substantial level, why not talk a bit about the !ioneerin artificialintelli ence !ro ram ?earsay II, whose very subtle architecture started e(ertin , only a year or two alter GEB came out, a vast im!act on my own com!uter models, and about which I already knew somethin way back in "#$D? 3hy not talk more about machine translation, and es!ecially its weaknesses? 3hy not have a whole cha!ter about the most !romisin develo!ments 4andIor e(a erated claims5 over the !ast two decades in artificial intelli encefeaturin my own research rou! as well as others? Er why not, as

some have su ested, come out with a 7:-;E/ with Escher !ictures and 0ach music on it, as well as recordin s of all of GEB's dialo ues as !erformed by to!-notch actors? 3ell, I can see the ar uments for any of these, but unfortunately, I .ust don1t buy them& The 7:-;E/ su estion, the one most often made to me, is the sim!lest to dismiss& I intended GEB as a book, not as a multimedia circus, and book it shall remain end of story& )s for the idea of revisin the te(t, however, that is more com!le(& 3here would one draw the line? 3hat would be sacrosanct? 3hat would survive, what would be tossed out? 3ere I to take that task on, I mi ht well wind u! rewritin every sin le sentenceand, let1s not for et, reverse-en ineerin old /r& T&&& Perha!s I1m .ust a cra<y !urist= !erha!s I1m .ust a la<y lout= but stubborn no doubt, and wouldn1t dream of chan in my book1s 5rte9t! That1s out2 Thus in my sternness, I won1t allow myself to add the names of two !eo!le:onald @ennedy and ?oward Edenber to my -3ords of Thanks-, des!ite the fact that for years, I1ve felt sad at havin inadvertently left them out& I won1t even correct the book1s ty!os 4and, to my cha rin, I did find, over the decades, that there are a few, aside from those listed e(!licitly under -ty!os- in the inde(52 3hy on earth am I such a stick-in-the-mud? 3hy not brin Gdel, Escher, Bach u! to date and make it a book worthy of usherin in the twenty-first centuryindeed, the third millennium?

8u9rendo In"enietis111
3ell, the only answer I can ive, other than that life is short, is that GEB was written in one sittin , so to s!eak& GEB was a clean and !ure vision that was dreamed by someone elsesomeone who, to be sure, was remarkably similar to yours truly, but someone who nonetheless had a sli htly different !ers!ective and a sli htly different a enda& GEB was that !erson1s labor of love, and as suchat least so say Iit should not be touched& Indeed, I somehow feel a stran e inner confidence that the true author of GEB, when one fine day he finally reaches m) ri!e a e, will tender to me the truest of thanks for not havin tam!ered with the vessel into which he !oured so much of his youn and ea er soulthe work that he even went so far as to call, in what some mi ht see as a cry!tic or even naPvely romantic remark, -a statement of my reli ion-& )t least I know what he meant& ;EMGIES7)T I8 7E8ST)8TI), E;6E, ;EP;QSE8T)TIE 7GPI:I )G7TE;IS ;EAI6IE8IS&

Contents
Everview Aist of Illustrations 3ords of Thanks Part I, 6E0 Introduction5 A Musico-Logical 0ffering Three%&art 3n#ention Cha!ter I5 The M&-!u::le Two%&art 3n#ention Cha!ter II5 Meaning and +orm in Mathematics -onata for 5naccompanied *chilles Cha!ter III5 +igure and Ground 8ontracrostipunctus Cha!ter I45 Consistency) Com!leteness) and Geometry +ittle ,armonic +a()rinth Cha!ter 45 $ecursi"e Structures and ,rocesses 8anon () 3nter#allic *ugmentation Cha!ter 4I5 The Location of Meaning 8hromatic 7antas), *nd 7eud Cha!ter 4II5 The ,ro!ositional Calculus 8ra( 8anon Cha!ter 4III5 Ty!ogra!hical ;umber Theory * 'u :ffering Cha!ter I<5 Mumon and G#del C+ DL D# $# *B #* "+" ""B ""# "C+ "DB "** "#% B"B B"L B%% B%D BDC B$$ BL %" %D

Part II, E60 &relude!!! Cha!ter <5 Le"els of escri!tion) and Com!uter Systems *nt 7ugue Cha!ter <I5 Brains and Thoughts English 7rench German -uite Cha!ter <II5 Minds and Thoughts *ria with .i#erse <ariations Cha!ter <III5 Bloo, and +loo, and Gloo, *ir on G's -tring Cha!ter <I45 0n +ormally &ndecidable ,ro!ositions of T;T and $elated Systems Birthda) 8antatatata!!! Cha!ter <45 3um!ing out of the System Edif)ing Thoughts of a To(acco -moker Cha!ter <4I5 Self-$ef and Self-$e! The 'agnificra(, 3ndeed Cha!ter <4II5 Church) Turing) Tars.i) and 0thers -,=.+5, To) of 'an's .esigning Cha!ter <4III5 Artificial Intelligence5 $etros!ects 8ontrafactus Cha!ter <I<5 Artificial Intelligence5 ,ros!ects -loth 8anon Cha!ter <<5 Strange Loo!s) 0r Tangled /ierarchies -i9%&art =icercar 8otes 0iblio ra!hy 7redits Inde( %+L %"C %%* %DB %*# %#B C"B CBD CL" CL# C*" C*L C## L"C LDD L$D D+" D+* DCD DLC D#B D#L $B# $L" $LC $D$ $D#

Everview
Part I, 6E0
Introduction5 A Musico-Logical 0ffering1 The book o!ens with the story of 0ach1s /usical Efferin & 0ach made an im!rom!tu visit to @in 'rederick the 6reat of Prussia, and was requested to im!rovise u!on a theme !resented by the @in & ?is im!rovisations formed the basis of that reat work& The /usical Efferin and its story form a theme u!on which I -im!rovise- throu hout the book, thus makin a sort of -/etamusical Efferin -& Self-reference and the inter!lay between different levels in 0ach are discussed, this leads to a discussion of !arallel ideas in Escher1s drawin s and then 6>del1s Theorem& ) brief !resentation of the history of lo ic and !arado(es is iven as back round for 6>del1s Theorem& This leads to mechanical reasonin and com!uters, and the debate about whether )rtificial Intelli ence is !ossible& I close with an e(!lanation of the ori ins of the book!articularly the why and wherefore of the :ialo ues& Three%&art 3n#ention! 0ach wrote fifteen three-!art inventions& In this three-!art :ialo ue, the Tortoise and )chillesthe main fictional !rota onists in the :ialo uesare -invented- by 9eno 4as in fact they were, to illustrate 9eno1s !arado(es of motion5& Gery short, it sim!ly ives the flavor of the :ialo ues to come& Cha!ter I5 The M&-!u::le1 ) sim!le formal system 4the /IA1-system5 is !resented, and the reader is ur ed to work out a !u<<le to ain familiarity with formal systems in eneral& ) number of fundamental notions are introduced, strin , theorem, a(iom, rule of inference, derivation, formal system, decision !rocedure, workin insideIoutside the system& Two%&art 3n#ention! 0ach also wrote fifteen two-!art inventions& This two-!art :ialo ue was written not by me, but by Aewis 7arroll in "*#L& 7arroll borrowed )chilles and the Tortoise from 9eno, and I in turn borrowed them from 7arroll& The to!ic is the relation between reasonin , reasonin about reasonin , reasonin about reasonin about reasonin , and so on& It !arallels, in a way, 9eno1s !arado(es about the im!ossibility of motion, seemin to show, by usin infinite re ress, that reasonin is im!ossible& It is a beautiful !arado(, and is referred to several times later in the book& Cha!ter II5 Meaning and +orm in Mathematics1 ) new formal system 4the !q-system5 is !resented, even sim!ler than the /IF-system of 7ha!ter I& )!!arently meanin less at first, its symbols are suddenly revealed to !ossess meanin by virtue of the form of the theorems they a!!ear in& This revelation is the first im!ortant insi ht into meanin , its dee! connection to isomor!hism& Garious issues related to meanin are then discussed, such as truth, !roof, symbol mani!ulation, and the elusive conce!t, -form-& -onata for 5naccompanied *chilles& ) :ialo ue which imitates the 0ach Sonatas for unaccom!anied violin& In !articular, )chilles is the only s!eaker, since it is a transcri!t of one end of a tele!hone call, at the far end of which is the Tortoise& Their conversation concerns the conce!ts of -fi ure- and - round- in various conte(tse& &, Escher1s art& The :ialo ue itself forms an e(am!le of the distinction, since )chilles1 lines form a -fi ure-, and the Tortoise1s lines im!licit in )chilles1 linesform a - round-&

Cha!ter III5 +igure and Ground1 The distinction between fi ure and round in art is com!ared to the distinction between theorems and nontheorems in formal systems& The question -:oes a fi ure necessarily contain the same information as its round?- leads to the distinction between recursively enumerable sets and recursive sets& 8ontracrostipunctus& This :ialo ue is central to the book, for it contains a set of !ara!hrases of 6>del1s self-referential construction and of his Incom!leteness Theorem& Ene of the !ara!hrases of the Theorem says, -'or each record !layer there is a record which it cannot !lay&- The :ialo ue1s title is a cross between the word -acrostic- and the word -contra!unctus-, a Aatin word which 0ach used to denote the many fu ues and canons makin u! his *rt of the 7ugue& Some e(!licit references to the )rt of the 'u ue are made& The :ialo ue itself conceals some acrostic tricks& Cha!ter I45 Consistency) Com!leteness) and Geometry1 The !recedin :ialo ue is e(!licated to the e(tent it is !ossible at this sta e& This leads back to the question of how and when symbols in a formal system acquire meanin & The history of Euclidean and non-Euclidean eometry is iven, as an illustration of the elusive notion of -undefined terms-& This leads to ideas about the consistency of different and !ossibly -rival- eometries& Throu h this discussion the notion of undefined terms is clarified, and the relation of undefined terms to !erce!tion and thou ht !rocesses is considered& +ittle ,armonic +a()rinth& This is based on the 0ach or an !iece by the same name& It is a !layful introduction to the notion of recursivei&e&, nested structures& It contains stories within stories& The frame story, instead of finishin as e(!ected, is left o!en, so the reader is left dan lin without resolution& Ene nested story concerns modulation in music!articularly an or an !iece which ends in the wron key, leavin the listener dan lin without resolution& Cha!ter 45 $ecursi"e Structures and ,rocesses1 The idea of recursion is !resented in many different conte(ts, musical !atterns, lin uistic !atterns, eometric structures, mathematical functions, !hysical theories, com!uter !ro rams, and others& 8anon () 3nter#allic *ugmentation& )chilles and the Tortoise try to resolve the question, -3hich contains more informationa record, or the !hono ra!h which !lays it? This odd question arises when the Tortoise describes a sin le record which, when !layed on a set of different !hono ra!hs, !roduces two quite different melodies, 0-)-7-? and 7-)-6-E& It turns out, however, that these melodies are -the same-, in a !eculiar sense& Cha!ter 4I5 The Location of Meaning1 ) broad discussion of how meanin is s!lit amon coded messa e, decoder, and receiver& E(am!les !resented include strands of :8), undeci!hered inscri!tions on ancient tablets, and !hono ra!h records sailin out in s!ace& The relationshi! of intelli ence to -absolute- meanin is !ostulated& 8hromatic 7antas), *nd 7eud& ) short :ialo ue bearin hardly any resemblance, e(ce!t in title, to 0ach1s 8hromatic 7antas) and 7ugue& It concerns the !ro!er way to mani!ulate sentences so as to !reserve truth-and in !articular the question of whether there e(ist rules for the usa e of the word -and-& This :ialo ue has much in common with the :ialo ue by Aewis 7arroll&

Cha!ter 4II5 The ,ro!ositional Calculus1 It is su ested how words such as -and- can be overned by formal rules& Ence a ain, the ideas of isomor!hism and automatic acquisition of meanin by symbols in such a system are brou ht u!& )ll the e(am!les in this 7ha!ter, incidentally, are -9entences--sentences taken from 9en kRans& This is !ur!osefully done, somewhat ton ue-incheek, since 9en kRans are deliberately illo ical stories& 8ra( 8anon& ) :ialo ue based on a !iece by the same name from the 'usical :ffering& 0oth are so named because crabs 4su!!osedly5 walk backwards& The 7rab makes his first a!!earance in this :ialo ue& It is !erha!s the densest :ialo ue in the book in terms of formal trickery and level!lay& 6>del, Escher, and 0ach are dee!ly intertwined in this very short :ialo ue& Cha!ter 4III5 Ty!ogra!hical ;umber Theory1 )n e(tension of the Pro!ositional 7alculus called -T8T- is !resented& In T8T, number-theoretical reasonin can be done by ri id symbol mani!ulation& :ifferences between formal reasonin and human thou ht are considered& * 'u :ffering& This :ialo ue foreshadows several new to!ics in the book& Estensibly concerned with 9en 0uddhism and kRans, it is actually a thinly veiled discussion of theoremhood and nontheoremhood, truth and falsity, of strin s in number theory& There are fleetin references to molecular biolo y!articularly the 6enetic 7ode& There is no close affinity to the 'usical :ffering, other than in the title and the !layin of self-referential ames& Cha!ter I<5 Mumon and G#del1 )n attem!t is made to talk about the stran e ideas of 9en 0uddhism& The 9en monk /umon, who ave well known commentaries on many kRans, is a central fi ure& In a way, 9en ideas bear a meta!horical resemblance to some contem!orary ideas in the !hiloso!hy of mathematics& )fter this -9ennery-, 6>del1s fundamental idea of 6>delnumberin is introduced, and a first !ass throu h 6>del1s Theorem is made&

Part II, E60

&relude&&& This :ialo ue attaches to the ne(t one& They are based on !reludes and fu ues from 0ach1s >ell%Tempered 8la#ier& )chilles and the Tortoise brin a !resent to the 7rab, who has a uest, the )nteater& The !resent turns out to be a recordin of the 3&T&7&= it is immediately !ut on& )s they listen to a !relude, they discuss the structure of !reludes and fu ues, which leads )chilles to ask how to hear a fu ue, as a whole, or as a sum of !arts? This is the debate between holism and reductionism, which is soon taken u! in the )nt 'u ue& Cha!ter <5 Le"els of escri!tion) and Com!uter Systems1 Garious levels of seein !ictures, chessboards, and com!uter systems are discussed& The last of these is then e(amined in detail& This involves describin machine lan ua es, assembly lan ua es, com!iler lan ua es, o!eratin systems, and so forth& Then the discussion turns to com!osite systems of other ty!es, such as s!orts teams, nuclei, atoms, the weather, and so forth& The question arises as to how many intermediate levels e(istor indeed whether any e(ist&

S*nt 7ugue! )n imitation of a musical fu ue, each voice enters with the same statement& The themeholism versus reductionismis introduced in a recursive !icture com!osed of words com!osed of smaller words& etc& The words which a!!ear on the four levels of this stran e !icture are -?EAIS/-, -;E:F7TIE8IS/-, and -/A-& The discussion veers off to a friend of the )nteater1s )unt ?illary, a conscious ant colony& The various levels of her thou ht !rocesses are the to!ic of discussion& /any fu al tricks are ensconced in the :ialo ue& )s a hint to the reader, references are made to !arallel tricks occurrin in the fu ue on the record to which the foursome is listenin & )t the end of the )nt 'u ue, themes from the Prelude return& transformed considerably& Cha!ter <I5 Brains and Thoughts1 -?ow can thou hts he su!!orted by the hardware of the brain is the to!ic of the 7ha!ter& )n overview of the lar e scale and small-scale structure of the brain is first iven& Then the relation between conce!ts and neural activity is s!eculatively discussed in some detail& English 7rench German -uite! )n interlude consistin of Aewis 7arroll1s nonsense !oem -Jabberwocky11 to ether with two translations, one into 'rench and one into 6erman, both done last century& Cha!ter <II5 Minds and Thoughts1 The !recedin !oems brin u! in a forceful way the question of whether lan ua es, or indeed minds, can be -ma!!ed- onto each other& ?ow is communication !ossible between two se!arate !hysical brains, 3hat do all human brains have in common? ) eo ra!hical analo y is used to su est an answer& The question arises, -7an a brain be understood, in some ob.ective sense, by an outsider?*ria with .i#erse <ariations& ) :ialo ue whose form is based on 0ach1s 6oldber Gariations, and whose content is related to number-theoretical !roblems such as the 6oldbach con.ecture& This hybrid has as its main !ur!ose to show how number theory1s subtlety stems from the fact that there are many diverse variations on the theme of searchin throu h an infinite s!ace& Some of them lead to infinite searches, some of them lead to finite searches, while some others hover in between& Cha!ter <III5 Bloo, and +loo, and Gloo,1 These are the names of three com!uter lan ua es& 0looP !ro rams can carry out only !redictably finite searches, while 'looP !ro rams can carry out un!redictable or even infinite searches& The !ur!ose of this 7ha!ter is to ive an intuition for the notions of !rimitive recursive and eneral recursive functions in number theory, for they are essential in 6>del1s !roof& *ir on G's -tring& ) :ialo ue in which 6>del1s self-referential construction is mirrored in words& The idea is due to 3& G& E& Muine& This :ialo ue serves as a !rototy!e for the ne(t 7ha!ter& Cha!ter <I45 0n +ormally &ndecidable ,ro!ositions of T;T and $elated Systems1 This 7ha!ter1s title is an ada!tation of the title of 6>del1s "#%" article, in which his Incom!leteness Theorem was first !ublished& The two ma.or !arts of 6>del1s !roof are one throu h carefully& It is shown how the assum!tion of consistency of T8T forces one to conclude that T8T 4or any similar system5 is incom!lete& ;elations to Euclidean and non-Euclidean eometry are discussed& Im!lications for the !hiloso!hy of mathematics are one into with some care&

0irthday 7antatatata&&& In which )chilles cannot convince the wily and ske!tical Tortoise that today is his 4)chilles15 birthday& ?is re!eated but unsuccessful tries to do so foreshadow the re!eatability of the 6>del ar ument& Cha!ter <45 3um!ing out of the System1 The re!eatability of 6>del1s ar ument is shown, with the im!lication that T8T is not only incom!lete, but -essentially incom!lete&- The fairly notorious ar ument by J& ;& Aucas, to the effect that 6>del1s Theorem demonstrates that human thou ht cannot in any sense be -mechanical-, is analy<ed and found to be wantin & Edif)ing Thoughts of a To(acco -moker& ) :ialo ue treatin of many to!ics, with the thrust bein !roblems connected with self-re!lication and self-reference& Television cameras filmin television screens, and viruses and other subcellular entities which assemble themselves, are amon the e(am!les used& The title comes from a !oem by J& S& 0ach himself, which enters in a !eculiar way& Cha!ter <4I5 Self-$ef and Self-$e!1 This 7ha!ter is about the connection between self-reference in its various uises, and self-re!roducin entities e& &, com!uter !ro rams or :8) molecules5& The relations between a self-re!roducin entity and the mechanisms e(ternal to it which aid it in re!roducin itself 4e& &, a com!uter or !roteins5 are discussed!articularly the fu<<iness of the distinction& ?ow information travels between various levels of such systems is the central to!ic of this 7ha!ter& The 'agnificra(, 3ndeed! The title is a !un on 0ach1s /a nificat in :& The tale is about the 7rab, who ives the a!!earance of havin a ma ical !ower of distin uishin between true and false statements of number theory by readin them as musical !ieces, !layin them on his flute, and determinin whether they are -beautiful- or not& Cha!ter <4II5 Church) Turing) Tars.i) and 0thers1 The fictional 7rab of the !recedin :ialo ue is re!laced by various real !eo!le with ama<in mathematical abilities& The 7hurch-Turin Thesis, which relates mental activity to com!utation, is !resented in several versions of differin stren ths& )ll are analy<ed, !articularly in terms of their im!lications for simulatin human thou ht mechanically, or !ro rammin into a machine an ability to sense or create beauty& The connection between brain activity and com!utation brin s u! some other to!ics, the haltin !roblem of Turin , and Tarski1s Truth Theorem& -,=.+5, To) of 'an's .esigning& This :ialo ue is lifted out of an article by Terry 3ino rad on his !ro ram S?;:AF, only a few names have been chan ed& In it& a !ro ram communicates with a !erson about the so-called -blocks world- in rather im!ressive En lish& The com!uter !ro ram a!!ears to e(hibit some real understandin in its limited world& The :ialo ue1s title is based on ?esu, ?o) of 'an's .esiring, one movement of 0ach1s 7antata "C$& Cha!ter <4III5 Artificial Intelligence5 $etros!ects1 This 7ha!ter o!ens with a discussion of the famous -Turin test-a !ro!osal by the com!uter !ioneer )lan Turin for a way to detect the !resence or absence of -thou ht- in a machine& 'rom there, we o on to an abrid ed history of )rtificial Intelli ence& This covers !ro rams that canto some de ree!lay ames, !rove theorems, solve !roblems, com!ose music, do mathematics, and use -natural lan ua e- 4e& &, En lish5&

8ontrafactus& )bout how we unconsciously or ani<e our thou hts so that we can ima ine hy!othetical variants on the real world all the time& )lso about aberrant variants of this ability such as !ossessed by the new character, the Sloth, an avid lover of 'rench fries, and rabid hater of counterfactuals& Cha!ter <I<5 Artificial Intelligence5 ,ros!ects1 The !recedin :ialo ue tri ers a discussion of how knowled e is re!resented in layers of conte(ts& This leads to the modern )I idea of -frames-& ) frame-like way of handlin a set of visual !attern !u<<les is !resented, for the !ur!ose of concreteness& Then the dee! issue of the interaction of conce!ts in eneral is discussed, which leads into some s!eculations on creativity& The 7ha!ter concludes with a set of !ersonal -Muestions and S!eculations- on )I and minds in eneral& -loth 8anon& ) canon which imitates a 0ach canon in which one voice !lays the same melody as another, only u!side down and twice as slowly, while a third voice is free& ?ere, the Sloth utters the same lines as the Tortoise does, only ne ated 4in a liberal sense of the term5 and twice as slowly, while )chilles is free& Cha!ter <<5 Strange Loo!s) 0r Tangled /ierarchies1 ) rand windu! of many of the ideas about hierarchical systems and self-reference& It is concerned with the snarls which arise when systems turn back on themselvesfor e(am!le, science !robin science, overnment investi atin overnmental wron doin , art violatin the rules of art, and finally, humans thinkin about their own brains and minds& :oes 6>del1s Theorem have anythin to say about this last -snarl-? )re free will and the sensation of consciousness connected to 6>del1s Theorem? The 7ha!ter ends by tyin 6>del, Escher, and 0ach to ether once a ain& -i9%&art =icercar! This :ialo ue is an e(uberant ame !layed with many of the ideas which have !ermeated the book& It is a reenactment of the story of the /usical Efferin , which be an the book= it is simultaneously a -translation- into words of the most com!le( !iece in the /usical Efferin , the Si(-Part ;icercar& This duality imbues the :ialo ue with more levels of meanin than any other in the book& 'rederick the 6reat is re!laced by the 7rab, !ianos by com!uters, and so on& /any sur!rises arise& The :ialo ue1s content concerns !roblems of mind, consciousness, free will, )rtificial Intelli ence, the Turin test, and so forth, which have been introduced earlier& It concludes with an im!licit reference to the be innin of the book, thus makin the book into one bi self-referential loo!, symboli<in at once 0ach1s music, Escher1s drawin s, and 6>del1s Theorem&

Aist of Illustrations
7over, ) -6E0- and an -E60- tri!-let sus!ended in s!ace, castin their symbolic shadows on three !lanes that meet at the corner of a room& 4-Tri!-let- is the name which I have iven to blocks sha!ed in such a way that their shadows in three ortho onal directions are three different letters& The tri!-let idea came to me in a flash one evenin as I was tryin to think how best to symboli<e the unity of 6odel, Escher, and 0ach by somehow fusin their names in a strikin desi n& The two tri!-lets shown on the cover were desi ned and made by me, usin mainly a band saw, with an end mill for the holes= they are redwood, and are .ust under C inches on a side&5 'acin 3ords of Thanks, The be innin of 6enesis, in ancient ?ebrew, 9#iii Part I, The -6E0- tri!-let castin its three ortho onal shadows& "& Johann Sebastian 0ach, by Elias 6ottlieb ?aussmann& B& 7lute 8oncert in -anssouci, by )dol!h von /en<el& %& The ;oyal Theme& C& 0ach1s acrostic on 1;I7E;7);1& L& 3aterfall, by /& 7& Escher& D& )scendin and :escendin , by /& 7& Escher& $& ?and with ;eflectin 6lobe, by /& 7& Escher& *& 'etamorphosis 33, by /& 7& Escher& #& @urt 6>del& "+& '(ius -trip 3, by /& 7& Escher& ""& -Tree- of all theorems of the /IF-system& "B& Sky 7astle, by /& 7& Escher& "%& Aiberation, by /& 7& Escher& "C& /osaic II, by /& 7& Escher& "L& -/)IA 0ET-& "D& Tilin of the !lane usin birds, by /& 7& Escher& "$& 'I6F;E-'I6F;E 'i ure, by Scott E& @im& "*& :ia ram of the relationshi!s between various classes of T8T strin s& "#& The last !a e of 0ach1s *rt of the 7ugue& B+& Gisual rendition of the !rinci!le underlyin 6>del1s Theorem& B"& Tower of Ba(el, by /& 7& Escher& BB& =elati#it), by /& 7& Escher& B%& 8on#e9 and 8onca#e, by /& 7& Escher& BC& =eptiles, by /& 7& Escher& BL& 7retan Aabyrinth& BD& The structure of the :ialo ue +ittle ,armonic +a()rinth& B$& ;ecursive Transition 8etworks for 0$;ATE ;0&; and +A;C= ;0&;& B*& The +A;C= ;0&; ;T8 with one node recursively e(!anded&

B L D $ "" "B "% "C "D %+ C+ CB L$ D" D$ D* D# $" *+ *C *# #* "+$ ""$ ""# "B# "%B "%C

B#& :ia rams 6 and ? re!resented im!licitly& %+& :ia ram 6 further e(!anded& %"& )n ;T8 for 'ibonacci numbers& %B& 6ra!h of the function I8T495& %%& Skeletons of I8T and 6!lot& %C& 6!lot, a recursive ra!h& %L& ) com!le( 'eynman dia ram& %D& 7ish and -cales, by /& 7& Escher& %$& Butterflies, by /& 7& Escher& %*& ) tic-tac-toe ame tree& %#& The ;osetta Stone& C+& ) colla e of scri!ts& C"& 0ase sequence for the chromosome of bacterio!ha e fT"$C& CB& 8ra( 8anon, by /& 7& Escher& C%& ) short section of one of the 7rab1s 6enes& CC& 8ra( 8anon from the /usical Efferin , by J& S& 0ach& CL& +a 'e@"uita, by /& 7& Escher& CD& Three >orlds, by /& 7& Escher& C$& .ewdrop, by /& 7& Escher& C*& *nother >orld, by /& 7& Escher& C#& .a) and Night, by /& 7& Escher& L+& =ind, by /& 7& Escher& L"& &uddle, by /& 7& Escher& LB& =ippled -urface, by /& 7& Escher& L%& Three -pheres 33, by /& 7& Escher& Part II, The -E60- tri!-let castin its three ortho onal shadows& LC& '(ius -trip 33, by /& 7& Escher& LL& Pierre de 'ermat& LD& 8u(e with 'agic =i((ons, by /& 7& Escher& L$& The idea of -chunkin -& L*& )ssemblers, com!ilers, and levels of com!uter lan ua es& L#& Intelli ence as built u! layer by layer& D+& The -/F-!icture-& D"& *nt 7ugue, by /& 7& Escher& DB& ) -crisscross- of two well-known names& D%& Photo of an ant brid e 4de 'ourmi, Aierre5&

"%L "%D "%D "%# "C" "C% "CL "C$ "C* "L" "DL "D* "$D "#* B+" B+B B%L BC$ BC# BL+ BLB BL% BLD BL$ BL*

B$D B$* B*B B** B#C %++ %"+ %B% %%" %%C

DC& ) ?EAIS/-;E:F7TIE8IS/ -!ro!eller-& DL& Schematic drawin of a neuron& DD& The human brain, seen from one side& D$& ;es!onses to !atterns by certain sam!le neurons& D*& Everla!!in neural !athways& D#& The construction of an arch by termite workers& $+& ) tiny !ortion of the author1s -semantic network-& $"& :rder and 8haos, by /& 7& Escher& $B& The structure of a call-less 0looP !ro ram& $%& 6eor 7antor& $C& *(o#e and Below, by /& 7& Escher& $L& -/ultifurcation- of T8T& $D& .ragon, by /& 7& Escher& $$& The -hadows, by ;enK /a ritte& $*& -tate of Grace, by ;enK /a ritte& $#& Tobacco /osaic Girus& *+& The 7air 8apti#e, by ;enK /a ritte& *"& Self-en ulfin TG screens& *B& The *ir and the -ong, by ;enK /a ritte& *%& E!imenides e(ecutin his own death sentence& *C& )n E!imenides-!arado( iceber & *L& ) Muine-sentence soa!cake& *D& ) self-re!roducin son & *$& The Ty!o enetic 7ode& **& The tertiary structure of a ty!oen<yme& *#& Table of bindin -!references for ty!oen<ymes& #+& The 7entral :o ma of Ty!o enetics& #"& The four constituent bases of :8)& #B& Aadder-like structure of :8)& #%& /olecular model of the :8) double heli(& #C& The 6enetic 7ode& #L& Secondary and tertiary structure of myo lobin& #D& ) section of m;8) !assin throu h a ribosome& #$& ) !olyribosome& #*& ) two-tiered molecular canon& ##& The 7entral :o ma!& "++& The 6>del 7ode& "+"& The TC bacterial virus&

%%L %%# %C" %CL %L$ %L* %$+ %## C"L CB% C%B CD$ C$C C*+ C*" C*L C*# C#+-" C#C C#D C#D C#$ L++ L"+ L"" L"" L"% L"L L"L L"D LB+ LB" LB% LBD LB$ L%% L%L L%$

"+B& "+%& "+C& "+L& "+D& "+$& "+*& "+#& ""+& """& ""B& ""%& ""C& ""L& ""D& ""$& ""*& ""#& "B+& "B"& "BB& "B%& "BC& "BL& "BD& "B$& "B*& "B#& "%+& "%"& "%B& "%%& "%C& "%L& "%D& "%$& "%*& "%#& "C+&

Infection of a bacterium by a virus& L%* The mor!ho enetic !athway of the TC virus& L%# 8astro#al#a, by /& 7& Escher& LL+ Srinivasa ;amanu.an and one of his stran e Indian melodies& LD% lsomor!hisms connectin natural numbers, calculators, and human brains& LD* 8eural and symbolic activity in the brain& L$" -Skimmin off- the to! level of the brain& L$% 7onflict between low and hi h levels of a brain& L$D The o!enin scene from a :ialo ue with S?;:AF& L*D ) later scene from the :ialo ue with S?;:AF& L*$ Ene last scene from the :ialo ue with S?;:AF& L*# )lan /athison Turin & L#C Pons )sinorum Proof& D+D 9eno1s endless oal tree& D"+ ) meanin ful story in )rabic& DBC 'ental *rithmetic, by ;enK /a ritte& DB$ Procedural re!resentation of -a red cube which su!!orts a !yramid- D%" 0on ard !roblem L"& DCD 0on ard !roblem C$& DC* 0on ard !roblem #"& DC# 0on ard !roblem C#& DL+ ) small !ortion of a conce!t network for 0on ard !roblems& DLB 0on ard !roblem %%& DLC 0on ard !roblems *L-*$& DLL 0on ard !roblem LL& DL* 0on ard !roblem BB& DL* 0on ard !roblem L*& DL# 0on ard !roblem D"& DL# 0on ard !roblems $+-$"& DD+ ) schematic dia ram of the :ialo ue 8ra( 8anon& DDD Two homolo ous chromosomes .oined at the center by a centromere& DD* --loth 8anon- from the 'usical :ffering, by J& S& 0ach& D*B )n authorshi! trian le& D*# .rawing ,ands, by /& 7& Escher& D#+ )bstract dia ram of /& 7& Escher1s .rawing ,ands& D#+ 8ommon -ense, by ;enK /a ritte& $++ The Two ')steries, by ;enK /a ritte& $+" -moke -ignal, by the author& $+B &ipe .ream, by the author& $+%

"C"& "CB& "C%& "CC& "CL& "CD& "C$& "C*& "C#& "L+& "L"& "LB&

The ,uman 8ondition 3, by ;enK /a ritte& &rint Galler), by /& 7& Escher& )bstract dia ram of /& 7& Escher1s &rint Galler)& ) colla!sed version of 'i ure "C%& 'urther colla!se of 'i ure "C%& )nother way of colla!sin 'i ure "C%& 0ach1s Endlessly ;isin 7anon, !layed in She!ard tones, forms a Stran e Aoo!& Two com!lete cycles of a She!ard tone scale, notated for !iano& <er(um, by /& 7& Escher& 7harles 0abba e& The 7rab1s Theme& Aast !a e of the -i9%&art =icercar, from the ori inal edition of the 'usical :ffering, by J& S& 0ach&

$+L $"C $"L $"L $"D $"D $"$ $"* $%B $%% $C+ $C"

3ords of Thanks
This book was brewin in my mind over a !eriod of !robably nearly twenty yearsever since I was thirteen and thinkin about how I thou ht in En lish and 'rench& )nd even before that, there were clear si ns of my main-line interest& I remember that at some early a e, there was nothin more fascinatin to me than the idea of takin three %1s, o!eratin u!on % with itselfA I was sure that this idea was so subtle that it was inconceivable to anyone elsebut I dared ask my mother one day how bi it was anyway, and she answered -8ine-& I wasn1t sure she knew what I meant, thou h& Aater, my father initiated me into the mysteries of square roots, and i!!! I owe more to my !arents than to anyone else& They have been !illars of su!!ort I could rely on at all times& They have uided me, ins!ired me, encoura ed me, and sustained me& /ost of all, they have always believed in me& It is to them that this book is dedicated& To two friends of many years;obert 0oenin er, Peter JonesI owe s!ecial thanks, for they hel!ed mold a million ways of thou ht, and their influences and ideas are s!read all around in this book& To 7harles 0renner, I am much indebted for his havin tau ht me to !ro ram when we were both youn , and for his constant !ushin and !loddin im!licit !raiseand occasional criticism& I am !leased to acknowled e the immense influence of Ernest 8a el, a lon -time friend and mentor& I loved -8a el and 8ewman-, and I have learned much from many conversations, lon a o in Germont and more recently in 8ew Hork& ?oward :eAon , throu h his book, reawakened in me a lon -dormant love of the matters in this book& I truly owe him a reat debt& :avid Jonathan Justman tau ht me what it is to be a Tortoisean in enious, !ersistent, and humorous bein , with a fondness for !arado( and contradiction& I ho!e that he will read and en.oy this book, which owes a reat deal to him& Scott @im has e(erted a i antic influence on me& Ever since we met some two and a half years a o, the resonance between the two of us has been incredible& )side from his tan ible contributions of art, music, humor, analo ies, and so onincludin mucha!!reciated volunteer labor at crucial timesScott has contributed new !ers!ectives and insi hts which have chan ed my own views of my endeavor as it has evolved& If anyone understands this book, it is Scott& 'or advice on matters of lar e scale or small, I have turned re!eatedly to :on 0yrd, who knows this book forwards and backwards and every which w a y& & & ?e has an unerrin sense for its overall oals and structure, a n d time and a ain has iven me ood ideas which I have deli htedly incor!orated& /y only re ret is that I won1t be able to include all the future ideas :on will come u! with, once the book is in !rint& )nd let me not for et to thank :on for the marvelous fle(ibility-in-infle(ibility of his music-!rintin !ro ram, S/FT& ?e s!ent many lon days and arduous ni hts coa(in S/FT to sit u! and do !re!osterous tricks& Some of his results are included as fi ures in this book& 0ut :on1s influence is s!read throu hout, which !leases me reatly&

I could not !ossibly have written this book if it were not for the facilities of the Institute for /athematical Studies in the Social Sciences, at Stanford Fniversity& Its director, Pat Su!!es, is a lon -time friend and he was e(tremely enerous to me, in housin me in Gentura ?all, ivin me access to an outstandin com!uter system, and in eneral an e(cellent workin environment, for two whole yearsand then some& This brin s me to Pentti @anerva, the author of the te(t-editin !ro ram to which this book owes its e(istence& I have said to many !eo!le that it would have taken me twice as lon to write my book if I hadn1t been able to use -TGEdit-, that raceful !ro ram which is so sim!le in s!irit that only Pentti could have written it& It is also thanks to Pentti that I was able to do somethin which very few authors have ever done, ty!eset my own book& ?e has been a ma.or force in the develo!ment of com!uter ty!esettin at the I/SSS& Equally im!ortant to me, however, is Pentti1s rare quality, his sense of style& If my book looks ood, to Pentti @anerva is due much of the credit& It was in the )SSF Ty!esettin Sho! that this book was actually born& I would like to offer many hearty words of thanks to its director, 0everly ?endricks, and to her crew, for hel! in times of dire need, and for consistently ood s!irits in the face of one disaster after another& I would also like to thank 7ecille Taylor and 0arbara Aadda a, who did most of the actual work of runnin off alleys& Ever the years, my sister Aaura ?ofstadter has contributed much to my outlook on the world& ?er influence is !resent in both the form and content of this book& I would like to thank my new and old friends /arie )nthony, Sydney )rkowit<, 0en t Elle 0en tsson, 'eli( 0loch, 'rancisco 7laro, Persi :iaconis, 8ai-?ua :uan, John Ellis, ;obin 'reeman, :an 'riedman, Pranab 6hosh, /ichael 6oldhaber, )vril 6reenber , Eric ?ambur , ;obert ?erman, ;ay ?yman, :ave Jennin s, :ianne @anerva, Aauri @anerva, In a @arliner, Jonathan and Ellen @in , 6ayle Aandt, 0ill Aewis, Jos /arlowe, John /c7arthy, Jim /c:onald, Aouis /endelowit<, /ike /ueller, ;osemary 8elson, Steve Emohundro, Paul E!!enheimer, Peter E& Parks, :avid Policansky, Pete ;imbey, @athy ;osser, 3ilfried Sie , 6uy Steele, Aarry Tesler, 'rancois Gannucci, Phil 3adler, Terry 3ino rad, and 0ob 3olf for -resonatin with me at crucial times in my life, and thereby contributin in various and sundry ways to this book& I wrote this book twice& )fter havin written it once, I started all over a ain and rewrote it& The first o-round was when I was still a raduate student in !hysics at the Fniversity of Ere on, and four faculty members were mi hty indul ent concernin my aberrant ways, Paul 7sonka, ;udy ?wa, /ike /oravcsik, and 6re ory 3annier& I a!!reciate their understandin attitudes& In addition, Paul 7sonka read an entire early version and made hel!ful comments& Thanks to E& E& 3ilson for readin and commentin on an early version of my &relude, *nt 7ugue! Thanks to /arsha /eredith for bein the meta-author of a droll kRan&

Thanks to /arvin /insky for a memorable conversation one /arch day in his home, !arts of which the reader will find reconstructed herein& Thanks to 0ill @aufmann for advice on !ublication, and to Jeremy 0ernstein and )le( 6eor e for encoura in words when needed& Gery warm thanks to /artin @essler, /aureen 0ischoff, Gincent Torre, Aeon :orin, and all the other !eo!le at 0asic 0ooks, for undertakin this !ublishin venture which is unusual in quite a few ways& Thanks to Phoebe ?oss for doin well the difficult .ob of co!y editin , and to Aarry 0reed for valuable last-minute !roofreadin & Thanks to my many Imlac-roommates, who took so many !hone messa es over the years= also to the Pine ?all crew, who develo!ed and maintained much of the hardware and software that this book has so vitally de!ended on& Thanks to :ennis :avies of the Stanford Instructional Television 8etwork for his hel! in settin u! the -self-en ulfin televisions- which I s!ent hours !hoto ra!hin & Thanks to Jerry Pryke, 0ob Parks, Ted 0radshaw, and Ginnie )veni of the machine sho! in the ?i h Ener y Physics Aaboratory at Stanford, for enerously hel!in me make tri!-lets& Thanks to my uncle and aunt, Jimmy and 0etty 6ivan, for the 7hristmas !resent they never knew would so deli ht me, a -0lack 0o(- which had no other function than to turn itself off& 'inally, I would like to ive s!ecial thanks to my freshman En lish teacher, 0rent ?arold, who first s!ran 9en on me= to @ees 6u elot, who ave me a record of the 'usical :ffering one sad 8ovember lon a o= and to Etto 'risch, in whose office at 7ambrid e I first saw the ma ic of Escher& I have tried to remember all the !eo!le who have contributed to the develo!ment of this book, but I have undoubtedly failed to include all of them& In a way, this book is a statement of my reli ion& I ho!e that this will come throu h to my readers, and that my enthusiasm and reverence for certain ideas will infiltrate the hearts and minds of a few !eo!le& That is the best I could ask for& :& ;& ?& 0loomin ton and Stanford January, "#$#&

73G5=E B! ?ohann -e(astian Bach, in BCDE! 7rom a painting () Elias Gottlie( ,aussmann!

Introduction5 A Musico-Logical 0ffering

*uthor: ';E:E;I7@ T?E 6;E)T, @in of Prussia, came to !ower in "$C+& )lthou h he is remembered in history books mostly for his military astuteness, he was also devoted to the life of the mind and the s!irit& ?is court in Potsdam was one of the reat centers of intellectual activity in Euro!e in the ei hteenth century& The celebrated mathematician Aeonhard Euler s!ent twenty-five years there& /any other mathematicians and scientists came, as well as !hiloso!hersincludin Goltaire and Aa /ettrie, who wrote some of their most influential works while there& 0ut music was 'rederick1s real love& ?e was an avid flutist and com!oser& Some of his com!ositions are occasionally !erformed even to this day& 'rederick was one of the first !atrons of the arts to reco ni<e the virtues of the newly develo!ed -!iano-forte- 4-softloud-5& The !iano had been develo!ed in the first half of the ei hteenth century as a modification of the har!sichord& The !roblem with the har!sichord was that !ieces could only be !layed at a rather uniform loudness there was no way to strike one note more loudly than its nei hbors& The -soft-loud-, as its name im!lies, !rovided a remedy to this !roblem& 'rom Italy, where 0artolommeo 7ristofori had made the first one, the soft-loud idea had s!read widely& 6ottfried Silbermann, the foremost 6erman or an builder of the day, was endeavorin to make a -!erfect- !iano-forte& Fndoubtedly @in 'rederick was the reatest su!!orter of his effortsit is said that the @in owned as many as fifteen Silbermann !ianos2

Bach
'rederick was an admirer not only of !ianos, but also of an or anist and com!oser by the name of J& S& 0ach& This 0ach1s com!ositions were somewhat notorious& Some called them -tur id and confused-, while others claimed they were incom!arable master!ieces& 0ut no one dis!uted 0ach1s ability to im!rovise on the or an& In those days, bein an or anist not only meant bein able to !lay, but also to e(tem!ori<e, and 0ach was known far and wide for his remarkable e(tem!ori<ations& 4'or some deli htful anecdotes about 0ach1s e(tem!ori<ation, see The Bach =eader, by ?& T& :avid and )& /endel&5 In "$C$, 0ach was si(ty-two, and his fame, as well as one of his sons, had reached Potsdam, in fact, 7arl Phili!! Emanuel 0ach was the 7a!ellmeister 4choirmaster5 at the court of @in 'rederick& 'or years the @in had let it be known, throu h entle hints to Phili!! Emanuel, how !leased he would be to have the elder 0ach come and !ay him a visit= but this wish had never been reali<ed& 'rederick was !articularly ea er for 0ach to try out his new Silbermann !ianos, which he 4'rederick5 correctly foresaw as the reat new wave in music& It was 'rederick1s custom to have evenin concerts of chamber music in his court& Eften he himself would be the soloist in a concerto for flute& ?ere we have re!roduced a

!aintin of such an evenin by the 6erman !ainter )dol!h von /en<el, who, in the "*++1s, made a series of !aintin s illustratin the life of 'rederick the 6reat& )t the cembalo is 7& P& E& 0ach, and the fi ure furthest to the ri ht is Joachim Muant<, the @in 1s flute masterand the only !erson allowed to find fault with the @in 1s flute !layin & Ene /ay evenin in "$C$, an une(!ected uest showed u!& Johann 8ikolaus 'orkel, one of 0ach1s earliest bio ra!hers, tells the story as follows,
Ene evenin , .ust as he was ettin his flute ready, and his musicians were assembled, an officer brou ht him a list of the stran ers who had arrived& 3ith his flute in his hand he ran ever the list, but immediately turned to the assembled musicians, and said, with a kind of a itation, -6entlemen, old 0ach is come&- The flute was now laid aside, and old 0ach, who had ali hted at his son1s lod in s, was immediately summoned to the Palace& 3ilhelm 'riedemann, who accom!anied his father, told me this story, and I must say that I still think with !leasure on the manner in which he related it& )t that time it was the fashion to make rather !roli( com!liments& The first a!!earance of J& S& 0ach before so reat a @in , who did not even ive him time to chan e his travelin dress for a black chanter1s own, must necessarily be attended with many a!olo ies& I will not here dwell on these a!olo ies, but merely observe, that in 3ilhelm 'riedemann1s mouth they made a formal :ialo ue between the @in and the )!olo ist& 0ut what is mere im!ortant than this is that the @in ave u! his 7oncert for this evenin , and invited 0ach, then already called the Eld 0ach, to try his forte!ianos, made by Silbermann, which stood in several rooms of the !alace& U'orkel here inserts this footnote, -The !ianofortes manufactured by Silbermann, of 'reyber , !leased the @in so much, that he resolved to buy them all u!& ?e collected fifteen& I hear that they all now stand unfit for use in various corners of the ;oyal Palace&-V The musicians went with him from room to room, and 0ach was invited everywhere to try them and to !lay un!remeditated com!ositions& )fter he had ene on for some time, he asked the @in to ive him a sub.ect for a 'u ue, in order to e(ecute it immediately without any !re!aration& The @in admired the learned manner in which his sub.ect was thus e(ecuted e(tem!ore= and, !robably to see how far such art could be carried, e(!ressed a wish to hear a 'u ue with si( Ebli ato !arts& 0ut as it is not every sub.ect that is fit for such full harmony, 0ach chose one himself, and immediately e(ecuted it to the astonishment of all !resent in the same ma nificent and learned manner as he had done that of the @in & ?is /a.esty desired also to hear his !erformance on the or an& The ne(t day therefore 0ach was taken to all the or ans in Potsdam, as he had before been to Silbermann1s forte!ianos& )fter his return to Aei!<i , he com!osed the sub.ect, which he had received from the @in , in three and si( !arts, added several artificial !assa es in strict canon to it, and had it en raved, under the title of -/usikalisches E!fer- U 'usical :fferingV, and dedicated it to the Inventor&"

73G5=E F! 7lute 8oncert in -anssouci, () *dolph 'en@el BEGF/!

73G5=E H! The =o)al Theme!

3hen 0ach sent a co!y of his /usical Efferin to the @in , he included a dedicatory letter, which is of interest for its !rose style if nothin else rather submissive and flattersome& 'rom a modern !ers!ective it seems comical& )lso, it !robably ives somethin of the flavor of 0ach1s a!olo y for his a!!earance&
/EST 6;)7IEFS @I862 In dee!est humility I dedicate herewith to Hour /a.esty a musical offerin , the noblest !art of which derives from Hour /a.esty1s own au ust hand& 3ith awesome !leasure I still remember the very s!ecial ;oyal race when, some time a o, durin my visit in Potsdam, Hour /a.esty1s Self dei ned to !lay to me a theme for a fu ue u!on the clavier, and at the same time char ed me most raciously to carry it out in Hour /a.esty1s most au ust !resence& To obey Hour /a.esty1s command was my most humble duty& I noticed very soon, however, that, for lack of necessary !re!aration, the e(ecution of the task did not fare as well as such an e(cellent theme demanded& I resolved therefore and !rom!tly !led ed myself to work out this ri ht ;oyal theme more fully, and then make it known to the world& This resolve has now been carried out as well as !ossible, and it has none other than this irre!roachable intent, to lorify, if only in a small !oint, the fame of a monarch whose reatness and !ower, as in all the sciences of war and !eace, so es!ecially in music, everyone must admire and revere& I make bold to add this most humble request, may Hour /a.esty dei n to di nify the !resent modest labor with a racious acce!tance, and continue to rant Hour /a.esty1s most au ust ;oyal race to Hour /a.esty1s most humble and obedient servant, T?E )FT?E; Aei!<i , July $ "$C$

Some twenty-seven years later, when 0ach had been dead for twenty-four years, a 0aron named 6ottfried van Swietento whom, incidentally, 'orkel dedicated his bio ra!hy of 0ach, and 0eethoven dedicated his 'irst Sym!honyhad a conversation with @in 'rederick, which he re!orted as follows,
?e U'rederickV s!oke to me, amon other thin s, of music, and of a reat or anist named 0ach, who has been for a while in 0erlin& This artist U3ilhelm 'riedemann 0achV is endowed with a talent su!erior, in de!th of harmonic knowled e and !ower of e(ecution, to any I have heard or can ima ine, while those who knew his father claim that he, in turn, was even reater& The @in is of this o!inion, and to !rove it to me he san aloud a chromatic fu ue sub.ect which he had iven this old 0ach, who on the s!ot had made of it a fu ue in four !arts, then in five !arts, and finally in ei ht !arts&%

Ef course there is no way of knowin whether it was @in 'rederick or 0aron van Swieten who ma nified the story into lar er-than-life !ro!ortions& 0ut it shows how

!owerful 0ach1s le end had become by that time& To ive an idea of how e(traordinary a si(-!art fu ue is, in the entire 3ell-Tem!ered 7lavier by 0ach, containin forty-ei ht Preludes and 'u ues, only two have as many as five !arts, and nowhere is there a si(-!art fu ue2 Ene could !robably liken the task of im!rovisin a si(-!art fu ue to the !layin of si(ty simultaneous blindfold ames of chess, and winnin them all& To im!rovise an ei ht-!art fu ue is really beyond human ca!ability& In the co!y which 0ach sent to @in 'rederick, on the !a e !recedin the first sheet of music, was the followin inscri!tion,

73G5=E D!

4-)t the @in 1s 7ommand, the Son and the ;emainder ;esolved with 7anonic )rt&-5 ?ere 0ach is !unnin on the word -canonic-, since it means not only -with canons- but also -in the best !ossible way-& The initials of this inscri!tion are RICERCAR
an Italian word, meanin -to seek-& )nd certainly there is a reat deal to seek in the

'usical :ffering& It consists of one three-!art fu ue, one si(-!art fu ue, ten canons, and a trio sonata& /usical scholars have concluded that the three-!art fu ue must be, in essence, identical with the one which 0ach im!rovised for @in 'rederick& The si(-!art fu ue is one of 0ach1s most com!le( creations, and its theme is, of course, the ;oyal Theme& That theme, shown in 'i ure %, is a very com!le( one, rhythmically irre ular and hi hly chromatic 4that is, filled with tones which do not belon to the key it is in5& To write a decent fu ue of even two voices based on it would not be easy for the avera e musician2 0oth of the fu ues are inscribed -;icercar-, rather than -'u a-& This is another meanin of the word= -ricercar- was, in fact, the ori inal name for the musical form now known as -fu ue-& 0y 0ach1s time, the word -fu ue- 4or fu a, in Aatin and Italian5 had become standard, but the term -ricercar- had survived, and now desi nated an erudite kind of fu ue, !erha!s too austerely intellectual for the common ear& ) similar usa e survives in En lish today, the word -recherche- means, literally, -sou ht out-, but carries the same kind of im!lication, namely of esoteric or hi hbrow cleverness& The trio sonata forms a deli htful relief from the austerity of the fu ues and canons, because it is very melodious and sweet, almost danceable& 8evertheless, it too is based lar ely on the @in 1s theme, chromatic and austere as it is& It is rather miraculous that 0ach could use such a theme to make so !leasin an interlude& The ten canons in the /usical Efferin are amon the most so!histicated canons 0ach ever wrote& ?owever, curiously enou h, 0ach himself never wrote them out in full& This was deliberate& They were !osed as !u<<les to @in 'rederick& It was a familiar musical ame of the day to ive a sin le theme, to ether with some more or less tricky hints, and

to let the canon based on that theme be -discovered- by someone else& In order to know how this is !ossible, you must understand a few facts about canons&

Canons and +ugues

The idea of a canon is that one sin le theme is !layed a ainst itself& This is done by havin -co!ies- of the theme !layed by the various !artici!atin voices& 0ut there are many ways to do this& The most strai htforward of all canons is the round, such as -Three 0lind /ice-, -;ow, ;ow, ;ow Hour 0oat-, or -'rere Jacques-& ?ere, the theme enters in the first voice and, after a fi(ed time-delay, a -co!y- of it enters, in !recisely the same key& )fter the same fi(ed time-delay in the second voice, the third voice enters carryin the theme, and so on& /ost themes will not harmoni<e with themselves in this way& In order for a theme to work as a canon theme, each of its notes must be able to serve in a dual 4or tri!le, or quadru!le5 role, it must firstly be !art of a melody, and secondly it must be !art of a harmoni<ation of the same melody& 3hen there are three canonical voices, for instance, each note of the theme must act in two distinct harmonic ways, as well as melodically& Thus, each note in a canon has more than one musical meanin = the listener1s ear and brain automatically fi ure out the a!!ro!riate meanin , by referrin to conte(t& There are more com!licated sorts of canons, of course& The first escalation in com!le(ity comes when the -co!ies- of the theme are sta ered not only in time, but also in pitch= thus, the first voice mi ht sin the theme startin on 7, and the second voice, overla!!in with the first voice, mi ht sin the identical theme startin five notes hi her, on 6& ) third voice, startin on the : yet five notes hi her, mi ht overla! with the first two, and so on& The ne(t escalation in com!le(ity comes when the s!eeds of the different voices are not equal= thus, the second voice mi ht sin twice as quickly, or twice as slowly, as the first voice& The former is called diminution, the latter augmentation 4since the theme seems to shrink or to e(!and5& 3e are not yet done2 The ne(t sta e of com!le(ity in canon construction is to in#ert the theme, which means to make a melody which .um!s down wherever the ori inal theme .um!s up, and by e(actly the same number of semitones& This is a rather weird melodic transformation, but when one has heard many themes inverted, it be ins to seem quite natural& 0ach was es!ecially fond of inversions, and they show u! often in his work and the 'usical :ffering is no e(ce!tion& 4'or a sim!le e(am!le of inversion, try the tune -6ood @in 3enceslas-& 3hen the ori inal and its inversion are sun to ether, startin an octave a!art and sta ered with a time-delay of two beats, a !leasin canon results&5 'inally, the most esoteric of -co!ies- is the retro rade co!y where the theme is !layed backwards in time& ) canon which uses this trick is affectionately known as a cra( canon, because of the !eculiarities of crab locomotion& 0ach included a crab canon in the 'usical :ffering, needless to say& 8otice that every ty!e of -co!y- !reserves all the information in the ori inal theme, in the sense that the theme is fully recoverable from any of the co!ies& Such an information !reservin transformation is often called an isomorphism, and we will have much traffic with isomor!hisms in this book& Sometimes it is desirable to rela( the ti htness of the canon form& Ene way is to allow sli ht de!artures from !erfect co!yin , for the sake of more fluid harmony& )lso, some

canons have -free- voicesvoices which do not em!loy the canon1s theme, but which sim!ly harmoni<e a reeably with the voices that are in canon with each other& Each of the canons in the 'usical :ffering has for its theme a different variant of the @in 1s Theme, and all the devices described above for makin canons intricate are e(!loited to the hilt= in fact, they are occasionally combined& Thus, one three-voice canon is labeled -7anon !er )u mentationem, contrario /otu-= its middle voice is free 4in fact, it sin s the ;oyal Theme5, while the other two dance canonically above and below it, usin the devices of au mentation and inversion& )nother bears sim!ly the cry!tic label -Muaerendo invenietis- 4-0y seekin , you will discover-5& )ll of the canon !u<<les have been solved& The canonical solutions were iven by one of 0ach1s !u!ils, Johann Phili!! @irnber er& 0ut one mi ht still wonder whether there are more solutions to seek2 I should also e(!lain briefly what a fu ue is& ) fu ue is like a canon, in that it is usually based on one theme which ets !layed in different voices and different keys, and occasionally at different s!eeds or u!side down or backwards& ?owever, the notion of fu ue is much less ri id than that of canon, and consequently it allows for more emotional and artistic e(!ression& The telltale si n of a fu ue is the way it be ins, with a sin le voice sin in its theme& 3hen it is done, then a second voice enters, either five scale-notes u!, or four down& /eanwhile the first voice oes on, sin in the -countersub.ect-, a secondary theme, chosen to !rovide rhythmic, harmonic, and melodic contrasts to the sub.ect& Each of the voices enters in turn, sin in the theme, often to the accom!animent of the countersub.ect in some other voice, with the remainin voices doin whatever fanciful thin s entered the com!oser1s mind& 3hen all the voices have -arrived-, then there are no rules& There are, to be sure, standard kinds of thin s to do but not so standard that one can merely com!ose a fu ue by formula& The two fu ues in the 'usical :ffering are outstandin e(am!les of fu ues that could never have been -com!osed by formula-& There is somethin much dee!er in them than mere fu ality& )ll in all, the 'usical :ffering re!resents one of 0ach1s su!reme accom!lishments in counter!oint& It is itself one lar e intellectual fu ue, in which many ideas and forms have been woven to ether, and in which !layful double meanin s and subtle allusions are common!lace& )nd it is a very beautiful creation of the human intellect which we can a!!reciate forever& 4The entire work is wonderfully described in the book ?! -! Bach's 'usical :ffering, by ?& T& :avid&5

An Endlessly $ising Canon

There is one canon in the 'usical :ffering which is !articularly unusual& Aabeled sim!ly -7anon !er Tonos-, it has three voices& The u!!ermost voice sin s a variant of the ;oyal Theme, while underneath it, two voices !rovide a canonic harmoni<ation based on a second theme& The lower of this !air sin s its theme in 7 minor 4which is the key of the canon as a whole5, and the u!!er of the !air sin s the same theme dis!laced u!wards in !itch by an interval of a fifth& 3hat makes this canon different from any other, however, is that when it concludesor, rather, seems to concludeit is no lon er in the key of 7 minor, but now is in : minor& Somehow 0ach has contrived to modulate 4chan e keys5 ri ht under the listener1s nose& )nd it is so constructed that this -endin - ties smoothly

onto the be innin a ain= thus one can re!eat the !rocess and return in the key of E, only to .oin a ain to the be innin & These successive modulations lead the ear to increasin ly remote !rovinces of tonality, so that after several of them, one would e(!ect to be ho!elessly far away from the startin key& )nd yet ma ically, after e(actly si( such modulations, the ori inal key of 7 minor has been restored2 )ll the voices are e(actly one octave hi her than they were at the be innin , and here the !iece may be broken off in a musically a reeable way& Such, one ima ines, was 0ach1s intention= but 0ach indubitably also relished the im!lication that this !rocess could o on ad infinitum, which is !erha!s why he wrote in the mar in -)s the modulation rises, so may the @in 1s 6lory&- To em!hasi<e its !otentially infinite as!ect, I like to call this the -Endlessly ;isin 7anon-& In this canon, 0ach has iven us our first e(am!le of the notion of -trange +oops& The -Stran e Aoo!- !henomenon occurs whenever, by movin u!wards 4or downwards5 throu h the levels of some hierarchical system, we une(!ectedly find ourselves ri ht back where we started& 4?ere, the system is that of musical keys&5 Sometimes I use the term Tangled ,ierarch) to describe a system in which a Stran e Aoo! occurs& )s we o on, the theme of Stran e Aoo!s will recur a ain and a ain& Sometimes it will be hidden, other times it will be out in the o!en= sometimes it will be ri ht side u!, other times it will be u!side down, or backwards& -Muaerendo invenietis- is my advice to the reader&

Escher
To my mind, the most beautiful and !owerful visual reali<ations of this notion of Stran e Aoo!s e(ist in the work of the :utch ra!hic artist /& 7& Escher, who lived from "#+B to "#$B& Escher was the creator of some of the most intellectually stimulatin drawin s of all time& /any of them have their ori in in !arado(, illusion, or double-meanin & /athematicians were amon the first admirers of Escher1s drawin s, and this is understandable because they often are based on mathematical !rinci!les of symmetry or !attern&&& 0ut there is much more to a ty!ical Escher drawin than .ust symmetry or !attern= there is often an underlyin idea, reali<ed in artistic form& )nd in !articular, the Stran e Aoo! is one of the most recurrent themes in Escher1s work& Aook, for e(am!le, at the litho ra!h >aterfall 4'i & L5, and com!are its si(-ste! endlessly fallin loo! with the si(-ste! endlessly risin loo! of the -7anon !er Tonos-& The similarity of vision is remarkable& 0ach and Escher are !layin one sin le theme in two different -keys-, music and art&

73G5=E G! 3aterfall, () '! 8! Escher lithograph, BIJB/!

73G5=E J! )scendin and :escendin , () '! 8! Escher lithograph, BIJK/!

Escher reali<ed Stran e Aoo!s in several different ways, and they can be arran ed accordin to the ti htness of the loo!& The litho ra!h *scending and .escending 4'i & D5, in which monks trud e forever in loo!s, is the loosest version, since it involves so many ste!s before the startin !oint is re ained& ) ti hter loo! is contained in >aterfall, which, as we already observed, involves only si( discrete ste!s& Hou may be thinkin that there is some ambi uity in the notion of a sin le -ste!- for instance, couldn1t *scending and .escending be seen .ust as easily as havin four levels 4staircases5 as forty-five levels 4stairs5? It is indeed true that there is an inherent ha<iness in level-countin , not only in

73G5=E C! ?and with ;eflectin 6lobe! -elf%portrait 3n, '! 8! Escher lithograph, BIHG/!

73G5=E E! /etamor!hosis II, () '! 8! Escher woodcut BI!G cm! W DKK cm!, BIHI%DK/

Escher !ictures, but in hierarchical, many-level systems& 3e will shar!en our understandin of this ha<iness later on& 0ut let us not et too distracted now2 )s we ti hten our loo!, we come to the remarkable .rawing ,ands 4'i & "%L5, in which each of two hands draws the other, a two-ste! Stran e Aoo!& )nd finally, the ti htest of all Stran e Aoo!s is reali<ed in &rint Galler) 4'i & "CB5, a !icture of a !icture which contains itself& Er is it a !icture of a allery which contains itself? Er of a town which contains itself? Er a youn man who contains himself? 4Incidentally, the illusion underlyin *scending and .escending and >aterfall was not invented by Escher, but by ;o er Penrose, a 0ritish mathematician, in "#L*& ?owever, the theme of the Stran e Aoo! was already !resent in Escher1s work in "#C*, the year he drew .rawing ,ands& &rint Galler) dates from "#LD&5 Im!licit in the conce!t of Stran e Aoo!s is the conce!t of infinity, since what else is a loo! but a way of re!resentin an endless !rocess in a finite way? )nd infinity !lays a lar e role in many of Escher1s drawin s& 7o!ies of one sin le theme often fit into each other, formin visual analo ues to the canons of 0ach& Several such !atterns can be seen in Escher1s famous !rint 'etamorphosis 4'i & *5& It is a little like the -Endlessly ;isin 7anon-, wanderin further and further from its startin !oint, it suddenly is back& In the tiled !lanes of /etamor!hosis and other !ictures, there are already su estions of infinity& 0ut wilder visions of infinity a!!ear in other drawin s by Escher& In some of his drawin s, one sin le theme can a!!ear on different levels of reality& 'or instance, one level in a drawin mi ht clearly be reco ni<able as re!resentin fantasy or ima ination= another level would be reco ni<able as reality& These two levels mi ht be the only e(!licitly !ortrayed levels& 0ut the mere !resence of these two levels invites the viewer to look u!on himself as !art of yet another level= and by takin that ste!, the viewer cannot hel! ettin cau ht u! in Escher1s im!lied chain of levels, in which, for any one level, there is always another level above it of reater -reality-, and likewise, there is always a level below, -more ima inary- than it is& This can be mind-bo lin in itself& ?owever, what ha!!ens if the chain of levels is not linear, but forms a loo!? 3hat is real, then, and what is fantasy? The enius of Escher was that he could not only concoct, but actually !ortray, do<ens of half-real, half-mythical worlds, worlds filled with Stran e Aoo!s, which he seems to be invitin his viewers to enter&

G#del
In the e(am!les we have seen of Stran e Aoo!s by 0ach and Escher, there is a conflict between the finite and the infinite, and hence a stron sense of !arado(& Intuition senses that there is somethin mathematical involved here& )nd indeed in our own century a mathematical counter!art was discovered, with the most enormous re!ercussions& )nd, .ust as the 0ach and Escher loo!s a!!eal to very sim!le and ancient intuitions a musical scale, a staircaseso this discovery, by @& 6>del, of a Stran e Aoo! in mathematical systems has its ori ins in sim!le and ancient intuitions& In its absolutely barest form, 6>del1s discovery involves the translation of an ancient !arado( in !hiloso!hy into mathematical terms& That !arado( is the so-called Epimenides parado9, or liar parado9&

73G5=E I! ;urt Gdel!

E!imenides was a 7retan who made one immortal statement, -)ll 7retans are liars&- ) shar!er version of the statement is sim!ly -I am lyin -= or, -This statement is false-& It is that last version which I will usually mean when I s!eak of the E!imenides !arado(& It is a statement which rudely violates the usually assumed dichotomy of statements into true and false, because if you tentatively think it is true, then it immediately backfires on you and makes you think it is false& 0ut once you1ve decided it is false, a similar backfirin returns you to the idea that it must be true& Try it2 The E!imenides !arado( is a one-ste! Stran e Aoo!, like Escher1s &rint Galler)& 0ut how does it have to do with mathematics? That is what 6>del discovered& ?is idea was to use mathematical reasonin in e(!lorin mathematical reasonin itself& This notion of makin mathematics -intros!ective- !roved to be enormously !owerful, and !erha!s its richest im!lication was the one 6>del found, 6>del1s Incom!leteness Theorem& 3hat the Theorem states and how it is !roved are two different thin s& 3e shall discuss both in

quite some detail in this book& The Theorem can be likened to a !earl, and the method of !roof to an oyster& The !earl is !ri<ed for its luster and sim!licity= the oyster is a com!le( livin beast whose innards ive rise to this mysteriously sim!le em& 6>del1s Theorem a!!ears as Pro!osition GI in his "#%" !a!er -En 'ormally Fndecidable Pro!ositions in &rincipia 'athematica and ;elated Systems I&- It states, To every w-consistent recursive class L of formulae there corres!ond recursive classsi ns r, such that neither # 6en r nor 8e 4# 6en r5 belon s to 'l 4@5 4where # is the free #aria(le of r5& )ctually, it was in 6erman, and !erha!s you feel that it mi ht as well be in 6erman anyway& So here is a !ara!hrase in more normal En lish, )ll consistent a(iomatic formulations of number theory include undecidable !ro!ositions& This is the !earl& In this !earl it is hard to see a Stran e Aoo!& That is because the Stran e Aoo! is buried in the oysterthe !roof& The !roof of 6>del1s Incom!leteness Theorem hin es u!on the writin of a self-referential mathematical statement, in the same way as the E!imenides !arado( is a self-referential statement of lan ua e& 0ut whereas it is very sim!le to talk about lan ua e in lan ua e, it is not at all easy to see how a statement about numbers can talk about itself& In fact, it took enius merely to connect the idea of self-referential statements with number theory& Ence 6>del had the intuition that such a statement could be created, he was over the ma.or hurdle& The actual creation of the statement was the workin out of this one beautiful s!ark of intuition& 3e shall e(amine the 6>del construction quite carefully in 7ha!ters to come, but so that you are not left com!letely in the dark, I will sketch here, in a few strokes, the core of the idea, ho!in that what you see will tri er ideas in your mind& 'irst of all, the difficulty should be made absolutely clear& /athematical statementslet us concentrate on numbertheoretical onesare about !ro!erties of whole numbers& 3hole numbers are not statements, nor are their !ro!erties& ) statement of number theory is not about a& statement of number theory= it .ust is a statement of number theory& This is the !roblem= but 6>del reali<ed that there was more here than meets the eye& 6>del had the insi ht that a statement of number theory could be about a statement of number theory 4!ossibly even itself5, if only numbers could somehow stand for statements& The idea of a code, in other words, is at the heart of his construction& In the 6>del 7ode, usually called -6>del-numberin -, numbers are made to stand for symbols and sequences of symbols& That way, each statement of number theory, bein a sequence of s!eciali<ed symbols, acquires a 6>del number, somethin like a tele!hone number or a license !late, by which it can be referred to& )nd this codin trick enables statements of number theory to be understood on two different levels, as statements of number theory, and also as statements about statements of number theory&

Ence 6>del had invented this codin scheme, he had to work out in detail a way of trans!ortin the E!imenides !arado( into a number-theoretical formalism& ?is final trans!lant of E!imenides did not say, -This statement of number theory is false-, but rather, -This statement of number theory does not have any !roof-& ) reat deal of confusion can be caused by this, because !eo!le enerally understand the notion of -!roof- rather va uely& In fact, 6>del1s work was .ust !art of a lon attem!t by mathematicians to e(!licate for themselves what !roofs are& The im!ortant thin to kee! in mind is that !roofs are demonstrations within fi9ed s)stems of !ro!ositions& In the case of 6>del1s work, the fi(ed system of number-theoretical reasonin to which the word -!roof- refers is that of &rincipia 'athematica &!'!/, a iant o!us by 0ertrand ;ussell and )lfred 8orth 3hitehead, !ublished between "#"+ and "#"%& Therefore, the 6>del sentence 6 should more !ro!erly be written in En lish as, This statement of number theory does not have any !roof in the system of &rincipia 'athematica! Incidentally, this 6>del sentence 6 is not 6>del1s Theoremno more than the E!imenides sentence is the observation that -The E!imenides sentence is a !arado(&- 3e can now state what the effect of discoverin 6 is& 3hereas the E!imenides statement creates a !arado( since it is neither true nor false, the 6>del sentence 6 is un!rovable 4inside P&/&5 but true& The rand conclusion? That the system of &rincipia 'athematica is -incom!lete-there are true statements of number theory which its methods of !roof are too weak to demonstrate& 0ut if &rincipia 'athematica was the first victim of this stroke, it was certainly not the last2 The !hrase -and ;elated Systems- in the title of 6>del1s article is a tellin one, for if 6>del1s result had merely !ointed out a defect in the work of ;ussell and 3hitehead, then others could have been ins!ired to im!rove u!on P&/& and to outwit 6>del1s Theorem& 0ut this was not !ossible, 6>del1s !roof !ertained to any a(iomatic system which !ur!orted to achieve the aims which 3hitehead and ;ussell had set for themselves& )nd for each different system, one basic method did the trick& In short, 6>del showed that !rovability is a weaker notion than truth, no matter what a(iomatic system is involved& Therefore 6>del1s Theorem had an electrifyin effect u!on lo icians, mathematicians, and !hiloso!hers interested in the foundations of mathematics, for it showed that no fi(ed system, no matter how com!licated, could re!resent the com!le(ity of the whole numbers, +, ", B, %,&&& /odern readers may not be as non!lussed by this as readers of "#%" were, since in the interim our culture has absorbed 6>del1s Theorem, alon with the conce!tual revolutions of relativity and quantum mechanics, and their !hiloso!hically disorientin messa es have reached the !ublic, even if cushioned by several layers of translation 4and usually obfuscation5& There is a eneral mood of e(!ectation, these days, of -limitative- resultsbut back in "#%", this came as a bolt from the blue&

Mathematical Logic5 A Syno!sis

) !ro!er a!!reciation of 6>del1s Theorem requires a settin of conte(t& Therefore, I will now attem!t to summari<e in a short s!ace the history of mathematical lo ic !rior to "#%"an im!ossible task& 4See :eAon , @neebone, or 8a el and 8ewman, for ood !resentations of history&5 It all be an with the attem!ts to mechani<e the thou ht !rocesses of reasonin & 8ow our ability to reason has often been claimed to be what distin uishes us from other s!ecies= so it seems somewhat !arado(ical, on first thou ht, to mechani<e that which is most human& Het even the ancient 6reeks knew that reasonin is a !atterned !rocess, and is at least !artially overned by statable laws& )ristotle codified syllo isms, and Euclid codified eometry= but thereafter, many centuries had to !ass before !ro ress in the study of a(iomatic reasonin would take !lace a ain& Ene of the si nificant discoveries of nineteenth-century mathematics was that there are different, and equally valid, eometrieswhereby -a eometry- is meant a theory of !ro!erties of abstract !oints and lines& It had lon been assumed that eometry was what Euclid had codified, and that, althou h there mi ht be small flaws in Euclid1s !resentation, they were unim!ortant and any real !ro ress in eometry would be achieved by e(tendin Euclid& This idea was shattered by the rou hly simultaneous discovery of non-Euclidean eometry by several !eo!lea discovery that shocked the mathematics community, because it dee!ly challen ed the idea that mathematics studies the real world& ?ow could there be many different kinds of -!oints- and -lines- in one sin le reality? Today, the solution to the dilemma may be a!!arent, even to some nonmathematiciansbut at the time, the dilemma created havoc in mathematical circles& Aater in the nineteenth century, the En lish lo icians 6eor e 0oole and )u ustus :e /or an went considerably further than )ristotle in codifyin strictly deductive reasonin !atterns& 0oole even called his book "The +aws of Thought"surely an e(a eration, but it was an im!ortant contribution& Aewis 7arroll was fascinated by these mechani<ed reasonin methods, and invented many !u<<les which could be solved with them& 6ottlob 're e in Jena and 6iuse!!e Peano in Turin worked on combinin formal reasonin with the study of sets and numbers& :avid ?ilbert in 6>ttin en worked on stricter formali<ations of eometry than Euclid1s& )ll of these efforts were directed towards clarifyin what one means by -!roof-& In the meantime, interestin develo!ments were takin !lace in classical mathematics& ) theory of different ty!es of infinities, known as the theor) of sets, was develo!ed by 6eor 7antor in the "**+1s& The theory was !owerful and beautiful, but intuition-defyin & 0efore lon , a variety of set-theoretical !arado(es had been unearthed& The situation was very disturbin , because .ust as mathematics seemed to be recoverin from one set of !arado(esthose related to the theory of limits, in the calculusalon came a whole new set, which looked worse2 The most famous is ;ussell1s !arado(& /ost sets, it would seem, are not members of themselvesfor e(am!le, the set of walruses is not a walrus, the set containin only Joan of )rc is not Joan of )rc 4a set is not a !erson5and so on& In this res!ect, most sets are rather -run-of-the-mill-& ?owever, some -self-swallowin - sets do contain themselves as members, such as the set of all sets, or the set of all thin s e(ce!t Joan of )rc, and so

on& 7learly, every set is either run-of-the-mill or self-swallowin , and no set can be both& 8ow nothin !revents us from inventin ;, the set of all run%of%the%mill sets & )t first, ; mi ht seem a rather run-of-the-mill inventionbut that o!inion must be revised when you ask yourself, -Is ; itself -a run-of-the-mill set or a self-swallowin set?- Hou will find that the answer is, -; is neither run-of-the-mill nor self-swallowin , for either choice leads to !arado(&- Try it2 0ut if ; is neither run-of-the-mill nor self-swallowin , then what is it? )t the very least, !atholo ical& 0ut no one was satisfied with evasive answers of that sort& )nd so !eo!le be an to di more dee!ly into the foundations of set theory& The crucial questions seemed to be, -3hat is wron with our intuitive conce!t of 1set1? 7an we make a ri orous theory of sets which corres!onds closely with our intuitions, but which skirts the !arado(es?- ?ere, as in number theory and eometry, the !roblem is in tryin to line u! intuition with formali<ed, or a(iomati<ed, reasonin systems& ) startlin variant of ;ussell1s !arado(, called -6rellin 1s !arado(-, can be made usin ad.ectives instead of sets& :ivide the ad.ectives in En lish into two cate ories, those which are self-descri!tive, such as -!entasyllabic-, -awkwardnessful-, and -recherche-, and those which are not, such as -edible-, -incom!lete-, and -bisyllabic-& 8ow if we admit -non-selfdescri!tive- as an ad.ective, to which class does it belon ? If it seems questionable to include hy!henated words, we can use two terms invented s!ecially for this !arado(, autological 4X -self-descri!tive-5, and heterological 4X -nonself-descri!tive-5& The question then becomes, -Is 1heterolo ical1 heterolo ical?- Try it2 There seems to he one common cul!rit in these !arado(es, namely self-reference, or -Stran e Aoo!iness-& So if the oal is to ban all !arado(es, why not try bannin selfreference and anythin that allows it to arise? This is not so easy as it mi ht seem, because it can be hard to fi ure out .ust where self-reference is occurrin & It may be s!read out over a whole Stran e Aoo! with several ste!s, as in this -e(!anded- version of E!imenides, reminiscent of .rawing ,ands, The followin sentence is false& The !recedin sentence is true& Taken to ether, these sentences have the same effect as the ori inal E!imenides !arado(, yet se!arately, they are harmless and even !otentially useful sentences& The -blame- for this Stran e Aoo! can1t he !inned on either sentenceonly on the way they -!oint- at each other& In the same way, each local re ion of )scendin and :escendin is quite le itimate= it is only the way they are lobally !ut to ether that creates an im!ossibility& Since there are indirect as well as direct ways of achievin self-reference, one must fi ure out how to ban both ty!es at onceif one sees self-reference as the root of all evil&

Banishing Strange Loo!s

;ussell and 3hitehead did subscribe to this view, and accordin ly, &rincipia 'athematica was a mammoth e(ercise in e(orcisin Stran e Aoo!s from lo ic, set theory, and number theory& The idea of their system was basically this& ) set of the lowest

-ty!e- could contain only -ob.ects- as members, not sets& ) set of the ne(t ty!e u! could only contain ob.ects, or sets of the lowest ty!e& In eneral, a set of a iven ty!e could only contain sets of lower ty!e, or ob.ects& Every set would belon to a s!ecific ty!e& 7learly, no set could contain itself because it would have to belon to a ty!e hi her than its own ty!e& Enly -run-of-the-mill- sets e(ist in such a system= furthermore, old ;the set of all run-of-the-mill setsno lon er is considered a set at all, because it does not belon to any finite ty!e& To all a!!earances, then, this theor) of t)pes, which we mi ht also call the -theory of the abolition of Stran e Aoo!s-, successfully rids set theory of its !arado(es, but only at the cost of introducin an artificial-seemin hierarchy, and of disallowin the formation of certain kinds of setssuch as the set of all run-of-the-mill sets& Intuitively, this is not the way we ima ine sets& The theory of ty!es handled ;ussell1s !arado(, but it did nothin about the E!imenides !arado( or 6rellin 1s !arado(& 'or !eo!le whose interest went no further than set theory, this was quite adequatebut for !eo!le interested in the elimination of !arado(es enerally, some similar -hierarchi<ation- seemed necessary, to forbid loo!in back inside lan ua e& )t the bottom of such a hierarchy would be an o(Mect language& ?ere, reference could be made only to a s!ecific domainnot to as!ects of the ob.ect lan ua e itself 4such as its rammatical rules, or s!ecific sentences in it5& 'or that !ur!ose there would be a metalanguage& This e(!erience of two lin uistic levels is familiar to all learners of forei n lan ua es& Then there would be a metametalan ua e for discussin the metalan ua e, and so on& It would be required that every sentence should belon to some !recise level of the hierarchy& Therefore, if one could find no level in which a iven utterance fit, then the utterance would be deemed meanin less, and for otten& )n analysis can be attem!ted on the two-ste! E!imenides loo! iven above& The first sentence, since it s!eaks of the second, must be on a hi her level than the second& 0ut by the same token, the second sentence must be on a hi her level than the first& Since this is im!ossible, the two sentences are -meanin less-& /ore !recisely, such sentences sim!ly cannot be formulated at all in a system based on a strict hierarchy of lan ua es& This !revents all versions of the E!imenides !arado( as well as 6rellin 1s !arado(& 4To what lan ua e level could -heterolo ical- belon ?5 8ow in set theory, which deals with abstractions that we don1t use all the time, a stratification like the theory of ty!es seems acce!table, even if a little stran ebut when it comes to lan ua e, an all-!ervadin !art of life, such stratification a!!ears absurd& 3e don1t think of ourselves as .um!in u! and down a hierarchy of lan ua es when we s!eak about various thin s& ) rather matter-of-fact sentence such as, -In this book, I critici<e the theory of ty!es- would be doubly forbidden in the system we are discussin & 'irstly, it mentions -this book-, which should only be mentionable in a metabook-and secondly, it mentions mea !erson whom I should not be allowed to s!eak of at all2 This e(am!le !oints out how silly the theory of ty!es seems, when you im!ort it into a familiar conte(t& The remedy it ado!ts for !arado(estotal banishment of self-reference in any formis a real case of overkill, brandin many !erfectly ood constructions as meanin less& The ad.ective -meanin less-, by the way, would have to a!!ly to all discussions of the theory of lin uistic ty!es 4such as that of this very !ara ra!h5 for they clearly could not occur on any of the levelsneither ob.ect lan ua e, nor metalan ua e, nor metametalan ua e,

etc& So the very act of discussin the theory would be the most blatant !ossible violation of it2 8ow one could defend such theories by sayin that they were only intended to deal with formal lan ua esnot with ordinary, informal lan ua e& This may be so, but then it shows that such theories are e(tremely academic and have little to say about !arado(es e(ce!t when they cro! u! in s!ecial tailor-made systems& 0esides, the drive to eliminate !arado(es at any cost, es!ecially when it requires the creation of hi hly artificial formalisms, !uts too much stress on bland consistency, and too little on the quirky and bi<arre, which make life and mathematics interestin & It is of course im!ortant to try to maintain consistency, but when this effort forces you into a stu!endously u ly theory, you know somethin is wron & These ty!es of issues in the foundations of mathematics were res!onsible for the hi h interest in codifyin human reasonin methods which was !resent in the early !art of this century& /athematicians and !hiloso!hers had be un to have serious doubts about whether even the most concrete of theories, such as the study of whole numbers 4number theory5, were built on solid foundations& If !arado(es could !o! u! so easily in set theory a theory whose basic conce!t, that of a set, is surely very intuitively a!!ealin then mi ht they not also e(ist in other branches of mathematics? )nother related worry was that the !arado(es of lo ic, such as the E!imenides !arado(, mi ht turn out to be internal to mathematics, and thereby cast in doubt all of mathematics& This was es!ecially worrisome to thoseand there were a ood numberwho firmly believed that mathematics is sim!ly a branch of lo ic 4or conversely, that lo ic is sim!ly a branch of mathematics5& In fact, this very question-)re mathematics and lo ic distinct, or se!arate?-was the source of much controversy& This study of mathematics itself became known as metamathematicsor occasionally, metalo ic, since mathematics and lo ic are so intertwined& The most ur ent !riority of metamathematicians was to determine the true nature of mathematical reasonin & 3hat is a le al method of !rocedure, and what is an ille al one? Since mathematical reasonin had always been done in -natural lan ua e- 4e& &, 'rench or Aatin or some lan ua e for normal communication5, there was always a lot of !ossible ambi uity& 3ords had different meanin s to different !eo!le, con.ured u! different ima es, and so forth& It seemed reasonable and even im!ortant to establish a sin le uniform notation in which all mathematical work could be done, and with the aid of which any two mathematicians could resolve dis!utes over whether a su ested !roof was valid or not& This would require a com!lete codification of the universally acce!table modes of human reasonin , at least as far as they a!!lied to mathematics&

Consistency) Com!leteness) /ilbert%s ,rogram

This was the oal of &rincipia 'athematica, which !ur!orted to derive all of mathematics from lo ic, and, to be sure, without contradictions2 It was widely admired, but no one was sure if 4"5 all of mathematics really was contained in the methods delineated by ;ussell and 3hitehead, or 4B5 the methods iven were even self-consistent&

3as it absolutely clear that contradictory results could never be derived, by any mathematicians whatsoever, followin the methods of ;ussell and 3hitehead? This question !articularly bothered the distin uished 6erman mathematician 4and metamathematician5 :avid ?ilbert, who set before the world community of mathematicians 4and metamathematicians5 this challen e, to demonstrate ri orously !erha!s followin the very methods outlined by ;ussell and 3hiteheadthat the system defined in &rincipia 'athematica was both consistent 4contradiction-free5, and com!lete 4i&e&, that every true statement of, number theory could be derived within the framework drawn u! in P&/&5& This was a tall order, and one could critici<e it on the rounds that it was somewhat circular, how can you .ustify your methods of reasonin on the basis of those same methods of reasonin ? It is like liftin yourself u! by your own bootstra!s& 43e .ust don1t seem to be able to et away from these Stran e Aoo!s25 ?ilbert was fully aware of this dilemma, of course, and therefore e(!ressed the ho!e that a demonstration of consistency or com!leteness could be found which de!ended only on -finitistic- modes of reasonin & These were a small set of reasonin methods usually acce!ted by mathematicians& In this way, ?ilbert ho!ed that mathematicians could !artially lift themselves by their own bootstra!s, the sum total of mathematical methods mi ht be !roved sound, by invokin only a smaller set of methods& This oal may sound rather esoteric, but it occu!ied the minds of many of the reatest mathematicians in the world durin the first thirty years of this century& In the thirty-first year, however, 6>del !ublished his !a!er, which in some ways utterly demolished ?ilbert1s !ro ram& This !a!er revealed not only that there were irre!arable -holes- in the a(iomatic system !ro!osed by ;ussell and 3hitehead, but more enerally, that no a(iomatic system whatsoever could !roduce all number-theoretical truths, unless it were an inconsistent system2 )nd finally, the ho!e of !rovin the consistency of a system such as that !resented in &!'! was shown to be vain, if such a !roof could be found usin only methods inside &!'!, thenand this is one of the most mystifyin consequences of 6>del1s work&!'! itself would be inconsistent2 The final irony of it all is that the !roof of 6>del1s Incom!leteness Theorem involved im!ortin the E!imenides !arado( ri ht into the heart of &rincipia 'athematica, a bastion su!!osedly invulnerable to the attacks of Stran e Aoo!s2 )lthou h 6>del1s Stran e Aoo! did not destroy &rincipia 'athematica, it made it far less interestin to mathematicians, for it showed that ;ussell and 3hitehead1s ori inal aims were illusory&

Babbage) Com!uters) Artificial Intelligence111

3hen 6>del1s !a!er came out, the world was on the brink of develo!in electronic di ital com!uters& 8ow the idea of mechanical calculatin en ines had been around for a while& In the seventeenth century, Pascal and Aeibni< desi ned machines to !erform fi(ed o!erations 4addition and multi!lication5& These machines had no memory, however, and were not, in modern !arlance, !ro rammable& The first human to conceive of the immense com!utin !otential of machinery was the Aondoner 7harles 0abba e 4"$#B-"*$"5& ) character who could almost have ste!!ed out of the !a es of the &ickwick &apers, 0abba e was most famous durin his

lifetime for his vi orous cam!ai n to rid Aondon of -street nuisances-or an rinders above all& These !ests, lovin to et his oat, would come and serenade him at any time of day or ni ht, and he would furiously chase them down the street& Today, we reco ni<e in 0abba e a man a hundred years ahead of his time, not only inventor of the basic !rinci!les of modern com!uters, he was also one of the first to battle noise !ollution& ?is first machine, the -:ifference En ine-, could enerate mathematical tables of many kinds by the -method of differences-& 0ut before any model of the -:&E&- had been built, 0abba e became obsessed with a much more revolutionary idea, his -)nalytical En ine-& ;ather immodestly, he wrote, -The course throu h which I arrived at it was the most entan led and !er!le(ed which !robably ever occu!ied the human mind&-1 Fnlike any !reviously desi ned machine, the )&E& was to !ossess both a -store- 4memory5 and a -mill- 4calculatin and decision-makin unit5& These units were to be built of thousands of intricate eared cylinders interlocked in incredibly com!le( ways& 0abba e had a vision of numbers swirlin in and out of the mill tinder control of a program contained in !unched cardsan idea ins!ired by the Jacquard loom, a card-controlled loom that wove ama<in ly com!le( !atterns& 0abba e1s brilliant but ill-fated 7ountess friend, Aady )da Aovelace 4dau hter of Aord 0yron5, !oetically commented that -the )nalytical En ine wea#es alge(raic patterns .ust as the Jacquard-loom weaves flowers and leaves&Fnfortunately, her use of the !resent tense was misleadin , for no )&E& was ever built, and 0abba e died a bitterly disa!!ointed man& Aady Aovelace, no less than 0abba e, was !rofoundly aware that with the invention of the )nalytical En ine, mankind was flirtin with mechani<ed intelli ence !articularly if the En ine were ca!able of -eatin its own tail- 4the way 0abba e described the Stran e Aoo! created when a machine reaches in and alters its own stored !ro ram5& In an "*CB memoir,L she wrote that the )&E& -mi ht act u!on other thin s besides num(er-& 3hile 0abba e dreamt of creatin a chess or tic-tac-toe automaton, she su ested that his En ine, with !itches and harmonies coded into its s!innin cylinders, -mi ht com!ose elaborate and scientific !ieces of music of any de ree of com!le(ity or e(tent&- In nearly the same breath, however, she cautions that -The )nalytical En ine has no !retensions whatever to originate anythin & It can do whatever we know how to order it to perform&- Thou h she well understood the !ower of artificial com!utation, Aady Aovelace was ske!tical about the artificial creation of intelli ence& ?owever, could her keen insi ht allow her to dream of the !otential that would be o!ened u! with the tamin of electricity? In our century the time was ri!e for com!uterscom!uters beyond the wildest dreams of Pascal, Aeibni<, 0abba e, or Aady Aovelace& In the "#%+1s and "#C+1s, the first - iant electronic brains- were desi ned and built& They cataly<ed the conver ence of three !reviously dis!arate areas, the theory of a(iomatic reasonin , the study of mechanical com!utation, and the !sycholo y of intelli ence& These same years saw the theory of com!uters develo! by lea!s and bounds& This theory was ti htly linked to metamathematics& In fact, 6>del1s Theorem has a counter!art in the theory of com!utation, discovered by )lan Turin , which reveals the e(istence of ineluctable -holes- in even the most !owerful com!uter ima inable& Ironically, .ust as these somewhat eerie limits were bein ma!!ed out, real com!uters were bein built

whose !owers seemed to row and row beyond their makers1 !ower of !ro!hecy& 0abba e, who once declared he would ladly ive u! the rest of his life if he could come back in five hundred years and have a three-day uided scientific tour of the new a e, would !robably have been thrilled s!eechless a mere century after his deathboth by the new machines, and by their une(!ected limitations& 0y the early "#L+1s, mechani<ed intelli ence seemed a mere stone1s throw away= and yet, for each barrier crossed, there always cro!!ed u! some new barrier to the actual creation of a enuine thinkin machine& 3as there some dee! reason for this oal1s mysterious recession? 8o one knows where the borderline between non-intelli ent behavior and intelli ent behavior lies= in fact, to su est that a shar! borderline e(ists is !robably silly& 0ut essential abilities for intelli ence are certainly, to res!ond to situations very fle(ibly= to take advanta e of fortuitous circumstances= to make sense out of ambi uous or contradictory messa es= to reco ni<e the relative im!ortance of different elements of a situation= to find similarities between situations des!ite differences which may se!arate them= to draw distinctions between situations des!ite similarities may link them= to synthesi<e new conce!ts by takin old them to ether in new ways= to come u! with ideas which are novel& ?ere one runs u! a ainst a seemin !arado(& 7om!uters by their very nature are the most infle(ible, desireless, rule-followin of beasts& 'ast thou h they may be, they are nonetheless the e!itome of unconsciousness& ?ow, then, can intelli ent behavior be !ro rammed? Isn1t this the most blatant of contradictions in terms? Ene of the ma.or theses of this book is that it is not a contradiction at all& Ene of the ma.or !ur!oses of this book is to ur e each reader to confront the a!!arent contradiction head on, to savor it, to turn it over, to take it a!art, to wallow in it, so that in the end the reader mi ht emer e with new insi hts into the seemin ly unbreachable ulf between the formal and the informal, the animate and the inanimate, the fle(ible and the infle(ible& This is what )rtificial Intelli ence 4)I5 research is all about& )nd the stran e flavor of )I work is that !eo!le try to !ut to ether lon sets of rules in strict formalisms which tell infle(ible machines how to be fle(ible& 3hat sorts of -rules- could !ossibly ca!ture all of what we think of as intelli ent behavior, however? 7ertainly there must be rules on all sorts of different levels& There must be many -.ust !lain- rules& There must be -metarules- to modify the -.ust !lainrules= then -metametarules- to modify the metarules, and so on& The fle(ibility of intelli ence comes from the enormous number of different rules, and levels of rules& The reason that so many rules on so many different levels must e(ist is that in life, a creature is faced with millions of situations of com!letely different ty!es& In some situations, there are stereoty!ed res!onses which require -.ust !lain- rules& Some situations are mi(tures of stereoty!ed situationsthus they require rules for decidin which of the 1.ust !lain-

rules to a!!ly& Some situations cannot be classifiedthus there must e(ist rules for inventin new rules&&& and on and on& 3ithout doubt, Stran e Aoo!s involvin rules that chan e themselves, directly or indirectly, are at the core of intelli ence& Sometimes the com!le(ity of our minds seems so overwhelmin that one feels that there can be no solution to the !roblem of understandin intelli encethat it is wron to think that rules of any sort overn a creature1s behavior, even if one takes -rule- in the multilevel sense described above&

1 1 1and Bach
In the year "$LC, four years after the death of J& S& 0ach, the Aei!<i theolo ian Johann /ichael Schmidt wrote, in a treatise on music and the soul, the followin noteworthy !assa e,
8ot many years a o it was re!orted from 'rance that a man had made a statue that could !lay various !ieces on the 7leuttra#ersiere, !laced the flute to its li!s and took it down a ain, rolled its eyes, etc& 0ut no one has yet invented an ima e that thinks, or wills, or com!oses, or even does anythin at all similar& Aet anyone who wishes to be convinced look carefully at the last fu al work of the above-!raised 0ach, which has a!!eared in co!!er en ravin , but which was left unfinished because his blindness intervened, and let him observe the art that is contained therein= or what must strike him as even more wonderful, the 7horale which he dictated in his blindness to the !en of another, >enn wir in hchsten Nthen se)n & I am sure that he will soon need his soul if he wishes to observe all the beauties contained therein, let alone wishes to !lay it to himself or to form a .ud ment of the author& Everythin that the cham!ions of /aterialism !ut forward must fall to the round in view of this sin le e(am!le&D

Muite likely, the foremost of the -cham!ions of /aterialism- here alluded to was none other than Julien Effroy de la /ettrie!hiloso!her at the court of 'rederick the 6reat, author of +'homme machine 4-/an, the /achine-5, and /aterialist Par E(cellence& It is now more than B++ years later, and the battle is still ra in between those who a ree with Johann /ichael Schmidt, and those who a ree with Julien Effroy de la /ettrie& I ho!e in this book to ive some !ers!ective on the battle&

(G#del) Escher) Bach(

The book is structured in an unusual way, as a counter!oint between :ialo ues and 7ha!ters& The !ur!ose of this structure is to allow me to !resent new conce!ts twice, almost every new conce!t is first !resented meta!horically in a :ialo ue, yieldin a set of concrete, visual ima es= then these serve, durin the readin of the followin 7ha!ter, as an intuitive back round for a more serious and abstract !resentation of the same conce!t& In many of the :ialo ues I a!!ear to be talkin about one idea on the surface, but in reality I am talkin about some other idea, in a thinly dis uised way& Eri inally, the only characters in my :ialo ues were )chilles and the Tortoise, who came to me from 9eno of Elea, by way of Aewis 7arroll& 9eno of Elea, inventor of !arado(es, lived in the fifth century 0&7& Ene of his !arado(es was an alle ory, with

)chilles and the Tortoise as !rota onists& 9eno1s invention of the ha!!y !air is told in my first :ialo ue, Three%&art 3n#ention& In "*#L, Aewis 7arroll reincarnated )chilles and the Tortoise for the !ur!ose of illustratin his own new !arado( of infinity& 7arroll1s !arado(, which deserves to be far better known than it is, !lays a si nificant role in this book& Eri inally titled -3hat the Tortoise Said to )chilles-, it is re!rinted here as Two%&art 3n#ention& 3hen I be an writin :ialo ues, somehow I connected them u! with musical forms& I don1t remember the moment it ha!!ened= I .ust remember one day writin -'u ue- above an early :ialo ue, and from then on the idea stuck& Eventually I decided to !attern each :ialo ue in one way or another on a different !iece by 0ach& This was not so ina!!ro!riate& Eld 0ach himself used to remind his !u!ils that the se!arate !arts in their com!ositions should behave like -!ersons who conversed to ether as if in a select com!any-& I have taken that su estion !erha!s rather more literally than 0ach intended it= nevertheless I ho!e the result is faithful to the meanin & I have been !articularly ins!ired by as!ects of 0ach1s com!ositions which have struck me over and over, and which are so well described by :avid and /endel in The Bach =eader,
?is form in eneral was based on relations between se!arate sections& These relations ran ed from com!lete identity of !assa es on the one hand to the return of a sin le !rinci!le of elaboration or a mere thematic allusion on the other& The resultin !atterns were often symmetrical, but by no means necessarily so& Sometimes the relations between the various sections make u! a ma<e of interwoven threads that only detailed analysis can unravel& Fsually, however, a few dominant features afford !ro!er orientation at first si ht or hearin , and while in the course of study one may discover unendin subtleties, one is never at a loss to ras! the unity that holds to ether every sin le creation by 0ach&D

I have sou ht to weave an Eternal 6olden 0raid out of these three strands, 6>del, Escher, 0ach& I be an, intendin to write an essay at the core of which would be 6>del1s Theorem& I ima ined it would be a mere !am!hlet& 0ut my ideas e(!anded like a s!here, and soon touched 0ach and Escher& It took some time for me to think of makin this connection e(!licit, instead of .ust lettin it be a !rivate motivatin force& 0ut finally " reali<ed that to me, 6>del and Escher and 0ach were only shadows cast in different directions by some central solid essence& I tried to reconstruct the central ob.ect, and came u! with this book&

Three%&art 3n#ention
*chilles a Greek warrior, the fleetest of foot of all mortals/ and a Tortoise are standing together on a dust) runwa) in the hot sun! 7ar down the runwa), on a tall flagpole, there hangs a large rectangular flag! The flag is solid red, e9cept where a thin ring% shaped hole has (een cut out of it, through which one can see the sk)! *chilles, 3hat is that stran e fla down at the other end of the track? It reminds me somehow of a !rint by my favorite artist, /&7& Escher& Tortoise, That is 9eno1s fla & *chilles, 7ould it be that the hole in it resembles the holes in a />bian stri! Escher once drew? Somethin is wron about the fla , I can tell& Tortoise, The rin which has been cut from it has the sha!e of the numeral for <ero, which is 9eno1s favorite number& *chilles, The rin which hasn1t been invented yet2 It will only be invented by a ?indu mathematician some millennia hence& )nd thus, /r& T, my ar ument !roves that such a fla is im!ossible& Tortoise, Hour ar ument is !ersuasive, )chilles, and I must a ree that such a fla is indeed im!ossible& 0ut it is beautiful anyway, is it not? *chilles, Eh, yes, there is no doubt of its beauty& Tortoise, I wonder if its beauty is related to its im!ossibility& I don1t know, I1ve never had the time to analy<e 0eauty& It1s a 7a!itali<ed Essence, and I never seem to have time for 7a!itali<ed Essences& *chilles, S!eakin of 7a!itali<ed Essences, /r& T, have you ever wondered about the Pur!ose of Aife? Tortoise, Eh, heavens, no= *chilles, ?aven1t you ever wondered why we are here, or who invented us? Tortoise, Eh, that is quite another matter& 3e are inventions of 9eno 4as you will shortly see5 and the reason we are here is to have a footrace& *chilles, ) footrace? ?ow outra eous2 /e, the fleetest of foot of all mortals, versus you, the !loddin est of the !lodders2 There can be no !oint to such a race& Tortoise, Hou mi ht ive me a head start& *chilles, It would have to be a hu e one& Tortoise, I don1t ob.ect& *chilles, 0ut I will catch you, sooner or latermost likely sooner& Tortoise, 8ot if thin s o accordin to 9eno1s !arado(, you won1t& 9eno is ho!in to use our footrace to show that motion is im!ossible, you see& It is only in the mind that motion seems !ossible, accordin to 9eno& In truth, /otion Is Inherently Im!ossible& ?e !roves it quite ele antly&

7igure BK! />bius stri! () '!8! Escher wood%engra#ing printed from four (locks, BIJB/

*chilles, Eh, yes, it comes back to me now, the famous 9en kRan about 9en /aster 9eno& )s you say it is very sim!le indeed& Tortoise, 9en kRan? 9en /aster? 3hat do you mean? *chilles, It oes like this, Two monks were ar uin about a fla & Ene said, -The fla is movin &- The other said, -The wind is movin &- The si(th !atriarch, 9eno, ha!!ened to be !assin by& ?e told them, -8ot the wind, not the fla , mind is movin &Tortoise, I am afraid you are a little befuddled, )chilles& 9eno is no 9en master, far from it& ?e is in fact, a 6reek !hiloso!her from the town of Elea 4which lies halfway between !oints ) and 05& 7enturies hence, he will be celebrated for his !arado(es of motion& In one of those !arado(es, this very footrace between you and me will !lay a central role& *chilles, I1m all confused& I remember vividly how I used to re!eat over and over the names of the si( !atriarchs of 9en, and I always said, -The si(th !atriarch is 9eno, The si(th !atriarch is 9enoS- 4-uddenl) a soft warm (ree@e picks up! 5 Eh, look /r& Tortoiselook at the fla wavin 2 ?ow I love to watch the ri!!les shimmer throu h it1s soft fabric& )nd the rin cut out of it is wavin , too2

Tortoise, :on1t be silly& The fla is im!ossible, hence it can1t be wavin & The wind is wavin & *t this moment, Neno happens ()!/ Neno, ?allo2 ?ulloo2 3hat1s u!? 3hat1s new? *chilles, The fla is movin & Tortoise, The wind is movin & Neno, E friends, 'riends2 7ease your ar umentation2 )rrest your vitriolics2 )bandon your discord2 'or I shall resolve the issue for you forthwith& ?o2 )nd on such a fine day& *chilles, This fellow must be !layin the fool& Tortoise, 8o, wait, )chilles& Aet us hear what he has to say& Eh Fnknown Sir, do im!art to us your thou hts on this matter& Neno, /ost willin ly& 8ot the wind, not the fla neither one is movin , nor is anythin movin at all& 'or I have discovered a reat Theorem, which states= -/otion Is Inherently Im!ossible&- )nd from this Theorem follows an even reater Theorem 9eno1s Theorem, -/otion Fne(ists&*chilles, -9eno1s Theorem-? )re you, sir, by any chance, the !hiloso!her 9eno of Elea? Neno, I am indeed, )chilles& *chilles, scratching his head in pu@@lement5& 8ow how did he know my name? Neno, 7ould I !ossibly !ersuade you two to hear me out as to why this is the case? I1ve come all the way to Elea from !oint ) this afternoon, .ust tryin to find someone who1ll !ay some attention to my closely honed ar ument& 0ut they1re all hurryin hither and thither, and they don1t have time& Hou1ve no idea how disheartenin it is to meet with refusal after refusal& Eh, I1m sorry to burden you with my troubles, I1d .ust like to ask you one thin , 3ould the two of you humor a silly old !hiloso!her for a few momentsonly a few, I !romise youin his eccentric theories& *chilles, Eh, by all means2 Please do illuminate us2 I know I s!eak for both of us, since my com!anion, /r& Tortoise, was only moments a o s!eakin of you with reat venerationand he mentioned es!ecially your !arado(es& Neno, Thank you& Hou see, my /aster, the fifth !atriarch, tau ht me that reality is one, immutable, and unchan in = all !lurality, chan e, and motion are mere illusions of the senses& Some have mocked his views= but I will show the absurdity of their mockery& /y ar ument is quite sim!le& I will illustrate it with two characters of my own Invention, )chilles 4a 6reek warrior, the fleetest of foot of all mortals5, and a Tortoise& In my tale, they are !ersuaded by a !asserby to run a footrace down a runway towards a distant fla wavin in the bree<e& Aet us assume that, since the Tortoise is a much slower runner, he ets a head start of, say, ten rods& 8ow the race be ins& In a few bounds )chilles has reached the s!ot where the Tortoise started&

*chilles, ?ah2 Neno, )nd now the Tortoise is but a sin le rod ahead of )chilles& 3ithin only a moment, )chilles has attained that s!ot& *chilles, ?o ho2 Neno, Het, in that short moment, the Tortoise has mana ed to advance a sli ht amount& In a flash, )chilles covers that distance too& *chilles, ?ee hee hee2 Neno, 0ut in that very short flash, the Tortoise has mana ed to inch ahead by ever so little, and so )chilles is still behind& 8ow you see that in order for )chilles to catch the Tortoise, this ame of -try-to-catch-me- will have to be !layed an I8'I8ITE number of timesand therefore )chilles can 8EGE; catch u! with the Tortoise& Tortoise, ?eh heh heh heh2 *chilles, ?mmS ?mmS ?mmS ?mmS ?mmSThat ar ument sounds wron to me& )nd yes, I can1t quite make out what1s wron with it Neno, Isn1t it a teaser? It1s my favorite !arado(& Tortoise, E(cuse me, 9eno, but I believe your tale illustrates the wron !rinci!le, does it not? Hou have .ust told us what will come to known, centuries hence, as 9eno1s -)chilles !arado(- , which shows 4ahem25 that )chilles will never catch the Tortoise= but the !roof that /otion Is Inherently Im!ossible 4and thence that /otion Fne(ists5 is your -dichotomy !arado(-, isn1t that so? Neno, Eh, shame on me& Ef course, you1re ri ht& That1s the new one about how, in oin from ) to 0, one has to o halfway first Y and of that stretch one also has to o halfway, and so on and so forth& 0ut you see, both those !arado(es really have the same flavor& 'rankly, I1ve only had one 6reat Idea Y I .ust e(!loit it in different ways& *chilles, I swear, these ar uments contain a flaw& I don1t quite see where, but they cannot be correct& Neno, Hou doubt the validity of my !arado(? 3hy not .ust try it outZ? Hou see that red fla wavin down here, at the far end of the runway? *chilles, The im!ossible one, based on an Escher !rint? Neno, E(actly& 3hat do you say to you and /r& Tortoise racin for it, allowin /r& T a fair head start of, well, I don1t know Tortoise, ?ow about ten rods? Neno, Gery oodten rods& *chilles, )ny time& Neno, E(cellent2 ?ow e(citin 2 )n em!irical test of my ri orously !roven Theorem2 /r& Tortoise, will you !osition yourself ten rods u!wind? The Tortoise mo#es ten rods closer to the flag/ Tortoise and *chilles, ;eady2 Neno, En your mark2 6et set2 6o2

Cha!ter I

The M&-!u::le
+ormal Systems
E8E E' T?E most central notions in this book is that of a formal s)stem& The ty!e of formal system I use was invented by the )merican lo ician Emil Post in the "#B+1s, and is often called a -Post !roduction system-& This 7ha!ter introduces you to a formal system and moreover, it is my ho!e that you will want to e(!lore this formal system at least a little= so to !rovoke your curiosity, I have !osed a little !u<<le& -7an you !roduce M&?- is the !u<<le& To be in with, you will be su!!lied with a string 4which means a strin of letters5&[ 8ot to kee! you in sus!ense, that strin will be MI& Then you will be told some rules, with which you can chan e one strin into another& If one of those rules is a!!licable at some !oint, and you want to use it, you may, but there is nothin that will dictate which rule you should use, in case there are several a!!licable rules& That is left u! to youand of course, that is where !layin the ame of any formal system can become somethin of an art& The ma.or !oint, which almost doesn1t need statin , is that you must not do anythin which is outside the rules& 3e mi ht call this restriction the -;equirement of 'ormality-& In the !resent 7ha!ter, it !robably won1t need to be stressed at all& Stran e thou h it may sound, thou h, I !redict that when you !lay around with some of the formal systems of 7ha!ters to come, you will find yourself violatin the ;equirement of 'ormality over and over a ain, unless you have worked with formal systems before& The first thin to say about our formal systemthe '35-s)stemOis that it utili<es only three letters of the al!habet, M) I) && That means that the only strin s of the /IFsystem are strin s which are com!osed of those three letters& 0elow are some strin s of the /IF-system, M& &IM M&&M&& &II&MI&&IM&II&MI&&IM&II&
[

In this book, we shall em!loy the followin conventions when we refer to strin s& 3hen the strin is in the same ty!eface as the te(t, then it will be enclosed in sin le or double quotes& Punctuation which belon s to the sentence and not to the strin under discussion will o outside of the quotes, as lo ic dictates& 'or e(am!le, the first letter of this sentence is 1'1, while the first letter of 1this sentence1 is 1t1& 3hen the strin is in 8uadrata $oman, however, quotes will usually be left off, unless clarity demands them& 'or e(am!le, the first letter of 8uadrata is 8.

0ut althou h all of these are le itimate strin s, they are not strin s which are -in your !ossession-& In fact, the only strin in your !ossession so far is MI& Enly by usin the rules, about to be introduced, can you enlar e your !rivate collection& ?ere is the first rule, ;FAE I, If you !ossess a strin whose last letter is I, you can add on a & at the end& 0y the way, if u! to this !oint you had not uessed it, a fact about the meanin of -strin is that the letters are in a fi(ed order& 'or e(am!le, MI and IM are two different strin s& ) strin of symbols is not .ust a -ba - of symbols, in which the order doesn1t make any difference& ?ere is the second rule, ;FAE II, Su!!ose you have M9& Then you may add M99 to your collection& 3hat I mean by this is shown below, in a few e(am!les& 'rom MI&, you may et MI&I&& 'rom M&M, you may et M&M&M& 'rom M&, you may et M&&& So the letter 191 in the rule sim!ly stands for any strin = but once you have decided which strin it stands for, you have to stick with your choice 4until you use the rule a ain, at which !oint you may make a new choice5& 8otice the third e(am!le above& It shows how, once you !ossess M&, you can add another strin to your collection= but you have to et M& first2 I want to add one last comment about the letter 1 91, it is not !art of the formal system in the same way as the three letters 1 M1, 1I1, and 1&1 are& It is useful for us, thou h, to have some way to talk in eneral about strin s of the system, symbolicallyand that is the function of the 191, to stand for an arbitrary strin & If you ever add a strin containin an 191 to your -collection-, you have done somethin wron , because strin s of the /IFsystem never contain -9- 1s2 ?ere is the third rule, ;FAE III, If III occurs in one of the strin s in your collection, you may make a new strin with & in !lace of III& E(am!les, 'rom &MIIIM&, you could make &M&M&& 'rom MIIII, you could make MI& 4also M&I5& 'rom IIMII, you can1t et anywhere usin this rule& 4The three I1s have to be consecutive&5

'rom MIII, make M&& :on1t, under any circumstances, think you can run this rule backwards, as in the followin e(am!le, 'rom M&, make MIII ;ules are one-way& ?ere is the final rule& ;FAE IG, If && occurs inside one of your strin s, you can dro! it& 'rom &&&, et && 'rom M&&&III, et M&III& There you have it& 8ow you may be in tryin to make M&& :on1t worry you don1t et it& Just try it out a bitthe main thin is for you to et the flavor of this /F-!u<<le& ?ave fun& \ This is wron &

Theorems) A-ioms) $ules

The answer to the /F-!u<<le a!!ears later in the book& 'or now, what im!ortant is not findin the answer, but lookin for it& Hou !robably hay made some attem!ts to !roduce M&& In so doin , you have built u! your own !rivate collection of strin s& Such strin s, !roducible by the rules, are called theorems& The term -theorem- has, of course, a common usa e mathematics which is quite different from this one& It means some statement in ordinary lan ua e which has been !roven to be true by a ri orous ar ument, such as 9eno1s Theorem about the -une(istence- of motion, or Euclid1s Theorem about the infinitude of !rimes& 0ut in formal system theorems need not be thou ht of as statementsthey are merely strin s of symbols& )nd instead of bein pro#en, theorems are merely produced, as if by machine, accordin to certain ty!o ra!hical rules& To em!hasi<e this im!ortant distinction in meanin s for the word -theorem-, I will ado!t the followin convention in this book, when -theorem- is ca!itali<ed, its meanin will be the everyday onea Theorem is a statement in ordinary lan ua e which somebody once !roved to be true by some sort of lo ic ar ument& 3hen unca!itali<ed, -theorem- will have its technical meanin a strin !roducible in some formal system& In these terms, the /F-!u<<le asks whether M& is a theorem of the /IF-system& I ave you a theorem for free at the be innin , namely MI& Such -free- theorem is called an a9iomOthe technical meanin a ain bein quite different from the usual meanin & ) formal system may have <ero, or several, or even infinitely many a(ioms& E(am!les of all these ty!es will a!!ear in the book&

Every formal system has symbol-shuntin rules, such as the four rules of the /IFsystem& These rules are called either rules of production or rules of inference& I will use both terms& The last term which I wish to introduce at this !oint is derivation& Shown below is a derivation of the theorem M&II&, 4"5 MI 4B5 MII 4%5 MIII 4C5 MIIII& 4L5 M&I& 4D5 M&I&&I& 4$5 M&II& a(iom from 4"5 by rule II from 4B5 by rule II from 4%5 by rule I from 4C5 by rule III from 4L5 by rule II from 4D5 by rule IG

) derivation of a theorem is an e(!licit, line-by-line demonstration of how to !roduce that theorem accordin to the rules of the formal system& The conce!t of derivation is modeled on that of !roof, but a derivation is an austere cousin of a !roof& It would sound stran e to say that you had pro#en M&II&, but it does not sound so stran e to say you have deri#ed M&II&&

Inside and 0utside the System

/ost !eo!le o about the /F-!u<<le by derivin a number of theorems, quite at random, .ust to see what kind of thin turns u!& Pretty soon, they be in to notice some !ro!erties of the theorems they have made= that is where human intelli ence enters the !icture& 'or instance, it was !robably not obvious to you that all theorems would be in with M, until you had tried a few& Then, the !attern emer ed, and not only could you see the !attern, but you could understand it by lookin at the rules, which have the !ro!erty that they make each new theorem inherit its first letter from an earlier theorem= ultimately, then, all theorems1 first letters can be traced back to the first letter of the sole a(iom MIand that is a !roof that theorems of the /IF-system must all be in with M& There is somethin very si nificant about what has ha!!ened here& It shows one difference between !eo!le and machines& It would certainly be !ossiblein fact it would be very easyto !ro ram a com!uter to enerate theorem after theorem of the /IFsystem= and we could include in the !ro ram a command to sto! only u!on eneratin && Hou now know that a com!uter so !ro rammed would never sto!& )nd this does not ama<e you& 0ut what if you asked a friend to try to enerate &? It would not sur!rise you if he came back after a while, com!lainin that he can1t et rid of the initial M, and therefore it is a wild oose chase& Even if a !erson is not very bri ht, he still cannot hel! makin some observations about what he is doin , and these observations ive him ood insi ht into the taskinsi ht which the com!uter !ro ram, as we have described it, lacks&

8ow let me be very e(!licit about what I meant by sayin this shows a difference between !eo!le and machines& I meant that it is possi(le to !ro ram a machine to do a routine task in such a way that the machine will never notice even the most obvious facts about what it is doin = but it is inherent in human consciousness to notice some facts about the thin s one is doin & 0ut you knew this all alon & If you !unch -"- into an addin machine, and then add " to it, and then add " a ain, and a ain, and a ain, and continue doin so for hours and hours, the machine will never learn to antici!ate you, and do it itself, althou h any !erson would !ick u! the !ick u! the re!etitive behavior very quickly& Er, to take a silly e(am!le, a car will never !ick u! the idea, no matter how much or how well it is driven, that it is su!!osed to avoid other cars and obstacles on the road= and it will never learn even the most frequently traveled routes of its owner& The difference, then, is that it is possi(le for a machine to act unobservant= it is im!ossible for a human to act unobservant& 8otice I am not sayin that all machines are necessarily inca!able of makin so!histicated observations= .ust that some machines are& 8or am I sayin that all !eo!le are always makin so!histicated observations= !eo!le, in fact, are often very unobservant& 0ut machines can be made to be totally unobservant= many !eo!le cannot& )nd in fact, most machines made so far are !retty close to bein totally unobservant& Probably for this reason, the !ro!erty of bein = unobservant seems to be the characteristic feature of machines, to most !eo!le& 'or e(am!le, if somebody says that some task is -mechanical-, is does not mean that !eo!le are inca!able of doin the task= it im!lies thou h, that only a machine could do it over and over without ever com!lainin , or feelin bored&

3um!ing out of the System

It is an inherent !ro!erty of intelli ence that it can .um! out of the task which it is !erformin , and survey what it has done= it is always lookin for and often findin , !atterns& 8ow I said that an intelli ence can .um! out of its task, but that does not mean that it always will& ?owever, a little !rom!tin will often suffice& 'or e(am!le, a human bein who is readin a book may row slee!y& Instead of continuin to read until the book is finished he is .ust as likely to !ut the book aside and turn off the li ht& ?e has ste!!ed -out of the system- and yet it seems the most natural thin in the world to us& Er, su!!ose !erson ) is watchin television when !erson 0 comes in the room, and shows evident dis!leasure with the situation& Person ) may think he understands the !roblem, and try to remedy it by e(itin the !resent system 4that television !ro ram5, and fli!!in the channel knob, lookin for a better show& Person 0 may have a more radio conce!t of what it is to -e(it the system-namely to turn the television off& Ef course, there are cases where only a rare individual will have the vision to !erceive a system which overns many !eo!les lives, a system which has never before even been reco ni<ed as a system= then such !eo!le often devote their lives to convincin other !eo!le that the system really is there and that it ou ht to be e(ited from2 ?ow well have com!uters been tau ht to .um! out of the system? I will cite one e(am!le which sur!rised some observers& In a com!uter chess tournament not lon a o in 7anada, one !ro ramthe weakest of all the com!etin oneshad the unusual feature

of quittin lon before the ame was over& It was not a very ood chess !layer, but it at least had the redeemin quality of bein able to s!ot a ho!eless !osition, and to resi n then and there, instead of waitin for the other !ro ram to o throu h the borin ritual of checkmatin & )lthou h it lost every ame it !layed, it did it in style& ) lot of local chess e(!erts were im!ressed& Thus, if you define -the system- as -makin moves in a chess ame-, it is clear that this !ro ram had a so!histicated, !re!ro rammed ability to e(it from the system& En the other hand, if you think of -the system- as bein -whatever the com!uter had been !ro rammed to do-, then there is no doubt that the com!uter had no ability whatsoever to e(it from that system& It is very im!ortant when studyin formal systems to distin uish workin within the system from makin statements or observations a(out the system& I assume that you be an the /F-!u<<le, as do most !eo!le, by workin within the system= and that you then radually started ettin an(ious, and this an(iety finally built u! to the !oint where without any need for further consideration, you e(ited from the system, tryin to take stock of what you had !roduced, and wonderin why it was that you had not succeeded in !roducin M&& Perha!s you found a reason why you could not !roduce M&= that is thinkin about the system& Perha!s you !roduced MI& somewhere alon the way= that is workin within the system& 8ow I do not want to make it sound as if the two modes are entirely incom!atible= I am sure that every human bein is ca!able to some e(tent of workin inside a system and simultaneously thinkin about what he is doin & )ctually, in human affairs, it is often ne(t to im!ossible to break thin s neatly u! into -inside the system- and -outside the system-= life is com!osed of so many interlockin and interwoven and often inconsistent -systems- that it may seem sim!listic to think of thin s in those terms& 0ut it is often im!ortant to formulate sim!le ideas very clearly so that one can use them as models in thinkin about more com!le( ideas& )nd that is why I am showin you formal systems= and it is about time we went back to discussin the /IFsystem&

M-Mode) I-Mode) &-Mode

The /F-!u<<le was stated in such a way that it encoura ed some amount of e(!loration within the /IF-system-derivin theorems& 0ut it was also stated in a way so as not to im!ly that stayin inside the system would necessarily yield fruit& Therefore it encoura ed some oscillation between the two modes of work& Ene way to se!arate these two modes would be to have two sheets of !a!er= on one sheet, you work -in your ca!acity as a machine-, thus fillin it with nothin but M1s, I1s, and &1s= on the second sheet, you work -in your ca!acity as a thinkin bein -, and are allowed to do whatever your intelli ence su estswhich mi ht involve usin En lish, sketchin ideas, workin backwards, usin shorthand 4such as the letter 1915, com!ressin several ste!s into one, modifyin the rules of the system to see what that ives, or whatever else you mi ht dream u!& Ene thin you mi ht do is notice that the numbers % and B !lay an im!ortant role, since I1s are otten rid of in three1s, and &1s in two1sand doublin of len th 4e(ce!t for the M5 is allowed by rule II& So the second sheet mi ht also have some fi urin on it&

3e will occasionally refer back to these two modes of dealin with a formal system, and we will call them the 'echanical mode '%mode5 and the 3ntelligent mode 3%mode5& To round out our mode with one for each letter of the /IF-system, I will also mention a final modethe 5n%mode 5%mode5, which is the 9en way of a!!roachin thin s& /ore about this in a few 7ha!ters&

ecision ,rocedures
)n observation about this !u<<le is that it involves rules of two o!!osite tendenciesthe lengthening rules and the shortening rules& Two rules 4I and II5 allow you to increase the si<e of strin s 4but only in very ri id, !rescribed ways, of course5= and two others allow you to shrink strin s somewhat 4a ain in very ri id ways5& There seems to be an endless variety to the order in which these different ty!es of rules mi ht be a!!lied, and this ives ho!e that one way or another, M& could be !roduced& It mi ht involve len thenin the strin to some i antic si<e, and then e(tractin !iece after !iece until only two symbols are left= or, worse yet, it mi ht involve successive sta es of len thenin and then shortenin and then len thenin and then shortenin , and so on& 0ut there is no uarantee of it& )s a matter of fact, we already observed that & cannot be !roduced at all and it will make no difference if you len then and shorten till kin dom come& Still, the case of & and the case of M& seem quite different& It is by very su!erficial feature of & that we reco ni<e the im!ossibility of !roducin it, it doesn1t be in with an M 4whereas all theorems must5& It is very convenient to have such a sim!le way to detect nontheorems& ?owever who says that that test will detect all nontheorems? There may be lots of strin s which be in with M but are not !roducible& /aybe M& is one of them& That would mean that the -first-letter test- is of limited usefulness able only to detect a !ortion of the nontheorems, but missin others& 0ut there remains the !ossibility of some more elaborate test which discriminates !erfectly between those strin s which can be !roduced by the rules and those which cannot& ?ere we have to face the question, -3hat do mean by a test?- It may not be obvious why that question makes sense, of im!ortant, in this conte(t& 0ut I will ive an e(am!le of a -test- which somehow seems to violate the s!irit of the word& Ima ine a enie who has all the time in the world, and who en.oys usin it to !roduce theorems of the /IF-system, in a rather methodical way& ?ere, for instance, is a !ossible way the enie mi ht o about it Ste! ", )!!ly every a!!licable rule to the a(iom MI& This yields two new theorems MI&, MII& Ste! B, )!!ly every a!!licable rule to the theorems !roduced in ste! "& This yields three new theorems, MII&, MI&I&, MIIII&

Ste! %, )!!ly every a!!licable rule to the theorems !roduced in ste! B& This yields five new theorems, MIIII&, MII&II&, MI&I&I&I&, MIIIIIIII, M&I& This method !roduces every sin le theorem sooner or later, because the rules are a!!lied in every conceivable order& 4See 'i & ""&5 )ll of the len thenin -shortenin alternations which we mentioned above eventually et carried out& ?owever, it is not clear how lon to wait for a iven strin to a!!ear on this list, since theorems are listed accordin to the

73G5=E BB! * s)stematicall) constructed "tree" of all the theorems of the '35%s)stem! The 8th le#el down contains those theorems whose deri#ations contain e9actl) 8 steps! The encircled num(ers tell which rule was emplo)ed! 3s M& an)where in this tree$

shortness of their derivations& This is not a very useful order, if you are interested in a s!ecific strin 4such as M&5, and you don1t even know if it has any derivation, much less how lon that derivation mi ht be& 8ow we state the !ro!osed -theoremhood-test-, 3ait until the strin in question is !roduced= when that ha!!ens, you know it is a theoremand if it never ha!!ens, you know that it is not a theorem& This seems ridiculous, because it !resu!!oses that we don1t mind waitin around literally an infinite len th of time for our answer& This ets to the cru( of the matter of what should count as a -test-& Ef !rime im!ortance is a uarantee that we will et our answer in a finite len th of time& If there is a test for theoremhood, a test which does always terminate in a finite amount of time, then that test is called a decision procedure for the iven formal system& 3hen you have a decision !rocedure, then you have a very concrete characteri<ation of the nature of all theorems in the system& Effhand, it mi ht seem that the rules and a(ioms of the formal system !rovide no less com!lete a characteri<ation of

the theorems of the system than a decision !rocedure would& The tricky word here is -characteri<ation-& 7ertainly the rules of inference and the a(ioms of the /IF-system do characteri<e, im!licitly, those strin s that are theorems& Even more im!licitly, they characteri<e those strin s that are not theorems& 0ut im!licit characteri<ation is not enou h, for many !ur!oses& If someone claims to have a characteri<ation of all theorems, but it takes him infinitely lon to deduce that some !articular strin is not a theorem, you would !robably tend to say that there is somethin lackin in that characteri<ationit is not quite concrete enou h& )nd that is why discoverin that a decision !rocedure e(ists is a very im!ortant ste!& 3hat the discovery means, in effect, is that you can !erform a test for theoremhood of a strin , and that, even if the test is com!licated, it is guaranteed to terminate& In !rinci!le, the test is .ust as easy, .ust as mechanical, .ust as finite, .ust as full of certitude, as checkin whether the first letter of the strin is M& ) decision !rocedure is a -litmus test- for theoremhood2 Incidentally, one requirement on formal systems is that the set of a9ioms must be characteri<ed by a decision !rocedurethere must be a litmus test for a(iomhood& This ensures that there is no !roblem in ettin off the round at the be innin , at least& That is the difference between the set of a(ioms and the set of theorems, the former always has a decision !rocedure, but the latter may not& I am sure you will a ree that when you looked at the /IF-system for the first time, you had to face this !roblem e(actly& The lone a(iom was known, the rules of inference were sim!le, so the theorems had been im!licitly characteri<edand yet it was still quite unclear what the consequences of that characteri<ation were& In !articular, it was still totally unclear whether M& is, or is not, a theorem&

73G5=E BF! Sky 7astle, () '! 8!: Escher woodcut, BIFE/!

Two-Part Invention
or, >hat the Tortoise -aid to *chilles by Aewis 7arroll"
)chilles had overtaken the Tortoise, and had seated himself comfortably on its back& -So you1ve ot to the end of our race-course?- said the Tortoise& -Even thou h it :EES consist of an infinite series of distances? I thou ht some wiseacre or other had !roved that the thin couldn1t be done?-It 7)8 be done,- said )chilles& -It ?)S been done2 -ol#itur am(ulando& Hou see the distances were constantly :I/I8IS?I86= and so-0ut if they had been constantly I87;E)SI86?- the Tortoise interru!ted& -?ow then?-Then I shouldn1t be here,- )chilles modestly re!lied= -and HEF would have ot several times round the world, by this time2-Hou flatter me'A)TTE8, I mean,- said the Tortoise= -for you );E a heavy wei ht, and 8E mistake2 3ell now, would you like to hear of a race-course, that most !eo!le fancy they can et to the end of in two or three ste!s, while it ;E)AAH consists of an infinite number of distances, each one lon er than the !revious one?-Gery much indeed2- said the 6recian warrior, as he drew from his helmet 4few 6recian warriors !ossessed PE7@ETS in those days5 an enormous note-book and !encil& -Proceed2 )nd s!eak SAE3AH, !lease2 S?E;T?)8: isn1t invented yet2-That beautiful 'irst Pro!osition by Euclid2- the Tortoise murmured dreamily& -Hou admire Euclid?-Passionately2 So far, at least, as one 7)8 admire a treatise that won1t be !ublished for some centuries to come2-3ell, now, let1s take a little bit of the ar ument in that 'irst Pro!osition .ust T3E ste!s, and the conclusion drawn from them& @indly enter them in your note-book& )nd in order to refer to them conveniently, let1s call them ), 0, and 9, 4)5 Thin s that are equal to the same are equal to each other& 405 The two sides of this Trian le are thin s that are equal to the same& 495 The two sides of this Trian le are equal to each other& ;eaders of Euclid will rant, I su!!ose, that 9 follows lo ically from ) and 0, so that any one who acce!ts ) and 0 as true, /FST acce!t 9 as true?-Fndoubtedly2 The youn est child in a ?i h Schoolas soon as ?i h Schools are invented, which will not be till some two thousand years laterwill rant T?)T&-)nd if some reader had 8ET yet acce!ted ) and 0 as true, he mi ht still acce!t the SEMFE87E as a G)AI: one, I su!!ose?-

-8o doubt such a reader mi ht e(ist& ?e mi ht say, 1I acce!t as true the hy!othetical !ro!osition that, I' ) and 0 be true, 9 must be true= but I :E81T acce!t ) and ) as true&1 Such a reader would do wisely in abandonin Euclid, and takin to football&-)nd mi ht there not )ASE be some reader who would say 1I acce!t ) and 0 as true, but I :E81T acce!t the ?y!othetical1?-7ertainly there mi ht& ?E, also, had better take to football&-)nd 8EIT?E; of these readers,- the Tortoise continued, -is )S HET under any lo ical necessity to acce!t 9 as true?-Muite so,- )chilles assented& -3ell, now, I want you to consider /E as a reader of the SE7E8: kind, and to force me, lo ically, to acce!t 9 as true&-) tortoise !layin football would be- )chilles was be innin & -an anomaly, of course,- the Tortoise hastily interru!ted& -:on1t wander from the !oint& Aet1s have 9 first, and football afterwards2-I1m to force you to acce!t 9, am I?- )chilles said musin ly& -)nd your !resent !osition is that you acce!t ) and 0, but you :E81T acce!t the ?y!othetical-Aet1s call it 7,- said the Tortoise& -but you :E81T acce!t 475 If ) and 0 are true, 9 must be true&-That is my !resent !osition,- said the Tortoise& -Then I must ask you to acce!t 7&-I1ll do so,- said the Tortoise, -as soon as you1ve entered it in that notebook of yours& 3hat else have you ot in it?-Enly a few memoranda,- said )chilles, nervously flutterin the leaves, -a few memoranda ofof the battles in which I have distin uished myself2-Plenty of blank leaves, I see2- the Tortoise cheerily remarked& -3e shall need them )AA2- 4)chilles shuddered&5 -8ow write as I dictate, 4)5 Thin s that are equal to the same are equal to each other& 405 The two sides of this Trian le are thin s that are equal to the same& 475 If ) and 0 are true, 9 must be true& 495 The two sides of this Trian le are equal to each other&-Hou should call it :, not 9,- said )chilles& -It comes 8ETT to the other three& If you acce!t ) and 0 and 7, you /FST acce!t 9& -)nd why must I?-0ecause it follows AE6I7)AAH from them& If ) and 0 and 7 are true, 9 /FST be true& Hou can1t dis!ute T?)T, I ima ine?-If ) and 0 and 7 are true, 9 /FST be true,- the Tortoise thou htfully re!eated& -That1s )8ET?E; ?y!othetical, isn1t it? )nd, if I failed to see its truth, I mi ht acce!t ) and 0 and 7, and STIAA not acce!t 9, mi htn1t I?-

-Hou mi ht,- the candid hero admitted= -thou h such obtuseness would certainly be !henomenal& Still, the event is PESSI0AE& So I must ask you to rant E8E more ?y!othetical&-Gery ood, I1m quite willin to rant it, as soon as you1ve written it down& 3e will call it 4:5 If ) and 0 and 7 are true, 9 must be true& ?ave you entered that in your note-book?-I ?)GE2- )chilles .oyfully e(claimed, as he ran the !encil into its sheath& -)nd at last we1ve ot to the end of this ideal race-course2 8ow that you acce!t ) and 0 and 7 and :, E' 7EF;SE you acce!t 9&-:o I?- said the Tortoise innocently& -Aet1s make that quite clear& I acce!t ) and 0 and 7 and :& Su!!ose I STIAA refused to acce!t 9?-Then Ao ic would take you by the throat, and 'E;7E you to do it2- )chilles trium!hantly re!lied& -Ao ic would tell you, 1Hou can1t hel! yourself& 8ow that you1ve acce!ted ) and 0 and 7 and :, you /FST acce!t 921 So you1ve no choice, you see&-3hatever AE6I7 is ood enou h to tell me is worth 3;ITI86 :E38,- said the Tortoise& -So enter it in your book, !lease& 3e will call it 4E5 If ) and 0 and 7 and : are true, 9 must be true& Fntil I1ve ranted T?)T, of course I needn1t rant 9& So it1s quite a 8E7ESS);H ste!, you see?-I see,- said )chilles= and there was a touch of sadness in his tone& ?ere the narrator, havin !ressin business at the 0ank, was obli ed to leave the ha!!y !air, and did not a ain !ass the s!ot until some months afterwards& 3hen he did so, )chilles was still seated on the back of the much-endurin Tortoise, and was writin in his notebook, which a!!eared to be nearly full& The Tortoise was sayin , -?ave you ot that last ste! written down? Fnless I1ve lost count, that makes a thousand and one& There are several millions more to come& )nd 3EFA: you mind, as a !ersonal favor, considerin what a lot of instruction this colloquy of ours will !rovide for the Ao icians of the 8ineteenth 7entury3EFA: you mind ado!tin a !un that my cousin the /ockTurtle will then make, and allowin yourself to be renamed T)F6?T-FS?-)s you !lease,- re!lied the weary warrior, in the hollow tones of des!air, as he buried his face in his hands& -Provided that HEF, for HEF; !art, will ado!t a !un the /ock-Turtle never made, and allow yourself to be re-named ) @IAA-E)SE2-

C/A,TE$ II

Meaning and +orm in Mathematics

T?IS Two%&art 3n#ention was the ins!iration for my two characters& Just as Aewis 7arroll took liberties with 9eno1s Tortoise and )chilles, so have I taken liberties with Aewis 7arroll1s Tortoise and )chilles& In 7arroll1s dialo ue, the same events take !lace over and over a ain, only each time on a hi her and hi her level= it is a wonderful analo ue to 0ach1s Ever-;isin 7anon& The 7arrollian :ialo ue, with its wit subtracted out, still leaves a dee! !hiloso!hical !roblem, .o words and thoughts follow formal rules, or do the) not$ That !roblem is the !roblem of this book& In this 7ha!ter and the ne(t, we will look at several new formal systems& This will ive us a much wider !ers!ective on the conce!t of formal system& 0y the end of these two 7ha!ters, you should have quite a ood idea of the !ower of formal systems, and why they are of interest to mathematicians and lo icians&

The !q-System
The formal system of this 7ha!ter is called the p"%s)stem& It is not im!ortant to mathematicians or lo iciansin fact, it is .ust a sim!le invention of mine& Its im!ortance lies only in the fact that it !rovides an e(cellent e(am!le of many ideas that !lay a lar e role in this book& There are three distinct symbols of the !q-system, ! q

the letters !, q, and the hy!hen& The !q-system has an infinite number of a(ioms& Since we can1t write them all down, we have to have some other way of describin what they are& )ctually, we want more than .ust a descri!tion of the a(ioms= we want a way to tell whether some iven strin is an a(iom or not& ) mere descri!tion of a(ioms mi ht characteri<e them fully and yet weakly which was the !roblem with the way theorems in the /IF-system were characteri<ed& 3e don1t want to have to stru le for an indeterminate!ossibly infinite len th of time, .ust to find out if some strin is an a(iom or not& Therefore, we will define a(ioms in such a way that there is an obvious decision !rocedure for a(iomhood of a strin com!osed of !1s, q1s, and hy!hens&

:E'I8ITIE8, 9!-q9 is an a(iom, whenever 9 is com!osed of hy!hens only& 8ote that 191 must stand for the same strin of hy!hens in both occurrences 'or e(am!le, --!-q--- is an a(iom& The literal e(!ression 19!-q9-1 is not an a(iom, of course 4because 191 does not belon to the !q-system5= it is more like a mold in which all a(ioms are cast and it is called an a(iom schema& The !q-system has only one rule of !roduction, ;FAE, Su!!ose 9, ), and @ all stand for !articular strin s containin only hy!hens& )nd su!!ose that 9!)q@ is known to be a theorem& Then 9!)-q@- is a theorem& 'or e(am!le, take 9 to be 1--1, ) to be 1---1, and @ to be 1-1& The rule tells us, If --!---q- turns out to be a theorem, then so will --!----q--& )s is ty!ical of rules of !roduction, the statement establishes a causal connection between the theoremhood of two strin s, but without assertin theoremhood for either one on its own& ) most useful e(ercise for you is to find a decision !rocedure for the theorems of the !q-system& It is not hard= if you !lay around for a while you will !robably !ick it u!& Try it&

The

ecision ,rocedure

I !resume you have tried it& 'irst of all, thou h it may seem too obvious to mention, I would like to !oint out that every theorem of the !q-system has three se!arate rou!s of hy!hens, and the se!aratin elements are one !, and one q, in that order& 4This can be shown by an ar ument based on -heredity-, .ust the way one could show that all /IFsystem theorems had to be in with M&5 This means that we can rule out, from its form alone, a strin such as --!--!--!--q--------& 8ow, stressin the !hrase -from its form alone- may seem silly= what else is there to a strin e(ce!t its form? 3hat else could !ossibly !lay a role in determinin its !ro!erties? 7learly nothin could& 0ut bear this in mind as the discussion of formal systems oes on= the notion of -form- will start to et rather more com!licated and abstract, and we will have to think more about the meanin of the word -form-& In any case, let us ive the name well formed string to any strin which be ins with a hy!henrou!, then has one !, then has a second hy!hen- rou!, then a q, and then a final hy!henrou!& 0ack to the decision !rocedure&&& The criterion for theoremhood is that the first two hy!hen- rou!s should add u!, in len th, to the third hy!hen- rou!& for instance , --!--

q---- is a theorem, since B !lus B equals C, whereas --!--q-is not, since B !lus B is not "& To see why this is the !ro!er criterion, look first at the a(iom schema& Ebviously, it only manufactures a(ioms which satisfy the addition criterion& Second, look at the rule of !roduction& If the first strin satisfies the addition criterion, so must the second oneand conversely, if the first strin does not satisfy the addition criterion, then neither does the second strin & The rule makes the addition criterion into a hereditary !ro!erty of theorems, any theorem !asses the !ro!erty on to its offs!rin & This shows why the addition criterion is correct& There is, incidentally, a fact about the !q-system which would enable us to say with confidence that it has a decision !rocedure, even before findin the addition criterion& That fact is that the !q-system is not com!licated by the o!!osin currents of lengthening and shortening rules= it has only len thenin rules& )ny formal system which tells you how to make lon er theorems from shorter ones, but never the reverse, has ot to have a decision !rocedure for its theorems& 'or su!!ose you are iven a strin & 'irst check whether it1s an a(iom or not 4I am assumin that there is a decision !rocedure for a(iomhoodotherwise, thin s are ho!eless5& If it is an a(iom, then it is by definition a theorem, and the test is over& So su!!ose instead that it1s not an a(iom& Then, to be a theorem, it must have come from a shorter strin , via one of the rules& 0y oin over the various rules one by one, you can !in!oint not only the rules that could conceivably !roduce that strin , but also e(actly which shorter strin s could be its forebears on the -family tree-& In this way, you -reduce- the !roblem to determinin whether any of several new but shorter strin s is a theorem& Each of them can in turn be sub.ected to the same test& The worst that can ha!!en is a !roliferation of more and more, but shorter and shorter, strin s to test& )s you continue inchin your way backwards in this fashion, you must be ettin closer to the source of all theoremsthe a(iom schemata& Hou .ust can1t et shorter and shorter indefinitely= therefore, eventually either you will find that one of your short strin s is an a(iom, or you1ll come to a !oint where you1re stuck, in that none of your short strin s is an a(iom, and none of them can be further shortened by runnin some rule or other backwards& This !oints out that there really is not much dee! interest in formal systems with len thenin rules only= it is the inter!lay of len thenin and shortenin rules that ives formal systems a certain fascination&&

Bottom-u! #s& To!-do'n

The method above mi ht be called a top%down decision !rocedure, to be contrasted with a (ottom%up decision !rocedure, which I ive now& It is very reminiscent of the enie1s systematic theorem- eneratin method for the /IF-system, but is com!licated by the !resence of an a(iom schema& 3e are oin to form a -bucket- into which we throw theorems as they are enerated& ?ere is how it is done, 4"a5 Throw the sim!lest !ossible a(iom (-!-q--5 into the bucket& 4"b5 )!!ly the rule of inference to the item in the bucket, and !ut the result into the bucket& 4Ba5 Throw the second-sim!lest a(iom into the bucket&

4Bb5 )!!ly the rule to each item in the bucket, and throw all results into the bucket& 4%a5 Throw the third-sim!lest a(iom into the bucket& 4%b5 )!!ly the rule to each item in the bucket, and throw all results into the bucket& etc&, etc& ) moment1s reflection will show that you can1t fail to !roduce every theorem of the !qsystem this way& /oreover, the bucket is ettin filled with lon er and lon er theorems, as time oes on& It is a ain a consequence of that lack of shortenin rules& So if you have a !articular strin , such as --!---q-----, which you want to test for theoremhood, .ust follow the numbered ste!s, checkin all the while for the strin in question& If it turns u! theorem2 If at some !oint everythin that oes into the bucket is lon er than the strin in question, for et itit is not a theorem& This decision !rocedure is bottom-u! because it is workin its way u! from the basics, which is to say the a(ioms& The !revious decision !rocedure is to!-down because it does !recisely the reverse, it works its way back down towards the basics&

Isomor!hisms Induce Meaning

8ow we come to a central issue of this 7ha!terindeed of the book& Perha!s you have already thou ht to yourself that the !q-theorems are like additions& The strin --!---q--- is a theorem because B !lus % equals L& It could even occur to you that the theorem --!---q-is a statement, written in an odd notation, whose meaning is that B !lus % is L& Is this a reasonable way to look at thin s? 3ell, I deliberately chose 1 !1 to remind you of 1!lus1, and 1q1 to remind you of 1equals1&&& So, does the strin --!---q---- actually mean -B !lus % equals L-? 3hat would make us feel that way? /y answer would be that we have !erceived an isomorphism between !q-theorems and additions& In the Introduction, the word -isomor!hism- was defined as an information !reservin transformation& 3e can now o into that notion a little more dee!ly, and see it from another !ers!ective& The word -isomor!hism1 a!!lies when two com!le( structures can be ma!!ed onto each other, in such a way that to each !art of one structure there is a corres!ondin !art in the other structure, where -corres!ondin - means that the two !art !lay similar roles in their res!ective structures& This usa e of the word -isomor!hism- is derived from a more !recise notion in mathematics& It is cause for .oy when a mathematician discovers an isomor!hism between two structures which he knows& It is often a -bolt from the blue-, and a source of wonderment& The !erce!tion of an isomor!hism between two known structures is a si nificant advance in knowled eand I claim that it is such !erce!tions of isomor!hism which create meanings in the minds of !eo!le& ) final word on the !erce!tion of isomor!hisms, since they come in many sha!es and si<es, fi uratively s!eakin , it is not always totally clear when you really have found an isomor!hism& Thus, -isomor!hism- is

a word with all the usual va ueness of wordswhich is a defect but an advanta e as well& In this case, we have an e(cellent !rototy!e for the conce!t of isomor!hism& There is a -lower level- of our isomor!hismthat is, a ma!!in between the !arts of the two structures, ! q --- !lus equals one two three etc&

This symbol-word corres!ondence has a name, interpretation& Secondly, on a hi her level, there is the corres!ondence between true statements and theorems& 0utnote carefullythis hi her-level corres!ondence could not be !erceived without the !rior choice of an inter!retation for the symbols& Thus it would be more accurate to describe it as a corres!ondence between true statements and interpreted theorems& In any case we have dis!layed a two-tiered corres!ondence, which is ty!ical of all isomor!hisms& 3hen you confront a formal system you know nothin of, and if you ho!e to discover some hidden meanin in it, your !roblem is how to assi n inter!retations to its symbols in a meanin ful waythat is, in such a way that a hi her-level corres!ondence emer es between true statements and theorems& Hou may make several tentative stabs in the dark before findin a ood set of words to associate with the symbols& It is very similar to attem!ts to crack a code, or to deci!her inscri!tions in an unknown lan ua e like Ainear 0 of 7rete, the only way to !roceed is by trial and error, based on educated uesses& 3hen you hit a ood choice, a -meanin ful- choice, all of a sudden thin s .ust feel ri ht, and work s!eeds u! enormously& Pretty soon everythin falls into !lace& The e(citement of such an e(!erience is ca!tured in The .ecipherment of +inear B by John 7hadwick& 0ut it is uncommon, to say the least, for someone to be in the !osition of -decodin - a formal system turned u! in the e(cavations of a ruined civili<ation2 /athematicians 4and more recently, lin uists, !hiloso!hers, and some others5 are the only users of formal systems, and they invariably have an inter!retation in mind for the formal systems which they use and !ublish& The idea of these !eo!le is to set u! a formal system whose theorems reflect some !ortion of reality isomor!hically& In such a case, the choice of symbols is a hi hly motivated one, as is the choice of ty!o ra!hical rules of !roduction& 3hen I devised the !q-system, I was in this !osition& Hou see why I chose the symbols I chose& It is no accident theorems are isomor!hic to additions= it ha!!ened because I deliberately sou ht out a way to reflect additions ty!o ra!hically&

Meaningless and Meaningful Inter!retations

Hou can choose inter!retations other than the one I chose& Hou need make every theorem come out true& 0ut there would be very little reason make an inter!retation in which, say, all theorems came out false, certainly even less reason to make an inter!retation under which there is no correlation at all, !ositive or ne ative, between theoremhood and truth& Aet us therefore make a distinction between two ty!es of inter!retations a formal system& 'irst, we can have a meaningless inter!retation, one in which we fail to see any isomor!hic connection between theorems of system, and reality& Such inter!retations aboundany random choice a will do& 'or instance, take this one, ! horse q ha!!y - a!!le 8ow -!-q-- acquires a new inter!retation, -a!!le horse a!!le hat a!!le a!!le-a !oetic sentiment, which mi ht a!!eal to horses, and mi ht even lead them to favor this mode of inter!retin !q-strin s2 ?owever, this inter!retation has very little -meanin fulness-= under inter!retation, theorems don1t sound any truer, or any better, than nontheorems& ) horse mi ht en.oy -ha!!y ha!!y ha!!y a!!le horse- 4ma!!ed onto q q q-!5 .ust as much as any inter!reted theorem& The other kind of inter!retation will be called meaningful& Fnder such an inter!retation, theorems and truths corres!ondthat is, an isomor!hism e(ists between theorems and some !ortion of reality& That is why it is ood to distin uish between interpretations and meanings& )ny old word can be used as an inter!retation for 1 !1, but 1!lus1 is the only meaningful choice we1ve come u! with& In summary, the meanin of 1 !1 seems to be 1!lus1 thou h it can have a million different inter!retations&

Acti"e #s! ,assi"e Meanings

Probably the most si nificant fact of this 7ha!ter, if understood dee!ly this, the !qsystem seems to force us into reco ni<in that s)m(ols of a formal s)stem, though initiall) without meaning, cannot a#oid taking on "meaning" of sorts at least if an isomorphism is found! The difference between meanin it formal system and in a lan ua e is a very im!ortant one, however& It is this, in a lan ua e, when we have learned a meanin for a word, we then make new statements based on the meanin of the word& In a sense the meanin becomes acti#e, since it brin s into bein a new rule for creatin sentences& This means that our command of lan ua e is not like a finished !roduct, the rules for makin sentences increase when we learn new meanin s& En the other hand, in a formal system, the theorems are !redefined, by the rules of !roduction& 3e can choose -meanin s- based on an isomor!hism 4if we can find one5 between theorems and true statements& 0ut this does not ive us the license to o out and add new theorems to the

established theorems& That is what the ;equirement of 'ormality in 7ha!ter I was warnin you of& In the /IF-system, of course, there was no tem!tation to o beyond the four rules, because no inter!retation was sou ht or found& 0ut here, in our new system, one mi ht be seduced by the newly found -meanin - of each symbol into thinkin that the strin --!--!--!--q------is a theorem& )t least, one mi ht wish that this strin were a theorem& 0ut wishin doesn1t chan e the fact that it isn1t& )nd it would be a serious mistake to think that it -must- be a theorem, .ust because B !lus B !lus B !lus B equals *& It would even be misleadin to attribute it any meanin at all, since it is not well-formed, and our meanin ful inter!retation is entirely derived from lookin at well-formed strin s& In a formal system, the meanin must remain passi#e= we can read each strin accordin to the meanin s of its constituent symbols, but we do not have the ri ht to create new theorems !urely on the basis of the meanin s we1ve assi ned the symbols& Inter!reted formal systems straddle the line between systems without meanin , and systems with meanin & Their strin s can be thou ht of as -e(!ressin - thin s, but this must come only as a consequence of the formal !ro!erties of the system&

ouble-Entendre*
)nd now, I want to destroy any illusion about havin found the meanin s for the symbols of the !q-system& 7onsider the followin association, ! equals q taken from - one -- two etc& 8ow, --!---q----- has a new inter!retation, -B equals % taken from L-& Ef course it is a true statement& )ll theorems will come out true under this new inter!retation& It is .ust as meanin ful as the old one& Ebviously, it is silly to ask, -0ut which one is the meanin of the strin ?- )n inter!retation will be meanin ful to the e(tent that it accurately reflects some isomor!hism to the real world& 3hen different as!ects of the real world are isomor!hic to each other 4in this case, additions and subtractions5, one sin le formal system can be isomor!hic to both, and therefore can take on two !assive meanin s& This kind of double-valuedness of symbols at strin s is an e(tremely im!ortant !henomenon& ?ere it seems trivial, curious, annoyin & 0ut it will come back in dee!er conte(ts and brin with it a reat richness of ideas&

?ere is a summary of our observations about the !q-system& Fnder either of the two meanin ful inter!retations iven, every well-formed strin has a rammatical assertion for its counter!artsome are true, some false& The idea of well%formed strings in any formal system is that they are those strin s which, when inter!reted symbol for symbol, yield grammatical sentences& 4Ef course, it de!ends on the inter!retation, but usually, there one in mind&5 )mon the well-formed strin s occur the theorems& The, are defined by an a(iom schema, and a rule of !roduction& /y oal in inventin the !qsystem was to imitate additions, I wanted every theorem to e(!ress a true addition under inter!retation= conversely, I wanted every true addition of !recisely two !ositive inte ers to be translatable into a strin , which would be a theorem& That oal was achieved& 8otice, therefore, that all false additions, such as -B !lus % equals D-, are ma!!ed into strin s which are well-formed, but which are not theorems&

+ormal Systems and $eality

This is our first e(am!le of a case where a formal system is based u!on !ortion of reality, and seems to mimic it !erfectly, in that its theorems are isomor!hic to truths about that !art of reality& ?owever, reality and the formal systems are inde!endent& 8obody need be aware that there is a isomor!hism between the two& Each side stands by itselfone !lus or equals two, whether or not we know that -!-q-- is a theorem= and -!-q-- is still a theorem whether or not we connect it with addition& Hou mi ht wonder whether makin this formal system, or any formal system, sheds new li ht on truths in the domain of its inter!retation& ?ave we learned any new additions by !roducin !q-theorems? 7ertainly not= but we have learned somethin about the nature of addition as a !rocessnamely, that it is easily mimicked by a ty!o ra!hical rule overnin meanin less symbols& This still should not be a bi sur!rise since addition is such a sim!le conce!t& It is a common!lace that addition can be ca!tured in the s!innin ears of a device like a cash re ister& 0ut it is clear that we have hardly scratched the surface, as far formal systems o= it is natural to wonder about what !ortion of reality can be imitated in its behavior by a set of meanin less symbols overned by formal rules& 7an all of reality be turned into a formal system? In a very broad sense, the answer mi ht a!!ear to be yes& Ene could su est, for instance, that reality is itself nothin but one very com!licated formal system& Its symbols do not move around on !a!er, but rather in a three-dimensional vacuum 4s!ace5= they are the elementary !articles of which everythin is com!osed& 4Tacit assum!tion, that there is an end to the descendin chain of matter, so that the e(!ression -elementary !articles- makes sense&5 The -ty!o ra!hical rules- are the laws of !hysics, which tell how, iven the !ositions and velocities of all !articles at a iven instant, to modify them, resultin in a new set of !ositions and velocities belon in to the -ne(t- instant& So the theorems of this rand formal system are the !ossible confi urations of !articles at different times in the history of the universe& The sole a(iom is 4or !erha!s, was5 the ori inal confi uration of all the !articles at the -be innin of time-& This is so randiose a conce!tion, however, that it has only the most theoretical interest= and besides, quantum mechanics 4and other !arts of !hysics5 casts at least some

doubt on even the theoretical worth of this idea& 0asically, we are askin if the universe o!erates deterministically, which is an o!en question&

Mathematics and Symbol Mani!ulation

Instead of dealin with such a bi !icture, let1s limit ourselves to mathematics as our -real world-& ?ere, a serious question arises, ?ow can we be sure, if we1ve tried to model a formal system on some !art of mathematics, that we1ve done the .ob accurately es!ecially if we1re not one hundred !er cent familiar with that !ortion of mathematics already? Su!!ose the oal of the formal system is to brin us new knowled e in that disci!line& ?ow will we know that the inter!retation of every theorem is true, unless we1ve !roven that the isomor!hism is !erfect? )nd how will we !rove that the isomor!hism is !erfect, if we don1t already know all about the truths in the disci!line to be in with? Su!!ose that in an e(cavation somewhere, we actually did discover some mysterious formal system& 3e would try out various inter!retations and !erha!s eventually hit u!on one which seemed to make every theorem come out true, and every nontheorem come out false& 0ut this is somethin which we could only check directly in a finite number of cases& The set of theorems is most likely infinite& ?ow will we know that all theorems e(!ress truths under this inter!retation, unless we know everythin there is to know about both the formal system and the corres!ondin domain of inter!retation? It is in somewhat this odd !osition that we will find ourselves when we attem!t to match the reality of natural numbers 4i&e&, the nonne ative inte ers, +, ", B,&&&5 with the ty!o ra!hical symbols of a formal system& 3e will try to understand the relationshi! between what we call -truth- in number theory and what we can et at by symbol mani!ulation& So let us briefly look at the basis for callin some statements of number theory true, and others false& ?ow much is "B times "B? Everyone knows it is "CC& 0ut how many of the !eo!le who ive that answer have actually at any time in their lives drawn a "B by "B rectan le, and then counted the little squares in it? /ost !eo!le would re ard the drawin and countin unnecessary& They would instead offer as !roof a few marks on !a!er, such as are shown below, "B W "B ]]] BC "B ]]] "CC

)nd that would be the -!roof-& 8early everyone believes that if you counted the squares, you would et "CC of them= few !eo!le feel that outcome is in doubt& The conflict between the two !oints of view comes into shar!er focus when you consider the !roblem of determinin the value #*$DLC%B" ( "B%CLD$*#& 'irst of all, it is virtually im!ossible to construct the a!!ro!riate rectan le= and what is worse, even if it were constructed and hu e armies of !eo!le s!ent centuries countin the little squares, only a very ullible !erson would be willin to believe their final answer& It is .ust too likely that somewhere, somehow, somebody bobbled .ust a little bit& So is it ever !ossible to know what the answer is? If you trust the symbolic !rocess which involves mani!ulatin di its accordin to certain sim!le rules, yes& That !rocess is !resented to children as a device which ets ri ht answer= lost in the shuffle, for many children, are the rhyme reason of that !rocess& The di it-shuntin laws for multi!lication are based mostly on a few !ro!erties of addition and multi!lication which are assumed to hold for all numbers&

The Basic La's of Arithmetic

The kind of assum!tion I mean is illustrated below& Su!!ose that you lay down a few sticks,

I II II II

8ow you count them& )t the same time, somebody else counts them, startin from the other end& Is it clear that the two of you will et the same answer? The result of a countin !rocess is inde!endent of the way in which it is done& This is really an assum!tion about what countin is& It would be senseless to try to !rove it, because it is so basic= either you see it or you don1tbut in the latter case, a !roof won1t hel! you a bit& 'rom this kind of assum!tion, one can et to the commutativity and associativity of addition 4i&e&, first that ( P c Q c P ( always, and second that ( P c P d/ Q ( P c/ P d always5& The same assum!tion can also you to the commutativity and associativity of multi!lication= .ust think of many cubes assembled to form a lar e rectan ular solid& /ulti!licative commutativity and associativity are .ust the assum!tions that when you rotate the solid in various ways, the number of cubes will not chan e& 8ow these assum!tions are not verifiable in all !ossible cases, because the number of such cases is infinite& 3e take them for ranted= we believe them 4if we ever think about them5 as dee!ly as we could believe anythin & The amount of money in our !ocket will not chan e as we walk down the street, .ostlin it u! and down= the number of books we have will not chan e if we !ack them u! in a bo(, load them into our car, drive one hundred miles, unload the bo(, un!ack it, and !lace the books in a new shelf& )ll of this is !art of what we mean by num(er& There are certain ty!es of !eo!le who, as soon as some undeniable fact is written down, find it amusin to show why that -fact- is false after all& I am such a !erson, and as soon as I had written down the e(am!les above involvin sticks, money, and books, I invented situations in which they were wron & Hou may have done the same& It oes to

show that numbers as abstractions are really quite different from the everyday numbers which we use& Peo!le en.oy inventin slo ans which violate basic arithmetic but which illustrate -dee!er- truths, such as -" and " make "- 4for lovers5, or -" !lus " !lus " equals "- 4the Trinity5& Hou can easily !ick holes in those slo ans, showin why, for instance, usin the !lus-si n is ina!!ro!riate in both cases& 0ut such cases !roliferate& Two raindro!s runnin down a window!ane mer e= does one !lus one make one? ) cloud breaks u! into two cloudsmore evidence for the same? It is not at all easy to draw a shar! line between cases where what is ha!!enin could be called -addition-, and where some other word is wanted& If you think about the question, you will !robably come u! with some criterion involvin se!aration of the ob.ects in s!ace, and makin sure each one is clearly distin uishable from all the others& 0ut then how could one count ideas? Er the number of ases com!risin the atmos!here? Somewhere, if you try to look it u!, you can !robably find a statement such as, -There are "$ lan ua es in India, and CDB dialects&There is somethin stran e about !recise statements like that, when the conce!ts -lan ua e- and -dialect- are themselves fu<<y&

Ideal ;umbers
8umbers as realities misbehave& ?owever, there is an ancient and innate sense in !eo!le that numbers ou ht not to misbehave& There is somethin clean and !ure in the abstract notion of number, removed from countin beads, dialects, or clouds= and there ou ht to be a way of talkin about numbers without always havin the silliness of reality come in and intrude& The hard-ed ed rules that overn -ideal- numbers constitute arithmetic, and their more advanced consequences constitute number theory& There is only one relevant question to be asked, in makin the transition from numbers as !ractical thin s to numbers as formal thin s& Ence you have decided to try to ca!suli<e all of number theory in an ideal system, is it really !ossible to do the .ob com!letely? )re numbers so clean and crystalline and re ular that their nature can be com!letely ca!tured in the rules of a formal system? The !icture +i(eration 4'i & "%5, one of Escher1s most beautiful, is a marvelous contrast between the formal and the informal, with a fascinatin transition re ion& )re numbers really as free as birds? :o they suffer as much from bein crystalli<ed into a rule-obeyin system? Is there a ma ical transition re ion between numbers in reality and numbers on !a!er? 3hen I s!eak of the !ro!erties of natural numbers, I don1t .ust mean !ro!erties such as the sum of a !articular !air of inte ers& That can be found out by countin , and anybody who has rown u! in this century cannot doubt the mechani<ability of such !rocesses as countin , addin , multi!lyin , and so on& I mean the kinds of !ro!erties which mathematicians are interested in e(!lorin , questions for which no countin !rocess is sufficient to !rovide the answernot even theoretically sufficient& Aet us take a classic e(am!le of such a !ro!erty of natural numbers& The statement is, -There are infinitely many !rime numbers&- 'irst of all, there is no countin !rocess which will ever be able to confirm, or refute, this assertion& The best we could do would be to count !rimes for a while and concede that there are -a lot-& 0ut no amount of countin alone

would ever resolve the question of whether the number of !rimes is finite or infinite& There could always be more& The statementand it is called -Euclid1s Theorem- 4notice the ca!ital -T-5is quite unobvious& It may seem reasonable, or a!!ealin , but it is not obvious& ?owever, mathematicians since Euclid have always called it true& 3hat is the reason?

73G5=E BH! Aiberation, () '!8! Escher lithograph, BIGG/!

Euclid%s ,roof
The reason is that reasoning tells them it is so& Aet us follow the reasonin involved& 3e will look at a variant of Euclid1s !roof& This !roof works by showin that whatever number you !ick, there is a !rime lar er than it& Pick a number N& /ulti!ly all the !ositive inte ers startin with " and endin with N= in other words, form the factorial of N, written -N2-& 3hat you et is divisible by every number u! to N& 3hen you add " to N2, the result can1t be a multi!le of B 4because it leaves " over, when you divide by B5= can1t be a multi!le of % 4because it leaves I over, when you divide by %5= can1t be a multi!le of C 4because it leaves " over, when you divide by C5=

. . .
can1t be a multi!le of N 4because it leaves " over, when you divide by N5= In other words, N2 ^ ", if it is divisible at all 4other than by " and itself only is divisible by numbers reater than N& So either it is itself !rime, or !rime divisors are reater than N! 0ut in either case we1ve shown the must e(ist a !rime above N& The !rocess holds no matter what number is& 3hatever 8 is, there is a !rime reater than N& )nd thus ends the demonstration of the infinitude of the !rimes& This last ste!, incidentally, is called generali@ation, and we will meet a ain later in a more formal conte(t& It is where we !hrase an ar ument terms of a sin le number 4N5, and then !oint out that N was uns!ecified and therefore the ar ument is a eneral one& Euclid1s !roof is ty!ical of what constitutes -real mathematics-& It is sim!le, com!ellin , and beautiful& It illustrates that by takin several rash short ste!s one can et a lon way from one1s startin !oint& In our case, the startin !oints are basic ideas about multi!lication and division and forth& The short ste!s are the ste!s of reasonin & )nd thou h every individual ste! of the reasonin seems obvious, the end result is not obvious& 3e can never check directly whether the statement is true or not, yet we believe it, because we believe in reasonin & If you acce!t reasonin , there seems to be no esca!e route= once you a ree to hear Euclid out, you1ll have to a ree with his conclusion& That1s most fortunatebecause it means that mathematicians will always a ree on what statements to label -true- and what statements to label -false-& This !roof e(em!lifies an orderly thou ht !rocess& Each statement related to !revious ones in an irresistible way& This is why it is called -!roof- rather than .ust - ood evidence-& In mathematics the oal always to ive an ironclad !roof for some unobvious

statement& The very fact of the ste!s bein linked to ether in an ironclad way su ests that there may be a patterned structure bindin these statements to ether& This structure can best be e(!osed by findin a new vocabularya styli<ed vocabulary, consistin of symbolssuitable only for e(!ressin statements about numbers& Then we can look at the !roof as it e(ists in its translated version& It will be a set of statements which are related, line by line, in some detectable way& 0ut the statements, since they1re re!resented by means a small and styli<ed set of symbols, take on the as!ect of !atterns& In other words, thou h when read aloud, they seem to be statements about numbers and their !ro!erties, still when looked at on !a!er, they seem to be abstract !atternsand the lineby-line structure of the !roof may start to look like slow transformation of !atterns accordin to some few ty!o ra!hical rules&

Getting Around Infinity

)lthou h Euclid1s !roof is a !roof that all numbers have a certain !ro!erty it avoids treatin each of the infinitely many cases se!arately& It ets around it by usin !hrases like -whatever N is-, or -no matter what number N is-& 3e could also !hrase the !roof over a ain, so that it uses the !hrase -all N-& 0y knowin the a!!ro!riate conte(t and correct ways of usin such !hrases, we never have to deal with infinitely many statements& 3e deal with .ust two or three conce!ts, such as the word -all-which, thou h themselves finite, embody an infinitude= and by usin them, we sideste! the a!!arent !roblem that there are an infinite number of facts we want to !rove& 3e use the word -all- in a few ways which are defined by the thou ht !rocesses of reasonin & That is, there are rules which our usa e of -all- obeys& 3e may be unconscious of them, and tend to claim we o!erate on the basis of the meaning of the word= but that, after all, is only a circumlocution for sayin that we are uided by rules which we never make e(!licit& 3e have used words all our lives in certain !atterns, and instead of callin the !atterns -rules-, we attribute the courses of our thou ht !rocesses to the -meanin s- of words& That discovery was a crucial reco nition in the lon !ath towards the formali<ation of number theory& If we were to delve into Euclid1s !roof more and more carefully, we would see that it is com!osed of many, many smallalmost infinitesimal ste!s& If all those ste!s were written out line after line, the !roof would a!!ear incredibly com!licated& To our minds it is clearest when several ste!s are telesco!ed to ether, to form one sin le sentence& If we tried to look at the !roof in slow motion, we would be in to discern individual frames& In other words, the dissection can o only so far, and then we hit the -atomic- nature of reasonin !rocesses& ) !roof can be broken down into a series of tiny but discontinuous .um!s which seem to flow smoothly when !erceived from a hi her vanta e !oint& In 7ha!ter GIII, I will show one way of breakin the !roof into atomic units, and you will see how incredibly many ste!s are involved& Perha!s it should not sur!rise you, thou h& The o!erations in Euclid1s brain when he invented the !roof must have involved millions of neurons 4nerve cells5, many of which fired several hundred times in a sin le second& The mere utterance of a sentence involves hundreds of thousands of neurons& If Euclid1s thou hts were that com!licated, it makes sense for his

!roof to contain a hu e number of ste!s2 4There may be little direct connection between the neural actions in his brain, and a !roof in our formal system, but the com!le(ities of the two are com!arable& It is as if nature wants the com!le(ity of the !roof of the infinitude of !rimes to be conserved, even when the systems involved are very different from each other&5 In 7ha!ters to come, we will lay out a formal system that 4"5 includes a styli<ed vocabulary in which all statements about natural numbers can be e(!ressed, and 4B5 has rules corres!ondin to all the ty!es of reasonin which seem necessary& ) very im!ortant question will be whether the rules for symbol mani!ulation which we have then formulated are really of equal !ower 4as far as number theory is concerned5 to our usual mental reasonin abilitiesor, more enerally, whether it is theoretically !ossible to attain the level of our thinkin abilities, by usin some formal system&

Sonata for Unaccompanied Achilles

The telephone ringsR *chilles picks it up!
*chilles, ?ello, this is )chilles& *chilles, Eh, hello, /r& T& ?ow are you? *chilles, ) torticollis? Eh, I1m sorry to hear it& :o you have any idea what caused it? *chilles, ?ow lon did you hold it in that !osition? *chilles, 3ell, no wonder it1s stiff, then& 3hat on earth induced you kee! your neck twisted that way for so lon ? *chilles, 3ondrous many of them, eh? 3hat kinds, for e(am!le? *chilles, 3hat do you mean, -!hantasma orical beasts-?
73G5=E BD! /osaic II, () '! 8! Escher lithograph, BIGC/!

*chilles, 3asn1t it terrifyin to see so many of them at the same time? *chilles, ) uitar2? Ef all thin s to be in the midst of all those weird creatures& Say, don1t you !lay the uitar? *chilles, Eh, well, it1s all the same to me&

*chilles, Hou1re ri ht= I wonder why I never noticed that difference between fiddles and uitars before& S!eakin of fiddlin , how would you like to come over and listen to one of the sonatas for unaccom!anied violin by your favorite com!oser, J& S& 0ach? I .ust bou ht a marvelous recordin of them& I still can1t et over the way 0ach uses a sin le violin to create a !iece with such interest& *chilles, ) headache too? That1s a shame& Perha!s you should .ust o to bed& *chilles, I see& ?ave you tried countin shee!? *chilles, Eh, oh, I see& Hes, I fully know what you mean& 3ell, if it1s T?)T distractin , !erha!s you1d better tell it to me, and let me try to work on it, too& *chilles, ) word with the letters 1)1, 1:1, 1)1, 171 consecutively inside it&&& ?mm&&& 3hat about -abracadabra-? *chilles, True, -):)7- occurs backwards, not forwards, in that word& *chilles, ?ours and hours? It sounds like I1m in for a lon !u<<le, then& 3here did you hear this infernal riddle? *chilles, Hou mean he looked like he was meditatin on esoteric 0uddhist matters, but in reality he was .ust tryin to think u! com!le( word !u<<les? *chilles, )ha2the snail knew what this fellow was u! to& 0ut how did you come to talk to the snail? *chilles, Say, I once heard a word !u<<le a little bit like this one& :o you want to hear it? Er would it .ust drive you further into distraction? *chilles, I a reecan1t do any harm& ?ere it is, 3hat1s a word that be ins with the letters -?E- and also ends with -?E-? *chilles, Gery in eniousbut that1s almost cheatin & It1s certainly not what I meant2 *chilles, Ef course you1re ri htit fulfills the conditions, but it1s a sort of -de eneratesolution& There1s another solution which I had in mind& *chilles, That1s e(actly it2 ?ow did you come u! with it so fast? *chilles, So here1s a case where havin a headache actually mi ht have hel!ed you, rather than hinderin you& E(cellent2 0ut I1m still in the dark on your -):)7- !u<<le& *chilles, 7on ratulations2 8ow maybe you1ll be able to et to slee!2 So tell me, what is the solution? *chilles, 3ell, normally I don1t like hints, but all ri ht& 3hat1s your hint? *chilles, I don1t know what you mean by -fi ure- and - round- in this case& *chilles, 7ertainly I know 'osaic 332 I know )AA of Escher1s works& )fter all, he1s my favorite artist& In any case, I1ve ot a !rint of 'osaic 33 han in on my wall, in !lain view from here& *chilles, Hes, I see all the black animals& *chilles, Hes, I also see how their -ne ative s!ace-what1s left outdefines the white animals& *chilles, So T?)T1S what you mean by -fi ure- and - round-& 0ut what does that have to do with the -):)7- !u<<le? *chilles, Eh, this is too tricky for me& I think I1/ startin to et a headache& *chilles, Hou want to come over now? 0ut I thou ht *chilles, Gery well& Perha!s by then I1ll have thou ht of the ri ht answer to HEF; !u<<le, usin your fi ure- round hint, relatin it to /H !u<<le&

*chilles, I1d love to !lay them for you& *chilles, Hou1ve invented a theory about them? *chilles, )ccom!anied by what instrument? *chilles, 3ell, if that1s the case, it seems a little stran e that he would have written out the har!sichord !art, then, and had it !ublished as well& *chilles, I seesort of an o!tional feature& Ene could listen to them either waywith or without accom!animent& 0ut how would one know what the accom!animent is su!!osed to sound like? *chilles, )h, yes, I uess that it is best, after all, to leave it to the listener1s ima ination& )nd !erha!s, as you said, 0ach never even had accom!animent in mind at all& Those sonatas seem to work very indeed as they are& *chilles, ;i ht& 3ell, I1ll see you shortly& *chilles, 6ood-bye, /r& T&

C/A,TE$ III

+igure and Ground

,rimes #s! Com!osites
T?E;E IS ) stran eness to the idea that conce!ts can be ca!tured by sim!le ty!o ra!hical mani!ulations& The one conce!t so far ca!tured is that of addition, and it may not have a!!eared very stran e& 0ut su!!ose the oal were to create a formal system with theorems of the form ,9, the letter 191 standin for a hy!hen-strin , and where the only such theorems would be ones in which the hy!hen-strin contained e(actly a !rime number of hy!hens& Thus, ,--- would be a theorem, but ,---- would not& ?ow could this be done ty!o ra!hically? 'irst, it is im!ortant to s!ecify clearly what is meant by t)pographical o!erations& The com!lete re!ertoire has been !resented in the /IF-system and the !q-system, so we really only need to make a list of the kinds of thin s we have !ermitted, 4"5 4B5 4%5 4C5 4L5 4D5 readin and reco ni<in any of a finite set of symbols= writin down any symbol belon in to that set= co!yin any of those symbols from one !lace to another= erasin any of those symbols= checkin to see whether one symbol is the same as another= kee!in and usin a list of !reviously enerated theorems&

The list is a little redundant, but no matter& 3hat is im!ortant is that it clearly involves only trivial abilities, each of them far less than the ability to distin uish !rimes from non!rimes& ?ow, then, could we com!ound some of these o!erations to make a formal system in which !rimes are distin uished from com!osite numbers?

The tq-System
) first ste! mi ht be to try to solve a sim!ler, but related, !roblem& 3e could try to make a system similar to the !q-system, e(ce!t that it re!resents multi!lication, instead of addition& Aet1s call it the t"%s)stem, 1t1 for 1times1& /ore s!ecifically, su!!ose S, Y, and N are, res!ectively, the numbers of hy!hens in the hy!hen-strin s 9, ), and @& 48otice I am takin s!ecial !ains to distin uish between a strin and the number of hy!hens it contains&5 Then we wish the strin 9 t ) q @ to be a theorem if and only if S times Y equals N& 'or instance, --t---q----- should be a theorem because B times % equals D, but --t--q--should not be a theorem& The tq-system can be characteri<ed .ust about as easily as the !q-system namely, by usin .ust one a(iom schema and one rule of inference,

)TIE/ S7?E/), 9 t - q 9 is an a(iom, whenever 9 is a hy!hen strin & ;FAE E' I8'E;E87E, Su!!ose that 9, ), and @ are all hy!hen-strin s& )nd su!!ose that 9 t ) q @ is an old theorem& Then, 9 t ) - q @ 9 is a new theorem& 0elow is the derivation of the theorem --t---q-----4"5 4B5 4%5 --t-q---t--q-----t---q------4a(iom5 4by rule of inference, usin line 4"5 as the old theorem5 4by rule of inference, usin line 4B5 as the old theorem5

8otice how the middle hy!hen-strin rows by one hy!hen each time the rule of inference is a!!lied= so it is !redictable that if you want a theorem with ten hy!hens in the middle, you a!!ly the rule of inference nine times in a row&

Ca!turing Com!ositeness
/ulti!lication, a sli htly trickier conce!t than addition, has now been -ca!turedty!o ra!hically, like the birds in Escher1s +i(eration& 3hat about !rimeness? ?ere1s a !lan that mi ht seem smart, usin the tq-system define a new set of theorems of the form C(, which characteri<e composite numbers, as follows, ;FAE, Su!!ose 9, ), and @ are hy!hen-strin s& If 9 - t ) - q @ is a theorem then C@ is a theorem& This works by sayin that N 4the number of hy!hens in @5 is com!osite as lon as it is the !roduct of two numbers reater than "namely, S ^ " 4the number of hy!hens in 9-5, and Y ^ " 4the number of hy!hens in )-5& I am defendin this new rule by ivin you some -Intelli ent mode- .ustifications for it& That is because you are a human bein , and want to know wh) there is such a rule& If you were o!eratin e(clusively in the -/echanical mode-, you would not need any .ustification, since /-mode workers .ust follow the rules mechanically and ha!!ily, never questionin = them2 0ecause you work in the I-mode, you will tend to blur in your mind the distinction between strin s and their inter!retations& Hou see, thin s can become quite confusin as soon as you !erceive -meanin - in the symbol which you are mani!ulatin & Hou have to fi ht your own self to kee! from thinkin that the string 1%%%1 is the num(er %& The ;equirement of 'ormality, which in 7ha!ter I !robably seemed !u<<lin 4because it seemed so obvious5, here becomes tricky, and crucial& It is the essential thin which kee!s you from mi(in u! the I-mode with the /-mode= or said another way, it kee!s you from mi(in u! arithmetical facts with ty!o ra!hical theorems&

Illegally Characteri:ing ,rimes

It is very tem!tin to .um! from the 7-ty!e theorems directly to P-ty!e theorems, by !ro!osin a rule of the followin kind, P;EPESE: ;FAE, Su!!ose 9 is a hy!hen-strin & If C9 is not a theorem, then ,9 is a theorem& The fatal flaw here is that checkin whether C9 is not a theorem is not an e(!licitly ty!o ra!hical o!eration& To know for sure that M& is not a theorem of the /IF-system, you have to o outside of the system&&& and so it is with this Pro!osed ;ule& It is a rule which violates the whole idea of formal systems, in that it asks you to o!erate informally that is, outside the system& Ty!o ra!hical o!eration 4D5 allows you to look into the stock!ile of !reviously found theorems, but this Pro!osed ;ule is askin you to look into a hy!othetical -Table of 8ontheorems-& 0ut in order to enerate such a table, you would have to do some reasonin outside the s)stemOreasonin which shows why various strin s cannot be enerated inside the system& 8ow it may well be that there is another formal system which can enerate the -Table of 8ontheorems-, by !urely ty!o ra!hical means& In fact, our aim is to find .ust such a system& 0ut the Pro!osed ;ule is not a ty!o ra!hical rule, and must be dro!!ed& This is such an im!ortant !oint that we mi ht dwell on it a bit more& In our 8% s)stem 4which includes the tq-system and the rule which defines 7-ty!e theorems5, we have theorems of the form C9, with 191 standin , as usual, for a hy!hen-strin & There are also nontheorems of the form C9& 4These are what I mean when I refer to -nontheorems-, althou h of course t t - C q q and other ill-formed messes are also nontheorems&5 The difference is that theorems have a com!osite number of hy!hens, nontheorems have a !rime number of hy!hens& 8ow the theorems all have a common -form-, that is, ori inate from a common set of ty!o ra!hical rules& :o all nontheorems also have a common -form-, in the same sense? 0elow is a list of 7-ty!e theorems, shown without their derivations& The !arenthesi<ed numbers followin them sim!ly count the hy!hens in them& C---- 4C5 C------ 4D5 C-------- 4*5 C--------- 4#5 C---------- 4"+5 C------------ 4"B5 C-------------- 4"C5 C--------------- 4"L5 C---------------- 4"D5 C------------------ 4"*5

...

The -holes- in this list are the nontheorems& To re!eat the earlier question, :o the holes also have some -form- in common? 3ould it be reasonable to say that merely by virtue of bein the holes in this list, they share a common form? Hes and no& That they share some ty!o ra!hical quality is undeniable, but whether we want to call it -form- is unclear& The reason for hesitatin is that the holes are only negati#el) definedthey are the thin s that are left out of a list which is positi#el) defined&

+igure and Ground

This recalls the famous artistic distinction between figure and ground& 3hen a fi ure or -!ositive s!ace- 4e& &, a human form, or a letter, or a still life5 is drawn inside a frame, an unavoidable consequence is that its com!lementary sha!ealso called the - round-, or -back round-, or -ne ative s!ace-has also been drawn& In most drawin s, however, this fi ure round relationshi! !lays little role& The artist is much less interested in round than in the fi ure& 0ut sometimes, an artist will take interest in round as well& There are beautiful al!habets which !lay with this fi ure- round distinction& ) messa e written in such an al!habet is shown below& )t first it looks like a collection of somewhat random blobs, but if you ste! back a ways and stare at it for a while, all of a sudden, you will see seven letters a!!ear in this&&&

73G5=E BG!

'or a similar effect, take a look at my drawin -moke -ignal 4'i & "%#5& )lon these lines, you mi ht consider this !u<<le, can you somehow create a drawin containin words in both the fi ure and the round? Aet us now officially distin uish between two kinds of fi ures, cursi#el) drawa(le ones, and recursi#e ones 4by the way, these are my own terms are not in common usa e5& ) cursi#el) drawa(le fi ure is one whose round is merely an accidental by-!roduct of the drawin act& ) recursi#e fi ure is one whose round can be seen as a fi ure in its own ri ht& Fsually this is quite deliberate on the !art of the artist& The -re- in -recursive- re!resents the fact that both fore round and back round are cursively drawablethe fi ure is -twice-cursive-& Each fi ure- round boundary in a recursive fi ure is a double-ed ed sword& /& 7& Escher was a master at drawin recursive fi ures see, for instance, his beautiful recursive drawin of birds 4'i & "D5&

73G5=E BJ! Tilin of the !lane usin birds, () '! 8! Escher from a BIDF note(ook/!

Eur distinction is not as ri orous as one in mathematics, for who can definitively say that a !articular round is not a fi ure? Ence !ointed out, almost any round has interest of its own& In that sense, every fi ure is recursive& 0ut that is not what I intended by the term& There is a natural and intuitive notion of reco ni<able forms& )re both the fore round and back round reco ni<able forms? If so, then the drawin is recursive& If you look at the rounds of most line drawin s, you will find them rather unreco ni<able& This demonstrates that There e(ist reco ni<able forms whose ne ative s!ace is not any reco ni<able form& In more -technical- terminolo y, this becomes, There e(ist cursively drawable fi ures which are not recursive& Scott @im1s solution to the above !u<<le, which I call his -'I6F;E-'I6F;E 'i ure-, is shown in 'i ure "$& If you read both black and white, you will see -'I6F;E-

everywhere, but -6;EF8:- nowhere2 It is a !ara on of recursive fi ures& In this clever drawin , there are two nonequivalent ways of characteri<in the black re ions, 4"5 as the negati#e space to the white re ions= 4B5 as altered copies of the white re ions 4!roduced by colorin and shiftin each white re ion5& 4In the s!ecial case of the 'I6F;E-'I6F;E 'i ure, the two characteri<ations are equivalent but in most black-and-white !ictures, they would not be&5 8ow in 7ha!ter GIII, when we create our Ty!o ra!hical 8umber Theory 4T8T5, it will be our ho!e that the set of all false statements of number theory can be characteri<ed in two analo ous ways, 4"5 as the ne ative s!ace to the set of all T8T-theorems= 4B5 as altered co!ies of the set of all T8T-theorems 4!roduced by ne atin each T8T-theorem5& 0ut this ho!e will be dashed, because, 4"5 4B5 inside the set of all nontheorems are found some truths outside the set of all ne ated theorems are found some falsehoods&

Hou will see why and how this ha!!ens, in 7ha!ter TIG& /eanwhile, !onder over a !ictorial re!resentation of the situation 4'i & "*5&

+igure and Ground in Music

Ene may also look for fi ures and rounds in music& Ene analo ue is the distinction between melody and accom!animentfor the melody is always in the forefront of our attention, and the accom!animent is subsidiary, in some sense& Therefore it is sur!risin when we find, in the lower lines of a !iece of music, reco ni<able melodies& This does not ha!!en too often in !ost-baroque music& Fsually the harmonies are not thou ht of as fore round& 0ut in baroque musicin 0ach above all-the distinct lines, whether hi h or low or in between, all act as -fi ures-& In this sense, !ieces by 0ach can be called -recursive-&

73G5=E BC! 'I6F;E-'I6F;E 'i ure, () -cott E! ;im BICG/!

)nother fi ure- round distinction e(ists in music, that between on-beat and offbeat& If you count notes in a measure -one-and, two-and, three-and, four-and-, most melody-notes will come on numbers, not on -and- 1s& 0ut sometimes, a melody will be deliberately !ushed onto the -and-1s, for the sheer effect of it& This occurs in several etudes for the !iano by 7ho!in, for instance& It also occurs in 0ach!articularly in his Sonatas and Partitas for unaccom!anied violin, and his Suites for unaccom!anied cello& There, 0ach mana es to et two or more musical lines oin simultaneously& Sometimes he does this by havin the solo instrument !lay -double sto!s-two notes at once& Ether times, however, he !uts one voice on the on-beats, and the other voice on the off-beats,

73G5=E BE! 8onsidera(le #isual s)m(olism is featured in this diagram of the relation (etween #arious classes of TNT strings! The (iggest (o9 represents the set of all TNT strings The ne9t%(iggest (o9 represents the set of all well%formed TNT strings! >ithin it is found the set of all sentences of TNT! Now things (egin to get interesting! The set of theorems is pictured as a tree growing out of a trunk representing the set of a9ioms/! The tree%s)m(ol chosen (ecause of the recursi#e growth pattern which it e9hi(its: new (ranches theorems constantl) sprouting from old ones! The fingerlike (ranches pro(e into the corners of constraining region the set of truths/, )et can ne#er full) occup) it! The (oundar) (eta the set of truths and the set of falsities is meant to suggest a randoml) meandering coastline which, no matter how closel) )ou e9amine it, alwa)s has finer le#els of structure, an conse"uentl) impossi(le to descri(e e9actl) in an) finite wa)! -ee B! 'andel(rot's (ook 'ractals!/ The reflected tree represents the set of ne ations of theorems: all of them false )et una(le collecti#el) to span the space of false statements! U.rawing () the author!V

so the ear se!arates them and hears two distinct melodies weavin in and out, and harmoni<in with each other& 8eedless to say, 0ach didn1t sto! at this level of com!le(ity&&&

$ecursi"ely Enumerable Sets #s! $ecursi"e Sets

8ow let us carry back the notions of fi ure and round to the domain formal systems& In our e(am!le, the role of !ositive s!ace is !layed by 7-ty!e theorems, and the role of ne ative s!ace is !layed by strin s with a !rime number of hy!hens& So far, the only way we have found to re!resent !rime numbers ty!o ra!hically is as a ne ative s!ace& Is there, however, some wayI don1t care how com!licatedof re!resentin the !rimes as a positi#e s!acethat is, as a set of theorems of some formal system? :ifferent !eo!le1s intuitions ive different answers here& I remember quite vividly how !u<<led and intri ued I was u!on reali<in the difference between a !ositive characteri<ation and a ne ative characteri<ation& I was quite convinced that not only the !rimes, but any set of numbers which could be re!resented ne atively, could also be re!resented !ositively& The intuition underlyin my belief is re!resented by the question, -,ow could a figure and its ground not carr) e9actl) the same information ?- They seemed to me to embody the same information, .ust coded in two com!lementary ways& 3hat seems ri ht to you? It turns out I was ri ht about the !rimes, but wron in eneral& This astonished me, and continues to astonish me even today& It is a fact that, There e(ist formal systems whose ne ative s!ace 4set of nontheorems5 is not the !ositive s!ace 4set of theorems5 of any formal system& This result, it turns out, is of de!th equal to 6>del1s Theoremso it is not sur!risin that my intuition was u!set& I, .ust like the mathematicians of the early twentieth century, e(!ected the world of formal systems and natural numbers to be more !redictable than it is& In more technical terminolo y, this becomes, There e(ist recursively enumerable sets which are not recursive& The !hrase recursi#el) enumera(le 4often abbreviated -r&e&-5 is the mathematical counter!art to our artistic notion of -cursively drawable-and recursi#e is the counter!art of -recursive-& 'or a set of strin s to be -r&e&- means that it can be enerated accordin to ty!o ra!hical rulesfor e(am!le, the set of 7-ty!e theorems, the set of theorems of the /IF-systemindeed, the set of theorems of any formal system& This could be com!ared with the conce!tion of a -fi ure- as -a set of lines which can be enerated accordin to artistic rules- 4whatever that mi ht mean25& )nd a -recursive setis like a fi ure whose round is also a fi ure-not only is it r&e&, but its com!lement is also r&e& It follows from the above result that, There e(ist formal systems for which there is no ty!o ra!hical decision !rocedure&

?ow does this follow? Gery sim!ly& ) ty!o ra!hical decision !rocedure is a method which tells theorems from nontheorems& The e(istence of such a test allows us to enerate all nontheorems systematically, sim!ly by oin down a list of all strin s and !erformin the test on them one at a time, discardin ill-formed strin s and theorems alon the way& This amounts to a ty!o ra!hical method for eneratin the set of nontheorems& 0ut accordin to the earlier statement 4which we here acce!t on faith5, for some systems this is not !ossible& So we must conclude that ty!o ra!hical decision !rocedures do not e(ist for all formal systems& Su!!ose we found a set 7 of natural numbers 4171 for 1'i ure15 which we could enerate in some formal waylike the com!osite numbers& Su!!ose its com!lement is the set G 4for 16round15like the !rimes& To ether 7 and G make u! all the natural numbers, and we know a rule for makin all the numbers in set 7, but we know no such rule for makin all the numbers in set G& It is im!ortant to understand that if the members of were always enerated in order of increasing si@e, then we could always characteri<e G& The !roblem is that many r&e& sets are enerated by methods which throw in elements in an arbitrary order, so you never know if a number which has been ski!!ed over for a lon time will et included if you .ust wait a little lon er& 3e answered no to the artistic question, -)re all fi ures recursive?- 3e have now seen that we must likewise answer no to the analo ous question in mathematics, -)re all sets recursive?- 3ith this !ers!ective, let us now come back to the elusive word -form-& Aet us take our fi ure-set and our round-set G a ain& 3e can a ree that all the numbers in set 7 have some common -form-but can the same be said about numbers in set G? It is a stran e question& 3hen we are dealin with an infinite set to start withthe natural numbersthe holes created by removin some subs may be very hard to define in any e(!licit way& )nd so it may be that they are not connected by any common attribute or -form-& In the last analysis it is a matter of taste whether you want to use the word -form-but .ust thinkin about it is !rovocative& Perha!s it is best not to define -form-, but to leave it with some intuitive fluidity& ?ere is a !u<<le to think about in connection with the above matter 7an you characteri<e the followin set of inte ers 4or its ne ative s!ace5 " % $ "B "* BD %L CL LD D#&&&

?ow is this sequence like the 'I6F;E-'I6F;E 'i ure?

,rimes as +igure $ather than Ground

'inally, what about a formal system for eneratin !rimes? ?ow is it done? The trick is to ski! ri ht over multi!lication, and to o directly to nondi#isi(ilit) as the thin to re!resent !ositively& ?ere are an a(iom schema and rule for !roducin theorems which re!resent the notion that one number does not di#ide 4 ; 5 another number e(actly, )TIE/ S7?E/), 9 ) ; 9 where 9 and ) are hy!hen-strin s&

'or e(am!le ----

--, where 9 has been re!laced by 1Y1 and ) by 1---1&

;FAE, If 9 ; ) is a theorem, then so is 9 ; 9 ) & If you use the rule twice, you can enerate this theorem, ----- ; ------------which is inter!reted as -L does not divide "B-& 0ut --- ; ------------ is not a theorem& 3hat oes wron if you try to !roduce it? 8ow in order to determine that a iven number is !rime, we have to build u! some knowled e about its nondivisibility !ro!erties& In !articular, we want to know that it is not divisible by B or % or C, etc&, all the way u! to " less than the number itself& 0ut we can1t be so va ue in formal systems as to say -et cetera-& 3e must s!ell thin s out& 3e would like to have a way of sayin , in the lan ua e of the system, -the number 9 is di#isor free u! to T-, meanin that no number between B and T divides 9& This can be done, but there is a trick to it& Think about it if you want& ?ere is the solution, ;FAE, If -- ; @ is a theorem, so is @ + --& ;FAE, If @ + 9 is a theorem and also 9- ; @ is a theorem, @ + 9 - is a theorem& These two rules ca!ture the notion of divisor freeness& )ll we need to do is to say that !rimes are numbers which are divisor-free u! to " less than themselves, ;FAE, If @- + @ is a theorem, then , @ - is a theorem& Ehlet1s not for et that B is !rime2 )TIE/, ,--. )nd there you have it& The !rinci!le of re!resentin !rimality formally is that there is a test for divisibility which can be done without any backtrackin & Hou march steadily u!ward, testin first for divisibility by B, then by %, and so on& It is this -monotonicity- or unidirectionalitythis absence of cross-!lay between len thenin and shortenin , increasin and decreasin that allows !rimality to be ca!tured& )nd it is this !otential com!le(ity of formal systems to involve arbitrary amounts of backwards-forwards interference that is res!onsible for such limitative results as 6>del1s Theorem, Turin 1s ?altin Problem, and the fact that not all recursively enumerable sets are recursive&

Contracrostipunctus
*chilles has come to #isit his friend and Mogging companion, the Tortoise, at his home *chilles, ?eavens, you certainly have an admirable boomeran collection Tortoise, Eh, !shaw& 8o better than that of any other Tortoise& )nd now would you like to ste! into the !arlor? *chilles, 'ine& 4>alks to the corner of the room &5 I see you also have a lar e collection of records& 3hat sort of music do you en.oy? Tortoise, Sebastian 0ach isn1t so bad, in my o!inion& 0ut these days, I must say, I am develo!in more and more of an interest in a rather s!eciali<ed sort of music& *chilles, Tell me, what kind of music is that? Tortoise, ) ty!e of music which you are most unlikely to have heard of& I call it -music to break !hono ra!hs by-& *chilles, :id you say -to break !hono ra!hs by-? That is a curious conce!t& I can .ust see you, sled ehammer in hand, whackin one !hono ra!h after another to !ieces, to the strains of 0eethoven1s heroic master!iece >ellington's <ictor)& Tortoise, That1s not quite what this music is about& ?owever, you mi ht find its true nature .ust as intri uin & Perha!s I should ive you a brief descri!tion of it? *chilles, E(actly what I was thinkin & Tortoise, ;elatively few !eo!le are acquainted with it& It all be an when my friend the 7rabhave you met him, by the way?!aid me a visit& *chilles, 1twould be a !leasure to make his acquaintance, I1m sure& Thou h I1ve heard so much about him, I1ve never met him& Tortoise, Sooner or later I1ll et the two of you to ether& Hou1d hit it off s!lendidly& Perha!s we could meet at random in the !ark on day&&& *chilles, 7a!ital su estion2 I1ll be lookin forward to it& 0ut you were oin to tell me about your weird -music to smash !hono ra!hs by-, weren1t you? Tortoise, Eh, yes& 3ell, you see, the 7rab came over to visit one day& Hou must understand that he1s always had a weakness for fancy ad ets, and at that time he was quite an aficionado for, of all thin s, record !layers& ?e had .ust bou ht his first record !layer, and bein somewhat ullible, believed every word the salesman had told him about itin !articular, that it was ca!able of re!roducin any and all sounds& In short, he was convinced that it was a Perfect !hono ra!h& *chilles, 8aturally, I su!!ose you disa reed& Tortoise, True, but he would hear nothin of my ar uments& ?e staunchly maintained that any sound whatever was re!roducible on his machine& Since I couldn1t convince him of the contrary, I left it at that& 0ut not lon after that, I returned the visit, takin with me a record of a son which I had myself com!osed& The son was called -I 7annot 0e Played on ;ecord Player "-& *chilles, ;ather unusual& 3as it a !resent for the 7rab? Tortoise, )bsolutely& I su ested that we listen to it on his new !hono ra!h, and he was very lad to obli e me& So he !ut it on& 0ut unfortunately, after only a few notes,

the record !layer be an vibratin rather severely, and then with a loud -!o!-, broke into a lar e number of fairly small !ieces, scattered all about the room& The record was utterly destroyed also, needless to say& *chilles, 7alamitous blow for the !oor fellow, I1d say& 3hat was the matter with his record !layer? Tortoise, ;eally, there was nothin the matter, nothin at all& It sim!ly couldn1t re!roduce the sounds on the record which I had brou ht him, because they were sounds that would make it vibrate and break& *chilles, Edd, isn1t it? I mean, I thou ht it was a Perfect !hono ra!h& That1s what the salesman had told him, after all& Tortoise, Surely, )chilles, you don1t believe everythin that salesmen tell you2 )re you as naPve as the 7rab was? *chilles, The 7rab was naPver by far2 I know that salesmen are notorious !revaricators& I wasn1t born yesterday2 Tortoise, In that case, maybe you can ima ine that this !articular salesman had somewhat e(a erated the quality of the 7rab1s !iece of equi!ment&&& !erha!s it was indeed less than Perfect, and could not re!roduce every !ossible sound& *chilles, Perha!s that is an e(!lanation& 0ut there1s no e(!lanation for the ama<in coincidence that your record had those very sounds on it&&& Tortoise, Fnless they ot !ut there deliberately& Hou see, before returnin the 7rab1s visit, I went to the store where the 7rab had bou ht his machine, and inquired as to the make& ?avin ascertained that, I sent off to the manufacturers for a descri!tion of its desi n& )fter receivin that by return mail, I analy<ed the entire construction of the !hono ra!h and discovered a certain set of sounds which, if they were !roduced anywhere in the vicinity, would set the device to shakin and eventually to fallin a!art& *chilles, 8asty fellow2 Hou needn1t s!ell out for me the last details, that you recorded those sounds yourself, and offered the dastardly item as a ift&&& Tortoise, 7lever devil2 Hou .um!ed ahead of the story2 0ut that wasn1t the end of the adventure, by any means, for the 7rab did not believe that his record !layer was at fault& ?e was quite stubborn& So he went out and bou ht a new record !layer, this one even more e(!ensive, and this time the salesman !romised to ive him double his money back in case the 7rab found a sound which it could not re!roduce e(actly& So the 7rab told me e(citedly about his new model, and I !romised to come over and see it& *chilles, Tell me if I1m wron I bet that before you did so, you on a ain wrote the manufacturer, and com!osed and recorded a new son called -I 7annot 0e Played on ;ecord Player B-, based on the construction of the new model& Tortoise, Ftterly brilliant deduction, )chilles& Hou1ve quite ot the s!irit& *chilles, So what ha!!ened this time? Tortoise, )s you mi ht e(!ect, !recisely the same thin & The !hono ra!h fell into innumerable !ieces, and the record was shattered& *chilles, 7onsequently, the 7rab finally became convinced that there could be no such thin as a Perfect record !layer&

Tortoise, ;ather sur!risin ly, that1s not quite what ha!!ened& ?e was sure that the ne(t model u! would fill the bill, and havin twice the money, he *chilles, EhoI have an idea2 ?e could have easily outwitted you, by obtainin a AE3fidelity !hono ra!hone that was not ca!able of re!roducin the sounds which would destroy it& In that way, he would avoid your trick& Tortoise, Surely, but that would defeat the ori inal !ur!osenamely, to have a !hono ra!h which could re!roduce any sound whatsoever, even its own selfbreakin sound, which is of course im!ossible& *chilles, That1s true& I see the dilemma now& If any record !layersay ;ecord Player T is sufficiently hi h-fidelity, then when attem!ts to !lay the son -I 7annot 0e Played on ;ecord Player T-, it will create .ust those vibrations which will cause to break&&& So it fails to be Perfect& )nd yet, the only way to et around that trickery, namely for ;ecord Player T to be of lower fidelity, even more directly ensures that it is not Perfect& It seems that every record !layer is vulnerable to one or the other of these frailties, and hence all record !layers are defective& Tortoise, I don1t see why you call them -defective-& It is sim!ly an inherent fact about record !layers that they can1t do all that you mi ht wish them to be able to do& 0ut if there is a defect anywhere, is not in T?E/, but in your e(!ectations of what they should be able to do2 )nd the 7rab was .ust full of such unrealistic e(!ectations& *chilles, 7om!assion for the 7rab overwhelms me& ?i h fidelity or low fidelity, he loses either way& Tortoise, )nd so, our little ame went on like this for a few more rounds, and eventually our friend tried to become very smart& ?e ot wind of the !rinci!le u!on which I was basin my own records, and decided to try to outfo( me& ?e wrote to the !hono ra!h makers, and described a device of his own invention, which they built to s!ecification& ?e called it -;ecord Player Eme a-& It was considerably more so!histicated than an ordinary record !layer& *chilles, Aet me uess how, :id it have no movin !arts? Er was it made of cotton? Er Tortoise, Aet me tell you, instead& That will save some time& In the first !lace, ;ecord Player Eme a incor!orated a television camera whose !ur!ose it was to scan any record before !layin it& This camera was hooked u! to a small built-in com!uter, which would determine e(actly the nature of the sounds, by lookin at the roove!atterns& *chilles, Hes, so far so ood& 0ut what could ;ecord Player Eme a do with this information? Tortoise, 0y elaborate calculations, its little com!uter fi ured out what effects the sounds would have u!on its !hono ra!h& If it deduced that the sounds were such that they would cause the machine in its !resent confi uration to break, then it did somethin very clever& Eld Eme a contained a device which could disassemble lar e !arts of its !hono ra!h subunit, and rebuild them in new ways, so that it could, in effect, chan e its own structure& If the sounds were -dan erous-, a new confi uration was chosen, one to which the sounds would !ose no threat, and this new confi uration would then be built by the rebuildin subunit, under direction

of the little com!uter& Enly after this rebuildin o!eration would ;ecord Player Eme a attem!t to !lay the record& *chilles, )ha2 That must have s!elled the end of your tricks& I bet you were a little disa!!ointed& Tortoise, 7urious that you should think so&&& I don1t su!!ose that you know 6>del1s Incom!leteness Theorem backwards and forwards, do you? *chilles, @now 3?ESE Theorem backwards and forwards? I1ve never heard of anythin that sounds like that& I1m sure it1s fascinatin , but I1d rather hear more about -music to break records by-& It1s an amusin little story& )ctually, I uess I can fill in the end& Ebviously, there was no !oint in oin on, and so you shee!ishly admitted defeat, and that was that& Isn1t that e(actly it? Tortoise, 3hat2 It1s almost midni ht2 I1m afraid it1s my bedtime& I1d love to talk some more, but really I am rowin quite slee!y& *chilles, )s am I& 3ell, I1ll be on my way& 4 *s he reaches the door, he suddenl) stops, and turns around&5 Eh, how silly of me2 I almost for ot& I brou ht you a little !resent& ?ere& 4,ands the Tortoise a small neatl) wrapped package&5 Tortoise, ;eally, you shouldn1t have2 3hy, thank you very much indeed& I think I1ll o!en it now& 4Eagerl) tears open the package, and inside disco#ers a glass go(let&5 Eh, what an e(quisite oblet2 :id you know that I am quite an aficionado for, of all thin s, lass oblets? *chilles, :idn1t have the fo iest& 3hat an a reeable coincidence2 Tortoise, Say, if you can kee! a secret, I1ll let you in on somethin , I tryin to find a Perfect oblet, one havin no defects of a sort in its sha!e& 3ouldn1t it be somethin if this obletlet1s call it -6-were the one? Tell me, where did you come across 6oblet 6? *chilles, Sorry, but that1s /H little secret& 0ut you mi ht like to know who its maker is& Tortoise, Pray tell, who is it? *chilles, Ever hear of the famous lassblower Johann Sebastian 0ach? 3ell, he wasn1t e(actly famous for lassblowin but he dabbled at the art as a hobby, thou h hardly a soul knows itand this oblet is the last !iece he blew& Tortoise, Aiterally his last one? /y racious& If it truly was made by 0ach its value is inestimable& 0ut how are you sure of its maker? *chilles, Aook at the inscri!tion on the insidedo you see where the letters 101, 1)1, 171, 1?1 have been etched? Tortoise, Sure enou h2 3hat an e(traordinary thin & 46entl) sets Go(let G down on a shelf!5 0y the way, did you know that each of the four letters in 0ach1s name is the name of a musical note? *chilles, 1tisn1t !ossible, is it? )fter all, musical notes only o from _)1 throu h 161& Tortoise, Just so= in most countries, that1s the case& 0ut in 6ermany, 0ach1s own homeland, the convention has always been similar, e(ce!t that what we call 101, they call 1?1, and what we call 10-flat1, they call 101& 'or instance, we talk about 0ach1s -/ass in 0 /inor-, whereas they talk about his -?-moll /esse-& Is that clear?

*chilles,&&& hmm&&& I uess so& It1s a little confusin , ? is 0, and 0 0-flat& I su!!ose his name actually constitutes a melody, then& Tortoise, Stran e but true& In fact, he worked that melody subtly into one of his most elaborate musical !iecesnamely, the final 8ontrapunctus in his *rt of the 7ugue& It was the last fu ue 0ach ever wrote& 3hen I heard it for the first time, I had no idea how would end& Suddenly, without warnin , it broke off& )nd then&&& dead silence& I reali<ed immediately that was where 0ach died& It is an indescribably sad moment, and the effect it had on me wasshatterin & In any case, 0-)-7-? is the last theme of that fu ue& It is hidden inside the !iece& 0ach didn1t !oint it out e(!licitly, but if you know about it, you can find it without much trouble& )h, methere are so many clever ways of hidin thin s in music&&& *chilles, &&&or in !oems& Poets used to do very similar thin s, you know 4thou h it1s rather out of style these days5& 'or instance, Aewis 7arroll often hid words and names in the first letters 4or characters5 of the successive lines in !oems he wrote& Poems which conceal messa es that way are called -acrostics-& Tortoise, 0ach, too, occasionally wrote acrostics, which isn1t sur!risin & )fter all, counter!oint and acrostics, with their levels of hidden meanin , have quite a bit in common& /ost acrostics, however, have only one hidden levelbut there is no reason that one couldn1t make a double-deckeran acrostic on to! of an acrostic& Er one could make a -contracrostic-where the initial letters, taken in reverse order, form a messa e& ?eavens2 There1s no end to the !ossibilities inherent in the form& /oreover, it1s not limited to !oets= anyone could write acrosticseven a dialo ician&

73G5=E BI! The last page of Bach's )rt of the 'u ue! 3n the original manuscript, in the handwriting of Bach's son 8arl &hilipp Emanuel, is written: "N!B! 3n the course of this fugue, at the point where the name B!*!8!,! was (rought in as countersu(Mect, the composer died!" B%*%8%, in (o9!/ 3 ha#e let this final page of Bach's last fugue ser#e as an epitaph! U'usic &rinted () .onald B)rd's program "-'5T", de#eloped at 3ndiana 5ni#ersit)V

*chilles, ) dial-a-lo ician? That1s a new one on me& Tortoise, 7orrection, I said -dialo ician-, by which I meant a writer of dialo ues& ?mm&&& somethin .ust occurred to me& In the unlikely event that a dialo ician should write a contra!untal acrostic in homa e to J& S& 0ach, do you su!!ose it would be more !ro!er for him to acrostically embed his E38 nameor that of 0ach? Eh, well, why worry about such frivolous matters? )nybody who wanted to write such a !iece could make u! his own mind& 8ow ettin back to 0ach1s melodic name, did you know that the melody 0-)-7-?, if !layed u!side down and backwards, is e(actly the same as the ori inal? *chilles, ?ow can anythin be !layed u!side down? 0ackwards, I can see-you et ?-7)-0but u!side down? Hou must be !ullin my le & Tortoise, 1!on my word, you1re quite a ske!tic, aren1t you? 3ell, I uess I1ll have to ive you a demonstration& Aet me .ust o and fetch my fiddle >alks into the ne9t room, and returns in a Miff) with an ancient%looking #iolin!/ and !lay it for you forwards and backwards and every which way& Aet1s see, now&&& &laces his cop) of the *rt of the 7ugue on his music stand and opens it to the last page!/ &&&here1s the last 7ontra!unctus, and here1s the last theme&&& The Tortoise (egins to pla): B%*%8% O (ut as he (ows the final ,, suddenl), without warning, a shattering sound rudel) interrupts his performance! Both he and *chilles spin around, Must in time to catch a glimpse of m)riad fragments of glass tinkling to the floor from the shelf where Go(let G had stood, onl) moments (efore! *nd then!!! dead silence!

Cha!ter I4

Consistency) Com!leteness) and Geometry

Im!licit and E-!licit Meaning
I8 7?)PTE; II, we saw how meanin at least in the relatively sim!le conte(t of formal systemsarises when there is an isomor!hism between rule- overned symbols, and thin s in the real world& The more com!le( the isomor!hism, in eneral, the more -equi!ment-both hardware and softwareis required to e(tract the meanin from the symbols& If an isomor!hism is very sim!le 4or very familiar5, we are tem!ted to say that the meanin which it allows us to see is e(!licit& 3e see the meanin without seein the isomor!hism& The most blatant e(am!le is human lan ua e, where !eo!le often attribute meanin to words in themselves, without bein in the sli htest aware of the very com!le( -isomor!hism- that imbues them with meanin s& This is an easy enou h error to make& It attributes all the meanin to the ob.ect 4the word5, rather than to the link between that ob.ect and the real world& Hou mi ht com!are it to the naPve belief that noise is a necessary side effect of any collision of two ob.ects& This is a false belief= if two ob.ects collide in a vacuum, there will be no noise at all& ?ere a ain, the error stems from attributin the noise e(clusively to the collision, and not reco ni<in the role of the medium, which carries it from the ob.ects to the ear& )bove, I used the word -isomor!hism- in quotes to indicate that it must be taken with a rain of salt& The symbolic !rocesses which underlie the understandin of human lan ua e are so much more com!le( than the symbolic !rocesses in ty!ical formal systems, that, if we want to continue thinkin of meanin as mediated by isomor!hisms, we shall have to ado!t a far more fle(ible conce!tion of what isomor!hisms can be than we have u! till now& In my o!inion, in fact, the key element in answerin the question -3hat is consciousness?- will be the unravelin of the nature of the -isomor!hismwhich underlies meanin &

E-!licit Meaning of the 8ontracrostipunctus

)ll this is by way of !re!aration for a discussion of the 8ontracrostipunctusOa study in levels of meanin & The :ialo ue has both e(!licit and im!licit meanin s& Its most e(!licit meanin is sim!ly the story which was related& This -e(!licit- meanin is, strictly s!eakin , e(tremely implicit, in the sense that the brain !rocesses required to understand the events in the story, iven only the black marks on !a!er, are incredibly com!le(& 8evertheless, we shall consider the events in the story to be the e(!licit meanin of the :ialo ue, and assume that every reader of En lish uses more or less the same -isomor!hism- in suckin that meanin from the marks on the !a!er&

Even so, I1d like to be a little more e(!licit about the e(!licit meanin of the story& 'irst I1ll talk about the record !layers and the records& The main !oint is that there are two levels of meanin for the rooves in the records& Aevel Ene is that of music& 8ow what is -music-a sequence of vibrations in the air, or a succession of emotional res!onses in a brain? It is both& 0ut before there can be emotional res!onses, there have to be vibrations& 8ow the vibrations et -!ulled- out of the rooves by a record !layer, a relatively strai htforward device= in fact you can do it with a !in, .ust !ullin it down the rooves& )fter this sta e, the ear converts the vibrations into firin s of auditory neurons in the brain& Then ensue a number of sta es in the brain, which radually transform the linear sequence of vibrations into a com!le( !attern of interactin emotional res!onsesfar too com!le( for us to o into here, much thou h I would like to& Aet us therefore content ourselves with thinkin of the sounds in the air as the -Aevel Ene- meanin of the rooves& 3hat is the Aevel Two meanin of the rooves? It is the sequence of vibrations induced in the record !layer& This meanin can only arise after the Aevel Ene meanin has been !ulled out of the rooves, since the vibrations in the air cause the vibrations in the !hono ra!h& Therefore, the Aevel Two meanin de!ends u!on a chain of two isomor!hisms, 4"5 Isomor!hism between arbitrary roove !atterns and air vibrations= 4B5 Isomor!hism between ra!h vibrations, arbitrary air vibrations and !hono ra!h vibrations This chain of two isomor!hisms is de!icted in 'i ure B+& 8otice that isomor!hism I is the one which ives rise to the Aevel Ene meanin & The Aevel Two meanin is more im!licit than the Aevel Ene meanin , because it is mediated by the chain of two isomor!hisms& It is the Aevel Two meanin which -backfires-, causin the record !layer to break a!art& 3hat is of interest is that the !roduction of the Aevel Ene meanin forces the !roduction of the Aevel Two meanin simultaneouslythere is no way to have Aevel Ene without Aevel Two& So it was the im!licit meanin of the record which turned back on it, and destroyed it& Similar comments a!!ly to the oblet& Ene difference is that the ma!!in from letters of the al!habet to musical notes is one more level of isomor!hism, which we could call -transcri!tion-& That is followed by -translation-conversion of musical notes into musical sounds& Thereafter, the vibrations act back on the oblet .ust as they did on the escalatin series of !hono ra!hs&

73G5=E FK! <isual rendition of the principle underl)ing Gdel's Theorem: two (ack%to%(ack mappings which ha#e an une9pected (oomeranging effect! The first is from groo#e patterns to sounds, carried out () a phonograph! The secondOfamiliar, (ut usuall) ignoredOis from sounds to #i(rations of the phonograph! Note that the second mapping e9ists independentl) of the first one, for an) sound in the #icinit), not Must ones produced () the phonograph itself, will cause such #i(rations! The paraphrase of Gdel's Theorem sa)s that for an) record pla)er, there are records which it cannot pla) (ecause the) will cause its indirect self%destruction! U.rawing () the author!V

Im!licit Meanings of the Contracrosti!unctus

3hat about im!licit meanin s of the :ialo ue? 4Hes, it has more than one of these&5 The sim!lest of these has already been !ointed out in the !ara ra!hs abovenamely, that the events in the two halves of the dialo ue are rou hly isomor!hic to each other, the !hono ra!h becomes a violin, the Tortoise becomes )chilles, the 7rab becomes the Tortoise, the rooves become the etched auto ra!h, etc& Ence you notice this sim!le isomor!hism, you can o a little further& Ebserve that in the first half of the story, the Tortoise is the !er!etrator of all the mischief, while in the second half, he is the victim& 3hat do you know, but his own method has turned around and backfired on him2 ;eminiscent of the backfirin of the records1 musicor the oblet1s inscri!tionor !erha!s of the Tortoise1s boomeran collection? Hes, indeed& The story is about backfirin on two levels, as follows&&& Aevel Ene, 6oblets and records which backfire= Aevel Two, The Tortoise1s devilish method of e(!loitin im!licit meanin to cause backfireswhich backfires& Therefore we can even make an isomor!hism between the two levels of the story, in which we equate the way in which the records and oblet boomeran back to destroy themselves, with the way in which the Tortoise1s own fiendish method boomeran s back to et him in the end& Seen this way, the story itself is an e(am!le of the backfirin s which it discusses& So we can think of the 7ontracrosti!unctus as referrin to itself indirectly, in that its own structure is isomor!hic to the events it !ortrays& 4E(actly as the oblet and records refer im!licitly to themselves via the back-to-back mor!hisms of

!layin and vibration-causin &5 Ene may read the :ialo ue without !erceivin this fact, of coursebut it is there all the time&

Ma!!ing Bet'een the 8ontracrostipunctus and G#del%s Theorem

8ow you may feel a little di<<ybut the best is yet to come& 4)ctually, levels of im!licit meanin will not even be discussed herethey will be left for you to ferret out&5 The dee!est reason for writin this :ialo ue was to illustrate 6>del1s Theorem, which, as I said in the Introduction, relies heavily on two different levels of meanin of statements of number theory& Each of the two halves of the 7ontracrosti!unctus is an -isomor!hic co!y- of 6>del1s Theorem& 0ecause this ma!!in is the central idea of the :ialo ue and is rather elaborate, I have carefully charted it out below&
Phono ra!h a(iomatic system for number theory low-fidelity !hono ra!h -weak- a(iomatic system hi h-fidelity !hono ra!h -stron - a(iomatic system -Perfect- !hono ra!h- com!lete system for number theory -blue!rint- of !hono ra!h a(ioms and rules of formal system record strin of the formal system !layable record theorem of the a(iomatic system un!layable record nontheorem of the a(iomatic system sound true statement of number theory re!roducible sound inter!reted theorem of the system unre!roducible sound true statement which isn1t a theorem, son title im!licit meanin of 6>del1s strin , -I 7annot 0e Played -I 7annot 0e :erived on ;ecord Player Tin 'ormal System T-

This is not the full e(tent of the isomor!hism between 6>del1s theorem and the 7ontracrosti!unctus, but it is the core of it& Hou need not worry if you don1t fully ras! 6>del1s Theorem by nowthere are still a few 7ha!ters to o before we reach it2 8evertheless, havin read this :ialo ue you have already tasted some of the flavor of 6>del1s Theorem without necessarily bein aware of it& I now leave you to look for any other ty!es of im!licit meanin in the 7ontracrosti!unctus& -Muaerendo invenietis2-

The Art of the +ugue

) few words on the *rt of the 7ugue&&& 7om!osed in the last year of 0ach1s life, it is a collection of ei hteen fu ues all based on one theme& )!!arently, writin the 'usical :ffering was an ins!iration to 0ach& ?e decided to com!ose another set of fu ues on a much sim!ler theme, to demonstrate the full ran e of !ossibilities inherent in the form& In the *rt of the 7ugue, 0ach uses a very sim!le theme in the most com!le( !ossible ways& The whole work is in a sin le key& /ost of the fu ues have four voices, and they radually increase in com!le(ity and de!th of e(!ression& Toward the end, they soar to such hei hts of intricacy that one sus!ects he can no lon er maintain them& Het he does&&&until the last 8ontrapunctus& The circumstances which caused the break-off of the *rt of the 7ugue 4which is to say, of 0ach1s life5 are these, his eyesi ht havin troubled him for years, 0ach wished to have an o!eration& It was done= however, it came out quite !oorly, and as a consequence, he lost his si ht for the better !art of the last year of his life& This did not kee! him from vi orous work on his monumental !ro.ect, however& ?is aim was to construct a com!lete e(!osition of fu al writin , and usa e of multi!le themes was one im!ortant facet of it& In what he !lanned as the ne(t-to-last fu ue, he inserted his own name coded into notes as the third theme& ?owever, u!on this very act, his health became so !recarious that he was forced to abandon work on his cherished !ro.ect& In his illness, he mana ed to dictate to his son-in-law a final chorale !relude, of which 0ach1s bio ra!her 'orkel wrote, -The e(!ression of !ious resi nation and devotion in it has always affected me whenever I have !layed it= so that I can hardly say which I would rather missthis 7horale, or the end of the last fu ue&Ene day, without warnin , 0ach re ained his vision& 0ut a few hours later, he suffered a stroke= and ten days later, he died, leavin it for others to s!eculate on the incom!leteness of the *rt of the 7ugue& 7ould it have been caused by 0ach1s attainment of self-reference?

,roblems Caused by G#del%s $esult

The Tortoise says that no sufficiently !owerful record !layer can be !erfect, in the sense of bein able to re!roduce every !ossible sound from a record& 6>del says that no sufficiently !owerful formal system can be !erfect, in the sense of re!roducin every sin le true statement as a theorem& 0ut as the Tortoise !ointed out with res!ect to !hono ra!hs, this fact only seems like a defect if you have unrealistic e(!ectations of what formal systems should be able to do& 8evertheless, mathematicians be an this century with .ust such unrealistic e(!ectations, thinkin that a(iomatic reasonin was the cure to all ills& They found out otherwise in "#%"& The fact that truth transcends theoremhood, in any iven formal system, is called -incom!leteness- of that system& ) most !u<<lin fact about 6>del1s method of !roof is that he uses reasonin methods which seemin ly cannot be -enca!sulated-they resist bein incor!orated into any formal system& Thus, at first si ht, it seems that 6>del has unearthed a hitherto unknown, but dee!ly si nificant, difference between human reasonin and mechanical

reasonin & This mysterious discre!ancy in the !ower of livin and nonlivin systems is mirrored in the discre!ancy between the notion of truth, and that of theoremhood or at least that is a -romantic- way to view the situation&

The Modified !q-System and Inconsistency

In order to see the situation more realistically, it is necessary to see in more de!th why and how meanin is mediated, in formal systems, by isomor!hisms& )nd I believe that this leads to a more romantic way to view the situation& So we now will !roceed to investi ate some further as!ects of the relation between meanin and form& Eur first ste! is to make a new formal system by modifyin our old friend, the !q-system, very sli htly& 3e add one more a(iom schema 4retainin the ori inal one, as well as the sin le rule of inference5, )TIE/ S7?E/) II, If 9 is a hy!hen-strin , then 9!-q9 is an a(iom& 7learly, then, --!-q-- is a theorem in the new system, and so --!--q---& )nd yet, their inter!retations are, res!ectively, -B !lus " equals B-, and -B !lus B equals %-& It can be seen that our new system contain a lot of false statements 4if you consider strin s to be statement& Thus, our new system is inconsistent with the e9ternal world! )s if this weren1t bad enou h, we also have internal !roblems with our new system, since it contains statements which disa ree with one another such as -!-q-- 4an old a(iom5 and -!-q- 4a new a(iom5& So our system is inconsistent in a second sense, internally& 3ould, therefore, the only reasonable thin to do at this !oint be dro! the new system entirely? ?ardly& I have deliberately !resented the -inconsistencies- in a wool!ullin manner, that is, I have tried to !ress fu<<y-headed ar uments as stron ly as !ossible, with the !ur!ose of misleadin & In fact, you may well have detected the fallacies in what I have said& The crucial fallacy came when I unquestionin ly ado!ted the very same inter!retin words for the new system as I had for the old one& ;emember that there was only one reason for ado!tin those words in in the last 7ha!ter, and that reason was that the s)m(ols acted isomorphicall) to the concepts which they were matched with, by the inter!retation& 0ut when you modify the rules overnin the system, you are bound to dama e the isomor!hism& It .ust cannot be hel!ed& Thus all the !roblems which were lamented over in !recedin !ara ra!hs were bo us !roblems= they can made to vanish in no time, by suita(l) reinterpreting some of the s)m(ols of s)stem& 8otice that I said -some-= not necessarily all symbols will have to ma!!ed onto new notions& Some may very well retain their -meanin s,- while others chan e&

$egaining Consistency
Su!!ose, for instance, that we reinter!ret .ust the symbol q, leavin all the others constant= in !articular, inter!ret q by the !hrase -is reater than or equal to-& 8ow, our

-contradictory- theorems -!-q- and -!-q-- come out harmlessly as, -" !lus " is reater than or equal to "-, and -" !lus " is reater than or equal to B-& 3e have simultaneously otten rid of 4"5 the inconsistency with the e(ternal world, and 4B5 the internal inconsistency& )nd our new inter!retation is a meanin ful inter!retation= of course the ori inal one is meaningless& That is, it is meanin less for the new s)stem= for the ori inal !q-system, it is fine& 0ut it now seems as !ointless and arbitrary to a!!ly it to the new !qsystem as it was to a!!ly the -horse-a!!le-ha!!y- inter!retation to the old !q-system&

The /istory of Euclidean Geometry

)lthou h I have tried to catch you off uard and sur!rise you a little, this lesson about how to inter!ret symbols by words may not seem terribly difficult once you have the han of it& In fact, it is not& )nd yet it is one of the dee!est lessons of all of nineteenth century mathematics2 It all be ins with Euclid, who, around %++ 0&7&, com!iled and systemati<ed all of what was known about !lane and solid eometry in his day& The resultin work, Euclid1s Elements, was so solid that it was virtually a bible of eometry for over two thousand yearsone of the most endurin works of all time& 3hy was this so? The !rinci!al reason was that Euclid was the founder of ri or in mathematics& The Elements be an with very sim!le conce!ts, definitions, and so forth, and radually built u! a vast body of results or ani<ed in such a way that any iven result de!ended only on fore oin results& Thus, there was a definite !lan to the work, an architecture which made it stron and sturdy& 8evertheless, the architecture was of a different ty!e from that of, say, a skyscra!er& 4See 'i & B"&5 In the latter, that it is standin is !roof enou h that its structural elements are holdin it u!& 0ut in a book on eometry, when each !ro!osition is claimed to follow lo ically from earlier !ro!ositions, there will be no visible crash if one of the !roofs is invalid& The irders and struts are not !hysical, but abstract& In fact, in Euclid1s Elements, the stuff out of which !roofs were constructed was human lan ua ethat elusive, tricky medium of communication with so many hidden !itfalls& 3hat, then, of the architectural stren th of the Elements? Is it certain that it is held u! by solid structural elements, or could it have structural weaknesses? Every word which we use has a meanin to us, which uides us in our use of it& The more common the word, the more associations we have with it, and the more dee!ly rooted is its meanin & Therefore, when someone ives a definition for a common word in the ho!es that we will abide by that definition, it is a fore one conclusion that we will not do so but will instead be uided, lar ely unconsciously, by what our minds find in their associative stores& I mention this because it is the sort of !roblem which Euclid created in his Elements, by attem!tin to ive definitions of ordinary, common words such as -!oint-, -strai ht line-, -circle-, and so forth& ?ow can you define somethin of which everyone already has a clear conce!t? The only way is if you can make it clear that your

73G5=E FB! Tower of 0abel, () '! 8! Escher woodcut, BIFE/!

word is su!!osed to be a technical term, and is not to be confused with the everyday word with the same s!ellin & Hou have to stress that the connection with the everyday word is only su estive& 3ell, Euclid did not do this, because he felt that the !oints and lines of his Elements were indeed the !oints and lines of the real world& So by not makin sure that all associations were dis!elled, Euclid was invitin readers to let their !owers of association run free&&& This sounds almost anarchic, and is a little unfair to Euclid& ?e did set down a(ioms, or !ostulates, which were su!!osed to be used in the !roofs of !ro!ositions& In fact, nothin other than those a(ioms and !ostulates was su!!osed to be used& 0ut this is where he sli!!ed u!, for an inevitable consequence of his usin ordinary words was that some of the ima es con.ured u! by those words cre!t into the !roofs which he created& ?owever, if you read !roofs in the Elements, do not by any means e(!ect to find larin -.um!s- in the reasonin & En the contrary, they are very subtle, for Euclid was a !enetratin thinker, and would not have made any sim!leminded errors& 8onetheless, a!s are there, creatin sli ht im!erfections in a classic work& 0ut this is not to be com!lained about& Ene should merely ain an a!!reciation for the difference between absolute ri or and relative ri or& In the lon run, Euclid1s lack of absolute ri or was the cause of some of the most fertile !ath-breakin in mathematics, over two thousand years after he wrote his work& Euclid ave five !ostulates to be used as the - round story- of the infinite skyscra!er of eometry, of which his Elements constituted only the first several hundred stories& The first four !ostulates are rather terse and ele ant, 4"5 ) strai ht line se ment can be drawn .oinin any two !oints& 4B5 )ny strai ht line se ment can be e(tended indefinitely in a strai ht line& 4%5 6iven any strai ht line se ment, a circle can be drawn havin the se ment as radius and one end !oint as center& 4C5 )ll ri ht an les are con ruent& The fifth, however, did not share their race, 4L5 If two lines are drawn which intersect a third in such a way that the sum of the inner an les on one side is less than two ri ht an les, then the two lines inevitably must intersect each other on that side if e(tended far enou h&

Thou h he never e(!licitly said so, Euclid considered this !ostulate to be somehow inferior to the others, since he mana ed to avoid usin it in the !roofs of the first twentyei ht !ro!ositions& Thus, the first twenty-ei ht !ro!ositions belon to what mi ht be called -four-!ostulate eometry- that !art of eometry which can be derived on the basis of the first to !ostulates of the Elements, without the hel! of the fifth !ostulate& 4It is also often called a(solute geometr)&5 7ertainly Euclid would have found it far !referable to !rove this u ly ducklin , rather than to have to assume it& 0ut he found no !roof, and therefore ado!ted it& 0ut the disci!les of Euclid were no ha!!ier about havin to assume this fifth !ostulate& Ever the centuries, untold numbers of !eo!le ave untold years of their lives in attem!tin to !rove that the fifth !ostulate was itself !art of four-!ostulate eometry& 0y "$D%, at least twenty-ei ht deficient !roofs had been !ublishedall erroneous2 4They were all critici<ed in the dissertation of one 6& S& @l` el&5 )ll of these erroneous !roofs involve a confusion between everyday intuition and strictly formal !ro!erties& It safe to say that today, hardly any of these -!roofs- holds any mathematical or historical interest but there are certain e(ce!tions&

The Many +aces of ;oneuclid

6irolamo Saccheri 4"DD$-"$%%5 lived around 0ach1s time& ?e had the ambition to free Euclid of every flaw& 0ased on some earlier work he had done in lo ic, he decided to try a novel a!!roach to the !roof of the famous fifth, su!!ose you assume its o!!osite= then work with that as your fifth !ostulate&&& Surely after a while you will create a contradiction& Since no mathematical system can su!!ort a contradiction, you will have shown the unsoundness of your own fifth !ostulate, and therefore the soundness of Euclid1s fifth !ostulate& 3e need not o into details here& Suffice it to say that with reat skill, Saccheri worked out !ro!osition after !ro!osition -Saccherian eometry- and eventually became tired of it& )t one !oint, decided he had reached a !ro!osition which was -re!u nant to the nature of the strai ht line-& That was what he had been ho!in for to his mind, it was the lon -sou ht contradiction& )t that !oint, he !ublished his work under the title Euclid 7reed of E#er) 7law, and then e(!ired& 0ut in so doin , he robbed himself of much !osthumous lory, since he had unwittin ly discovered what came later to be known as -hy!erbolic eometry-& 'ifty years after Saccheri, J& ?& Aambert re!eated the -near miss-, this time comin even closer, if !ossible& 'inally, forty years after Aambert, and ninety years after Saccheri, nonEuclidean eometry was reco ni<ed for what it wasan authentic new brand of eometry, a bifurcation in the hitherto sin le stream of mathematics& In "*B%, non% Euclidean geometr) was discovered simultaneously, in one of those ine(!licable coincidences, by a ?un arian mathematician, Janos 4or Johann5 0olyai, a e twenty-one, and a ;ussian mathematician, 8ikolay Aobachevskiy, a e thirty& )nd, ironically, in that same year, the reat 'rench mathematician )drien-/arie Ae endre came u! with what he was sure was a !roof of Euclid1s fifth !ostulate, very much alon the lines of Saccheri&

Incidentally, 0olyai1s father, 'arkas 4or 3olf an 5 0olyai, a close friend of the reat 6auss, invested much effort in tryin to !rove Euclid1s fifth !ostulate& In a letter to his son Janos, he tried to dissuade him from thinkin about such matters,
Hou must not attem!t this a!!roach to !arallels& I know this way to its very end& I have traversed this bottomless ni ht, which e(tin uished all li ht and .oy of my life& I entreat you, leave the science of !arallels alone&&& I thou ht I would sacrifice myself for the sake of the truth& I was ready to become a martyr who would remove the flaw from eometry and return it !urified to mankind& I accom!lished monstrous, enormous labors= my creations are far better than those of others and yet I have not achieved com!lete satisfaction& 'or here it is true that si paullum a summo discessit, #ergit ad imum! I turned back when I saw that no man can reach the bottom of this ni ht& I turned back unconsoled, !ityin myself and all mankind&&& I have traveled !ast all reefs of this infernal :ead Sea and have always come back with broken mast and torn sail& The ruin of my dis!osition and my fall date back to this time& I thou htlessly risked my life and ha!!inessaut 8aesar aut nihil!"

0ut later, when convinced his son really -had somethin -, he ur ed him to !ublish it, antici!atin correctly the simultaneity which is so frequent in scientific discovery,
3hen the time is ri!e for certain thin s, these thin s a!!ear in different !laces in the manner of violets comin to li ht in early s!rin &B

?ow true this was in the case of non-Euclidean eometry2 In 6ermany, 6auss himself and a few others had more or less inde!endently hit u!on non-Euclidean ideas& These included a lawyer, '& @& Schweikart, who in "*"* sent a !a e describin a new -astral- eometry to 6auss= Schweikart1s ne!hew, '& )& Taurinus, who did non-Euclidean tri onometry= and '& A& 3achter, a student of 6auss, who died in "*"$, a ed twenty-five, havin found several dee! results in non-Euclidean eometry& The clue to non-Euclidean eometry was -thinkin strai ht- about the !ro!ositions which emer e in eometries like Saccheri1s and Aambert1s& The Saccherian !ro!ositions are only -re!u nant to the nature of the strai ht line- if you cannot free yourself of !reconceived notions of what -strai ht line- must mean& If, however, you can divest yourself of those !reconceived ima es, and merely let a -strai ht line- be somethin which satisfies the new !ro!ositions, then you have achieved a radically new view!oint&

&ndefined Terms
This should be in to sound familiar& In !articular, it harks back to the !q-system, and its variant, in which the symbols acquired !assive meanin s by virtue of their roles in theorems& The symbol q is es!ecially interestin , since its -meanin - chan ed when a new a(iom schema was added& In that very same way, one can let the meanings of "point", "line", and so on (e determined () the set of theorems or propositions/ in which the) occur& This was the reat reali<ation of the discoverers of non-Euclidean eometry& They found different sorts of non-Euclidean eometries by denyin Euclid1s fifth

!ostulate in different ways and followin out the consequences& Strictly s!eakin , they 4and Saccheri5 did not deny the fifth !ostulate directly, but rather, they denied an equivalent !ostulate, called the parallel postulate, which runs as follows, 6iven any strai ht line, and a !oint not on it, there e(ists one, and only one, strai ht line which !asses throu h that !oint and never intersects the first line, no matter how far they are e(tended& The second strai ht line is then said to be !arallel to the first& If you assert that no such line e(ists, then you reach elliptical geometr)= if you assert that, at least two such lines e(ist, you reach h)per(olic geometr)& Incidentally, the reason that such variations are still called - eometries- is that the core elementabsolute, or four-!ostulate, eometryis embedded in them& It is the !resence of this minimal core which makes it sensible to think of them as describin !ro!erties of some sort of eometrical s!ace, even if the s!ace is not as intuitive as ordinary s!ace& )ctually, elli!tical eometry is easily visuali<ed& )ll -!oints-, -lines- and so forth are to be !arts of the surface of an ordinary s!here& Aet us write - PEI8T- when the technical term is meant, and -!oint- when the everyday sense is desired& Then, we can say that a PEI8T consists of a !air of diametrically o!!osed !oints of the s!here1s surface& ) AI8E is a reat circle on the s!here 4a circle which, like the equator, has its center at the center of the s!here5& Fnder these inter!retations, the !ro!ositions of elli!tical eometry, thou h they contain words like -PEI8T- and -AI8E s!eak of the oin s-on on a s!here, not a !lane& 8otice that two AI8ES always intersect in e(actly two anti!odal !oints of the s!here1s surface that is, in e(actly one sin le PEI8T2 )nd .ust as two AI8ES determine PEI8T, so two PEI8TS determine a AI8E& 0y treatin words such as -PEI8T- and -AI8E- as if they had only the meanin instilled in them by the !ro!ositions in which they occur, we take ste! towards com!lete formali<ation of eometry& This semiformal version still uses a lot of words in En lish with their usual meanin s 4words such -the-, 1 if -, -and-, -.oin-, -have-5, althou h the everyday meanin has been drained out of s!ecial words like - PEI8T- and -AI8E-, which are consequently called undefined terms& Fndefined terms, like the ! and q of the !qsystem, do et defined in a sense, implicitl)by the totality of all !ro!ositions in which they occur, rather than e(!licitly, in a definition& Ene could maintain that a full definition of the undefined terms resides in the !ostulates alone, since the !ro!ositions which follow from them are im!licit in the !ostulates already& This view would say that the !ostulates are im!licit definitions of all the undefined terms, all of the undefined terms bein defined in terms of the others&

The ,ossibility of Multi!le Inter!retations

) full formali<ation of eometry would take the drastic ste! of makin every term undefinedthat is, turnin every term into a -meanin less- symbol of a formal system& I !ut quotes around -meanin less- because, as you know, the symbols automatically !ick u! !assive meanin s in accordance with the theorems they occur in& It is another question, thou h, whether !eo!le discover those meanin s, for to do so requires findin a set of conce!ts which can be linked by an isomor!hism to the symbols in the formal system& If one be ins with the aim of formali<in eometry, !resumably one has an intended inter!retation for each symbol, so that the !assive meanin s are built into the system& That is what I did for ! and q when I first created the !q-system& 0ut there may be other !assive meanin s which are !otentially !erce!tible, which no one has yet noticed& 'or instance, there were the sur!rise inter!retations of ! as -equals- and q as -taken from-, in the ori inal !q-system& )lthou h this is rather a trivial e(am!le, it contains the essence of the idea that symbols may have many meanin ful inter!retationsit is u! to the observer to look for them& 3e can summari<e our observations so far in terms of the word -consistency-& 3e be an our discussion by manufacturin what a!!eared to be an inconsistent formal systemone which was internally inconsistent, as well as inconsistent with the e(ternal world& 0ut a moment later we took it all back, when we reali<ed our error, that we had chosen unfortunate inter!retations for the symbols& 0y chan in the inter!retations, we re ained consistency2 It now becomes clear that consistenc) is not a propert) of a formal s)stem per se, (ut depends on the interpretation which is proposed for it & 0y the same token, inconsistency is not an intrinsic !ro!erty of any formal system&

4arieties of Consistency
3e have been s!eakin of -consistency- and -inconsistency- all alon , without definin them& 3e have .ust relied on ood old everyday notions& 0ut now let us say e(actly what is meant by consistenc) of a formal system 4to ether with an inter!retation5, that every theorem, when inter!reted, becomes a true statement& )nd we will say that inconsistency occurs when there is at least one false statement amon the inter!reted theorems& This definition a!!ears to be talkin about inconsistency with the e(ternal world what about internal inconsistencies? Presumably, a system would be internally inconsistent if it contained two or more theorems whose inter!retations were incom!atible with one another, and internally consistent if all inter!reted theorems were com!atible with one another& 7onsider, for e(am!le, a formal system which has only the followin three theorems, Tb2, 2bE, and EbT& If T is inter!reted as -the Tortoise-, 2 as -9eno-, E as -E bert-, and 9b) as -9 beats ) in chess always-, then we have the followin inter!reted theorems,

The Tortoise always beats 9eno at chess& 9eno always beats E bert at chess& E bert always beats the Tortoise at chess& The statements are not incom!atible, althou h they describe a rather bi<arre circle of chess !layers& ?ence, under this inter!retation, the form= system in which those three strin s are theorems is internally consistent althou h, in !oint of fact, none of the three statements is true2 Internal consistency does not require all theorems to come out true, but merely that they come out compati(le with one another& 8ow su!!ose instead that 9b) is to be inter!reted as -9 was invented by )-& Then we would have, The Tortoise was invented by 9eno& 9eno was invented by E bert& E bert was invented by the Tortoise& In this case, it doesn1t matter whether the individual statements are true or falseand !erha!s there is no way to know which ones are true, and which are not& 3hat is nevertheless certain is that not all three can (e true at once! Thus, the inter!retation makes the system internally inconsistent& The internal inconsistency de!ends not on the inter!retations of the three ca!ital letters, but only on that of b, and on the fact that the three ca!itals are cyclically !ermuted around the occurrences of b& Thus, one can have internal inconsistency without havin inter!reted all of the symbols of the formal system& 4In this case it sufficed to inter!ret a sin le symbol&5 0y the time sufficiently many symbols have been iven inter!retations, it may be clear that there is no way that the rest of them can be inter!reted so that all theorems will come out true& 0ut it is not .ust a question of truthit is question of !ossibility& )ll three theorems would come out false if the ca!itals were inter!reted as the names of real !eo!lebut that is not why we would call the system internally inconsistent= our rounds for doin so would be the circularity, combined with the inter!retation of the letter b& 40y the way, you1ll find more on this -authorshi! trian le- in 7ha!ter TT&5

/y!othetical >orlds and Consistency

3e have iven two ways of lookin at consistency, the first says that system-!lusinter!retation is consistent with the e9ternal world if every theorem comes out true when inter!reted= the second says that a system-!lus-inter!retation is internall) consistent if all theorems come out mutually com!atible when inter!reted& 8ow there is a close relationshi! between these two ty!es of consistency& In order to determine whether several statements at mutually com!atible, you try to ima ine a world in which all of them could be simultaneously true& Therefore, internal consistency de!ends u!on consistency with the e(ternal worldonly now, -the e(ternal world- is allowed to be an) imagina(le world, instead of the one we live in& 0ut this is an e(tremely va ue,

unsatisfactory conclusion& 3hat constitutes an -ima inable- world? )fter all, it is !ossible to ima ine a world in which three characters invent each other cyclically& Er is it? Is it !ossible to ima ine a world in which there are square circles? Is a world ima inable in which 8ewton1s laws, and not relativity, hold? Is it !ossible to ima ine a world in which somethin can be simultaneously reen and not reen? Er a world in which animals e(ist which are not made of cells? In which 0ach im!rovised an ei ht-!art fu ue on a theme of @in 'rederick the 6reat? In which mosquitoes are more intelli ent than !eo!le? In which tortoises can !lay footballor talk? ) tortoise talkin football would be an anomaly, of course& Some of these worlds seem more ima inable than others, since some seem to embody logical contradictionsfor e(am!le, reen and not reenwhile some of them seem, for want of a better word, -!lausible-such as 0ach im!rovisin an ei ht-!art fu ue, or animals which are not made of cells& Er even, come to think of it, a world in which the laws of !hysics are different&&& ;ou hly, then, it should be !ossible to establish different brands of consistency& 'or instance, the most lenient would be -lo ical consistency-, !uttin no restraints on thin s at all, e(ce!t those of lo ic& /ore s!ecifically, a system-!lus-inter!retation would be logicall) consistent .ust as lon as no two of its theorems, when inter!reted as statements, directly contradict each other= and mathematicall) consistent .ust as lon as inter!reted theorems do not violate mathematics= and ph)sicall) consistent .ust as lon as all its inter!reted theorems are com!atible with !hysical law= then comes biolo ical consistency, and so on& In a biolo ically consistent system, there could be a theorem whose inter!retation is the statement -Shakes!eare wrote an o!era-, but no theorem whose inter!retation is the statement -7ell-less animals e(ist-& 6enerally s!eakin , these fancier kinds of inconsistency are not studied, for the reason that they are very hard to disentan le from one another& 3hat kind of inconsistency, for e(am!le, should one say is involved in the !roblem of the three characters who invent each other cyclically? Ao ical? Physical? 0iolo ical? Aiterary? Fsually, the borderline between uninterestin and interestin is drawn between !hysical consistency and mathematical consistency& 4Ef course, it is the mathematicians and lo icians who do the drawin hardly an im!artial crew&&&5 This means that the kinds of inconsistency which -count-, for formal systems, are .ust the lo ical and mathematical kinds& )ccordin to this convention, then, we haven1t yet found an inter!retation which makes the trio of theorems Tb2, 2bE, EbT inconsistent& 3e can do so by inter!retin b as -is bi er than-& 3hat about T and 2 and E? They can be inter!reted as natural numbers for e(am!le, 2 as +, T as B, and E as ""& 8otice that two theorems come out true this way, one false& If, instead, we had inter!reted 2 as %, there would have been two falsehoods and only one truth& 0ut either way, we1d have had inconsistency& In fact, the values assi ned to T, 2, and E are irrelevant, as lon as it is understood that they are restricted to natural numbers& Ence a ain we see a case where only some of the inter!retation is needed, in order to reco ni<e internal inconsistency&

Embedding of 0ne +ormal System In Another

The !recedin e(am!le, in which some symbols could have inter!retations while others didn1t, is reminiscent of doin eometry in natural lan ua e usin some words as undefined terms& In such a case, words are divide into two classes, those whose meanin is fi(ed and immutable, and, those whose meanin is to be ad.usted until the system is consistent 4these are the undefined terms5& :oin eometry in this way requires that meanin s have already been established for words in the first class, somewhere outside of eometry& Those words form a ri id skeleton, ivin an underlyin structure to the system= fillin in that skeleton comes other material, which can vary 4Euclidean or nonEuclidean eometry5& 'ormal systems are often built u! in .ust this ty!e of sequential, or hierarchical, manner& 'or e(am!le, 'ormal System I may be devised, with rules and a(ioms that ive certain intended !assive meanin s to its symbols& Then 'ormal System I is incor!orated fully into a lar er system with more symbols'ormal System II& Since 'ormal System I1s a(ioms and rules are !art of 'ormal System II, the !assive meanin s of 'ormal System I symbols remain valid= they form an immutable skeleton which then !lays a lar e role in the determination of the !assive meanin s of the new symbols of 'ormal System II& The second system may in turn !lay the role of skeleton with res!ect to a third system, and so on& It is also !ossibleand eometry is a ood e(am!le of thisto have a system 4e& &, absolute eometry5 which partl) !ins down the !assive meanin s of its undefined terms, and which can be su!!lemented by e(tra rules or a(ioms, which then further restrict the !assive meanin s of the undefined terms& This the case with Euclidean versus nonEuclidean eometry&

Layers of Stability in 4isual ,erce!tion

In a similar, hierarchical way, we acquire new knowled e, new vocabulary or !erceive unfamiliar ob.ects& It is !articularly interestin in the case understandin drawin s by Escher, such as =elati#it) 4'i & BB5, in which there occur blatantly im!ossible ima es& Hou mi ht think that we would seek to reinter!ret the !icture over and over a ain until we came to an inter!retation of its !arts which was free of contradictions but we don1t do that at all& 3e sit there amused and !u<<led by staircases which o every which way, and by !eo!le oin in inconsistent directions on a sin le staircase& Those staircases are -islands of certainty- u!on which we base our inter!retation of the overall !icture& ?avin once identified them, we try to e(tend our understandin , by seekin to establish the relationshi! which they bear to one another& )t that sta e, we encounter trouble& 0ut if we attem!ted to backtrackthat is, to question the -islands of certainty-we would also encounter trouble, of another sort& There1s no way of backtrackin and -undecidin - that they are staircases& They are not fishes, or whi!s, or handsthey are .ust staircases& 4There is, actually, one other outto leave all the lines of the !icture totally uninter!reted, like the -meanin less symbols- of a formal system& This ultimate esca!e route is an e(am!le of a -F-mode- res!onsea 9en attitude towards symbolism&5

73G5=E FF! ;elativity, () '! 8! Escher lithograph, BIGH/!

So we are forced, by the hierarchical nature of our !erce!tive !rocesses, to see either a cra<y world or .ust a bunch of !ointless lines& ) similar analysis could be made of do<ens of Escher !ictures, which rely heavily u!on the reco nition of certain basic forms, which are then !ut to ether in nonstandard ways= and by the time the observer sees the !arado( on a hi h level, it is too latehe can1t o back and chan e his mind about how to inter!ret the lower-level ob.ects& The difference between an Escher drawin and nonEuclidean eometry is that in the latter, com!rehensible inter!retations can be found for the undefined terms, resultin in a com!rehensible total system, whereas for the former, the end result is not reconcilable with one1s conce!tion of the world, no matter how lon one stares at the !ictures& Ef course, one can still manufacture hy!othetical worlds, in which Escherian events can ha!!en&&& but in such worlds, the laws of biolo y, !hysics,

mathematics, or even lo ic will be violated on one level, while simultaneously bein obeyed on another, which makes them e(tremely weird worlds& 4)n e(am!le of this is in >aterfall 4'i & L5, where normal ravitation a!!lies to the movin water, but where the nature s!ace violates the laws of !hysics&5

Is Mathematics the Same in E"ery Concei"able >orld7

3e have stressed the fact, above, that internal consistency of a form= system 4to ether with an inter!retation5 requires that there be some imagina(le worldthat is, a world whose only restriction is that in it, mathematics and lo ic should be the same as in our worldin which all the inter!reted theorems come out true& E9ternal consistency, howeverconsistency with the e(ternal worldrequires that all theorems come of true in the real world& 8ow in the s!ecial case where one wishes to create consistent formal system whose theorems are to be inter!reted as statements of mathematics, it would seem that the difference between the two ty!es of consistency should fade away, since, accordin to what we sat above, all imagina(le worlds ha#e the same mathematics as the real world& Thus, in every conceivable world, " !lus " would have to be B= likewise, there would have to be infinitely many !rime numbers= furthermore, in every conceivable world, all ri ht an les would have to be con ruent= and of course throu h any !oint not on a iven line there would have to be e(actly one !arallel line&&& 0ut wait a minute2 That1s the !arallel !ostulateand to assert its universality would be a mistake, in li ht of what1s .ust been said& If in all conceivable worlds the !arallel !ostulate is obeyed, then we are assertin that non-Euclidean eometry is inconceivable, which !uts us back in the same mental state as Saccheri and Aambert surely an unwise move& But what, then, if not all of mathematics, must all concei#a(le worlds share? 7ould it be as little as lo ic itself? Er is even lo ic sus!ect? 7ould there be worlds where contradictions are normal !arts of e(istenceworlds where contradictions are not contradictions? 3ell, in some sense, by merely inventin the conce!t, we have shown that such worlds are indeed conceivable= but in a dee!er sense, they are also quite inconceivable& 4This in itself is a little contradiction&5 Muite seriously, however, it seems that if we want to be able to communicate at all, we have to ado!t some common base, and it !retty well has to include lo ic& 4There are belief systems which re.ect this !oint of viewit is too lo ical& In !articular, 9en embraces contradictions and non-contradictions with equal ea erness& This may seem inconsistent, but then bein inconsistent is !art of 9en, and so&&&what can one say?5

Is ;umber Theory the Same In All Concei"able >orlds7

If we assume that logic is !art of every conceivable world 4and note that we have not defined lo ic, but we will in 7ha!ters to come5, is that all? Is it really conceivable that, in some worlds, there are not infinitely many !rimes? 3ould it not seem necessary that numbers should obey the same laws in all conceivable worlds? Er&&& is the conce!t -natural number- better thou ht of as an undefined term, like -PEI8T- or -AI8E-? In that case, number theory would be a bifurcated theory, like eometry, there would be standard and nonstandard number theories& 0ut there would have to be some counter!art to absolute eometry, a -core- theory, an invariant in redient of all number theories which identified them as number theories rather than, say, theories about cocoa or rubber or bananas& It seems to be the consensus of most modern mathematicians and !hiloso!hers that there is such a core number theory, which ou ht to be included, alon with lo ic, in what we consider to be -conceivable worlds-& This core of number theory, the counter!art to absolute eometryis called &eano arithmetic, and we shall formali<e it in 7ha!ter GIII& )lso, it is now well establishedas a matter of fact as a direct consequence of 6>del1s Theoremthat number theory is a bifurcated theory, with standard and nonstandard versions& Fnlike the situation in eometry, however, the number of -brandsof number theory is infinite, which makes the situation of number theory considerably more com!le(& 'or !ractical !ur!oses, all number theories are the same& In other words, if brid e buildin de!ended on number theory 4which in a sense it does5, the fact that there are different number theories would not matter, since in the as!ects relevant to the real world, all number theories overla!& The same cannot be said of different eometries= for e(am!le, the sum of the an les in a trian le is "*+ de rees only in Euclidean eometry= it is reater in elli!tic eometry, less in hy!erbolic& There is a story that 6auss once attem!ted to measure the sum of the an les in a lar e trian le defined by three mountain !eaks, in order to determine, once and for all, which kind of eometry really rules our universe& It was a hundred years later that Einstein ave a theory 4 eneral relativity5 which said that the eometry of the universe is determined by its content of matter, so that no one eometry is intrinsic to s!ace itself& Thus to the question, -3hich eometry is true?- nature ives an ambi uous answer not only in mathematics, but also in !hysics& )s for the corres!ondin question, ->hich num(er theor) is true$-, we shall have more to say on it after oin throu h 6>del1s Theorem in detail&

Com!leteness
If consistency is the minimal condition under which symbols acquire !assive meanin s, then its com!lementary notion, com!leteness, is the ma(imal confirmation of those !assive meanin s& 3here consistency is the !ro!erty that, Everythin !roduced by the system is true-, com!leteness is the other way round, -Every true statement is !roduced by the system-& 8ow to refine the notion sli htly& 3e can1t mean every true statement in the worldwe mean only those which belon to the domain which we at attem!tin to

re!resent in the system& Therefore, com!leteness means, -Every true statement which can be e(!ressed in the notation of the system is a theorem&7onsistency, when every theorem, u!on inter!retation, comes out true 4in some ima inable world5& 7om!leteness, when all statements which are true 4in some ima inable world5, and which can be e(!ressed as well-formed strin s of the system, are theorems& )n e(am!le of a formal system which is com!lete on its own mode level is the ori inal !q-system, with the ori inal inter!retation& )ll true additions of two !ositive inte ers are re!resented by theorems of the system& 3e mi ht say this another way, -)ll true additions of two !ositive inte ers are pro#a(le within the system&- 43arnin , 3hen we start usin the term -!rovable statements- instead of -theorems-, it shows that we at be innin to blur the distinction between formal systems and their inter!retations& This is all ri ht, !rovided we are very conscious of the blurrin that is takin !lace, and !rovided that we remember that multi!le inter!retations are sometimes !ossible&5 The !q-system with the ori in inter!retation is complete= it is also consistent, since no false statement is to use our new !hrase!rovable within the system& Someone mi ht ar ue that the system is incom!lete, on the rounds that additions of three !ositive inte ers 4such as B ^ % ^ C X#5 are not re!resented by theorems of the !q-system, des!ite bein translatable into the notation of the system 4e& &, --!---!---q---------5. ?owever, this strin is not well-formed, and hence should be considered to be .ust as devoid of meanin as is !q!---q!q& Tri!le additions are sim!ly not e9pressi(le in the notation of the systemso the com!leteness of the system is !reserved& :es!ite the com!leteness of the !q-system under this inter!retation, it certainly falls far short of ca!turin the full notion of truth in number theory& 'or e(am!le, there is no way that the !q-system tells us how many !rime numbers there are& 6>del1s Incom!leteness Theorem says that any system which is -sufficiently !owerful- is, by virtue of its !ower, incom!lete, in the sense that there are well-formed strin s which e(!ress true statements of number theory, but which are not theorems& 4There are truths belon in to number theory which are not !rovable within the system&5 Systems like the !q-system, which are com!lete but not very !owerful, are more like low-fidelity !hono ra!hs= they are so !oor to be in with that it is obvious that they cannot do what we would wish them to donamely tell us everythin about number theory&

/o' an Inter!retation May Ma.e or Brea. Com!leteness

3hat does it mean to say, as I did above, that -com!leteness is the ma(imal confirmation of !assive meanin s-? It means that if a system is consistent but incom!lete, there is a mismatch between the symbols and their inter!retations& The system does not have the !ower to .ustify bein inter!reted that way& Sometimes, if the inter!retations are -trimmed- a little, the system can become com!lete& To illustrate this idea, let1s look at the modified !q-system 4includin )(iom Schema II5 and the inter!retation we used for it& )fter modifyin the !q-system, we modified the inter!retation for q from -equalsto -is reater than or equal to-& 3e saw that the modified !q-system was consistent under this inter!retation= yet somethin about the new inter!retation is not very satisfyin & The !roblem is sim!le, there are now many e(!ressible truths which are not theorems& 'or instance, -B !lus % is reater than or equal to "- is e(!ressed by the nontheorem --!---q-. The inter!retation is .ust too slo!!y2 It doesn1t accurately reflect what the theorems in the system do& Fnder this slo!!y inter!retation, the !q-system is not com!lete& 3e could re!air the situation either by 4"5 adding new rules to the system, makin it more !owerful, or by 4B5 tightening up the interpretation& In this case, the sensible alternative seems to be to ti hten the inter!retation& Instead of inter!retin q as -is reater than or equal to-, we should say -equals or e(ceeds by "-& 8ow the modified !q-system becomes both consistent and com!lete& )nd the com!leteness confirms the a!!ro!riateness of the inter!retation&

Incom!leteness of +ormali:ed ;umber Theory

In number theory, we will encounter incom!leteness a ain= but there, to remedy the situation, we will be !ulled in the other direction-towards addin new rules, to make the system more !owerful& The irony is that we think, each time we add a new rule, that we surely have made the system com!lete now2 The nature of the dilemma can be illustrated by the followin alle ory&&& 3e have a record !layer, and we also have a record tentatively labeled -7anon on 0-)-7-?-& ?owever, when we !lay the record on the record !layer, the feedback-induced vibrations 4as caused by the Tortoise1s records5 interfere so much that we do not even reco ni<e the tune& 3e conclude that something is defectiveeither our record, or our record !layer& In order to test our record, we would have to !lay it on friends1 record !layers, and listen to its quality& In order to test our phonograph, we would have to !lay friends1 records on it, and see if the music we hear a rees with the labels& If our record !layer !asses its test, then we will say the record was defective= contrariwise, if the record !asses its test, then we will say our record !layer was defective& 3hat, however, can we conclude when we find out that (oth !ass their res!ective tests? That is the moment to remember the chain of two isomor!hisms 4'i & B+5, and think carefully2

Little Harmonic Labyrinth

The Tortoise and *chilles are spending a da) at 8one) 3sland! *fter (u)ing a couple of cotton candies, the) decide to take a ride on the 7erris wheel! Tortoise, This is my favorite ride& Ene seems to move so far, and reality one ets nowhere& *chilles, I can see why it would a!!eal to you& )re you all stra!!ed in? Tortoise, Hes, I think I1ve ot this buckle done& 3ell, here we o& *chilles, Hou certainly are e(uberant today& Tortoise, I have ood reason to be& /y aunt, who is a fortune-teller told me that a stroke of 6ood 'ortune would befall me today& So I am tin lin with antici!ation& *chilles, :on1t tell me you believe in fortune-tellin 2 Tortoise, 8o&&& but they say it works even if you don1t believe it& *chilles, 3ell, that1s fortunate indeed& Tortoise, )h, what a view of the beach, the crowd, the ocean, the city&&& *chilles, Hes, it certainly is s!lendid& Say, look at that helico!ter there& It seems to be flyin our way& In fact it1s almost directly above us now& Tortoise, Stran ethere1s a cable dan lin down from it, which is very close to us& It1s comin so close we could !ractically rab it *chilles, Aook2 )t the end of the line there1s a iant hook, with a note& ,e reaches out and snatches the note! The) pass () and are on their wa) down!/ Tortoise, 7an you make out what the note says? *chilles, Hesit reads, -?owdy, friends& 6rab a hold of the hook time around, for an Fne(!ected Sur!rise&Tortoise, The note1s a little corny but who knows where it mi ht lead, Perha!s it1s ot somethin to do with that bit of 6ood 'ortune due me& 0y all means, let1s try it2 *chilles, Aet1s2 :n the trip up the) un(uckle their (uckles, and at the crest of the ride, gra( for the giant hook! *ll of a sudden the) are whooshed up () the ca(le which "uickl) reels them sk)ward into the ho#ering helicopter! * large strong hand helps them in!/ <oice, 3elcome aboardSuckers& *chilles, 3h-who are you?

<oice, )llow me to introduce myself& I am ?e(achloro!hene J& 6oodfortune, @idna!!er )t-Aar e, and :evourer of Tortoises !ar E(cellence, at your service& Tortoise, 6ul!2 *chilles, 4whispering to his friend5, Fh-ohI think that this -6oodfortune- is not e(actly what we1d antici!ated& 4To Goodfortune/ )hif I may be so boldwhere are you s!iritin us off to? Goodfortune, ?o ho2 To my all-electric kitchen-in-the-sky, where I will !re!are T?IS tasty morsel4leering at the Tortoise as he sa)s this 5in a delicious !ie-in-the-sky2 )nd make no mistakeit1s all .ust for my obblin !leasure2 ?o ho ho2 *chilles, )ll I can say is you1ve ot a !retty fiendish lau h& Goodfortune, 4laughing fiendishl)/: ?o ho ho2 'or that remark, my friend, you will !ay dearly& ?o ho2 *chilles, 6ood riefI wonder what he means by that2 Goodfortune, Gery sim!leI1ve ot a Sinister 'ate in store for both of you2 Just you wait2 ?o ho ho2 ?o ho ho2 *chilles, Hikes2 Goodfortune, 3ell, we have arrived& :isembark, my friends, into my fabulous all-electric kitchen-in-the-sky& The) walk inside!/ Aet me show you around, before I !re!are your fates& ?ere is my bedroom& ?ere is my study& Please wait here for me for a moment& I1ve ot to o shar!en my knives& 3hile you1re waitin , hel! yourselves to some !o!corn& ?o ho ho2 Tortoise !ie2 Tortoise !ie2 /y favorite kind of !ie2 4E9it&5 *chilles, Eh, boy!o!corn2 I1m oin to munch my head off2 Tortoise, )chilles2 Hou .ust stuffed yourself with cotton candy2 0esides, how can you think about food at a time like this? *chilles, 6ood ravyoh, !ardon meI shouldn1t use that turn of !hrase, should I? I mean in these dire circumstances&&& Tortoise, I1m afraid our oose is cooked& *chilles, Saytake a ander at all these books old 6oodfortune has in his study& Muite a collection of esoterica, Bird(rains 3 ,a#e ;nownR 8hess and 5m(rella%Twirling 'ade Eas)R 8oncerto for Tapdancer and :rchestra&&& ?mmm& Tortoise, 3hat1s that small volume lyin o!en over there on the desk, ne(t to the dodecahedron and the o!en drawin !ad? *chilles, This one? 3hy, its title is &ro#ocati#e *d#entures of *chilles and the Tortoise Taking &lace in -undr) -pots of the Glo(e& Tortoise, ) moderately !rovocative title& *chilles, Indeedand the adventure it1s o!ened to looks !rovocative& It1s called -:.inn and Tonic-& Tortoise, ?mm&&& I wonder why& Shall we try readin it? I could take the Tortoise1s !art, and you could take that of )chilles&

*chilles, I1m ame& ?ere oes nothin &&& The) (egin reading ".Minn and Tonic"!/ *chilles has in#ited the Tortoise o#er to see his collection of prints () his fa#orite artist, '! 8! Escher!/ Tortoise, These are wonderful !rints, )chilles& *chilles, I knew you would en.oy seein them& :o you have any !articular favorite? Tortoise, Ene of my favorites is 8on#e9 and 8onca#e, where two internally consistent worlds, when .u(ta!osed, make a com!letely inconsistent com!osite world& Inconsistent worlds are always fun !laces to visit, but I wouldn1t want to live there& *chilles, 3hat do you mean, -fun to visit-? Inconsistent worlds don1t ETIST, so how can you visit one? Tortoise, I be your !ardon, but weren1t we .ust a reein that in this Escher !icture, an inconsistent world is !ortrayed? *chilles, Hes, but that1s .ust a two-dimensional worlda fictitious worlda !icture& Hou can1t visit that world& Tortoise, I have my ways&&& *chilles, ?ow could you !ro!el yourself into a flat !icture-universe? Tortoise, 0y drinkin a little lass of PFS?I86-PETIE8& That does the trick& *chilles, 3hat on earth is !ushin -!otion? Tortoise, It1s a liquid that comes in small ceramic !hials, and which, when drunk by someone lookin at a !icture, -!ushes11 him ri ht into the world of that !icture& Peo!le who aren1t aware of the !owers of !ushin -!otion often are !retty sur!rised by the situations they wind u! in& *chilles, Is there no antidote? Ence !ushed, is one irretrievably lost? Tortoise, In certain cases, that1s not so bad a fate& 0ut there is, in fact, another !otionwell, not a !otion, actually, but an eli(irno, not an eli(ir, but aa Tortoise, ?e !robably means -tonic-& *chilles, Tonic? Tortoise, That1s the word I was lookin for2 -PEPPI86-TE8I7- is what it1s called, and if you remember to carry a bottle of it in your ri ht hand as you swallow the !ushin -!otion, it too will be !ushed into the !icture= then, whenever you et a hankerin to -!o!- back out into real life, you need only take a swallow of !o!!in -tonic, and !resto2 Hou1re back in the real world, e(actly where you were before you !ushed yourself in& *chilles, That sounds very interestin & 3hat would ha!!en it you took some !o!!in -tonic without havin !reviously !ushed yourself into a !icture?

Tortoise, I don1t !recisely know, )chilles, but I would be rather wary of horsin around with these stran e !ushin and !o!!in liquids& Ence I had a friend, a 3easel, who did !recisely what you su estedand no one has heard from him since& *chilles, That1s unfortunate& 7an you also carry alon the bottle of !ushin !otion with you? Tortoise, Eh, certainly& Just hold it in your left hand, and it too will et !ushed ri ht alon with you into the !icture you1re lookin at& *chilles, 3hat ha!!ens if you then find a !icture inside the !icture which you have already entered, and take another swi of !ushin -!otion? Tortoise, Just what you would e(!ect, you wind u! inside that !icture-in-a!icture& *chilles, I su!!ose that you have to !o! twice, then, in order to e(tricate yourself from the nested !ictures, and re-emer e back in real life& Tortoise, That1s ri ht& Hou have to !o! once for each !ush, since a !ush takes you down inside a !icture, and a !o! undoes that& *chilles, Hou know, this all sounds !retty fishy to me&&& )re you sure you1re not .ust testin the limits of my ullibility? Tortoise, I swear2 Aookhere are two !hials, ri ht here in my !ocket& 4=eaches into his lapel pocket, and pulls out two rather large unla(eled phials, in one of which one can hear a red li"uid sloshing around, and in the other of which one can hear a (lue li"uid sloshing around&5 If you1re willin , we can try them& 3hat do you say? *chilles, 3ell, I uess, ahm, maybe, ahm&&& Tortoise, 6ood2 I knew you1d want to try it out& Shall we !ush ourselves into the world of Escher1s 8on#e9 and 8onca#e? *chilles, 3ell, ah,&&& Tortoise, Then it1s decided& 8ow we1ve ot to remember to take alon this flask of tonic, so that we can !o! back out& :o you want to take that heavy res!onsibility, )chilles? *chilles, If it1s all the same to you, I1m a little nervous, and I1d !refer lettin you, with your e(!erience, mana e the o!eration& Tortoise, Gery well, then& -o sa)ing, the Tortoise pours two small portions of pushing%potion! Then he picks up the flask of tonic and grasps it firml) in his right hand, and (oth he and *chilles lift their glasses to their lips!/ Tortoise, 0ottoms u!2 The) swallow!/

73G5=E FH! 7onve( and 7oncave, () '! 8! Escher lithograph, BIGG/!

*chilles, That1s an e(ceedin ly stran e taste& Tortoise, Ene ets used to it& *chilles, :oes takin the tonic feel this stran e? Tortoise, Eh, that1s quite another sensation& 3henever you taste the tonic, you feel a dee! sense of satisfaction, as if you1d been waitin to taste it all your life& *chilles, Eh, I1m lookin forward to that& Tortoise, 3ell, )chilles, where are we? *chilles, 4taking cogni@ance of his surroundings5, 3e1re in a little ondola, lidin down a canal2 I want to et out& /r& 6ondolier, !lease let us out here& The gondolier pa)s no attention to this re"uest!/ Tortoise, ?e doesn1t s!eak En lish& If we want to et out here, we1d better .ust clamber out quickly before he enters the sinister -Tunnel of Aove-= .ust ahead of us&

*chilles, his face a little pale scram(les out in a split second and then pulls his slower friend out!/ *chilles, I didn1t like the sound of that !lace, somehow& I1m lad we ot out here& Say, how do you know so much about this !lace, anyway? ?ave you been here before? Tortoise, /any times, althou h I always came in from other Escher !ictures& They1re all connected behind the frames, you know& Ence you1re in one, you can et to any other one& *chilles, )ma<in 2 3ere I not here, seein these thin s with my own eyes, I1m not sure I1d believe you& 4They wander out throu h a little arch&5 Eh, look at those two cute li<ards2 Tortoise, 7ute? They aren1t cuteit makes me shudder .ust to think of them2 They are the vicious uardians of that ma ic co!!er lam! han in from the ceilin over there& ) mere touch of their ton ues, and any mortal turns to a !ickle& *chilles, :ill, or sweet? Tortoise, :ill& *chilles, Eh, what a sour fate2 0ut if the lam! has ma ical !owers, I would like to try for it& Tortoise, It1s a foolhardy venture, my friend& I wouldn1t risk it& *chilles, I1m oin to try .ust once& ,e stealthil) approaches the lamp, making sure not to awaken the sleeping lad near()! But suddenl), he slips on a strange shell%like indentation in the floor, and lunges out into space! +urching cra@il), he reaches for an)thing, and manages somehow to gra( onto the lamp with one hand! -winging wildl), with (oth li@ards hissing and thrusting their tongues #iolentl) out at him, he is left dangling helplessl) out in the middle of space!/ *chilles, ?e-e-e-el!2 ,is cr) attracts the attention of a woman who rushes downstairs and awakens the sleeping (o)! ,e takes stock of the situation, and, with a kindl) smile on his face, gestures to *chilles that all will (e well! ,e shouts something in a strange guttural tongue to a pair of trumpeters high up in windows, and immediatel), weird tones (egin ringing out and making (eats each other! The sleep) )oung lad points at the li@ards, and *chilles sees that the music is ha#ing a strong soporific effect on them! -oon, the) are completel) unconscious! Then the helpful lad shouts to two companions clim(ing up ladders! The) (oth pull their ladders up and then e9tend them out into space Must underneath the stranded *chilles, forming a sort of (ridge! Their gestures make it clear that *chilles should hurr) and clim(

on! But (efore he does so, *chilles carefull) unlinks the top link of the chain holding the lamp, and detaches the lamp! Then he clim(s onto the ladder%(ridge and the three )oung lads pull him in to safet)! *chilles throws his arms around them and hugs them gratefull)!/ *chilles, Eh, /r& T, how can I re!ay them? Tortoise, I ha!!en to know that these valiant lads .ust love coffee, and down in the town below, there1s a !lace where they make an incom!arable cu! of es!resso& Invite them for a cu! of es!resso2 *chilles, That would hit the s!ot& *nd so, () a rather comical series of gestures, smiles, and words, *chilles manages to con#e) his in#itation to the )oung lads, and the part) of fi#e walks out and down a steep staircase descending into the town! The) reach a charming small cafe, sit down outside, and order fi#e espressos! *s the) sip their drinks, *chilles remem(ers he has the lamp with him!/ *chilles, I for ot, /r& TortoiseI1ve ot this ma ic lam! with me2 0ut what1s ma ic about it? Tortoise, Eh, you know, .ust the usuala enie& *chilles, 3hat? Hou mean a enie comes out when you rub it, and rants you wishes? Tortoise, ;i ht& 3hat did you e(!ect? Pennies from heaven? *chilles, 3ell, this is fantastic2 I can have any wish want, eh? I1ve always wished this would ha!!en to me&&& *nd so *chilles gentl) ru(s the large letter '+' which is etched on the lamp's copper surface!!! -uddenl) a huge puff of smoke appears, and in the forms of the smoke the fi#e friends can make out a weird, ghostl) figure towering a(o#e them!/ Genie, ?ello, my friends and thanks ever so much for rescuin my Aam! from the evil Ai<ard-:uo& *nd so sa)ing, the Genie picks up the +amp, and stuffs it into a pocket concealed among the folds of his long ghostl) ro(e which swirls out of the +amp!/ )s a si n of ratitude for your heroic deed, I would like to offer you, on the !art of my Aam!, the o!!ortunity to have any three of your wishes reali<ed& *chilles, ?ow stu!efyin 2 :on1t you think so, /r& T? Tortoise, I surely do& 6o ahead, )chilles, take the first wish&

*chilles, 3ow2 0ut what should I wish? Eh, I know2 It1s what I thou ht of the first time I read the )rabian 8i hts 4that collection of silly 4and nested5 tales5I wish that I had a ?F8:;E: wishes, instead of .ust three2 Pretty clever, eh, /r& T? I bet HEF never would have thou ht of that trick& I always wondered why those do!ey !eo!le in the stories never tried it themselves& Tortoise, /aybe now you1ll find out the answer& Genie, I am sorry, )chilles, but I don1t rant metawishes& *chilles, I wish you1d tell me what a -meta-wish- is2 Genie, 0ut T?)T is a meta-meta-wish, )chillesand I don1t rant them, either& *chilles, 3haaat? I don1t follow you at all& Tortoise, 3hy don1t you re!hrase your last request, )chilles? *chilles, 3hat do you mean? 3hy should I? Tortoise, 3ell, you be an by sayin -I wish-& Since you1re .ust askin for information, why don1t you .ust ask a question? *chilles, )ll ri ht, thou h I don1t see why& Tell me, /r& 6eniewhat is a meta-wish? Genie, It is sim!ly a wish about wishes& I am not allowed to rant metawishes& It is only within my !urview to rant !lain ordinary wishes, such as wishin for ten bottles of beer, to have ?elen of Troy on a blanket, or to have an all-e(!enses-!aid weekend for two at the 7o!acabana& Hou knowsim!le thin s like that& 0ut meta-wishes I cannot rant& 6E: won1t !ermit me to& *chilles, 6E:? 3ho is 6E:? )nd why won1t he let you rant meta-wishes? That seems like such a !uny thin com!ared to the others you mentioned& Genie, 3ell, it1s a com!licated matter, you see& 3hy don1t you .ust o ahead and make your three wishes? Er at least make one of them& I don1t have all I time in the world, you know&&& *chilles, Eh, I feel so rotten& I was ;E)AAH ?EPI86 wish for a hundred wishes&&& Genie, 6ee, I hate to see anybody so disa!!ointed that& )nd besides, metawishes are my favorite kind of wish& Aet me .ust see if there isn1t anythin I can do about this& This1ll .ust take one moment The Genie remo#es from the wisp) folds of his ro(e an o(Mect which looks Must like the copper +amp he had put awa), e9cept that this one is made of sil#erR and where the pre#ious one had '+' etched on it, this one has ''+' in smaller letters, so as to co#er the same area!/ *chilles, )nd what is that? Genie, This is my /eta-Aam!&&&

,e ru(s the 'eta%+amp, and a huge puff of smoke appears! 3n the (illows of smoke, the) can all make out a ghostl) form towering a(o#e them!/ 'eta%Genie, I am the /eta-6enie& Hou summoned me, E 6enie? 3hat is your wish? Genie, I have a s!ecial wish to make of you, E :.inn and of 6E:& I wish for !ermission for tem!orary sus!ension of all ty!e-restrictions on wishes, for duration of one Ty!eless 3ish& 7ould you !lease rant this wish for me? 'eta%Genie, I1ll have to send it throu h 7hannels, of course& Ene half a moment, !lease& *nd, twice as "uickl) as the Genie did, this 'eta%Genie remo#es from the wisp) folds of her ro(e an o(Mect which looks Must like the sil#er 'eta% +amp, e9cept that it is made of goldR and where the pre#ious one had ''+' etched on it, this one has '''+' in smaller letters, so as to co#er the same area!/ *chilles, 4his #oice an octa#e higher than (efore5, )nd what is that? 'eta%Genie, This is my /eta-/eta-Aam!&&& -he ru(s the 'eta%'eta%+amp, and a hugs puff of smoke appears! 3n the (illows of smoke, the) can all make out a ghostl) fore towering a(o#e them!/ 'eta%'eta%Genie, I am the /eta-/eta-6enie& Hou summoned me, E /eta-6enie? 3hat is your wish? 'eta%Genie, I have a s!ecial wish to make of you, E :.inn, and of 6E:& I wish for !ermission for tem!orary sus!ension of all ty!e-restrictions on wishes, for the duration of one Ty!eless 3ish& 7ould you !lease rant this wish for me? 'eta%'eta%Genie, I1ll have to send it throu h 7hannels, of course& Ene quarter of a moment, !lease&

*nd, twice as "uickl) as the 'eta%Genie did, this 'eta%'eta% Genie remo#es from the folds of his ro(e an o(Mect which looks Must like the gold 'eta+amp, e9cept that it is made of!!!/

TG:.U

!!!swirls (ack into the 'eta%'eta%'eta%+amp, which the 'eta% 'eta%Genie then folds (ack into his ro(e, half as "uickl) as the 'eta%'eta%'eta%Genie did!/ Hour wish is ranted, E /eta-6enie& 'eta%Genie, Thank you, E :.inn, and 6E:& *nd the 'eta%'eta%Genie, as all the higher ones (efore him, swirls (ack into the 'eta%'eta%+amp, which the 'eta%Genie then folds (ack into her ro(e, half as "uickl) as the 'eta%'eta%Genie did!/ Hour wish is ranted, E 6enie& Genie, Thank you, E :.inn, and 6E:& *nd the 'eta%Genie, as all the higher ones (efore her, swirls (ack into the 'eta%+amp, which the Genie folds (ack into his ro(e, half as "uickl) as the 'eta%Genie did!/ Hour wish is ranted, )chilles& *nd one precise moment has elapsed since he "This will Must take one moment!"/ *chilles, Thank you, E :.inn, and 6E:& Genie, I am !leased to re!ort, )chilles, that you may have e(actly one 4"5 Ty!eless 3ishthat is to say wish, or a meta-wish, or a meta-metawish, as many -meta- 1s as you wisheven infinitely many 4if you wish5&

*chilles, Eh, thank you so very much, 6enie& 0ut curiosity is !rovoked& 0efore I make my wish, would you mind tellin me whoor what 6E: is? Genie, 8ot at all& -6E:- is an acronym which stands -6E: Ever :.inn-& The word -:.inn- is used desi nate 6enies, /eta-6enies, /eta-/eta6enies, etc& It is a Ty!eless word& *chilles, 0ut-but-how can -6E:- be a word in own acronym? That doesn1t make any sense2 Genie, Eh, aren1t you acquainted with recursive acronyms? I thou ht everybody knew about them& Hou see, -6E:- stands for -6E: Ever :.inn-which can be e(!anded as -6E: Ever :.inn, Ever :.inn- and that can, in turn, be e(!anded to -6E: Ever :.inn, Ever :.inn, Ever :.inn-which can its turn, be further e(!anded&&& Hou can o as far as you like& *chilles, 0ut I1ll never finish2 Genie, Ef course not& Hou can never totally e(!and 6E:& *chilles, ?mm&&& That1s !u<<lin & 3hat did you me when you said to the /eta-6enie, -I have a s!ecial wish to make of you, E :.inn, and of 6E:-? Genie, I wanted not only to make a request of /eta-6enie, but also of all the :.inns over her& The recursive acronym method accom!lishes this quite naturally& Hou see, when the /eta-6enie received my request, she then had to !ass it u!wards to her 6E:& So she forwarded a similar messa e to the /eta-/eta-6enie, who then did likewise to the /eta-/eta-/eta-6enie&&& )scendin the chain this way transmits the messa e to 6E:& *chilles, I see& Hou mean 6E: sits u! at the to! of the ladder of d.inns? Genie, 8o, no, no2 There is nothin -at the to!-, for there is no to!& That is why 6E: is a recursive acronym& 6E: is not some ultimate d.inn= 6E: is the tower of d.inns above any iven d.inn& Tortoise, It seems to me that each and every d.inn would have a different conce!t of what 6E: is, then, since to any d.inn, 6E: is the set of d.inns above him or her, and no two d.inns share that set& Genie, Hou1re absolutely ri htand since I am the lowest d.inn of all, my notion of 6E: is the most e(alted one& I !ity the hi her d.inns, who fancy themselves somehow closer to 6E:& 3hat blas!hemy2 *chilles, 0y um, it must have taken enies to invent 6E:& Tortoise, :o you really believe all this stuff about 6E:, )chilles? *chilles, 3hy certainly, I do& )re you atheistic, /r& T? Er are you a nostic? Tortoise, I don1t think I1m a nostic& /aybe I1m meta-a nostic& *chilles, 3haaat? I don1t follow you at all& Tortoise, Aet1s see&&& If I were meta-a nostic, I1d be confused over whether I1m a nostic or notbut I1m not quite sure if I feel T?)T way= hence I must be meta-meta-a nostic 4I uess5& Eh, well& Tell me, 6enie, does any

d.inn ever make a mistake, and arble u! a messa e movin u! or down the chain? Genie, This does ha!!en= it is the most common cause for Ty!eless 3ishes not bein ranted& Hou see, the chances are infinitesimal, that a arblin will occur at any P);TI7FA); link in the chainbut when you !ut an infinite number of them in a row, it becomes virtually certain that a arblin will occur SE/E3?E;E& In fact, stran e as it seems, an infinite number of arblin s usually occur, althou h they are very s!arsely distributed in the chain& *chilles, Then it seems a miracle that any Ty!eless 3ish ever ets carried out& Genie, 8ot really& /ost arblin s are inconsequential, and many arblin s tend to cancel each other out& 0ut occasionallyin fact, rather seldom the nonfulfillment of a Ty!eless 3ish can be traced back to a sin le unfortunate d.inn1s arblin & 3hen this ha!!ens, the uilty d.inn is forced to run an infinite 6auntlet and et !addled on his or her rum!, by 6E:& It1s ood fun for the !addlers, and quite harmless for the !addlee& Hou mi ht be amused by the si ht& *chilles, I would love to see that2 0ut it only ha!!ens when a Ty!eless 3ish oes un ranted? Genie, That1s ri ht& *chilles, ?mm&&& That ives me an idea for my wish& Tortoise, Eh, really? 3hat is it? *chilles, I wish my wish would not be ranted2 *t that moment, an e#entOor is "e#ent" the word for it$Otakes place which cannot (e descri(ed, and hence no attempt will (e made to descri(e it!/ *chilles, 3hat on earth does that cry!tic comment mean? Tortoise, It refers to the Ty!eless 3ish )chilles made& *chilles, 0ut he hadn1t yet made it& Tortoise, Hes, he had& ?e said, -I wish my wish would not be ranted-, and the 6enie took T?)T to be his wish& *t that moment, some footsteps are heard coming down the hallwa) in their direction!/ *chilles, Eh, my2 That sounds ominous& The footsteps stopR then the) turn around and fade awa)!/ Tortoise, 3hew2 *chilles, 0ut does the story o on, let1s see& or is that the end? Turn the !a e and let1s see&

The Tortoise turns the page of ".Minn and Tonic", where the) find that the stor) goes on!!!/ *chilles, ?ey2 3hat ha!!ened? 3here is my 6enie? /y lam!? /y cu! of es!resso? 3hat ha!!ened to our youn friends from the 7onve( and 7oncave worlds? 3hat are all those little li<ards doin here? Tortoise, I1m afraid our conte(t ot restored incorrectly, )chilles& *chilles, 3hat on earth does that cry!tic comment mean? Tortoise, I refer to the Ty!eless 3ish you made& *chilles, 0ut I hadn1t yet made it& Tortoise, Hes, you had& Hou said, -I wish my wish would not be ranted-, and the 6enie took T?)T to be your wish& *chilles, Eh, my2 That sounds ominous& Tortoise, It s!ells P);):ET& 'or that Ty!eless wish to be ranted, it had to be deniedyet not to rant it would be to rant it& *chilles, So what ha!!ened? :id the earth come to a standstill? :id the universe cave in? Tortoise, 8o& The System crashed& *chilles, 3hat does that mean? Tortoise, It means that you and I, )chilles, were suddenly and instantaneously trans!orted to Tumbolia& *chilles, To where? Tortoise, Tumbolia, the land of dead hiccu!s and e(tin uished li ht bulbs& It1s a sort of waitin room, where dormant software waits for its host hardware to come back u!& 8o tellin how lon the System was down, and we were in Tumbolia& It could have been moments, hours, dayseven years& *chilles, I don1t know what software is, and I don1t know what hardware is& 0ut I do know that I didn1t et to make my wishes2 I want my 6enie back2 Tortoise, I1m sorry, )chillesyou blew it& Hou crashed the System, and you should thank your lucky stars that we1re back at all& Thin s could have come out a lot worse& 0ut I have no idea where we are& *chilles, I reco ni<e it nowwe1re inside another of Escher1s !ictures& This time it1s =eptiles& Tortoise, )ha2 The System tried to save as much of our conte(t as it could before it crashed, and it ot as far as recordin that it was an Escher !icture with li<ards before it went down& That1s commendable& *chilles, )nd lookisn1t that our !hial of !o!!in -tonic over there on the table, ne(t to the cycle of li<ards?

Tortoise, It certainly is, )chilles& I must say, we are very lucky indeed& The System was very kind to us, in ivin us back our !o!!in tonicit1s !recious stuff2 *chilles, I1ll say2 8ow we can !o! back out of the Escher world, into my house& Tortoise, There are a cou!le of books on the desk, ne(t to the tonic& I wonder what they are& 4,e picks up the smaller one, which is open to a random page!5 This looks like a moderately !rovocative book& *chilles, Eh, really? 3hat is its title? Tortoise, &ro#ocati#e *d#entures of the Tortoise and *chilles Taking &lace in -undr) &arts of the Glo(e& It sounds like an interestin book to read out of&

73G5=E FD! ;e!tiles, () '! 8! Escher lithograph, BIDH5&

*chilles, 3ell, HEF can read it if you want, but as for me, I1m not oin to take any chances with that !o!!in -tonicone of the li<ards mi ht knock it off the table, so I1m oin to et it ri ht now2

,e dashes o#er to the ta(le and reaches for the popping%tonic, (ut in his haste he somehow (umps the flask of tonic, and it tum(les off the desk and (egins rolling!/ Eh, no2 /r& Tlook2 I accidentally knocked the tonic onto the floor, and it1s rollin towards- towards- the stairwell2 Muick before it falls2 The Tortoise, howe#er, is completel) wrapped up in the thin #olume which he has in his hands!/ Tortoise 4muttering5, Eh? This story looks fascinatin & *chilles, /r& T, /r& T, hel!2 ?el! catch the tonic-flask2 Tortoise, 3hat1s all the fuss about? *chilles, The tonic-flask-I knocked it down from the desk, and now it1s rollin and *t that instant it reaches the (rink of the stairwell, and plummets o#er!!! / Eh no2 3hat can we do? /r& Tortoisearen1t you alarmed? 3e1re losin our tonic2 It1s .ust fallen down the stairwell2 There1s only one thin to do2 3e1ll have to o down one story2 Tortoise, 6o down one story? /y !leasure& 3on1t you .oin me? ,e (egins to read aloud, and *chilles, pulled in two directions at once, finall) sta)s, taking the role of the Tortoise!/ *chilles, It1s very dark here, /r& T& I can1t see a thin & Eof2 I bum!ed into a wall& 3atch out2 Tortoise, ?ereI have a cou!le of walkin sticks& 3hy don1t you take one of them? Hou can hold it out in front of you so that you don1t ban into thin s& *chilles, 6ood idea& 4,e takes the stick&5 :o you et the sense that this !ath is curvin ently to the left as we walk? Tortoise, Gery sli htly, yes& *chilles, I wonder where we are& )nd whether we1ll ever see the li ht of day a ain& I wish I1d never listened to you, when you su ested I swallow some of that -:;I8@ /E- stuff& Tortoise, I assure you, it1s quite harmless& I1ve done it scads of times, and not a once have I ever re retted it& ;ela( and en.oy bein small& *chilles, 0ein small? 3hat is it you1ve done to me, /r& T? Tortoise, 8ow don1t o blamin me& Hou did it of your own free will&

*chilles, ?ave you made me shrink? So that this labyrinth we1re in is actually some teeny thin that someone could STEP on?

73G5=E FG! 8retan +a()rinth 3talian engra#ingR -chool of 7iniguerra/! U7rom N 'atthews, /a<es and Aabyrinths, Their ?istory and :evelo!ment New York: .o#er &u(lications, BICK/!

Tortoise, Aabyrinth? Aabyrinth? 7ould it )re we in the notorious Aittle ?armonic Aabyrinth of the dreaded /a.otaur? *chilles, Hiikes2 3hat is that? Tortoise, They sayalthou h I !erson never believed it myselfthat an Evil /a.otaur has created a tiny labyrinth and sits in a !it in the middle of it, waitin for innocent victims to et lost in its fearsome com!le(ity& Then, when they wander and da<ed into the center, he lau hs and lau hs at themso hard, that he lau hs them to death2 *chilles, Eh, no2 Tortoise, 0ut it1s only a myth& 7oura e, )chilles& *nd the dauntless pair trudge on!/ *chilles, 'eel these walls& They1re like corru ated tin sheets, or somethin & 0ut the corru ations have different si<es&

To emphasi@e his point, he sticks out his walking stick against the wall surface as he walks! *s the stick (ounces (ack and forth against the corrugations, strange noises echo up and down the long cur#ed corridor the) are in!/ Tortoise alarmed/, 3hat was T?)T? *chilles, Eh, .ust me, rubbin my walkin stick a ainst the wall& Tortoise, 3hew2 I thou ht for a moment it was the bellowin of the ferocious /a.otaur2 *chilles, I thou ht you said it was all a myth& Tortoise, Ef course it is& 8othin to be afraid of& *chilles puts his walking stick (ack against the wall, and continues walking! *s he does so, some musical sounds are heard, coming from the point where his stick is scraping the wall!/ Tortoise, Fh-oh& I have a bad feelin , )chilles& That Aabyrinth may not be a myth, after all& *chilles, 3ait a minute& 3hat makes you chan e your mind all of a sudden? Tortoise, :o you hear that music? To hear more clearl), *chilles lowers the stick, and the strains of melod) cease!/ ?ey2 Put that back2 I want to hear the end of this !iece2 8onfused, *chilles o(e)s, and the music resumes!/ Thank you& 8ow as I was about to say, I have .ust fi ured out where we are& *chilles, ;eally? 3here are we? Tortoise, 3e are walkin down a s!iral roove of a record in its .acket& Hour stick scra!in a ainst the stran e sha!es in the wall acts like a needle runnin down the roove, allowin us to hear the music& *chilles, Eh, no, oh, no&&& Tortoise, 3hat? )ren1t you over.oyed? ?ave you ever had the chance to be in such intimate contact with music before? *chilles, ?ow am I ever oin to win footraces a ainst full-si<ed !eo!le when I am smaller than a flea, /r& Tortoise? Tortoise, Eh, is that all that1s botherin you? That1s nothin to fret about, )chilles&

*chilles, The way you talk, I et the im!ression that you never worry at all& Tortoise, I don1t know& 0ut one thin for certain is that I don1t worry about bein small& Es!ecially not when faced with the awful dan er of the dreaded /a.otaur2 *chilles, ?orrors2 )re you tellin me Tortoise, I1m afraid so, )chilles& The music ave it away& *chilles, ?ow could it do that? Tortoise, Gery sim!le& 3hen I heard melody 0-)-7-? in the to! voice, I immediately reali<ed that the rooves we1re walkin throu h could only be +ittle ,armonic +a()rinth, one of 0ach1s lesser known or an !ieces& It is so named because of its di<<yin ly frequent modulations& *chilles, 3h-what are they? Tortoise, 3ell, you know that most music !ieces are written in a key, or tonality, as 7 ma.or, which is the key of this one& *chilles, I had heard the term before& :o that mean that 7 is the note you want to end on? Tortoise, Hes, 7 acts like a home base, in a way& )ctually, the usual word is -tonic-& *chilles, :oes one then stray away from tonic with the aim of eventually returnin ? Tortoise, That1s ri ht& )s the !iece develo!s, ambi uous chords and melodies are used which lead away from the tonic& Aittle by little, tension builds u!you feel an increasin desire to return home, to hear the tonic& *chilles, Is that why, at the end of a !ie always feel so satisfied, as if I had waitin my whole life to hear the tonic? Tortoise, E(actly& The com!oser has uses knowled e of harmonic !ro ressions to mani!ulate your emotions, and to build u! ho!es in you to hear that tonic& *chilles, 0ut you were oin to tell me about modulations& Tortoise, Eh, yes& Ene very im!ortant thin a com!oser can do is to -modulate- !artway throu h a !iece, which means that he sets u! a tem!orary oal other than resolution into the tonic& *chilles, I see&&& I think& :o you mean that some sequence of chords shifts the harmonic tension somehow so that I actually desire to resolve in a new key? Tortoise, ;i ht& This makes the situation more com!le(, for althou h in the short term you want to resolve in the new key, all the while at the back of your mind you retain the lon in to hit that ori inal oal-in this case, 7 ma.or& )nd when the subsidiary oal is reached, there is

*chilles 4suddenl) gesturing enthusiasticall)5, Eh, listen to the or eous u!ward-swoo!in chords which mark the end of this +ittle ,armonic +a()rinthA Tortoise, 8o, )chilles, this isn1t the end& It1s merely *chilles, Sure it is2 3ow2 3hat a !owerful, stron endin 2 3hat a sense of relief2 That1s some resolution2 6ee2 *nd sure enough, at that moment the music stops, as the) emerge into an open area with no walls!/ Hou see, it IS over& 3hat did I tell you? Tortoise, Somethin is very wron & This record is a dis race to the world of music& *chilles, 3hat do you mean? Tortoise, It was e(actly what I was tellin you about& ?ere 0ach had modulated from 7 into 6, settin u! a secondary oal of hearin 6& This means that you e(!erience two tensions at oncewaitin for resolution into 6, but also kee!in in mind that ultimate desireto resolve trium!hantly into 7 /a.or& *chilles, 3hy should you have to kee! anythin in mind when listenin to a !iece of music? Is music only an intellectual e(ercise? Tortoise, 8o, of course not& Some music is hi hly intellectual, but most music is not& )nd most of the time your ear or brain does the -calculation- for you, and lets your emotions know what they want to hear& Hou don1t have to think about it consciously in this !iece, 0ach was !layin tricks ho!in to lead you astray& )nd in your case, )chilles, he succeeded& *chilles, )re you tellin me that I res!onded to a resolution in a subsidiary key? Tortoise, That1s ri ht& *chilles, It still sounded like an endin to me& Tortoise, 0ach intentionally made it sound that way& Hou .ust fell into his tra!& It was deliberately contrived to sound like an endin but if you follow the harmonic !ro ression carefully, you will see that it is in the wron key& )!!arently not .ust you but this miserable record com!any fell for the same trickand they truncated the !iece early& *chilles, 3hat a dirty trick 0ach !layed on me2 Tortoise, That is his whole ameto make you lose your way in his Aabyrinth2 The Evil /a.otaur is in cahoots with 0ach, )nd if you don1t watch out, he will now lau h you to deathand !erha!s me with you2 *chilles, Eh, let us hurry u! and et here2 Muick2 Aet1s run backwards in the rooves, and esca!e on the outside of the record before the Evil /a.otaur finds us&

Tortoise, ?eavens, no2 /y sensibility is far too delicate to handle the bi<arre chord !ro ressions which occur when time is reversed& *chilles, Eh, /r& T, how will we ever et out of here, if we can1t .ust retrace our ste!s? Tortoise, That1s a very ood question& * little desperatel), *chilles starts running a(out aimlessl) in the dark! -uddenl) there is a slight gasp, and then a "thud"!/ )chillesare you all ri ht? *chilles, Just a bit shaken u! but otherwise fine& I fell into some bi hole& Tortoise, Hou1ve fallen into the !it of the Evil /a.otaur2 ?ere, I1ll come hel! you out& 3e1ve ot to move fast2 *chilles, 7areful, /r& TI don1t want HEF to fall in here, too&&& Tortoise, :on1t fret, )chilles& Everythin will be all -uddenl), there is a slight gasp, and then a "thud"!/ *chilles, /r& Tyou fell in, too2 )re you all ri ht? Tortoise, Enly my !ride is hurtotherwise I1m fine& *chilles, 8ow we1re in a !retty !ickle, aren1t we? -uddenl), a giant, (ooming laugh is heard, alarmingl) close to them!/ Tortoise, 3atch out, )chilles2 This is no lau hin matter& 'aMotaur, ?ee hee hee2 ?o ho2 ?aw haw haw2 *chilles, I1m startin to feel weak, /r& T&&& Tortoise, Try to !ay no attention to his lau h, *chilles& That1s your only ho!e& *chilles, I1ll do my best& If only my stomach weren1t em!ty2 Tortoise, Say, am I smellin thin s, or is there a bowl of hot buttered !o!corn around here? *chilles, I smell it, too& 3here is it comin from? Tortoise, Ever here, I think& Eh2 I .ust ran into a bi bowl of the stuff& Hes, indeedit seems to be a bowl of !o!corn2 *chilles, Eh, boy!o!corn2 I1m oin to munch my head off2 Tortoise, Aet1s .ust ho!e it isn1t !ushcorn2 Pushcorn and !o!corn are e(traordinarily difficult to tell a!art& *chilles, 3hat1s this about Pushkin? Tortoise, I didn1t say a thin & Hou must be hearin thin s& *chilles, 6o- olly2 I ho!e not& 3ell, let1s di in2

*nd the two friends (egin munching popcornO or pushcorn$/Oand all at onceO&:&A 3 guess it was popcorn, after all!/ Tortoise, 3hat an amusin story& :id you en.oy it? *chilles, /ildly& Enly I wonder whether they ever ot out of that Evil /a.otaur1s !it or not& Poor )chilleshe wanted to be full-si<ed a ain& Tortoise, :on1t worrythey1re out, and he is full-si<ed a ain& That1s what the -PEP- was all about& *chilles, Eh, I couldn1t tell& 3ell, now I ;E)AAH want to find that bottle of tonic& 'or some reason, my li!s are burnin & )nd nothin would taste better than a drink of !o!!in -tonic& Tortoise, That stuff is renowned for its thirst quenchin !owers& 3hy, in some !laces !eo!le very nearly o cra<y over it& )t the turn of the century in Gienna, the Sch>nber food factory sto!!ed makin tonic, and started makin cereal instead& Hou can1t ima ine the u!roar that caused& *chilles, I have an inklin & 0ut let1s o look for the tonic& ?ey.ust a moment& Those li<ards on the deskyou see anythin funny about them? Tortoise, Fmm&&& not !articularly& 3hat do you see of such reat interest? *chilles, :on1t you see it? They1re emer in from that flat !icture without drinkin any !o!!in -tonic2 ?ow are they able to do that? Tortoise, Eh, didn1t I tell you? Hou can et out of a !icture by movin !er!endicularly to its !lane, if you have no !o!!in -tonic& The little li<ards have learned to climb FP when they want to et out of the two-dimensional sketchbook world& *chilles, 7ould we do the same thin to et Escher !icture we1re in? Tortoise, Ef course2 3e .ust need to o FP one story& :o you want to try it? *chilles, )nythin to et back to my house2 I1m tired of all these !rovocative adventures& Tortoise, 'ollow me, then, u! this way& *nd the) go up one stor)!/ *chilles, It1s ood to be back& 0ut somethin seems wron & This isn1t my house2 This is HEF; house, /r& Tortoise Tortoise, 3ell, so it isand am I lad for that2 I wasn1t lookin forward one whit to the lon walk back from your house& I am bushed, and doubt if I could have made it&

*chilles, I don1t mind walkin home, so I uess it1s lucky we ended u! here, after all& Tortoise, I1ll say2 This certainly is a !iece of 6ood 'ortune2

Cha!ter 4

$ecursi"e Structures and ,rocesses

>hat Is $ecursion7
3?)T IS ;E7F;SIE8? It is what was illustrated in the :ialo ue +ittle ,armonic +a()rinth, nestin , and variations on nestin & The conce!t is very eneral& 4Stories inside stories, movies inside movies, !aintin s inside !aintin s, ;ussian dolls inside ;ussian dolls 4even !arenthetical comments in& side !arenthetical comments25these are .ust a few of the charms of recursion&5 ?owever, you should he aware that the meanin of -recursive1 in this 7ha!ter is only faintly related to its meanin in 7ha!ter III& The relation should be clear by the end of this 7ha!ter& Sometimes recursion seems to brush !arado( very closely& 'or e(am!le, there are recursi#e definitions& Such a definition may ive the casual viewer the im!ression that somethin is bein defined in terms of itself! That would be circular and lead to infinite re ress, if not to !arado( !ro!er& )ctually, a recursive definition 4when !ro!erly formulated5 never leads to infinite re ress or !arado(& This is because a recursive definition never defines somethin in terms of itself, but always in terms of simpler #ersions of itself& 3hat I mean by this will become clearer shortly, when I show some e(am!les of recursive definitions& Ene of the most common ways in which recursion a!!ears in daily life is when you !ost!one com!letin a task in favor of a sim!ler task, often of the same ty!e& ?ere is a ood e(am!le& )n e(ecutive has a fancy tele!hone and receives many calls on it& ?e is talkin to ) when 0 calls& To ) he says, -3ould you mind holdin for a moment?- Ef course he doesn1t really care if ) minds= he .ust !ushes a button, and switches to 0& 8ow 7 calls& The same deferment ha!!ens to 0& This could o on indefinitely, but let us not et too bo ed down in our enthusiasm& So let1s say the call with 7 terminates& Then our e(ecutive -!o!s- back u! to 0, and continues& /eanwhile ) is sittin at the other end of the line, drummin his fin ernails a ain some table, and listenin to some horrible /u<ak !i!ed throu h the !hone lines to !lacate him&&& 8ow the easiest case is if the call with 0 sim!ly terminates, and the e(ecutive returns to ) finally& 0ut it could ha!!en that after the conversation with 0 is resumed, a new caller:calls& 0 is once a ain !ushed onto the stack of waitin callers, and : is taken care of& )fter : is done, back to 0, then back to )& This e(ecutive is ho!elessly mechanical, to be surebut we are illustratin recursion in its most !recise form&

,ushing) ,o!!ing) and Stac.s

In the !recedin e(am!le, I have introduced some basic terminolo y of recursionat least as seen throu h the eyes of com!uter scientists& The terms are push, pop, and stack 4or push%down stack, to be !recise5 and they are all related& They were introduced in the late "#L+1s as !art of IPA, one of the first lan ua es for )rtificial Intelli ence& Hou have already encountered -!ush- and -!o!- in the :ialo ue& 0ut I will s!ell thin s out anyway& To push means to sus!end o!erations on the task you1re currently workin on, without for ettin where you areand to take u! a new task& The new task is usually said to be -on a lower level- than the earlier task& To pop is the reverseit means to close o!erations on one level, and to resume o!erations e(actly where you left off, one level hi her& 0ut how do you remember e(actly where you were on each different level? The answer is, you store the relevant information in a stack& So a stack is .ust a table tellin you such thin s as 4"5 where you were in each unfinished task 4.ar on, the -return address-5, 4B5 what the relevant facts to know were at the !oints of interru!tion 4.ar on, the -variable bindin s-5& 3hen you !o! back u! to resume some task, it is the stack which restores your conte(t, so you don1t feel lost& In the tele!hone-call e(am!le, the stack tells you who is waitin on each different level, and where you were in the conversation when it was interru!ted& 0y the way, the terms -!ush-, -!o!-, and -stack- all come from the visual ima e of cafeteria trays in a stack& There is usually some sort of s!rin underneath which tends to kee! the to!most tray at a constant hei ht, more or less& So when you !ush a tray onto the stack, it sinks a littleand when you remove a tray from the stack, the stack !o!s u! a little& Ene more e(am!le from daily life& 3hen you listen to a news re!ort on the radio, oftentimes it ha!!ens that they switch you to some forei n corres!ondent& -3e now switch you to Sally Swum!ley in Peafo , En land&- 8ow Sally has ot a ta!e of some local re!orter interviewin someone, so after ivin a bit of back round, she !lays it& -I1m 8i el 7adwallader, here on scene .ust outside of Peafo , where the reat robbery took !lace, and I1m talkin with&&&- 8ow you are three levels down& It may turn out that the interviewee also !lays a ta!e of some conversation& It is not too uncommon to o down three levels in real news re!orts, and sur!risin ly enou h, we scarcely have any awareness of the sus!ension& It is all ke!t track of quite easily by our subconscious mind& Probably the reason it is so easy is that each level is e(tremely different in flavor from each other level& If they were all similar, we would et confused in no time flat& )n e(am!le of a more com!le( recursion is, of course, our :ialo ue& There, )chilles and the Tortoise a!!eared on all the different levels& Sometimes they were readin a story in which they a!!eared as characters& That is when your mind may et a little ha<y on what1s oin on, and you have to concentrate carefully to et thin s strai ht& -Aet1s see, the real )chilles and Tortoise are still u! there in 6oodfortune1s helico!ter, but the secondar) ones are in some Escher !ictureand then they found this book and are readin in it, so it1s the tertiar) )chilles and Tortoise who wanderin around inside the rooves of the +ittle ,armonic +a()rinth& 8o, wait a minuteI left out one level

somewhere&&&- Hou have to have a conscious mental stack like this in order to kee! track of the recursion in the :ialo ue& 4See 'i & BD&5

73G5=E FJ! .iagram of the structure of the .ialogue Aittle ?armonic Aabyrinth& <ertical descents are "pushes"R rises are "pops"! Notice the similarit) of this diagram to indentation pattern of the .ialogue! 7rom the diagram it is clear that the initial tension OGoodfortune's threatOne#er was resol#edR *chilles and the Tortoise were Must left dangling in the sk)! -ome readers might agoni@e o#er this unpopped push, while others might not (at an e)elash! 3n the stor), Bach's musical la()rinth likewise was cut off too soonO (ut *chilles didn't e#en notice an)thing funn)! :nl) the Tortoise was aware of the more glo(al dangling tension!

Stac.s in Music
3hile we1re talkin about the +ittle ,armonic +a()rinth, we should discuss somethin which is hinted at, if not stated e(!licitly in the :ialo ue, that we hear music recursivelyin !articular, that we maintain a mental stack of keys, and that each new modulation !ushes a new key onto the stack& The im!lication is further that we want to hear that sequence of keys retraced in reverse order!o!!in the !ushed keys off the stack, one by one, until the tonic is reached& This is an e(a eration& There is a rain of truth to it, however& )ny reasonably musical !erson automatically maintains a shallow stack with two keys& In that -short stack-, the true tonic key is held and also most immediate -!seudotonic- 4the key the com!oser is !retendin to be in5& In other words, the most lobal key and the most local key& That the listener knows when the true tonic is re ained, and feels a stron sense of -relief-& The listener can also distin uish 4unlike )chilles5 between a local easin of tensionfor e(am!le a resolution into the !seudotonicand a glo(al resolution& In fact, a !seudoresolution should hei hten the lobal tension, not relieve it, because it is a !iece of irony.ust like )chilles1 rescue from his !erilous !erch on the swin in lam!, when all the while you know he and the Tortoise are really awaitin their dire fates at the knife of /onsieur 6oodfortune& Since tension and resolution are the heart and soul of music, there are many, many e(am!les& 0ut let us .ust look at a cou!le in 0ach& 0ach wrote many !ieces in an -))00- formthat is, where there are two halves, and each one is re!eated& Aet1s take

the i ue from the 'rench Suite no& L, which is quite ty!ical of the form& Its tonic key is 6, and we hear a ay dancin melody which establishes the key of 6 stron ly& Soon, however, a modulation in the )-section leads to the closely related key of : 4the dominant5& 3hen the )-section ends, we are in the key of :& In fact, it sounds as if the !iece has ended in the key of :2 4Er at least it mi ht sound that way to )chilles&5 0ut then a stran e thin ha!!enswe abru!tly .um! back to the be innin , back to 6, and rehear the same transition into :& 0ut then a stran e thin ha!!enswe abru!tly .um! back to the be innin , back to 6, and rehear the same transition into :& Then comes the 0-section& 3ith the inversion of the theme for our melody, we be in in : as if that had always been the tonicbut we modulate back to 6 after all, which means that we !o! back into the tonic, and the 0-section ends !ro!erly& Then that funny re!etition takes !lace, .erkin us without warnin back into :, and lettin us return to 6 once more& Then that funny re!etition takes !lace, .erkin us without warnin back into :, and lettin us return to 6 once more& The !sycholo ical effect of all this key shiftin some .erky, some smoothis very difficult to describe& It is !art of the ma ic of music that we can automatically make sense of these shifts& Er !erha!s it is the ma ic of 0ach that he can write !ieces with this kind of structure which have such a natural race to them that we are not aware of e(actly what is ha!!enin & The ori inal +ittle ,armonic +a()rinth is a !iece by 0ach in which he tries to lose you in a labyrinth of quick key chan es& Pretty soon you are so disoriented that you don1t have any sense of direction leftyou don1t know where the true tonic is, unless you have !erfect !itch, or like Theseus, have a friend like )riadne who ives you a thread that allows you to retrace your ste!s& In this case, the thread would be a written score& This !ieceanother e(am!le is the Endlessly ;isin 7anon oes to show that, as music listeners, we don1t have very reliable dee! stacks&

$ecursion in Language
Eur mental stackin !ower is !erha!s sli htly stron er in lan ua e& The rammatical structure of all lan ua es involves settin u! quite elaborate !ush-down stacks, thou h, to be sure, the difficulty of understandin a sentence increases shar!ly with the number of !ushes onto the stack& The !roverbial 6erman !henomenon of the -verb-at-the-end-, about which droll tales of absentminded !rofessors who would be in a sentence, ramble on for an entire lecture, and then finish u! by rattlin off a strin of verbs by which their audience, for whom the stack had lon since lost its coherence, would be totally non!lussed, are told, is an e(cellent e(am!le of lin uistic !ushin and !o!!in & The confusion amon the audience out-of-order !o!!in from the stack onto which the !rofessor1s verbs been !ushed, is amusin to ima ine, could en ender& 0ut in normal s!oken 6erman, such dee! stacks almost never occurin fact, native s!eaker of 6erman often unconsciously violate certain conventions which force the verb to o to the end, in order to avoid the mental effort of kee!in track of the stack& Every lan ua e has constructions which involve stacks, thou h usually of a less s!ectacular nature than

6erman& 0ut there are always of re!hrasin sentences so that the de!th of stackin is minimal&

$ecursi"e Transition ;et'or.s

The syntactical structure of sentences affords a ood !lace to !resent a of describin recursive structures and !rocesses, the =ecursi#e Transition Network 4;T85& )n ;T8 is a dia ram showin various !aths which can be followed to accom!lish a !articular task& Each !ath consists of a number of nodes, or little bo(es with words in them, .oined by arcs, or lines with arrows& The overall name for the ;T8 is written se!arately at the left, and the first and last nodes have the words (egin and end in them& )ll the other nodes contain either very short e(!licit directions to !erform, or else name other ;T81s& Each time you hit a node, you are to carry out the directions inside it, or to .um! to the ;T8 named inside it, and carry it out& Aet1s take a sam!le ;T8, called 0$;ATE ;0&;, which tells how to construct a certain ty!e of En lish noun !hrase& 4See 'i & B$a&5 If we traverse 0$;ATE ;0&; !urely hori<ontally, we (egin, then we create an A$TICLE, an A 3ECTI4E, and a ;0&;, then we end& 'or instance, -the sham!oo- or -a thankless brunch-& 0ut the arcs show other !ossibilities such as ski!!in the article, or re!eatin the ad.ective& Thus we could construct -milk-, or -bi red blue reen snee<es-, etc& 3hen you hit the node ;0&;, you are askin the unknown black bo( called ;0&; to fetch any noun for you from its storehouse of nouns& This is known as a procedure call, in com!uter science terminolo y& It means you tem!orarily ive control to a procedure 4here, ;0&;5 which 4"5 does thin 4!roduces a noun5 and then 4B5 hands control back to you& In above ;T8, there are calls on three such !rocedures, A$TICLE, A 3ECTI4E and ;0&;& 8ow the ;T8 0$;ATE ;0&; could itself be called from some other ;T8for instance an ;T8 called SE;TE;CE& In this case, 0$;ATE ;0&; would !roduce a !hrase such as -the silly sham!oo- and then return to the !lace inside SE;TE;CE from which it had been called& It is quite reminiscent of the way in which you resume where you left off nested tele!hone calls or nested news re!orts& ?owever, des!ite callin this a -recursive transition network-, we have not e(hibited any true recursion so far& Thin s et recursiveand seemin ly circularwhen you o to an ;T8 such as the one in 'i ure B$b, for +A;C= ;0&;& )s you can see, every !ossible !athway in +A;C= ;0&; involves a call on 0$;ATE ;0&;, so there is no way to avoid ettin a noun of some sort or other& )nd it is !ossible to be no more ornate than that, comin out merely with -milk- or -bi red blue reen snee<es-& 0ut three of the !athways involve recursi#e calls on +A;C= ;0&; itself& It certainly looks as if somethin is bein defined in terms of itself& Is that what is ha!!enin , or not?

73G5=E FC! =ecursi#e Transition Networks for 0$;ATE ;0&; and +A;C= ;0&;!

The answer is -yes, but beni nly-& Su!!ose that, in the !rocedure SE;TE;CE, there is a node which calls +A;C= ;0&;, and we hit that node& This means that we commit to memory 4vi<&, the stack5 the location of that node inside SE;TE;CE, so we1ll know where to return tothen we transfer our attention to the !rocedure +A;C= ;0&;& 8ow we must choose a !athway to take, in order to enerate a +A;C= ;0&;& Su!!ose we choose the lower of the u!!er !athwaysthe one whose callin sequence oes, 0$;ATE ;0&;; $ELATI4E ,$0;0&;; +A;C= ;0&;; 4E$B. So we s!it out an 0$;ATE ;0&;, -the strange (agels-= a $ELATI4E ;0&;, -that-= and now we are suddenly asked for a +A;C= ;0&;& 0ut we are in the middle of +A;C= ;0&;2 Hes, but remember our e(ecutive who was in the middle of one !hone call when he ot another one& ?e merely stored the old !hone call1s status on a stack, and be an the new one as if nothin were unusual& So we shall do the same& 3e first write down in our stack the node we are at in the outer call on +A;C= ;0&;, so that we have a -return address-= then we .um! to the be innin of +A;C=

;0&; as if nothin were unusual& 8ow we have to choose a !athway a ain& 'or variety1s sake, let1s choose the lower !athway, 0$;ATE ;0&;; ,$E,0SITI0;; +A;C= ;0&;& That means we !roduce an 0$;ATE ;0&; 4say -the purple cow-5, then a ,$E,0SITI0; 4say -without-5, and once a ain, we hit the recursion& So we han onto our hats, and descend one more level& To avoid com!le(ity, let1s assume that this the !athway we take is the direct one .ust 0$;ATE ;0&;. 'or e(am!le, we mi ht et - horns-& 3e hit the node E; in this call on +A;C= ;0&; which amounts to !o!!in out, and so we o to our stack to find the return address& It tells us that we were in the middle of e(ecutin +A;C= ;0&; one level u!and so we resume there& This yields - the purple cow without horns-& En this level, too, we hit E8:, and so we !o! u! once more, this findin ourselves in need of a 4E$Bso let1s choose -go((led-& This ends hi hest-level call on +A;C= ;0&;, with the result that the !hrase "the strange (agels that the purple cow without horns go((led" will et !assed u!wards to the !atient SE;TE;CE, as we !o! for the last time& )s you see, we didn1t et into any infinite re ress& The reason is that at least one !athway inside the ;T8 +A;C= ;0&; does not involve recursive calls on +A;C= ;0&; itself& Ef course, we could have !erversely insisted on always choosin the bottom !athway inside +A;C= ;0&; then we would never have otten finished, .ust as the acronym -6E:- never ot fully e(!anded& 0ut if the !athways are chosen at random, an infinite re ress of that sort will not ha!!en&

(Bottoming 0ut( and /eterarchies

This is the crucial fact which distin uishes recursive definitions from circular ones& There is always some !art of the definition which avoids reference, so that the action of constructin an ob.ect which satisfies the definition will eventually -bottom out-& 8ow there are more oblique ways of achievin recursivity in ;T81s than by selfcallin & There is the analo ue of Escher's .rawing 4'i & "%L5, where each of two !rocedures calls the other, but not itself& 'or e(am!le, we could have an ;T8 named CLA&SE, which calls +A;C= ;0&; whenever it needs an ob.ect for a transitive verb, and conversely, the u!!er !ath of +A;C= ;0&; could call $ELATI4E ,$0;0&; and then CLA&SE whenever it wants a relative clause& This is an e(am!le of indirect recursion& It is reminiscent also of the two-ste! version of the E!imenides !arado(& 8eedless to say, there can be a trio of !rocedures which call one another, cyclicallyand so on& There can be a whole family of ;T81s which are all tan led u!, callin each other and themselves like cra<y& ) !ro ram which has such a structure in which there is no sin le -hi hest level-, or -monitor-, is called a heterarchy 4as distin uished from a hierarchy5& The term is due, I believe, to 3arren /c7ulloch, one of the first cyberneticists, and a reverent student of brains and minds&

E-!anding ;odes
Ene ra!hic way of thinkin about ;T81s is this& 3henever you are movin alon some !athway and you hit a node which calls on an ;T8, you -e(!and- that node, which means to re!lace it by a very small co!y of the ;T8 it calls 4see 'i & B*5& Then you !roceed into the very small ;T82

73G5=E FE! The +A;C= ;0&; =TN with one node recursi#el) e9panded!

3hen you !o! out of it, you are automatically in the ri ht !lace in the bi one& 3hile in the small one, you may wind u! constructin even more miniature ;T81s& 0ut by e(!andin nodes only when you come across them, you avoid the need to make an infinite dia ram, even when an ;T8 calls itself& E(!andin a node is a little like re!lacin a letter in an acronym by the word it stands for& The -6E:- acronym is recursive but has the defector advanta ethat you must re!eatedly e(!and the 161= thus it never bottoms out& 3hen an ;T8 is im!lemented as a real com!uter !ro ram, however, it always has at least one !athway which avoids recursivity 4direct or indirect5 so that infinite re ress is not created& Even the most heterarchical !ro ram structure bottoms outotherwise it couldn1t run2 It would .ust be constantly e(!andin node after node, but never !erformin any action&

iagram G and $ecursi"e Sequences

Infinite eometrical structures can be defined in .ust this waythat is by e(!andin node after node& 'or e(am!le, let us define an infinite dia ram called -:ia ram 6-& To do so, we shall use an im!licit re!resentation& In two nodes, we shall write merely the letter 161, which, however, will stand for an entire co!y of :ia ram 6& In 'i ure B#a, :ia ram 6 is !ortrayed im!licitly& 8ow if we wish to see :ia ram 6 more e(!licitly, we e(!and each of the two 61sthat is, we replace them () the same diagram, only reduced in scale 4see 'i & B#b5& This -second-order- version of :ia ram ives us an inklin of what the final, im!ossible-to-reali<e :ia ram 6 really looks like& In 'i ure %+ is shown a lar er !ortion of :ia ram 6, where all the nodes have been numbered from the bottom u!, and from left to ri ht& Two e(tra nodes-numbers" and Bhave been inserted at the bottom This infinite tree has some very curious mathematical !ro!erties ;unnin u! its ri ht-hand ed e is the famous sequence of 7i(onacci num(ers& ", ", B, %, L, *, "%, B", %C, LL, *#, "CC, B%%,&&& discovered around the year "B+B by Aeonardo of Pisa, son of 0onaccio, er o -'ilius 0onacci-, or -'ibonacci- for short& These numbers are best
73G5=E FI! a/ .iagram G, une9panded! (/ .iagram G, e9panded once! c/ .iagram ,, une9panded d/ .iagram ,, e9panded once

73G5=E HK! .iagram G, further e9panded and with num(ered nodes!

defined recursively by the !air of formulas 'I0E4n5 X 'I0E4n-"5 ^ 'I0E4n-B5 'I0E4l5 X 'I0E4B5 X " 8otice how new 'ibonacci numbers are defined in terms of !revious 'ibonacci numbers& 3e could re!resent this !air of formulas in an ;T8 4see 'i & %"5& for n a B

73G5=E HB! *n =TN for 7i(onacci num(ers!

Thus you can calculate 'I0E4"L5 by a sequence of recursive calls on the !rocedure defined by the ;T8 above& This recursive definition bottoms out when you hit 'I0E4"5 or 'I0E4B5 4which are iven e(!licitly5 after you have worked your way backwards throu h descendin values of n& It is sli htly awkward to work your way backwards, when you could .ust as well work your way forwards, startin with 'I0E4l5 and 'I0E4B5 and always addin the most recent two values, until you reach 'I0E4"L5& That way you don1t need to kee! track of a stack& 8ow :ia ram 6 has some even more sur!risin !ro!erties than this& Its entire structure can be coded u! in a sin le recursive definition, as follows,

64n5 X n - 6464n- "55 for n a + 64+5 X + ?ow does this function 64n5 code for the tree-structure? Muite sim!ly you construct a tree by !lacin 64n5 below n, for all values of n, you recreate :ia ram 6& In fact, that is how I discovered :ia ram 6 in the !lace& I was investi atin the function 6, and in tryin to calculate its values quickly, I conceived of dis!layin the values I already knew in a tree& To my sur!rise, the tree turned out to have this e(tremely orderly recursive eometrical descri!tion& 3hat is more wonderful is that if you make the analo ous tree function ?4n5 defined with one more nestin than 6 ?4n5 X n - ?4?4?4n - "555 ?4+5 X + then the associated -:ia ram ?- is defined im!licitly as shown in 'i ure B#c& The ri ht-hand trunk contains one more node= that is the difference& The first recursive e(!ansion of :ia ram ? is shown in 'i ure B#d& )nd so it oes, for any de ree of nestin & There is a beautiful re ularity to the recursive eometrical structures, which corres!onds !recisely to the recursive al ebraic definitions& ) !roblem for curious readers is, su!!ose you fli! :ia ram 6 around as if in a mirror, and label the nodes of the new tree so they increase left to ri ht& 7an you find a recursive alge(raic definition for this -fli!-tree-? 3hat about for the -fli!- of the ?-tree? Etc&? )nother !leasin !roblem involves a !air of recursively intertwined functions '4n5 and /4n5-married- functions, you mi ht saydefined this way, '4n5 X n - /4'4n-"55 /4n5 X n - '4/4n-"55 for n a +

'or n a +

'4+5 X ", and /4+5 X + The ;T81s for these two functions call each other and themselves as well& The !roblem is sim!ly to discover the recursive structures of :ia ram '= and :ia ram /& They are quite ele ant and sim!le&

A Chaotic Sequence
Ene last e(am!le of recursion in number theory leads to a small mystery& 7onsider the followin recursive definition of a function,

M4n5 X M4n-M4n-"55 ^ M4n-M4n-B55 for n a B M4"5 X M4B5 X "& It is reminiscent of the 'ibonacci definition in that each new value is a sum of two !revious values-but not of the immediatel) !revious two values& Instead, the two immediately !revious values tell how far to count back to obtain the numbers to be added to make the new value2 The first "$ M-numbers run as follows, ", ", B, %, %, C, L, L, D, D, D, *, *, *, "+, #, "+,&&& c c L ^ D X "" how far to mo#e to the left

To obtain the ne(t one, move leftwards 4from the three dots5 res!ectively "+ and # terms= you will hit a L and a D, shown by the arrows& Their sum""yields the new value, M4"*5& This is the stran e !rocess by which the list of known M-numbers is used to e(tend itself& The resultin sequence is, to !ut it mildly, erratic& The further out you o, the less sense it seems to make& This is one of those very !eculiar cases where what seems to be a somewhat natural definition leads to e(tremely !u<<lin behavior, chaos !roduced in a very orderly manner& Ene is naturally led to wonder whether the a!!arent chaos conceals some subtle re ularity& Ef course, by definition, there is re ularity, but what is of interest is whether there is another way of characteri<in this sequence-and with luck, a nonrecursive way&

T'o Stri.ing $ecursi"e Gra!hs

The marvels of recursion in mathematics are innumerable, and it is not my !ur!ose to !resent them all& ?owever, there are a cou!le of !articularly strikin e(am!les from my own e(!erience which I feel are worth !resentin & They are both ra!hs& Ene came u! in the course of some number-theoretical investi ations& The other came u! in the course of my Ph&:& thesis work, in solid state !hysics& 3hat is truly fascinatin is that the ra!hs are closely related& The first one 4'i & %B5 is a ra!h of a function which I call I8T4(5& It is !lotted here for ( between + and "& 'or ( between any other !air of inte ers n and n ^ ", you .ust find I8T4(-n5, then add n back& The structure of the !lot is quite .um!y, as you can see& It consists of an infinite number of curved !ieces, which et smaller and smaller towards the cornersand incidentally, less and less curved& 8ow if you look closely at each such !iece, you will find that it is actually a co!y of the full ra!h, merely curved2 The im!lications are wild& Ene of them is that the ra!h of I8T consists of nothin but co!ies of itself, nested down infinitely dee!ly& If you !ick u! any !iece of the ra!h, no matter how small, you are holdin a com!lete co!y of the whole ra!hin fact, infinitely many co!ies of it2

new term

The fact that I8T consists of nothin but co!ies of itself mi ht make you think it is too e!hemeral to e(ist& Its definition sounds too circular& ?ow does it ever et off the round? That is a very interestin matter& The main thin to notice is that, to describe I8T

73G5=E HF! Graph of the function 3NT 9/! There is a Mump discontinuit) at e#er) rational #alue of 9!

to someone who hasn1t see it will not suffice merely to say, -It consists of co!ies of itself&- The other half of the storythe nonrecursive halftells where those co!ies lie in the square, and how they have been deformed, relative to the full ra!h& Enly the combination of these two as!ects of I8T will s!ecify structure of I8T& It is e(actly as in the definition of 'ibonacci number where you need two linesone to define the recursion, the other to define the (ottom 4i&e&, the values at the be innin 5& To be very concrete, if make one of the bottom values % instead of ", you will !roduce a com!letely different sequence, known as the Aucas sequence, ", %, C, $, "", "*, B#, C$, $D, "B%,&&& B# ^ C$ X $D the "(ottom" same recursi#e rule as for the 7i(onacci num(ers

3hat corres!onds to the (ottom in the definition of I8T is a !icture 4'i & %%a5 com!osed of many bo(es, showin where the co!ies o, and how they are distorted& I call it the -skeleton- of I8T& To construct I8T from its skeleton, you do the followin & 'irst, for each bo( of the skeleton, you do two o!erations, 4"5 !ut a small curved co!y of the skeleton inside the bo(, usin the curved line inside it as a uide= 4B5 erase the containin bo( and its curved line& Ence this has been done for each bo( of the ori inal skeleton, you are left with many -baby- skeletons in !lace of one bi one& 8e(t you re!eat the !rocess one level down, with all the baby skeletons& Then a ain, a ain, and a ain&&& 3hat you a!!roach in the limit is an e(act ra!h of I8T, thou h you never et there& 0y nestin the skeleton inside itself over and over a ain, you radually construct the ra!h of I8T -from out of nothin -& 0ut in fact the -nothin - was not nothin it was a !icture& To see this even more dramatically, ima ine kee!in the recursive !art of the definition of I8T, but chan in the initial !icture, the skeleton& ) variant skeleton is shown in 'i ure %%b, a ain with bo(es which et smaller and smaller as they trail off to the four corners& If you nest this second skeleton inside itself over and over a ain, you will create the key ra!h from my Ph&:& thesis, which I call Gplot 4'i & %C5& 4In fact, some com!licated distortion of each co!y is needed as well-but nestin is the basic idea&5& 6!lot is thus a member of the I8T-family& It is a distant relative, because its skeleton is quite different fromand considerably more com!le( thanthat of I8T& ?owever, the recursive !art of the definition is identical, and therein lies the family tie& I should not kee! you too much in the dark about the ori in of these beautiful ra!hs& I8T-standin for -interchan e-comes from a !roblem involvin -Etasequences-, which are related to continued fractions& The basic idea behind I8T is that !lus and minus si ns are interchan ed in a certain kind of continued fraction& )s a consequence, I8T4I8T4955 X 9& I8T has the !ro!erty that if 9 is rational, so is I8T495= if 9 is quadratic, so is I8T495& I do not know if this trend holds for hi her al ebraic de rees& )nother lovely feature of I8T is that at all rational values of 9, it has a .um! discontinuity, but at all irrational values of (, it is continuous& 6!lot comes from a hi hly ideali<ed version of the question, -3hat are the allowed ener ies of electrons in a crystal in a ma netic field?- This !roblem is interestin because it is a cross between two very sim!le and fundamental !hysical situations, an electron in a !erfect crystal, and an electron in a homo eneous ma netic field& These two sim!ler !roblems are both well understood, and their characteristic solutions seem almost incom!atible with each other& Therefore, it is of quite some interest to see how nature mana es to reconcile the two& )s it ha!!ens, the crystal without-ma netic-field situation and the ma netic-field-without-crystal situation do have one feature in common, in each of them, the electron behaves !eriodically in time& It turns out that when the two situations are combined, the ratio of their two time !eriods is the key !arameter& In fact, that ratio holds all the information about the distribution of allowed electron ener ies but it only ives u! its secret u!on bein e(!anded into a continued fraction&

73G5=E HH a/ The skeleton from which 3NT can (e constructed () recursi#e su(stitutions! (/ The skeleton from which Gplot can (e constructed () recursi#e su(stitutions!

6!lot shows that distribution& The hori<ontal a(is re!resents ener y, and the vertical a(is re!resents the above-mentioned ratio of time !eriods, which we can call - -& )t the bottom, is <ero, and at the to! is unity& 3hen is <ero, there is no ma netic field& Each of the line se ments makin u! 6!lot is an -ener y band-that is, it re!resents allowed values of ener y& The em!ty swaths traversin 6!lot on all different si<e scales are therefore re ions of forbidden ener y& Ene of the most startlin !ro!erties of 6!lot is that when is rational 4say pV" in lowest terms5, there are e(actly " such bands 4thou h when " is even, two of them -kiss- in the middle5& )nd when is irrational, the bands shrink to !oints, of which there are infinitely many, very s!arsely distributed in a so-called -7antor set-another recursively defined entity which s!rin s u! in to!olo y& Hou mi ht well wonder whether such an intricate structure would ever show u! in an e(!eriment& 'rankly, I would be the most sur!rised !erson in the world if 6!lot came out of any e(!eriment& The !hysicality of 6!lot lies in the fact that it !oints the way to the !ro!er mathematical treatment of less ideali<ed !roblems of this sort& In other words, 6!lot is !urely a contribution to theoretical !hysics, not a hint to e(!erimentalists as to what to e(!ect to see2 )n a nostic friend of mine once was so struck by 6!lot1s infinitely many infinities that he called it -a !icture of 6od-, which I don1t think is blas!hemous at all&

$ecursion at the Lo'est Le"el of Matter

3e have seen recursion in the rammars of lan ua es, we have seen recursive eometrical trees which row u!wards forever, and we have seen one way in which recursion enters the theory of solid state !hysics& 8ow we are oin to see yet another way in which the whole world is built out of recursion& This has to do with the structure of elementary !articles, electrons, !rotons, neutrons, and the tiny quanta of electroma netic radiation called -!hotons-& 3e are oin to see that !articles arein a certain sense which can only be defined ri orously in relativistic quantum mechanics nested inside each other in a way which can be described recursively, !erha!s even by some sort of - rammar-& 3e be in with the observation that if !articles didn1t interact with each other, thin s would be incredibly sim!le& Physicists would like such a world because then they could calculate the behavior of all !articles easily 4if !hysicists in such a world e(isted, which is a doubtful !ro!osition5& Particles without interactions are called (are particles, and they are !urely hy!othetical creations= they don1t e(ist& 8ow when you -turn on- the interactions, then !articles et tan led u! to ether in the way that functions ' and / are tan led to ether, or married !eo!le are tan led to ether& These real !articles are said to be renormali@edOan u ly but intri uin term& 3hat ha!!ens is that no !article can even be defined without referrin to all other !articles, whose definitions in turn de!end on the first !articles, etc& ;ound and round, in a never-endin loo!&

73G5=E HD! GplotR a recursi#e graph, showing energ) (ands for electrons in an ideali@ed cr)stal in a magnetic field, representing magnetic field strength, runs #erticall) from K to B! Energ) runs hori@ontall)! The hori@ontal line segments are (ands of allowed electron energies!

Aet us be a little more concrete, now& Aet1s limit ourselves to only two kinds of !articles, electrons and photons& 3e1ll also have to throw in the electron1s anti!article, the positron& 4Photons are their own anti!articles&5 Ima ine first a dull world where a bare electron wishes to !ro!a ate from !oint ) to !oint 0, as 9eno did in my Three%&art 3n#ention! ) !hysicist would draw a !icture like this,

There is a mathematical e(!ression which corres!onds to this line and its end!oints, and it is easy to write down& 3ith it, a !hysicist can understand the behavior of the bare electron in this tra.ectory& 8ow let us -turn on- the electroma netic interaction, whereby electrons and !hotons interact& )lthou h there are no !hotons in the scene, there will nevertheless be !rofound consequences even for this sim!le tra.ectory& In !articular, our electron now becomes ca!able of emittin and then reabsorbin #irtual photonsO!hotons which flicker in and out of e(istence before they can be seen& Aet us show one such !rocess,

8ow as our electron !ro!a ates, it may emit and reabsorb one !hoton after another, or it may even nest them, as shown below,

The mathematical e(!ressions corres!ondin to these dia ramscalled -'eynman dia rams-are easy to write down, but they are harder to calculate than that for the bare electron& 0ut what really com!licates matters is that a !hoton 4real or virtual5 can decay for a brief moment into an electron-!ositron !air& Then these two annihilate each other, and, as if by ma ic, the ori inal !hoton rea!!ears& This sort of !rocess is shown below,

The electron has a ri ht-!ointin arrow, while the !ositron1s arrow !oints leftwards&

)s you mi ht have antici!ated, these virtual !rocesses can be inside each other to arbitrary de!th& This can ive rise to some com!licated-lookin drawin s, such as the one in 'i ure %L& In that 'eynman dia ram, a sin le electron enters on the left at ), does some an acrobatics, and then a sin le electron emer es on the ri ht at 0& To an outsider who can1t see the inner mess, it looks as if one electron !eacefully sailed from ) to 0& In the dia ram, you can see how electron lines can et arbitrarily embellished, and so can the !hoton lines& This dia ram would be ferociously hard to calculate&

73G5=E HG! * 7e)nman diagram showing the propagation of a renormali@ed electron from * to B! 3n this diagram, time increases to the right! Therefore, in the segments where the electron's arrow points leftwards, it is mo#ing "(ackwards in time"! * more intuiti#e wa) to sa) this is that an antielectron positron/ is mo#ing forwards in time! &hotons are their own antiparticlesR hence their lines ha#e no need of arrows!

There is a sort of - rammar- to these dia rams, that only certain !ictures to be reali<ed in nature& 'or instance, the one below is im!ossible,

Hou mi ht say it is not a -well-formed- 'eynman dia ram& The rammar is a result of basic laws of !hysics, such as conservation of ener y, conservation of electric char e, and so on& )nd, like the rammars of human lan ua es, this rammar has a recursive structure, in that it allows dee! nestin s of structures inside each other& It would be !ossible to draw a set of recursive transition networks definin the - rammar- of the electroma netic interaction& 3hen bare electrons and bare !hotons are allowed to interact in these arbitrarily tan led ways, the result is renormali@ed electrons and !hotons& Thus, to understand how a real, !hysical electron !ro!a ates from ) to 0, the !hysicist has to be able to take a sort of avera e of all the infinitely many different !ossible drawin s which involve virtual !articles& This is 9eno with a ven eance2

Thus the !oint is that a !hysical !articlea renormali<ed !article involves 4"5 a bare !article and 4B5 a hu e tan le of virtual !articles, ine(tricably wound to ether in a recursive mess& Every real !article1s e(istence therefore involves the e(istence of infinitely many other !articles, contained in a virtual -cloud- which surrounds it as it !ro!a ates& )nd each of the virtual !articles in the cloud, of course, also dra s alon its own virtual cloud, and so on ad infinitum& Particle !hysicists have found that this com!le(ity is too much to handle, and in order to understand the behavior of electrons and !hotons, they use a!!ro(imations which ne lect all but fairly sim!le 'eynman dia rams& 'ortunately, the more com!le( a dia ram, the less im!ortant its contribution& There is no known way of summin u! all of the infinitely many !ossible dia rams, to et an e(!ression for the behavior of a fully renormali<ed, !hysical electron& 0ut by considerin rou hly the sim!lest hundred dia rams for certain !rocesses, !hysicists have been able to !redict one value 4the socalled -factor of the muon5 to nine decimal !lacescorrectly2 ;enormali<ation takes !lace not only amon electrons and !hotons& 3henever any ty!es of !article interact to ether, !hysicists use the ideas of renormali<ation to understand the !henomena& Thus !rotons and neutrons, neutrinos, !i-mesons, quarksall the beasts in the subnuclear <oothey all have bare and renormali<ed versions in !hysical theories& )nd from billions of these bubbles within bubbles are all the beasts and baubles of the world com!osed&

Co!ies and Sameness

Aet us now consider 6!lot once a ain& Hou will remember that in the Introduction, we s!oke of different varieties of canons& Each ty!e of canon e(!loited some manner of takin an ori inal theme and co!yin it by an isomor!hism, or information-!reservin transformation& Sometimes the co!ies were u!side down, sometimes backwards, sometimes shrunken or e(!anded&&& In 6!lot we have all those ty!es of transformation, and more& The ma!!in s between the full 6!lot and the -co!iesof itself inside itself involve si<e chan es, skewin s, reflections, and more& )nd yet there remains a sort of skeletal identity, which the eye can !ick u! with a bit of effort, !articularly after it has !racticed with I8T& Escher took the idea of an ob.ect1s !arts bein co!ies of the ob.ect itself and made it into a !rint, his woodcut 7ishes and -cales 4'i & %D5& Ef course these fishes and scales are the same only when seen on a sufficiently abstract !lane& 8ow everyone knows that a fish1s scales aren1t really small co!ies of the fish= and a fish1s cells aren1t small co!ies of the fish= however, a fish1s :8), sittin inside each and every one of the fish1s cells, is a very convoluted -co!y- of the entire fishand so there is more than a rain of truth to the Escher !icture&

73G5=E HJ! 'ish and Scales, () '! 8! Escher woodcut, BIGI/!

3hat is there that is the -same- about all butterflies? The ma!!in from one butterfly to another does not ma! cell onto cell= rather, it ma!s functional !art onto functional !art, and this may be !artially on a macrosco!ic scale, !artially on a microsco!ic scale& The e(act !ro!ortions of !arts are not !reserved= .ust the functional relationshi!s between !arts& This is the ty!e of isomor!hism which links all butterflies in Escher1s wood en ravin Butterflies 4'i & %$5 to each other& The same oes for the more abstract butterflies of 6!lot, which are all linked to each other by mathematical ma!!in s that carry functional !art onto functional !art, but totally i nore e(act line !ro!ortions, an les, and so on& Takin this e(!loration of sameness to a yet hi her !lane of abstraction, we mi ht well ask, -3hat is there that is the 1same1 about all Escher drawin s?- It would be quite ludicrous to attem!t to ma! them !iece by !iece onto each other& The ama<in thin is that even a tiny section of an Escher drawin or a 0ach !iece ives it away& Just as a fish1s :8) is contained inside every tiny bit of the fish, so a creator1s -si nature- is contained inside every tiny section of his creations& 3e don1t know what to call it but -style-a va ue and elusive word&

73G5=E HC! 0utterflies, () '! 8! Escher wood%engra#ing, BIGK/!

3e kee! on runnin u! a ainst -sameness-in-differentness-, and the question 3hen are two thin s the same? It will recur over and over a ain in this book& 3e shall come at it from all sorts of skew an les, and in the end, we shall see how dee!ly this sim!le question is connected with the nature of intelli ence& That this issue arose in the 7ha!ter on recursion is no accident, for recursion is a domain where -sameness-in-differentness- !lays a central role& ;ecursion is based on the -same- thin ha!!enin on several different levels at once& 0ut the events on different levels aren1t e(actly the samerather, we find some invariant feature in them, des!ite many ways in which they differ& 'or e(am!le, in the +ittle ,armonic +a()rinth, all stories on different levels are quite unrelatedtheir -sameness- reside only two facts, 4"5 they are stories, and 4B5 they involve the Tortoise and )chilles& Ether than that, they are radically different from each other&

,rogramming and $ecursion5 Modularity) Loo!s) ,rocedures

Ene of the essential skills in com!uter !ro rammin is to !erceive when two !rocesses are the same in this e(tended sense, for that leads to modulari<ationthe breakin -u! of a task into natural subtasks& 'or instance, one mi ht want a sequence of many similar o!erations to be carried out one after another& Instead of writin them all out, one can write a loo!, which tells the com!uter to !erform a fi(ed set of o!erations and then loo! back and !erform them a ain, over and over, until some condition is satisfied& 8ow the (od) of the loo!the fi(ed set of instructions to re!eatedneed not actually be com!letely fi(ed& It may vary in some !redictable way& )n e(am!le is the most sim!le-minded test for the !rimality of natural number N, in which you be in by tryin to divide N by B, then %, C, L, etc& until N - "& If N has survived all these tests without be divisible, it1s !rime& 8otice that each ste! in the loo! is similar to, but not the same as, each other ste!& 8otice also that the number of ste!s varies with Nhence a loo! of fi(ed len th could never work as a eneral test !rimality& There are two criteria for -abortin - the loo!, 4"5 if so number divides N e(actly, quit with answer -8E-= 4B5 if N - " is reached as a test divisor and N survives, quit with answer -HES-& The eneral idea of loo!s, then, is this, !erform some series of related ste!s over and over, and abort the !rocess when s!ecific conditions are met& 8ow sometimes, the ma(imum number of ste!s in a loo! will be known in advance= other times, you .ust be in, and wait until it is aborted& The second ty!e of loo!which I call a free loo!is dan erous, because criterion for abortion may never occur, leavin the com!uter in a socalled -infinite loo!-& This distinction between (ounded loops and free loops is one the most im!ortant conce!ts in all of com!uter science, and we shall devote an entire 7ha!ter to it, -0looP and 'looP and 6looP-& 8ow loo!s may be nested inside each other& 'or instance, su!!ose that we wish to test all the numbers between " and L+++ for !rimality& 3e can write a second loo! which uses the above-described test over and over startin with N X " and finishin with N X L+++& So our !ro ram will have a -loo!-the-loo!- structure& Such !ro ram structures are ty!icalin fact they are deemed to be ood !ro rammin style& This kind of nested loo! also occurs in assembly instructions for common!lace items, and in such activities as knittin or crochetin -in which very small loo!s are re!eated several times in lar er loo!s, which in turn are carried out re!eatedly&&& 3hile the result of a low-level loo! mi ht be no more than cou!le of stitches, the result of a hi h-level loo! mi ht be a substantial !ortion of a !iece of clothin & In music, too, nested loo!s often occuras, for instance, when a scale 4a small loo!5 is !layed several times in a row, !erha!s dis!laced in !itch each new time& 'or e(am!le, the last movements of both the Prokofiev fifth !iano concerto and the ;achmaninoff second sym!hony contain e(tended !assa es in which fast, medium, and slow scale-loo!s are !layed simultaneously by different rou!s of instruments, to reat effect& The Prokofiev-scales o u!= the ;achmaninoff-scales, down& Take your !ick& ) more eneral notion than loo! is that of subroutine, or !rocedure, which we have already discussed somewhat& The basic idea here is that a rou! of o!erations are

lum!ed to ether and considered a sin le unit with a namesuch as the !rocedure 0$;ATE ;0&;& )s we saw in ;T81s, !rocedures can call each other by name, and thereby e(!ress very concisely sequences of o!erations which are to be carried out& This is the essence of modularity in !ro rammin & /odularity e(ists, of course, in hi-fi systems, furniture, livin cells, human societywherever there is hierarchical or ani<ation& /ore often than not, one wants a !rocedure which will act variably, accordin to conte(t& Such a !rocedure can either be iven a way of !eerin out at what is stored in memory and selectin its actions accordin ly, or it can be e(!licitly fed a list of !arameters which uide its choice of what actions to take& Sometimes both of these methods are used& In ;T8 terminolo y, choosin the sequence of actions to carry out amounts to choosin which pathwa) to follow& )n ;T8 which has been sou!ed u! with !arameters and conditions that control the choice of !athways inside it is called an *ugmented Transition Network 4)T85& ) !lace where you mi ht !refer )T81s to ;T81s is in !roducin sensibleas distin uished from nonsensicalEn lish sentences out of raw words, accordin to a rammar re!resented in a set of )T81s& The !arameters and conditions would allow you to insert various semantic constraints, so that random .u(ta!ositions like -a thankless brunch- would be !rohibited& /ore on this in 7ha!ter TGIII, however&

$ecursion in Chess ,rograms

) classic e(am!le of a recursive !rocedure with !arameters is one for choosin the -bestmove in chess& The best move would seem to be the one which leaves your o!!onent in the tou hest situation& Therefore, a test for oodness of a move is sim!ly this, !retend you1ve made the move, and now evaluate the board from the !oint of view of your o!!onent& 0ut how does your o!!onent evaluate the !osition? 3ell, he looks for his best move& That is, he mentally runs throu h all !ossible moves and evaluates them from what he thinks is your !oint of view, ho!in they will look bad to you& 0ut notice that we have now defined -best move- recursively, sim!ly usin the ma(im that what is best for one side is worst for the other& The !rocedure which looks for the best move o!erates by tryin a move and then calling on itself in the role of opponent2 )s such, it tries another move, and calls on itself in the role of its o!!onent1s o!!onentthat is, itself& This recursion can o several levels dee!but it1s ot to bottom out somewhere2 ?ow do you evaluate a board !osition without lookin ahead? There are a number of useful criteria for this !ur!ose, such as sim!ly the number of !ieces on each side, the number and ty!e of !ieces under attack, the control of the center, and so on& 0y usin this kind of evaluation at the bottom, the recursive move- enerator can !o! back u!wards and ive an evaluation at the to! level of each different move& Ene of the !arameters in the self-callin , then, must tell how many moves to look ahead& The outermost call on the !rocedure will use some e(ternally set value for this !arameter& Thereafter, each time the !rocedure recursively calls itself, it must decrease this look-ahead !arameter by "& That

way, when the !arameter reaches <ero, the !rocedure will follow the alternate !athway the non-recursive evaluation& In this kind of ame-!layin !ro ram, each move investi ated causes the eneration of a so-called -look-ahead tree-, with the move itself as the trunk, res!onses as main branches, counter-res!onses as subsidiary branches, and so on& In 'i ure %* I have shown a sim!le look-ahead tree, de!ictin the start of a tic-tac-toe ame& There is an art to fi urin out how to avoid e(!lorin every branch of a look-ahead tree out to its ti!& In chess trees, !eo!lenot com!utersseem to e(cel at this art= it is known that to!-level !layers look ahead relatively little, com!ared to most chess !ro ramsyet the !eo!le are far better2 In the early days of com!ute !eo!le used to estimate that it would be ten years until a com!uter 4or !ro ram5 was world cham!ion& 0ut after ten years had !assed, it seemed that the day a com!uter would become world cham!ion was still more than ten years away&&& This is .ust one more !iece of evidence for the rather recursive ,ofstadter's +aw, It always takes lon er than you e(!ect, even when you take into account ?ofstadter1s Aaw&
73G5=E HE! The (ranching tree of mo#es and countermo#es at the start of a game of tic%tac%toe!

$ecursion and &n!redictability

8ow what is the connection between the recursive !rocesses of this 7ha!ter, and the recursive sets of the !recedin 7ha!ter? The answer involves the notion of a recursi#el) enumera(le set& 'or a set to be r&e& means that it can be enerated from a set of startin !oints 4a(ioms5, by the re!eated a!!lication of rules of inference& Thus, the set rows and rows, each new element bein com!ounded somehow out of !revious elements, in a sort of -mathematical snowball-& 0ut this is the essence of recursionsomethin bein defined in terms of sim!ler versions of itself, instead of e(!licitly& The 'ibonacci numbers and the Aucas numbers are !erfect e(am!les of r&e& setssnowballin from two elements by a recursive rule into infinite sets& It is .ust a matter of convention to call an r&e& set whose com!lement is also r&e& -recursive-& ;ecursive enumeration is a !rocess in which new thin s emer e from old thin s by fi(ed rules& There seem to be many sur!rises in such !rocessesfor e(am!le the un!redictability of the M-sequence& It mi ht seem that recursively defined sequences of that ty!e !ossess some sort of inherently increasin com!le(ity of behavior, so that the further out you o, the less !redictable they et& This kind of thou ht carried a little further su ests that suitably com!licated recursive systems mi ht be stron enou h to break out of any !redetermined !atterns& )nd isn1t this one of the definin !ro!erties of intelli ence? Instead of .ust considerin !ro rams com!osed of !rocedures which can recursively call themselves, why not et really so!histicated, and invent !ro rams which can modif) themselves!ro rams which can act on !ro rams, e(tendin them, im!rovin them, enerali<in them, fi(in them, and so on? This kind of -tan led recursion- !robably lies at the heart of intelli ence&

Canon by Intervallic Au!mentation

*chilles and the Tortoise ha#e Must finished a delicious 8hinese (an"uet for two, at the (est 8hinese restaurant in town! *chilles, Hou wield a mean cho!stick, /r& T& Tortoise, I ou ht to& Ever since my youth, I have had a fondness for Eriental cuisine& )nd youdid you en.oy your meal, )chilles? *chilles, Immensely& I1d not eaten 7hinese food before& This meal was a s!lendid introduction& )nd now, are you in a hurry to o, or shall we .ust sit here and talk a little while? Tortoise, I1d love to talk while we drink our tea& 3aiter2 * waiter comes up!/ 7ould we have our bill, !lease, and some more tea? The waiter rushes off!/ *chilles, Hou may know more about 7hinese cuisine than I do, /r& T, but I1ll bet I know more about Ja!anese !oetry than you do& ?ave you ever read any haiku? Tortoise, I1m afraid not& 3hat is a haiku? *chilles, ) haiku is a Ja!anese seventeen-syllable !oemor mini!oem rather, which is evocative in the same way, !erha!s, as a fra rant !etal is, or a lily !ond in a li ht dri<<le& It enerally consists of rou!s of, of five, then seven, then five syllables& Tortoise, Such com!ressed !oems with seventeen syllables can1t have much meanin &&& *chilles, /eanin lies as much in the mind of the reader as in the haiku& Tortoise, ?mm&&& That1s an evocative statement& The waiter arri#es with their (ill, another pot of tea, and two fortune cookies!/ Thank you, waiter& 7are for more tea, )chilles? *chilles, Please& Those little cookies look delicious& 4 &icks one up, (ites into it and (egins to chew!5 ?ey2 3hat1s this funny thin inside? ) !iece of !a!er? Tortoise, That1s your fortune, )chilles& /any 7hinese restaurants ive out fortune cookies with their bills, as a way of softenin the blow& If you frequent 7hinese restaurants, you come to think of fortune cookies less as cookies than as messa e bearers Fnfortunately, you seem to have swallowed some of your fortune& 3hat does the rest say?

*chilles, It1s a little stran e, for all the letters are run to ether, with no s!aces in between& Perha!s it needs decodin in some way? Eh, now I see& If you !ut the s!aces back in where they belon , it says, -E8E 3); T3E E); E3E-& I can1t quite make head or tail of that& /aybe it was a haiku-like !oem, of which I ate the ma.ority of syllables& Tortoise, In that case, your fortune is now a mere LI"$-haiku& )nd a curious ima e it evokes& If LI"$-haiku is a new art form, then I1d say woe, E, woe are we&&& /ay I look at it? *chilles 4handing the Tortoise the small slip of paper5, 7ertainly& Tortoise, 3hy, when I -decode- it, )chilles, it comes out com!letely different2 It1s not a LI"$-haiku at all& It is a si(-syllable messa e which says, -E 8E3 );T 3EE );E 3E-& That sounds like an insi htful commentary on the new art form of LI"$-haiku& *chilles, Hou1re ri ht& Isn1t it astonishin that the !oem contains its own commentary2 Tortoise, )ll I did was to shift the readin frame by one unitthat is, shift all the s!aces one unit to the ri ht& *chilles, Aet1s see what your fortune says, /r& Tortoise& Tortoise 4deftl) splitting open his cookie, reads5, -'ortune lies as much in the hand of the eater as in the cookie&*chilles, Hour fortune is also a haiku, /r& Tortoiseat least it1s ot seventeen syllables in the L-$-L form& Tortoise, 6lory be2 I would never have noticed that, )chilles& It1s the kind of thin only you would have noticed& 3hat struck me more is what it sayswhich, of course, is o!en to inter!retation& *chilles, I uess it .ust shows that each of us has his own characteristic way of inter!retin messa es which we run across&&& 3dl), *chilles ga@es at the tea lea#es on the (ottom of his empt) teacup!/ Tortoise, /ore tea, )chilles? *chilles, Hes, thank you& 0y the way, how is your friend the 7rab? I have been thinkin about him a lot since you told me of your !eculiar !hono ra!h-battle& Tortoise, I have told him about you, too, and he is quite ea er to meet you& ?e is ettin alon .ust fine& In fact, he recently made a new acquisition in the record !layer line, a rare ty!e of .ukebo(& *chilles, Eh, would you tell me about it? I find .ukebo(es, with their flashin colored li hts and silly son s, so quaint and reminiscent of by one eras& Tortoise, This .ukebo( is too lar e to fit in his house, so he had a shed s!ecially built in back for it& *chilles, I can1t ima ine why it would be so lar e, unless it has an unusually lar e selection of records& Is that it? Tortoise, )s a matter of fact, it has e(actly one record&

*chilles, 3hat? ) .ukebo( with only one record? That1s a contradiction in terms& 3hy is the .ukebo( so bi , then? Is its sin le record i antictwenty feet in diameter? Tortoise, 8o, it1s .ust a re ular .ukebo(-style record& *chilles, 8ow, /r& Tortoise, you must be .oshin me& )fter all, what kind of a .ukebo( is it that has only a sin le son ? Tortoise, 3ho said anythin about a sin le son , )chilles? *chilles, Every .ukebo( I1ve ever run into obeyed the fundamental .ukebo(-a(iom, -Ene record, one son -& Tortoise, This .ukebo( is different, )chilles& The one record sits vertically, sus!ended, and behind it there is a small but elaborate network of overhead rails, from which han various record !layers& 3hen you !ush a !air of buttons, such as 0-", that selects one of the record !layers& This tri ers an automatic mechanism that starts the record !layer squeakily rollin alon the rusty tracks& It ets shunted alon side the recordthen it clicks into !layin !osition& *chilles, )nd then the record be ins s!innin and music comes outri ht? Tortoise, 8ot quite& The record stands stillit1s the record !layer which rotates& *chilles, I mi ht have known& 0ut how, if you have but one record to !lay can you et more than one son out of this cra<y contra!tion? Tortoise, I myself asked the 7rab that question& ?e merely su ested I try it out& So I fished a quarter from my !ocket 4you et three !lays for a quarter5, stuffed it in the slot, and hit buttons 0-", then 7-% then 0-"+all .ust at random& *chilles, So !hono ra!h 0-" came slidin down the rail, I su!!ose, !lu ed itself into the vertical record, and be an s!innin ? Tortoise, E(actly& The music that came out was quite a reeable, based the famous old tune 0-)-7-?, which I believe you remember&

*chilles, 7ould I ever for et it? Tortoise, This was record !layer 0-"& Then it finished, and was slowly rolled back into its han in !osition, so that 7-% could be slid into !osition& *chilles, 8ow don1t tell me that 7-% !layed another son ? Tortoise, It did .ust that& *chilles, )h, I understand& It !layed the fli! side of the first son , or another band on the same side& Tortoise, 8o, the record has rooves only on one side, and has only a sin le band&

*chilles, I don1t understand that at all& Hou 7)81T !ull different son s out of the same record2 Tortoise, That1s what I thou ht until I saw /r& 7rab1s .ukebo(& *chilles, ?ow did the second son o? Tortoise, That1s the interestin thin &&& It was a son based on the melody 7-)-6-E& *chilles, That1s a totally different melody2 Tortoise, True& *chilles, )nd isn1t John 7a e a com!oser of modern music? I seem to remember readin about him in one of my books on haiku& Tortoise, E(actly& ?e has com!osed many celebrated !ieces, such as D'HH", a threemovement !iece consistin of silences of different len ths& It1s wonderfully e(!ressiveif you like that sort of thin & *chilles, I can see where if I were in a loud and brash cafe I mi ht ladly !ay to hear 7a e1s D'HH" on a .ukebo(& It mi ht afford some relief2 Tortoise, ;i htwho wants to hear the racket of clinkin dishes and .an lin silverware? 0y the way, another !lace where D'HH" would come in handy is the ?all of 0i 7ats, at feedin time& *chilles, )re you su estin that 7a e belon s in the <oo? 3ell, I uess that makes some sense& 0ut about the 7rab1s .ukebo(&&& I am baffled& ?ow could both -0)7?- and -7)6E- be coded inside a sin le record at once? Tortoise, Hou may notice that there is some relation between the two, )chilles, if you ins!ect them carefully& Aet me !oint the way& 3hat do you et if you list the successive intervals in the melody 0-)-7-?? *chilles, Aet me see& 'irst it oes down one semitone, from 0 to ) 4where 0 is taken the 6erman way5= then it rises three semitones to 7= and finally it falls one semitone, to ?& That yields the !attern, -", ^%, -"& Tortoise, Precisely& 3hat about 7-)-6-E, now? *chilles, 3ell, in this case, it be ins by fallin three semitones, then rises ten semitones 4nearly an octave5, and finally falls three more semitones& That means the !attern is, -%, ^"+, -%& It1s very much like the other one, isn1t it? Tortoise, Indeed it is& They have e(actly the same -skeleton-, in a certain sense& Hou can make 7-)-6-E out of 0-)-7-? by multi!lyin all the intervals by %d, and takin the nearest whole number& *chilles, 3ell, blow me down and !ick me u!2 So does that mean that only some sort of skeletal code is !resent in the rooves, and that the various record !layers add their own inter!retations to that code?

Tortoise, I don1t know, for sure& The ca ey 7rab wouldn1t fill me in on the details& 0ut I did et to hear a third son , when record !layer 0-" swiveled into !lace& *chilles, ?ow did it o? Tortoise, The melody consisted of enormously wide intervals, and went 0-7-)-?&

The interval !attern in semitones was, -"+, ^%%, -"+&

It can be otten from the 7)6E !attern by yet another multi!lication by %d, and roundin to whole numbers& *chilles, Is there a name for this kind of interval multi!lication? Tortoise, Ene could call it -intervallic au mentation-& It is similar to the canonic device of tem!oral au mentation, where all the time values notes in a melody et multi!lied by some constant& There, the effect .ust to slow the melody down& ?ere, the effect is to e(!and the melodic ran e in a curious way& *chilles, )ma<in & So all three melodies you tried were intervallic au mentations of one sin le underlyin roove-!attern in the record, Tortoise, That1s what I concluded& *chilles, I find it curious that when you au ment 0)7? you et 7)6E and when you au ment 7)6E over a ain, you et 0)7? back, e(ce!t .umbled u! inside, as if 0)7? had an u!set stomach after !assin throu h the intermediate sta e of 7)6E& Tortoise, That sounds like an insi htful commentary on the new art form of 7a e&

C/A,TE$ 4I

The Location of Meaning

>hen Is 0ne Thing ;ot Al'ays the Same7
A)ST 7?)PTE;, 3E came u!on the question, -3hen are two thin s the same?- In this 7ha!ter, we will deal with the fli! side of that question, -3hen is one thin not always the same?- The issue we are broachin is whether meanin can be said to be inherent in a messa e, or whether meanin is always manufactured by the interaction of a mind or a mechanism with a messa eas in the !recedin :ialo ue& In the latter case, meanin could not said to be located in any sin le !lace, nor could it be said that a messa e has any universal, or ob.ective, meanin , since each observer could brin its own meanin to each messa e& 0ut in the former case, meanin would have both location and universality& In this 7ha!ter, I want to !resent the case for the universality of at least some messa es, without, to be sure, claimin it for all messa es& The idea of an -ob.ective meanin - of a messa e will turn out to be related, in an interestin way, to the sim!licity with which intelli ence can be described&

Information-Bearers and Information-$e"ealers

I1ll be in with my favorite e(am!le, the relationshi! between records, music, and record !layers& 3e feel quite comfortable with the idea that a record contains the same information as a !iece of music, because of the e(istence of record !layers, which can -read- records and convert the roove-!atterns into sounds& In other words, there is an isomor!hism between roove-!atterns and sounds, and the record !layer is a mechanism which !hysically reali<es that isomor!hism& It is natural, then, to think of the record as an information%(earer, and the record-!layer as an information%re#ealer& ) second e(am!le of these notions is iven by the !q-system& There, the -information-bearers- are the theorems, and the -information-revealer- is the inter!retation, which is so trans!arent that we don1t need any electrical machine to hel! us e(tract the information from !q-theorems& Ene ets the im!ression from these two e(am!les that isomor!hisms and decodin mechanisms 4i&e&, information-revealers5 sim!ly reveal information which is intrinsically inside the structures, waitin to be -!ulled out-& This leads to the idea that for each structure, there are certain !ieces of information which can be !ulled out of it, while there are other !ieces of information which cannot be !ulled out of it& 0ut what does this !hrase -!ull out- really mean? ?ow hard are you allowed to !ull? There are cases where by investin sufficient effort, you can !ull very recondite !ieces of information out of certain structures& In fact, the !ullin -out may involve such com!licated o!erations that it makes you feel you are !uttin in more information than you are !ullin out&

Genoty!e and ,henoty!e

Take the case of the enetic information commonly said to reside in the double heli( of deo(yribonucleic acid 4:8)5& ) molecule of :8)a genot)peOis converted into a !hysical or anisma phenot)peOby a com!le( !rocess, involvin the manufacture of !roteins, the re!lication the :8), the re!lication of cells, the radual differentiation of cell ty!es and so on& Incidentally, this unrollin of !henoty!e from enoty!e epigenesisO is the most tan led of tan led recursions, and in 7ha!ter TGI we shall devote our full attention to it& E!i enesis is uided by a set of enormously com!le( cycles of chemical reactions and feedback loo!s& 0y the time the full or anism has been constructed, there is not even the remotest similarity between its !hysical characteristics and its enoty!e& )nd yet, it is standard !ractice to attribute the !hysical structure of or anism to the structure of its :8), and to that alone& The first evidence for this !oint of view came from e(!eriments conducted by Eswald )very in "#CD, and overwhelmin corroborative evidence has since been amassed& )very1s e(!eriments showed that, of all the biolo ical molecules, only :8) transmits hereditary !ro!erties& Ene can modify other molecules in an or anism, such as !roteins, but such modifications will not be transmitted to later enerations& ?owever, when :8) is modified, all successive enerations inherit the modified :8)& Such e(!eriments show that the only way of chan in the instructions for buildin a new or anism is to chan e the :8)and this, in turn, im!lies that those instructions must be coded somehow in the structure of the :8)&

E-otic and ,rosaic Isomor!hisms

Therefore one seems forced into acce!tin the idea that the :8)1s structure contains the information of the !henoty!e1s structure, which is to say, the two are isomorphic& ?owever, the isomor!hism is an e(otic one, by which I mean that it is hi hly nontrivial to divide the !henoty!e and enoty!e into -!arts- which can be ma!!ed onto each other& Prosaic isomor!hisms, by contrast, would be ones in which the !arts of one structure are easily ma!!able onto the !arts of the other& )n e(am!le is the isomor!hism between a record and a !iece of music, where one knows that to any sound in the !iece there e(ists an e(act -ima e- in the !atterns etched into rooves, and one could !in!oint it arbitrarily accurately, if the need arose& )nother !rosaic isomor!hism is that between 6!lot and any of its internal butterflies& The isomor!hism between :8) structure and !henoty!e structure is anythin but !rosaic, and the mechanism which carries it out !hysically is awesomely com!licated& 'or instance, if you wanted to find some !iece of your :8) which accounts for the sha!e of your nose or the sha!e of your fin er!rint, you would have a very hard time& It would be a little like tryin to !in down the note in a !iece of music which is the carrier of the emotional meanin of the !iece& Ef course there is no such note, because the emotional meanin is carried on a very hi h level, by lar e -chunks- of the !iece, not by sin le notes& Incidentally, such -chunks- are not necessarily sets of conti uous notes= there may be disconnected sections which, taken to ether, carry some emotional meanin &

Similarly, - enetic meanin -that is, information about !henoty!e structureis s!read all throu h the small !arts of a molecule of :8), althou h nobody understands the lan ua e yet& 43arnin , Fnderstandin this -lan ua e- would not at all be the same as crackin the 6enetic 7ode, somethin which took !lace in the early "#D+1s& The 6enetic 7ode tells how to translate short !ortions of :8) into various amino acids& Thus, crackin the 6enetic 7ode is com!arable to fi urin out the !honetic values of the letters of a forei n al!habet, without fi urin out the rammar of the lan ua e or the meanin s of any of its words& The crackin of the 6enetic 7ode was a vital ste! on the way to e(tractin the meanin of :8) strands, but it was only the first on a lon !ath which is yet to be trodden&5

3u.ebo-es and Triggers

The enetic meanin contained in :8) is one of the best !ossible e(am!les of im!licit meanin & In order to convert enoty!e into !henoty!e, a set of mechanisms far more com!le( than the enoty!e must o!erate on the enoty!e& The various !arts of the enoty!e serve as triggers for those mechanisms& ) .ukebo( the ordinary ty!e, not the 7rab ty!e2!rovides a useful analo y here, a !air of buttons s!ecifies a very com!le( action to be taken by the mechanism, so that the !air of buttons could well be described as -tri erin - the son which is !layed& In the !rocess which converts enoty!e into !henoty!e, cellular .ukebo(esif you will !ardon the notion2acce!t -button-!ushin sfrom short e(cer!ts from a lon strand of :8), and the -son s- which they !lay are often !rime in redients in the creation of further -.ukebo(es-& It is as if the out!ut of real .ukebo(es, instead of bein love ballads, were son s whose lyrics told how to build more com!le( .ukebo(es&&& Portions of the :8) tri er the manufacture of !roteins= those !roteins tri er hundreds of new reactions= they in turn tri er the re!licatin -o!eration which, in several ste!s, co!ies the :8)and on and on&&& This ives a sense of how recursive the whole !rocess is& The final result of these many-tri ered tri erin s is the !henoty!ethe individual& )nd one says that the !henoty!e is the revelationthe -!ullin -out--of the information that was !resent in the :8) to start with, latently& 4The term -revelation- in this conte(t is due to Jacques /onod, one of the dee!est and most ori inal of twentieth-century molecular biolo ists&5 8ow no one would say that a son comin out of the louds!eaker of .ukebo( constitutes a -revelation- of information inherent in the !air of buttons which were !ressed, for the !air of buttons seem to be mere triggers, whose !ur!ose is to activate information-bearin !ortions of the .ukebo( mechanism& En the other hand, it seems !erfectly reasonable to call the e(traction of music from a record a -revelation- of information inherent in the record, for several reasons, 4"5 the music does not seem to be concealed in the mechanism of the record !layer= 4B5 it is !ossible to match !ieces of the in!ut 4the record5 with !ieces of the out!ut 4the music5 to an arbitrary de ree of accuracy=

4%5 it is !ossible to !lay other records on the same record !layer and et other sounds out= 4C5 the record and the record !layer are easily se!arated from one another& It is another question alto ether whether the fra ments of a smashed record contain intrinsic meanin & The ed es of the se!arate !ieces fit to ether and in that way allow the information to be reconstitutedbut somethin much more com!le( is oin on here& Then there is the question of the intrinsic meanin of a scrambled tele!hone call&&& There is a vast s!ectrum of de rees of inherency of meanin & It is interestin to try to !lace e!i enesis in this s!ectrum& )s develo!ment of an or anism takes !lace, can it be said that the information is bein -!ulled out- of its :8)? Is that where all of the information about the or anism1s structure resides?

;A and the ;ecessity of Chemical Conte-t

In one sense, the answer seems to be yes, thanks to e(!eriments like )very1s& 0ut in another sense, the answer seems to be no, because so much of the !ullin -out !rocess de!ends on e(traordinarily com!licated cellular chemical !rocesses, which are not coded for in the :8) itself& The :8) relies on the fact that they will ha!!en, but does not seem to contain a code which brin s them about& Thus we have two conflictin views on the nature of the information in a enoty!e& Ene view says that so much of the information is outside the .N* that it is not reasonable to look u!on the :8) as anythin more than a very intricate set of tri ers, like a sequence of buttons to be !ushed on a .ukebo(= another view says that the information is all there, but in a very im!licit form& 8ow it mi ht seem that these are .ust two ways of sayin the same thin , but that is not necessarily so& Ene view says that the :8) is quite meanin less out of conte(t= the other says that even if it were taken out conte(t, a molecule of :8) from a livin bein has such a compelling inner logic to its structure that its messa e could be deduced anyway& To !ut it as succinctly as !ossible, one view says that in order for :8) to have meanin , chemical conte9t is necessary= the other view says that only intelligence is necessary to reveal the -intrinsic meanin - of a strand of :8)&

An &nli.ely &+0
3e can et some !ers!ective on this issue by considerin a stran e hy!othetical event& ) record of :avid Eistrakh and Aev Eborin !layin 0ach1s sonata in ' /inor for violin and clavier is sent u! in a satellite& 'rom the satellite it is then launched on a course which will carry it outside of the solar system, !erha!s out of the entire ala(y .ust a thin !lastic !latter with a hole in the middle, swirlin its way throu h inter alactic s!ace& It has certainly lost its conte(t& ?ow much meanin does it carry? If an alien civili<ation were to encounter it, they would almost certainly be struck by its sha!e, and would !robably be very interested in it& Thus immediately its sha!e, actin as a tri er, has iven them some information, that it is an artifact, !erha!s an information-bearin artifact& This ideacommunicated, or tri ered, by the record itself

now creates a new conte(t in which the record will henceforth be !erceived& The ne(t ste!s in the decodin mi ht take considerably lon erbut that is very hard for us to assess& 3e can ima ine that if such a record had arrived on earth in 0ach1s time, no one would have known what to make of it, and very likely it would not have otten deci!hered& 0ut that does not diminish our conviction that the information was in !rinci!le there= we .ust know that human knowled e in those times was not very so!histicated with res!ect to the !ossibilities of stora e, transformation, and revelation of information&

Le"els of &nderstanding of a Message

8owadays, the idea of decodin is e(tremely wides!read= it is a si nificant !art of the activity of astronomers, lin uists, archaeolo ists, military s!ecialists, and so on& It is often su ested that we may be floatin in a sea of radio messa es from other civili<ations, messa es which we do not yet know how to deci!her& )nd much serious thou ht has been iven to the techniques of deci!herin such a messa e& Ene of the main !roblems!erha!s the dee!est !roblemis the question, -?ow will we reco ni<e the fact that there is a messa e at all? ?ow to identify a frame?- The sendin of a record seems to be a sim!le solutionits ross !hysical structure is very attention-drawin , and it is at least !lausible to us that it would tri er, in any sufficiently reat intelli ence, the idea of lookin for information hidden in it& ?owever, for technolo ical reasons, sendin of solid ob.ects to other star systems seems to be out of the question& Still, that does not !revent our thinkin about the idea& 8ow su!!ose that an alien civili<ation hit u!on the idea that the a!!ro!riate mechanism for translation of the record is a machine which converts the roove-!atterns into sounds& This would still be a far cry from a true deci!herin & 3hat, indeed, would constitute a successful deci!herin of such a record? Evidently, the civili<ation would have to be able to make sense out of the sounds& /ere !roduction of sounds is in itself hardly worthwhile, unless they have the desired tri erin effect in the brains 4if that is the word5 of the alien creatures& )nd what is that desired effect? It would be to activate structures in their brains which create emotional effects in them which are analo ous to the emotional effects which we e(!erience in hearin the !iece& In fact, the !roduction of sounds could even be by!assed, !rovided that they used the record in some other way et at the a!!ro!riate structures in their brains& 4If we humans had a way of tri erin the a!!ro!riate structures in our brains in sequential order, as music does, we mi ht be quite content to by!ass the soundsbut it seems e(traordinarily unlikely that there is any way to do that, other than via our ears& :eaf com!osers0eethoven, :voefk, 'aurKor musicians who can -hear- music by lookin at a score, do not ive the lie to this assertion, for such abilities are founded u!on !recedin decades of direct auditory e(!eriences&5 ?ere is where thin s become very unclear& 3ill bein s of an alien civili<ation have emotions? 3ill their emotionssu!!osin they have somebe ma!!able, in any sense, onto ours? If they do have emotions somewhat like ours, do the emotions cluster to ether in somewhat the same way as ours do? 3ill they understand such amal ams as

tra ic beauty or coura eous sufferin ? If it turns out that bein s throu hout the universe do share co nitive structures with us to the e(tent that even emotions overla!, then in some sense, the record can never be out of its natural conte(t= that conte(t is !art of the scheme of thin s, in nature& )nd if such is the case, then it is likely that a meanderin record, if not destroyed en route, would eventually et !icked u! by a bein or rou! of bein s, and et deci!hered in a way which we would consider successful&

(Imaginary S!acesca!e(
In askin about the meanin of a molecule of :8) above, I used the !hrase -com!ellin inner lo ic-= and I think this is a key notion& To illustrate this, let us sli htly modify our hy!othetical record-into-s!ace event by substitutin John 7a e1s -Ima inary Aandsca!e no& C- for the 0ach& This !iece is a classic of aleatoric, or chance, music music whose structure is chosen by various random !rocesses, rather than by an attem!t to convey a !ersonal emotion& In this case, twenty-four !erformers attach themselves to the twenty-four knobs on twelve radios& 'or the duration the !iece they twiddle their knobs in aleatoric ways so that each radio randomly ets louder and softer, switchin stations all the while& The total sound !roduced is the !iece of music& 7a e1s attitude is e(!ressed in "C own words, -to let sounds be themselves, rather than vehicles for manmade theories or e(!ressions of human sentiments&8ow ima ine that this is the !iece on the record sent out into s!ace& It would be e(traordinarily unlikelyif not downri ht im!ossiblefor an alien civili<ation to understand the nature of the artifact& They would !robably be very !u<<led by the contradiction between the frame messa e 4-I am a messa e= decode me-5, and the chaos of the inner structure& There are few -chunks- to sei<e onto in this 7a e !iece, few !atterns which could uide a deci!herer& En the other hand, there seems to be, in a 0ach !iece, much to sei<e onto!atterns, !atterns of !atterns, and so on& 3e have no way of knowin whether such !atterns are universally a!!ealin & 3e do not know enou h about the nature of intelli ence, emotions, or music to say whether the inner lo ic of a !iece by 0ach is so universally com!ellin that its meanin could s!an ala(ies& ?owever, whether 0ach in !articular has enou h inner lo ic is not the issue here= the issue is whether any messa e has, !er se, enou h com!ellin inner lo ic that its conte(t will be restored automatically whenever intelli ence of a hi h enou h level comes in contact with it& If some messa e did have that conte(t-restorin !ro!erty, then it would seem reasonable to consider the meanin of the messa e as an inherent !ro!erty of the messa e&

The /eroic

eci!herers

)nother illuminatin e(am!le of these ideas is the deci!herment of ancient te(ts written in unknown lan ua es and unknown al!habets& The intuition feels that there is information inherent in such te(ts, whether or not we succeed in revealin it& It is as stron a feelin as the belief that there is meanin inherent in a news!a!er written in

7hinese, even if we are com!letely i norant of 7hinese& Ence the scri!t or lan ua e of a te(t has been broken, then no one questions where the meanin resides, clearly it resides in the te(t, not in the method of deci!herment .ust as music resides in a record, not inside a record !layer2 Ene of the ways that we identify decodin mechanisms is by the fact that they do not add any meanin to the si ns or ob.ects which they take as in!ut= they merely reveal the intrinsic meanin of those si ns or ob.ects& ) .ukebo( is not a decodin mechanism, for it does not reveal any meanin belon in to its in!ut symbols= on the contrary, it su!!lies meanin concealed inside itself&

73G5=E HI! The =osetta -tone Ucourtes) of the British 'useum!V

8ow the deci!herment of an ancient te(t may have involved decades of labor by several rival teams of scholars, drawin on knowled e stored in libraries all over the world&&& :oesn1t this !rocess add information, too? Just how intrinsic is the meanin of a te(t, when such mammoth efforts are required in order to find the decodin rules? ?as one !ut

meanin into the te(t, or was that meanin already there? /y intuition says that the meanin was always there, and that des!ite the arduousness of the !ullin -out !rocess, no meanin was !ulled out that wasn1t in the te(t to start with& This intuition comes mainly from one fact, I feel that the result was inevitable= that, had the te(t not been deci!hered by this rou! at this time, it would have been deci!hered by that rou! at that time and it would have come out the same way& That is why the meanin is !art of the te(t itself= it acts u!on intelli ence in a !redictable way& 6enerally, we can say, meanin is !art of an ob.ect to the e(tent that it acts u!on intelli ence in a !redictable way& In 'i ure %# is shown the ;osetta stone, one of the most !recious of all historic discoveries& It was the key to the deci!herment of E y!tian hiero ly!hics, for it contains !arallel te(t in three ancient scri!ts, hiero ly!hic demotic characters, and 6reek& The inscri!tion on this basalt stele was first deci!hered in "*B" by Jean 'rancois 7ham!ollion, the -father of E y!tolo y-= it is a decree of !riests assembled at /em!his in favor of Ptolemy G E!i!hanes&

Three Layers of Any Message

In these e(am!les of deci!herment of out-of-conte(t messa es, we can se!arate out fairly clearly three levels of information, 4"5 the frame messa e= 4B5 the outer messa e= 4%5 the inner messa e& The one we are most familiar with is 4%5, the inner messa e= it is the messa e which is su!!osed to be transmitted, the emotional e(!eriences in music, the !henoty!e in enetics, the royalty and rites of ancient civili<ations in tablets, etc& To understand the inner messa e is to have e(tracted the meanin intended by the sender&&& The frame messa e is the messa e -I am a messa e= decode me if you can2-= and it is im!licitly conveyed by the ross structural as!ects of any information-bearer& To understand the frame messa e is to reco ni<e the need for a decodin -mechanism& If the frame messa e is reco ni<ed as such, then attention is switched to level 4B5, the outer messa e& This is information, im!licitly carried by symbol-!atterns and structures in the messa e, which tells how to decode the inner messa e& To understand the outer messa e is to build, or know how to build, the correct decodin mechanism for the inner messa e& This outer level is !erforce an im!licit messa e, in the sense that the sender cannot ensure that it will be understood& It would be a vain effort to send instructions which tell how to decode the outer messa e, for they would have to be !art of the inner messa e, which can only be understood once the decodin mechanism has been found& 'or this reason, the

outer messa e is necessarily a set of tri ers, rather than a messa e which can be revealed by a known decoder& The formulation of these three -layers- is only a rather crude be innin at analy<in how meanin is contained in messa es& There may be layers and layers of outer and inner messa es, rather than .ust one of each& Think, for instance, of how intricately tan led are the inner and outer messa es of the ;osetta stone& To decode a messa e fully, one would have to reconstruct the entire semantic structure which underlay its creation and thus to understand the sender in every dee! way& ?ence one could throw away the inner messa e, because if one truly understood all the finesses of the outer messa e, the inner messa e would be reconstructible& The book *fter Ba(el, by 6eor e Steiner, is a lon discussion of the interaction between inner and outer messa es 4thou h he never uses that terminolo y5& The tone of his book is iven by this quote,
3e normally use a shorthand beneath which there lies a wealth of subconscious, deliberately concealed or declared associations so e(tensive and intricate that they !robably equal the sum and uniqueness of our status as an individual !erson& "

Thou hts alon the same lines are e(!ressed by Aeonard 0& /eyer, in his book 'usic, the *rts, and 3deas:
The way of listenin to a com!osition by Elliott 7arter is radically different from the way of listenin a!!ro!riate to a work by John 7a e& Similarly, a novel by 0eckett must in a si nificant sense be read differently from one by 0ellow& ) !aintin by 3illem de @oonin and one by )ndy 3arhol require different !erce!tional-co nitive attitudes& B

Perha!s works of art are tryin to convey their style more than anythin else& In that case, if you could ever !lumb a style to its very bottom you could dis!ense with the creations in that style& -Style-, -outer messa e -decodin technique-all ways of e(!ressin the same basic idea&

Schrodinger%s A!eriodic Crystals

3hat makes us see a frame messa e in certain ob.ects, but none in other= 3hy should an alien civili<ation sus!ect, if they interce!t an errant record, that a messa e lurks within? 3hat would make a record any different from a meteorite? 7learly its eometric sha!e is the first clue that -somethin funny is oin on-& The ne(t clue is that, on a more microsco!ic scale, it consists of a very lon a!eriodic sequence of !atterns, arran ed in a s!iral& If we were to unwra! the s!iral, we would have one hu e linear sequence 4around B+++ feet lon 5 of minuscule symbols& This is not so different from a :8) molecule, whose symbols, drawn from a mea er -al!habet- of four different chemical bases, are arrayed in a one-dimensional sequence, and then coiled u! into a heli(& 0efore )very had established the connection between enes and :8), the !hysicist Erwin Schr>din er !redicted, on !urely theoretical rounds, that enetic information would have to be stored in -a!eriodic crystals-, in his influential book >hat 3s +ife? In fact, books themselves are

a!eriodic crystals contained inside neat eometric forms& These e(am!les su est that, where an a!eriodic crystal is found -!acka ed- inside a very re ular eometric structure, there may lurk a inner messa e& 4I don1t claim this is a com!lete characteri<ation of frame messa es= however, it is a fact that many common messa es have frame messa es of this descri!tion& See 'i ure C+ for some ood e(am!les&5

Languages for the Three Le"els

The three levels are very clear in the case of a messa e found in a bottle washed u! on a beach& The first level, the frame messa e, is found when one !icks u! the bottle and sees that it is sealed, and contains a dry !iece of !a!er& Even without seein writin , one reco ni<es this ty!e of artifact as an information-bearer, and at this !oint it would take an e(traordinaryalmost inhumanlack of curiosity, to dro! the bottle and not look further&

73G5=E DK! * collage of scripts! 5ppermost on the left is an inscription in the undeciphered (oustrophedonic writing s)stem from Easter 3sland, in which e#er) second line is upside down! The characters are chiseled on a wooden ta(let, D inches () HG inches! 'o#ing clockwise, we encounter #erticall) written 'ongolian: a(o#e, present%da) 'ongolian, (elow, a document dating from BHBD! Then we come to a poem in Bengali () =a(indranath Tagore in the (ottom righthand corner! Ne9t to it is a newspaper headline in 'ala)alam >est ;erala, southern 3ndia/, a(o#e which is the elegant cur#ilinear language Tamil East ;erala/! The smallest entr) is part of a folk tale in Buginese 8ele(es 3sland, 3ndonesia/! 3n the center of the collage is a paragraph in the Thai language, and a(o#e it a manuscript in =unic dating from the fourteenth centur), containing a sample of the pro#incial law of -cania south -weden/! 7inall), wedged in on the left is a section of the laws of ,ammura(i, written in *ss)rian cuneiform! *s an outsider, 3 feel a deep sense of m)ster) as 3 wonder how meanin is cloaked in the strange cur#es and angles of each of these (eautiful aperiodic cr)stals! 3n form, there is content! U7rom ,am ?ensen, Si n, Symbol, and Scri!t New York: G! &utnam's -ons, BIJI/, pp! EI cuneiform/, HHJ Easter 3sland/, HEJ, DBC 'ongolian/, GGF =unic/R from ;eno ;at@ner, The Aan ua es of the 3orld New York: 7unk W >agnalls, BICG/, pp! BIK Bengali/, FHC Buginese/R from 3! *! =ichards and 8hristine Gi(son, En lish Throu h Pictures New York: >ashington -"uare &ress, BIJK/, pp! CH Tamil/, EF Thai5&

8e(t, one o!ens the bottle and e(amines the marks on the !a!er& Perha!s, they are in Ja!anese= this can be discovered without any of the inner messa e bein understood it merely comes from a reco nition of the characters& The outer messa e can be stated as an En lish sentence, -I am in Ja!anese&- Ence this has been discovered, then one can !roceed to the inner messa e, which may be a call for hel!, a haiku !oem, a lover1s lament&&& It would be of no use to include in the inner messa e a translation of the sentence -This messa e is in Ja!anese-, since it would take someone who knew Ja!anese to read it& )nd before readin it, he would have to reco ni<e the fact that, as it is in Ja!anese, he can read it& Hou mi ht try to wri le out of this by includin translations of the statement -This messa e is in Ja!anese- into many different lan ua es& That would hel! in a !ractical sense, but in a theoretical sense the same difficulty is there& )n En lish-s!eakin !erson still has to reco ni<e the -En lishness- of the messa e= otherwise it does no ood& Thus one cannot avoid the !roblem that one has to find out how to deci!her the inner messa e from the outside= the inner messa e itself may !rovide clues and confirmations, but those are at best tri ers actin u!on the bottle finder 4or u!on the !eo!le whom he enlists to hel!5& Similar kinds of !roblem confront the shortwave radio listener& 'irst he has to decide whether the sounds he hears actually constitute a messa e, or are .ust static& The sounds in themselves do not ive the answer, not even in the unlikely case that the inner messa e is in the listener1s own native lan ua e, and is sayin , -These sounds actually constitute a messa e and are not .ust static2- If the listener reco ni<es a frame messa e in the sounds, then he tries to identify the lan ua e the broadcast is in and clearly, he is still on the outside= he acce!ts triggers from the radio, but they cannot e(!licitly tell him the answer& It is in the nature of outer messa es that they are not conveyed in any e(!licit lan ua e& To find an e(!licit lan ua e in which to convey outer messa es would not be a breakthrou hit would be a contradiction in terms2 It is always the listener1s burden to understand the outer messa e& Success lets him break throu h into the inside, at which !oint the ratio of tri ers to e(!licit meanin s shifts drastically towards the latter& 0y com!arison with the !revious sta es, understandin the inner messa e seems effortless& It is as if it .ust ets !um!ed in&

The (3u.ebo-( Theory of Meaning

These e(am!les may a!!ear to be evidence for the view!oint that no messa e has intrinsic meanin , for in order to understand any inner messa e, no matter how sim!le it is, one must first understand its frame messa e and its outer messa e, both of which are carried only by tri ers 4such as bein written in the Ja!anese al!habet, or havin s!iralin rooves, etc&5& It be ins to seem, then, that one cannot et away from a -.ukebo(- theory of meanin the doctrine that no message contains inherent meaning, because, before any messa e can be understood, it has to be used as the in!ut to some -.ukebo(-, which means that information contained in the -.ukebo(- must be added to the messa e before it acquires meanin &

This ar ument is very similar to the tra! which the Tortoise cau ht )chilles in, in Aewis 7arroll1s :ialo ue& There, the tra! was the idea that before you can use any rule, you have to have a rule which tells you how to use that rule= in other words, there is an infinite hierarchy of levels of rules, which !revents any rule from ever ettin used& ?ere, the tra! is the idea that before you can understand any messa e, you have to have a messa e which tells you how to understand that messa e= in other words, there is an infinite hierarchy of levels of messa es, which !revents any messa e from ever ettin understood& ?owever, we all know that these !arado(es are invalid, for rules do et used, and messa es do et understood& ?ow come?

Against the 3u.ebo- Theory

This ha!!ens because our intelli ence is not disembodied, but is instantiated in !hysical ob.ects, our brains& Their structure is due to the lon !rocess of evolution, and their o!erations are overned by the laws of !hysics& Since they are !hysical entities, our (rains run without (eing told how to run & So it is at the level where thou hts are !roduced by !hysical law that 7arroll1s rule-!arado( breaks down= and likewise, it is at the level where a brain inter!rets incomin data as a messa e that the messa e-!arado( breaks down& It seems that brains come equi!!ed with -hardware- for reco ni<in that certain thin s are messa es, and for decodin those messa es& This minimal inborn ability to e(tract inner meanin is what allows the hi hly recursive, snowballin !rocess of lan ua e acquisition to take !lace& The inborn hardware is like a .ukebo(, it su!!lies the additional information which turns mere tri ers into com!lete messa es&

Meaning Is Intrinsic If Intelligence Is ;atural

8ow if different !eo!le1s -.ukebo(es- had different -son s- in them, and res!onded to iven tri ers in com!letely idiosyncratic ways, then we would have no inclination to attribute intrinsic meanin to those tri ers& ?owever, human brains are so constructed that one brain res!onds in much the same way to a iven tri er as does another brain, all other thin s bein equal& This is why a baby can learn any lan ua e= it res!onds to tri ers in the same way as any other baby& This uniformity of -human .ukebo(esestablishes a uniform -lan ua e- in which frame messa es and outer messa es can be communicated& If, furthermore, we believe human intelli ence is .ust one e(am!le of a eneral !henomenon in naturethe emer ence of intelli ent bein s in widely varyin conte(tsthen !resumably the -lan ua e- in which frame messa es and outer messa es are communicated amon humans is a -dialect- of a uni#ersal lan ua e by which intelli ences can communicate with each other& Thus, there would be certain kinds of tri ers which would have -universal tri erin !ower-, in that all intelli ent bein s would tend to res!ond to them in the same way as we do& This would allow us to shift our descri!tion of where meanin located& 3e could ascribe the meanin s 4frame, outer, and inner5 of a messa e to the messa e itself, because of the fact that deci!herin mechanisms are themselves universalthat is, they are fundamental forms of nature which arise in the same way in diverse conte(ts& To make it

concrete, su!!ose that -)-L- tri ered the same son in all .ukebo(es and su!!ose moreover that .ukebo(es were not man-made artifacts, but widely occurrin natural ob.ects, like ala(ies or carbon atoms& Fnder such circumstances, we would !robably feel .ustified in callin the universal tri erin !ower of -)-L- its -inherent meanin -= also, -)-L- would merit the name of -messa e-, rather than -tri er-, and the son would indeed be a -revelation- of the inherent, thou h im!licit, meanin of -)-L-&

Earth Chau"inism
This ascribin of meanin to a messa e comes from the invariance of the !rocessin of the messa e by intelli ences distributed anywhere in the universe& In that sense, it bears some resemblance to the ascribin of mass to an ob.ect& To the ancients, it must have seemed that an ob.ect1s wei ht was an intrinsic !ro!erty of the ob.ect& 0ut as ravity became understood, it was reali<ed that wei ht varies with the ravitational field the ob.ect is immersed in& 8evertheless, there is a related quantity, the mass, which does not vary accordin to the ravitational field= and from this invariance came the conclusion that an ob.ect1s mass was an intrinsic !ro!erty of the ob.ect itself& If it turns out that mass is also a variable, accordin to conte(t, then we will backtrack and revise our o!inion that it is an intrinsic !ro!erty of an ob.ect& In the same way, we mi ht ima ine that there could e(ist other kinds of -.ukebo(es-intelli enceswhich communicate amon each other via messa es which we would never reco ni<e as messa es, and who also would never reco ni<e our messa es as messa es& If that were the case, then the claim that meanin is an intrinsic !ro!erty of a set of symbols would have to be reconsidered& En the other hand, how could we ever reali<e that such bein s e(isted? It is interestin to com!are this ar ument for the inherency of meanin with a !arallel ar ument for the inherency of wei ht& Su!!ose one defined an ob.ect1s wei ht as -the ma nitude of the downward force which the ob.ect e(erts when on the surface of the !lanet Earth-& Fnder this definition, the downward force which an ob.ect e(erts when on the surface of /ars would have to be iven another name than -wei ht-& This definition makes wei ht an inherent !ro!erty, but at the cost of eocentricity-Earth chauvinism-& It would be like -6reenwich chauvinism-refusin to acce!t local time anywhere on the lobe but in the 6/T time <one& It is an unnatural way to think of time& Perha!s we are unknowin ly burdened with a similar chauvinism with res!ect to intelli ence, and consequently with res!ect to meanin & In our chauvinism, we would call any bein with a brain sufficiently much like our own -intelli ent-, and refuse to reco ni<e other ty!es of ob.ects as intelli ent& To take an e(treme e(am!le, consider a meteorite which, instead of deci!herin the outer-s!ace 0ach record, !unctures it with colossal indifference, and continues in its merry orbit& It has interacted with the record in a way which we feel disre ards the record1s meanin & Therefore, we mi ht well feel tem!ted to call the meteorite -stu!id-& 0ut !erha!s we would thereby do the meteorite a disservice& Perha!s it has a -hi her intelli ence- which we in our Earth chauvinism cannot !erceive, and its interaction with the record was a manifestation of that hi her intelli ence& Perha!s, then, the record has a -hi her meanin - totally different from that

which we attribute to it= !erha!s its meanin de!ends on the ty!e of intelli ence !erceivin it& Perha!s& It would be nice if we could define intelli ence in some other way than -that which ets the same meanin out of a sequence of symbols as we do-& 'or if we can only define it this one way, then our ar ument that meanin is an intrinsic !ro!erty is circular, hence content-free& 3e should try to formulate in some inde!endent way a set of characteristics which deserve the name -intelli ence-& Such characteristics would constitute the uniform core of intelli ence, shared by humans& )t this !oint in history we do not yet have a well-defined list of those characteristics& ?owever, it a!!ears likely that within the ne(t few decades there will be much !ro ress made in elucidatin what human intelli ence is& In !articular, !erha!s co nitive !sycholo ists, workers in )rtificial Intelli ence, and neuroscientists will be able to synthesi<e their understandin s, and come u! with a definition of intelli ence& It may still be human-chauvinistic= there is no way around that& 0ut to counterbalance that, there may be some ele ant and beautiful and !erha!s even sim!leabstract ways of characteri<in the essence of intelli ence& This would serve to lessen the feelin of havin formulated an anthro!ocentric conce!t& )nd of course, if contact were established with an alien civili<ation from another star system, we would feel su!!orted in our belief that our own ty!e of intelli ence is not .ust a fluke, but an e(am!le of a basic form which rea!!ears in nature in diverse conte(ts, like stars and uranium nuclei& This in turn would su!!ort the idea of meanin bein an inherent !ro!erty& To conclude this to!ic, let us consider some new and old e(am!les, and discuss the de ree of inherent meanin which they have, by !uttin ourselves, to the e(tent that we can, in the shoes of an alien civili<ation which interce!ts a weird ob.ect&&&

T'o ,laques in S!ace

7onsider a rectan ular !laque made of an indestructible metallic alloy, on which are en raved two dots, one immediately above the another, the !recedin colon shows a !icture& Thou h the overall form of the mi ht su est that it is an artifact, and therefore that it mi ht conceal some messa e, two dots are sim!ly not sufficient to convey anythin & 47an you, before readin on, hy!othesi<e what they are su!!osed to mean?5 0ut su!!ose that we made a second !laque, containin more dots, as follows&

. . .. ... ..... ........ ............... .......................... .......................................

8ow one of the most obvious thin s to do so it mi ht seem to a terrestrial intelli ence at leastwould be to count the dots in the successive rows& The sequence obtained is, ", ", B, %, L, *, "%, B", %C& ?ere there is evidence of a rule overnin the !ro ression from one line to the ne(t& In fact, the recursive !art of the definition of the 'ibonacci numbers can be inferred, with some confidence, from this list& Su!!ose we think of the initial !air of values 4","5 as a - enoty!e- from which the -!henoty!e-the full 'ibonacci sequenceis !ulled out by a recursive rule& 0y sendin the enoty!e alonenamely the first version !laquewe fail to send the information which allows reconstitution !henoty!e& Thus, the enoty!e does not contain the full s!ecification of the !henoty!e& En the other hand, if we consider the second version of the !laque to be the enoty!e, then there is much better cause to su!!ose that the !henoty!e could actually be reconstituted& This new version of the enoty!ea -lon enoty!e-contains so much information that the mechanism () which phenot)pe is pulled out of genot)pe can (e inferred () intelligence from the genot)pe alone! Ence this mechanism is firmly established as the way to !ull !henoty!e from enoty!e, then we can o back to usin -short enoty!es-like the first !laque& 'or instance, the -short enoty!e- 4",%5 would yield the !henoty!e ", %, C, $, "", "*, B#, C$,&&&

the Aucas sequence& )nd for every set of two initial values that is, for every short enoty!ethere will be a corres!ondin !henoty!e& 0ut the short enoty!es, unlike the lon ones, are only tri ersbuttons to be !ushed on the .ukebo(es into which the recursive rule has been built& The lon enoty!es are informative enou h that they tri er, in an intelli ent bein , the reco nition of what kind of -.ukebo(- to build& In that sense, the lon enoty!es contain the information of the !henoty!e, whereas the short enoty!es do not& In other words, the lon enoty!e transmits not only an inner messa e, but also an outer messa e, which enables the inner messa e to be read& It seems that the clarity of the outer messa e resides in the sheer len th of the messa e& This is not une(!ected= it !arallels !recisely what ha!!ens in deci!herin ancient te(ts& 7learly, one1s likelihood of success de!ends crucially on the amount of te(t available&

Bach #s! Cage Again

0ut .ust havin a lon te(t may not be enou h& Aet us take u! once more the difference between sendin a record of 0ach1s music into s!ace, and a record of John 7a e1s music& Incidentally, the latter, bein a 7om!osition of )leatorically 6enerated Elements, mi ht be handily called a -7)6E-, whereas the former, bein a 0eautiful )!eriodic 7rystal of ?armony, mi ht a!tly be dubbed a -0)7?-& 8ow let1s consider what the meanin of a

7a e !iece is to ourselves& ) 7a e !iece has to be taken in a lar e cultural settin as a revolt a ainst certain kinds of traditions& Thus, if we want to transmit that meanin , we must not only send the notes of the !iece, but we must have earlier communicated an e(tensive history of 3estern culture& It is fair to say, then, that an isolated record of John 7a e1s music does not have an intrinsic meanin & ?owever, for a listener who is sufficiently well versed in 3estern and Eastern cultures, !articularly in the trends in 3estern music over the last few decades, it does carry meanin but such a listener is like a .ukebo(, and the !iece is like a !air of buttons& The meanin is mostly contained inside the listener to be in with= the music serves only to tri er it& )nd this -.ukebo(-, unlike !ure intelli ence, is not at all universal= it is hi hly earthbound, de!endin on idiosyncratic sequences of events all over our lobe for lon !eriod of time& ?o!in that John 7a e1s music will be understood by another civili<ation is like ho!in that your favorite tune, on a .ukebo( on the moon, will have the same buttons as in a saloon in Saskatoon& En the other hand, to a!!reciate 0ach requires far less cultural knowled e& This may seem like hi h irony, for 0ach is so much more com!le( and or ani<ed, and 7a e is so devoid of intellectuality& 0ut there stran e reversal here, intelli ence loves !atterns and balks at randomness& 'or most !eo!le, the randomness in 7a e1s music requires much e(!lanation= and even after e(!lanations, they may feel they are missin the messa ewhereas with much of 0ach, words are su!erfluous& In sense, 0ach1s music is more self-contained than 7a e1s music& Still, it is not clear how much of the human condition is !resumed by 0ach& 'or instance, music has three ma.or dimensions of structure 4melody, harmony, rhythm5, each of which can be further divided into small-scale, intermediate, and overall as!ects& 8ow in each of these dimensions, there is a certain amount of com!le(ity which our minds can handle before bo lin = clearly a com!oser takes this into account, mostly unconsciously, when writin a !iece& These -levels of tolerable com!le(ity- alon different dimensions are !robably very de!endent on the !eculiar conditions of our evolution as a s!ecies, and another intelli ent s!ecies mi ht have develo!ed music with totally different levels of tolerable com!le(ity alon these many dimensions& Thus a 0ach !iece mi ht conceivably have to be accom!anied by a lot of information about the human s!ecies, which sim!ly could not be inferred from the music1s structure alone& If we equate the 0ach music with a enoty!e, and the emotions which it is su!!osed to evoke with the !henoty!e, then what we are interested in is whether the enoty!e contains all the information necessary for the revelation of the !henoty!e&

/o' &ni"ersal Is

;A%s Message7

The eneral question which we are facin , and which is very similar to the questions ins!ired by the two !laques, is this, -?ow much of the conte(t necessary for its own understandin is a messa e ca!able of restorin ?- 3e can now revert to the ori inal biolo ical meanin s of - enoty!e- and -!henoty!e-:8) and a livin or anismand ask similar questions& :oes :8) have universal tri erin !ower? Er does it need a -bio.ukebo(- to reveal its meanin ? 7an :8) evoke a !henoty!e without bein

embedded in the !ro!er chemical conte(t? To this question the answer is no but a qualified no& 7ertainly a molecule of :8) in a vacuum will not create anythin at all& ?owever, if a molecule of :8) were set to seek its fortune in the universe, as we ima ined the 0)7? and the 7)6E were, it mi ht be interce!ted by an intelli ent civili<ation& They mi ht first of all reco ni<e its frame messa e& 6iven that, they mi ht o on to try to deduce from its chemical structure what kind of chemical environment it seemed to want, and then su!!ly such an environment& Successively more refined attem!ts alon these lines mi ht eventually lead to a full restoration of the chemical conte(t necessary for the revelation of :8)1s !henoty!ical meanin & This may sound a little im!lausible, but if one allows many millions of years for the e(!eriment, !erha!s the :8)1s meanin would finally emer e& En the other hand, if the sequence of bases which com!ose a strand of :8) were sent as abstract symbols 4as in 'i & C"5, not as a lon helical molecule, the odds are virtually nil that this, as an outer messa e, would tri er the !ro!er decodin mechanism which would enable the !henoty!e to be drawn out of the enoty!e& This would be a case of wra!!in an inner messa e in such an abstract outer messa e that the conte(t-restorin !ower of the outer messa e would be lost, and so in a very !ra matic sense, the set of symbols would have no intrinsic meanin & Aest you think this all sounds ho!elessly abstract and !hiloso!hical, consider that the e(act moment when !henoty!e can be said to be -available-, or -im!lied-, by enoty!e, is a hi hly char ed issue in our day, it is the issue of abortion&

73G5=E DB! This Giant *periodic 8r)stal is the (ase se"uence for the chromosome of (acteriophage fSBCD! 3t is the first complete genome e#er mapped out for an) organism! *(out F,KKK of these (oustrophedonic pages would (e needed to show the (ase se"uence of a single E! 8oli cell, and a(out one million pages to show the (ase se"uence of the .N* of a single human cell! The (ook now in )our hands contains roughl) the same amount of information as a molecular (lueprint for one measl) E! 8oli cell!

Chromatic "antasy# And "eud

,a#ing had a splendid dip in the pond, the Tortoise is Must crawling out and shaking himself dr), when who (ut *chilles walks ()! Tortoise, ?o there, )chilles& I was .ust thinkin of you as I s!lashed around in the !ond& *chilles, Isn1t that curious? I was .ust thinkin of you, too, while I meandered throu h the meadows& They1re so reen at this time of year& Tortoise, Hou think so? It reminds me of a thou ht I was ho!in to share with you& 3ould you like to hear it? *chilles, Eh, I would be deli hted& That is, I would be deli hted as lon you1re not oin to try to snare me in one of your wicked tra!s of lo ic, /r& T& Tortoise, 3icked tra!s? Eh, you do me wron & 3ould I do anythin wicked? I1m a !eaceful soul, botherin nobody and leadin a entle, herbivorous life& )nd my thou hts merely drift amon the oddities and quirks of how thin s are 4as I see them5& I, humble observer of !henomena, !lod alon and !uff my silly words into the air rather uns!ectacularly, I am afraid& 0ut to reassure you about my intentions, I was only !lannin to s!eak of my Tortoise-shell today, and as you know, those thin s have nothin nothin whatsoeverto do with lo ic2 *chilles, Hour words :E reassure me, /r& T& )nd, in fact, my curiosity quite !iqued& I would certainly like to listen to what you have to say, even if it is uns!ectacular& Tortoise, Aet1s see&&& how shall I be in? ?mm&&& 3hat strikes you most about my shell, )chilles? *chilles, It looks wonderfully clean2 Tortoise, Thank you& I .ust went swimmin and washed off several layers of dirt which had accumulated last century& 8ow you can see how reen my shell is& *chilles, Such a ood healthy reen shell, it1s nice to see it shinin in the sun& Tortoise, 6reen? It1s not reen& *chilles, 3ell, didn1t you .ust tell me your shell was reen? Tortoise, I did& *chilles, Then, we a ree, it is reen& Tortoise, 8o, it isn1t reen& *chilles, Eh, I understand your ame& Hou1re hintin to me that what you say isn1t necessarily true= that Tortoises !lay with lan ua e= that your statements and reality don1t necessarily match= that Tortoise, I certainly am not& Tortoises treat words as sacred& Tortoises revere accuracy& *chilles, 3ell, then, why did you say that your shell is reen, and that it is not reen also? Tortoise, I never said such a thin = but I wish I had& *chilles, Hou would have liked to say that? Tortoise, 8ot a bit& I re ret sayin it, and disa ree wholeheartedly with it& *chilles, That certainly contradicts what you said before2 Tortoise, 7ontradicts? 7ontradicts? I never contradict myself& It1s not !art of Tortoisenature&

*chilles, 3ell, I1ve cau ht you this time, you sli!!ery fellow, you& 7au ht you in a fullfled ed contradiction& Tortoise, Hes, I uess you did& *chilles, There you o a ain2 8ow you1re contradictin yourself more and more2 Hou are so stee!ed in contradiction it1s im!ossible to ar ue with you2 Tortoise, 8ot really& I ar ue with myself without any trouble at all& Perha!s the !roblem is with you& I would venture a uess that maybe you1re the one who1s contradictory, but you1re so tra!!ed in your own tan led web that you can1t see how inconsistent you1re bein & *chilles, 3hat an insultin su estion2 I1m oin to show you that you1re the contradictory one, and there are no two ways about it& Tortoise, 3ell, if it1s so, your task ou ht to be cut out for you& 3hat could be easier than to !oint out a contradiction? 6o aheadtry it out& *chilles, ?mm&&& 8ow I hardly know where to be in& Eh&&& I know& Hou first said that 4"5 your shell is reen, and then you went on to say that 4B5 your shell is not reen& 3hat more can I say? Tortoise, Just kindly !oint out the contradiction& Muit beatin around the bush& *chilles, 0ut-but-but&&& Eh, now I be in to see& 4Sometimes I am so slow-witted25 It must be that you and I differ as to what constitutes a contradiction& That1s the trouble& 3ell, let me make myself very clear, a contradiction occurs when somebody says one thin and denies it at the same time& Tortoise, ) neat trick& I1d like to see it done& Probably ventriloquists would e(cel at contradictions, s!eakin out of both sides of their mouth, as it were& 0ut I1m not a ventriloquist& *chilles, 3ell, what I actually meant is .ust that somebody can say one thin and deny it all within one sin le sentence2 It doesn1t literally have to be in the same instant& Tortoise, 3ell, you didn1t ive E8E sentence& Hou ave T3E& *chilles, Hestwo sentences that contradict each other2 Tortoise, I am sad to see the tan led structure of your thou hts becomin so e(!osed, )chilles& 'irst you told me that a contradiction is some thin which occurs in a sin le sentence& Then you told me that you found a contradiction in a !air of sentences I uttered& 'rankly, it1s .ust as I said& Hour own system of thou ht is so delusional that you mana e to avoid seein how inconsistent it is& 'rom the outside, however !lain as day& *chilles, Sometimes I et so confused by your diversionary tactics that I can1t quite tell if we1re ar uin about somethin utterly !etty, or somethin dee! and !rofound2 Tortoise, I assure you, Tortoises don1t s!end their time on the !etty& ?ence it1s the latter& *chilles, I am very reassured& Thank you& 8ow I have had a moment to reflect, and I see the necessary lo ical ste! to convince you that you contradicted yourself& Tortoise, 6ood, ood& I ho!e it1s an easy ste!, an indis!utable one& *chilles, It certainly is& Even you will a ree with it& The idea is that since you believed sentence " 4-/y shell is reen-5, )8: you believed sentence B 4-/y shell is not reen-5, you would believe one com!ound sentence in which both were combined, wouldn1t you?

Tortoise, Ef course& It would only be reasonable&&& !rovidin .ust that the manner of combination is universally acce!table& 0ut I1m sure we1ll a ree on that& *chilles, Hes, and then I1ll have you2 The combination I !ro!ose is Tortoise, 0ut we must be careful in combinin sentences& 'or instance you1d rant that -Politicians lie- is true, wouldn1t you? *chilles, 3ho could deny it? Tortoise, 6ood& Aikewise, -7ast-iron sinks- is a valid utterance, isn1t it? *chilles, Indubitably& Tortoise, Then, !uttin them to ether, we et -Politicians lie in cast iron sinks-& 8ow that1s not the case, is it? *chilles, 8ow wait a minute&&& -Politicians lie in cast-iron sinks?- 8o, but Tortoise, So, you see, combinin two true sentences in one is not a !olicy, is it? *chilles, 0ut youyou combined the twoin such a silly way2 Tortoise, Silly? 3hat have you ot to ob.ect to in the way I combined them? 3ould you have me do otherwise? *chilles, Hou should have used the word -and-, not -in-& Tortoise, I should have? Hou mean, if HEF1: had HEF; way, I should have& *chilles, 8oit1s the AE6I7)A thin to do& It1s ot nothin to do with !ersonally& Tortoise, This is where you always lose me, when you resort to your Ao ic and its hi hsoundin Princi!les& 8one of that for me today, !lease& *chilles, Eh, /r& Tortoise, don1t !ut me throu h all this a ony& Hou know very well that that1s what -and- means2 It1s harmless to combine true sentences with -and-2 Tortoise, -?armless-, my eye2 3hat all2 This is certainly a !ernicious !lot to entra! a !oor, innocent, bumblin Tortoise in a fatal contradiction& If it were so harmless, why would you be tryin so bloody hard to et me to do it? Eh? *chilles, Hou1ve left me s!eechless& Hou make me feel like a villain, where I really had only the most innocent of motivations& Tortoise, That1s what everyone believes of himself&&& *chilles, Shame on metryin to outwit you, to use words to snare you in a selfcontradiction& I feel so rotten& Tortoise, )nd well you should& I know what you were tryin to set u!& Hour !lan was to make me acce!t sentence %, to wit, -/y shell is reen and my shell is not reen-& )nd such a blatant falsehood is re!ellent to the Ton ue of a Tortoise& *chilles, Eh, I1m so sorry I started all this& Tortoise, Hou needn1t be sorry& /y feelin s aren1t hurt& )fter all, I1m used to the unreasonable ways of the folk about me& I en.oy your com!any, )chilles, even if your thinkin lacks clarity& *chilles, Hes&&& 3ell, I fear I am set in my ways, and will !robably continue to err and err a ain, in my quest for Truth& Tortoise, Today1s e(chan e may have served a little to ri ht your course& 6ood day, )chilles& *chilles, 6ood day, /r& T&

C/A,TE$ 4II

The ,ro!ositional Calculus

>ords and Symbols
T?E P;E7E:I86 :I)AE6FE is reminiscent of the Two%&art 3n#ention by Aewis 7arroll& In both, the Tortoise refuses to use normal, ordinary words in the normal, ordinary wayor at least he refuses to do so when it is not to his advanta e to do so& ) way to think about the 7arroll !arado( was iven last 7ha!ter& In this 7ha!ter we are oin to make symbols do what )chilles couldn1t make the Tortoise do with his words& That is, we are oin to make a formal system one of whose symbols will do .ust what )chilles wished the word 1and1 would do, when s!oken by the Tortoise, and another of whose symbols will behave the way the words 'if!!! then!!!1 ou ht to behave& There are only two other words which we will attem!t to deal with, 1or1 and 1not1& ;easonin which de!ends only on correct usa e of these words is termed propositional reasoning&

Al!habet and +irst $ule of the ,ro!ositional Calculus

I will !resent this new formal system, called the Pro!ositional 7alculus, like a !u<<le, not e(!lainin everythin at once, but lettin you thin s out to some e(tent& 3e be in with the list of symbols, < 8 [ > $ ]

'
~

The first rule of this system that I will reveal is the followin , ;FAE E' JEI8I86, If 9 and ) are theorems of the system, then so is the strin g 9) a& This rule takes two theorems and combines them into one& It should remind you of the :ialo ue&

>ell-+ormed Strings
There will be several other rules of inference, and they will all be !resented shortlybut first, it is im!ortant to define a subset of all strin s, namely the well%formed strin s& They will be defined in a recursive way& 3e be in with the

)TE/S, ,, 8, and $ are called atoms&& 8ew atoms are formed by a!!endin !rimes onto the ri ht of old atoms-thus, $1, 8-, ,111, etc& This ives an endless su!!ly of atoms& )ll atoms are well-formed& Then we have four recursive 'E;/)TIE8 ;FAES, If ( and y are well-formed, then the followin four strin s are also well-formed, 4"5 4B5 4%5 4C5 h9 g 9)a g 9)a g 9)a

'or e(am!le, all of the followin are well-formed, , h, hh, 81 h81 g,h81 a hg,h81 a hhg,h81 a ghg,h81 ahhg,h81 aa atom by 4"5 by 4"5 atom by 4"5 by 4B5 by 4"5 by 4C5 by 4%5

The last one may look quite formidable, but it is built u! strai htforwardly from two com!onentsnamely the two lines .ust above it& Each of them is in turn built u! from !revious lines&&& and so on& Every well-formed strin can in this way be traced back to its elementary constituentthat is, atoms& Hou sim!ly run the formation rules backwards until you can no more& This !rocess is uaranteed to terminate, since each formation rule 4when run forwards5 is a lengthening rule, so that runnin it backwards always drives you towards atoms& This method of decom!osin strin s thus serves as a check on the wellformedness of any strin & It is a top%down decision procedure for well-formedness& Hou can test your understandin of this decision !rocedure by checkin which of the followin strin s are well-formed, 4"5 4B5 4%5 4C5 4L5 g,a gh,a g,8$a g,8a gg,8a8h,aa

4D5 g,h,a 4$5 gg,g8$aagh,h$1aa 4*5 g,8ag 8,a 4)nswer, Those whose numbers are 'ibonacci numbers are not formed& The rest are wellformed&5

More $ules of Inference

8ow we come to the rest of the rules by which theorems of this system constructed& ) few rules of inference follow& In all of them, the symbols 1 9' and 1)1 are always to be understood as restricted to well formed strin s& ;FAE E' SEP);)TIE8, If g 9)a is a theorem, then both 9 and ) are theorems& Incidentally, you should have a !retty ood uess by now as to what conce!t the symbol 11 stands for& 4?int, it is the troublesome word from the !recedin :ialo ue&5 'rom the followin rule, you should be a fi ure out what conce!t the tilde 41h15 re!resents, :EF0AE-TIA:E ;FAE, The strin 1hh1 can be deleted from any theorem& It can also be inserted into any theorem, !rovided that the resultin strin is itself well-formed&

The +antasy $ule

8ow a s!ecial feature of this system is that it has no a9iomsOonly rules& If you think back to the !revious formal systems we1ve seen, you may wonder how there can be any theorems, then& ?ow does everythin et started? The answer is that there is one rule which manufactures theorems from out of thin airit doesn1t need an -old theorem- as in!ut& 4The rest of the do require in!ut&5 This s!ecial rule is called the fantas) rule& The reason I call it that is quite sim!le& To use the fantasy rule, the first thin you do is to write down an well-formed strin ( you like, and then -fantasi<e- by askin , ->hat if strin ( were an a(iom, or a theorem?- )nd then, you let the system ive an answer& That is, you o ahead and make a derivation with ( as the o!enin line= let us su!!ose y is the last line& 4Ef course the derivation must strictly follow the rules of the system&5 Everythin from ( to y 4inclusive5 is the fantas)= ( is the premise of the fantasy, and y is its outcome& The ne(t ste! is to Mump out of the fantas), havin learned from it that If 9 were a theorem, ) would be a theorem& Still, you mi ht wonder, where is the real theorem? The real theorem is the strin

g9)a 8otice the resemblance of this strin to the sentence !rinted above To si nal the entry into, and emer ence from, a fantasy, one uses the square brackets 1U1 and 1V1, res!ectively& Thus, whenever you see a left square bracket, you know you are -!ushin - into a fantasy, and the ne9t line will contain the fantasy1s premise& 3henever you see a ri ht square bracket, you know you are -!o!!in - back out, and the preceding line was the outcome& It is hel!ful 4thou h not necessary5 to indent those lines of a derivation which take !lace in fantasies& ?ere is an illustration of the fantasy rule, in which the strin , is taken as a !remise& 4It so ha!!ens that , is not a theorem, but that is of no im!ort= we are merely inquirin , -3hat if it were?-5 3e make the followin fantasy, U , hhh, V The fantasy shows that, If , were a theorem, so would hh, be one& 3e now -squee<e- this sentence of En lish 4the metalan ua e5 into the formal notation 4the ob.ect lan ua e5, g,hh,a& This, our first theorem of the Pro!ositional 7alculus, should reveal to you the intended inter!retation of the symbol 11& ?ere is another derivation usin the fantasy rule, U g,8a , 8 g8,a !ush !remise se!aration se!aration .oinin !o! fantasy rule enuine theorem, here !ush into fantasy !remise outcome 4by double tilde rule5 !o! out of fantasy

V gg,8ag8,aa

It is im!ortant to understand that only the last line is a everythin else is in the fantasy&

$ecursion and the +antasy $ule

)s you mi ht uess from the recursion terminolo y -!ush- and -!o!-, the fantasy rule can be used recursivelythus, there can be fantasies within fantasies, thrice-nested

fantasies, and so on& This means that there are all sorts of -levels of reality-, .ust as in nested stories or movies& 3hen you !o! out of a movie-within-a-movie, you feel for a moment as if you had reached the real world, thou h you are still one level away from the to!& Similarly, when you !o! out of a fantasy-within-a-fantasy, you are in a -realer- world than you had been, but you are still one level away from the to!& 8ow a -8o Smokin - si n inside a movie theater does not a!!ly to the characters in the moviethere is no carry-over from the real world into the fantasy world, in movies& 0ut in the Pro!ositional 7alculus, there is a carry-over from the real world into the fantasies= there is even carry-over from a fantasy to fantasies inside it& This is formali<ed by the followin rule, 7);;H-EGE; ;FAE, Inside a fantasy, any theorem from the -reality level hi her can be brou ht in and used& It is as if a -8o Smokin - si n in a theater a!!lied not only to a movie oers, but also to all the actors in the movie, and, by re!etition of the same idea, to anyone inside multi!ly nested movies2 43arnin , There is no carry-over in the reverse direction, theorems inside fantasies cannot be e(!orted to the e(terior2 If it weren1t for this fact, you could write anythin as the first line of a fantasy, and then lift it out into the real world as a theorem&5 To show how carry-over works, and to show how the fantasy rule can be used recursively, we !resent the followin derivation, U , U 8 , g,8a V g8g,8aa V g,g8g,8aaa !ush !remise of outer fantasy !ush a ain !remise of inner fantasy carry-over of , into inner fantasy .oinin !o! out of inner fantasy, re ain outer fantasy fantasy rule !o! out of outer fantasy, reach real world2 fantasy rule

8ote that I1ve indented the outer fantasy once, and the inner fantasy twice, to em!hasi<e the nature of these nested -levels of reality-& Ene to look at the fantasy rule is to say that an observation made a(out the system is inserted into the system& 8amely, the theorem g9)a which ets !roduced can be thou ht of as a re!resentation inside the system of the statement about the system -If 9 is a theorem, then ) is too-& To be s!ecific, the intended inter!retation for g,8a is -if ,, then 8 equivalently, -, im!lies 8-&

The Con"erse of the +antasy $ule

8ow Aewis 7arroll1s :ialo ue was all about -if-then- statements& In !articular, )chilles had a lot of trouble in !ersuadin the Tortoise to acce!t the second clause of an -if-thenstatement, even when the -if-then- state itself was acce!ted, as well as its first clause& The ne(t rule allows you to infer the second -clause- of a 11-strin , !rovided that the 11strin it a theorem, and that its first -clause- is also a theorem& ;FAE E' :ET)7?/E8T, If 9 and g9)a are both theorems, then ) is a theorem& Incidentally, this rule is often called -/odus Ponens-, and the fantasy rule is often called the -:eduction Theorem-&

The Intended Inter!retation of the Symbols

3e mi ht as well let the cat out of the ba at this !oint, and reveal the -meanin s- of the rest of the symbols of our new system& In case it is not yet a!!arent, the symbol 1)1 is meant to be actin isomor!hically to the normal, everyday word 1and1& The symbol 1h1 re!resents the word 1not1it is a formal sort of ne ation& The an le brackets 1g1 and 1a1 are rou!erstheir function bein very similar to that of !arentheses in ordinary al ebra& The main difference is that in al ebra, you have the freedom to insert !arentheses or to leave them out, accordin to taste and style, whereas in a formal system, such anarchic freedom is not tolerated& The symbol 1 1 re!resents the word 1or1 41vel1 is a Aatin word for 1or15& The 1or1 that is meant is the so-called inclusi#e 1or1, which means that the inter!retation of g9)a is -either 9 or )Oor both-& The only symbols we have not inter!reted are the atoms& )n atom has no sin le inter!retationit may be inter!reted by any sentence of En lish 4it must continue to be inter!reted by the same sentence if it occurs multi!ly within a strin or derivation5& Thus, for e(am!le, the well-formed strin g,h,a could be inter!reted by the com!ound sentence This mind is 0uddha, and this mind is not 0uddha& 8ow let us look at each of the theorems so far derived, and inter!ret them& The first one was g,hh,a& If we kee! the same inter!retation for ,, we have the followin inter!retation, If this mind is 0uddha, then it is not the case that this mind is not 0uddha& 8ote how I rendered the double ne ation& It is awkward to re!eat a ne ation in any natural lan ua e, so one ets around it by usin two different ways of e(!ressin

ne ation& The second theorem we derived was gg ,8ag8,aa& If we let 8 be inter!reted by the sentence -This fla( wei hs three !ounds-, then our theorem reads as follows, If this mind is 0uddha and this fla( wei hs three !ounds, then this fla( wei hs three !ounds and this mind is 0uddha& The third theorem was g,g8g,8aaa& This one oes into the followin nested -ifthen- sentence, If this mind is 0uddha, then, if this fla( wei hs three !ounds, then this mind is 0uddha and this fla( wei hs three !ounds& Hou !robably have noticed that each theorem, when inter!reted, says somethin absolutely trivial and self-evident& 4Sometimes they are so self-evident that they sound vacuous and!arado(ically enou hconfusin or even wron 25 This may not be very im!ressive, but .ust remember there are !lenty of falsities out there which could have been !roducedyet they weren1t& This systemthe Pro!ositional 7alculusste!s neatly from truth to truth, carefully avoidin all falsities, .ust as a !erson who is concerned with stayin dry will ste! carefully from one ste!!in -stone creek to the ne(t, followin the layout of ste!!in -stones no matter I twisted and tricky it mi ht be& 3hat is im!ressive is thatin the Pro!ositional 7alculusthe whole thin is done !urely t)pographicall)& There is nobody down -in there-, thinkin about the meaning of the strin s& It is all done mechanically, thou htlessly, ri idly, even stu!idly&

$ounding 0ut the List of $ules

3e have not yet stated all the rules of the Pro!ositional 7alculus& The com!lete set of rules is listed below, includin the three new ones& JEI8I86 ;FAE, If 9 and ) are theorems, then g9)a is a theorem& SEP);)TIE8 ;FAE, If g9)a is a theorem, then both 9 and ) are theorems& :EF0AE-TIA:E ;FAE, The strin 1hh1 can be deleted from any theorem& It can also be inserted into any theorem, !rovided that the result strin is itself well-formed& ')8T)SH ;FAE, If ) can be derived when 9 is assumed to be a theorem then g 9)a is a theorem& 7);;H-EGE; ;FAE, Inside a fantasy, any theorem from the -reality- one level hi her can be brou ht in and used& ;FAE E' :ET)7?/E8T, If 9 and g9)a are both theorems, then ) is a theorem&

7E8T;)PESITIGE ;FAE, g9)a and gh)h9a are interchan eable :E /E;6)81S ;FAE, gh9h)a and hg 9)a are interchan eable& S3IT7?E;EE ;FAE, g9)a and gh9)a are interchan eable& 4The Switcheroo rule is named after M& q& Switcheroo, an )lbanian railroad en ineer who worked in lo ic on the sidin &5 0y -interchan eable- in fore oin rules, the followin is meant, If an e(!ression of one form occurs as either a theorem or !art of a theorem, the other form may be substituted, and the resultin strin will also be a theorem& It must be ke!t in mind that the symbols _(1 and _y1 always stand for well-formed strin s of the system&

3ustifying the $ules

0efore we see these rules used inside derivations, let us look at some very short .ustifications for them& Hou can !robably .ustify them to yourself better than my e(am!leswhich is why I only ive a cou!le& The contra!ositive rule e(!resses e(!licitly a way of turnin around conditional statements which we carry out unconsciously& 'or instance, the -9entenceIf you are studyin it, then you are far from the 3ay /eans the same thin as If you are close to the 3ay, then you are not studyin it& :e /or an1s rule can be illustrated by our familiar sentence -The fla is not movin and the wind is not movin -& If , symboli<es -the fla is movin -, and 8 symboli<es -the wind is movin -, then the com!ound sentence is symboli<ed by gh,h8a, which, accordin to /or an1s law, is interchan eable with hg ,8a& whose inter!retation would be -It is not true that either the fla or the wind is movin -& )nd no one could deny that it is a 9ensible conclusion to draw& 'or the Switcheroo rule, consider the sentence -Either a cloud is han in over the mountain, or the moonli ht is !enetratin the waves of the lake,- which mi ht be s!oken, I su!!ose, by a wistful 9en master rememberin a familiar lake which he can visuali<e mentally but cannot see& 8ow han on to your seat, for the Switcheroo rule tells us that this is interchan eable with the thou ht -If a cloud is not han in over the mountain, then the moonli ht is !enetratin the waves of the lake&- This may not be enli htenment, but it is the best the Pro!ositional 7alculus has to offer&

,laying Around 'ith the System

8ow, let us a!!ly these rules to a !revious theorem, and see what we et, 'or instance, take the theorem g,hh,a, g,hh,a, ghhh,h,a, gh,h,a g,h,a old theorem contra!ositive double-tilde switcheroo

This new theorem, when inter!reted, says, Either this mind is 0uddha, or this mind is not 0uddha Ence a ain, the inter!reted theorem, thou h !erha!s less than mind bo true& lin , is at least

Semi-Inter!retations
It is natural, when one reads theorems of the Pro!ositional 7alculus out loud, to inter!ret everythin but the atoms& I call this semi-inter!retin & 'or e(am!le, the semiinter!retation of <,~,> would be , or not ,& :es!ite the fact that , is not a sentence, the above semisentence still sounds true, because you can very easily ima ine stickin any sentence in for ,and the form of the semiinter!reted theorem assures you that however you make your choice, the resultin sentence will be true& )nd that is the key idea of the Pro!ositional 7alculus, it !roduces theorems which, when semi-inter!reted, are seen to be -universally true semisentences-, by which is meant that no matter how you com!lete the inter!retation, the final result will be a true statement&

Gant%s A8ow we can do a more advanced e(ercise, based on a 9en kRan called -6antR1s )(-& ?ere is how it be an&
Ene day Tokusan told his student 6antR, -I have two monks who have been here for many years& 6o and e(amine them&- 6antR !icked u! an a( and went to the hut where the two monks were meditatin & ?e raised the a(, sayin -If you say a word, I will cut off your heads= and if you do not say a word, I will also cut off your heads&-"

If you say a word I will cut off this kRan, and if you do not say a word, I will also cut off this kRanbecause I want you to translate some of it into our notation& Aet us symboli<e -you say a word- by , and -I will cut off your heads- by 8& Then 6antR1s a( threat is symboli<ed by the strin gg,8agh,8aa& 3hat if this a( threat were an a(iom? ?ere is a fantasy to answer that question& 4"5 U 4B5 4%5 4C5 4L5 4D5 4$5 4*5 4#5 4"+5 4""5 4"B5 4"%5 4"C5 4"L5 4"D5 4"$5 4"*5 4"#5 4B+5 4B"5 4BB5 4B%5 4BC5 gg,8agh,8aa g,8a gh8h,a gh,8a gh8hh,a V h8 gh8h,a h, gh8hh,a hh, gh,hh,a gh,hh,a V gh8hg,h,aa gh,h,a8a U h, V gh,h,a g,h,a 8 V !ush 6antR1s a(iom se!aration contra!ositive se!aration contra!ositive !ush a ain !remise carry-over of line C detachment carry-over of line D detachment 4lines * and ""5 .oinin :e /or an !o! once fantasy rule contra!ositive !ush !remise 4also outcome5 !o! fantasy rule switcheroo detachment 4lines BB and "$5 !o! out

The !ower of the Pro!ositional 7alculus is shown in this e(am!le& 3hy, in but two do<en ste!s, we have deduced M, that the heads will be cut off2 4Eminously, the rule last invoked was -detachment-&&&5 It mi ht seem su!erfluous to continue the kRan now, since we know what must ensue&&& ?owever, I shall dro! my resolve to cut the kRan off= it is a true 9en kRan, after all& The rest of the incident is here related,
0oth monks continued their meditation as if he had not s!oken& 6antR dro!!ed the a( and said, -Hou are true 9en students&- ?e returned to Tokusan and related the incident& -I see your side well,- Tokusan a reed, -but tell me, how is their side?- -TR<an may admit them,- re!lied 6antR, -but they should not be admitted under Tokusan&-B

:o you see my side well? ?ow is the 9en side?

Is There a

ecision ,rocedure for Theorems7

The Pro!ositional 7alculus ives us a set of rules for !roducin statements which would be true in all conceivable worlds& That is why all of its theorems sound so sim!le-minded= it seems that they have absolutely no content2 Aooked at this way, the Pro!ositional 7alculus mi ht seem to be a waste of time, since what it tells us is absolutely trivial& En the other hand, it does it by s!ecifyin the form of statements that are universally true, and this throws a new kind of li ht onto the core truths of the universe, they are not only fundamental, but also regular, they can be !roduced by one set of ty!o ra!hical rules& To !ut it another way, they are all -cut from the same cloth-& Hou mi ht consider whether the same could be said about 9en kRans, could they all be !roduced by one set of ty!o ra!hical rules? It is quite relevant here to brin u! the question of a decision !rocedure& That is, does there e(ist any mechanical method to tell nontheorems from theorems? If so, that would tell us that the set of theorems of the Pro!ositional 7alculus is not only r&e&, but also recursive& It turns out that there is an interestin decision !rocedurethe method of truth tables& It would take us a bit afield to !resent it here= you can find it in almost any standard book on lo ic& )nd what about 9en kRans? 7ould there conceivably be a mechanical decision !rocedure which distin uishes enuine 9en kRans from other thin s?

o >e Kno' the System Is Consistent7

F! till now, we have only presumed that all theorems, when inter!reted as indicated, are true statements& 0ut do we know that that is the case? 7ould we !rove it to be? This is .ust another way of askin whether the intended inter!retations 41and1 for 1 1, etc&5 merit bein called the -!assive meanin s- of the symbols& Ene can look at this issue from two very different !oints of view, which mi ht be called the -!rudent- and -im!rudent- !oints I will now !resent those two sides as I see them, !ersonifyin their holders as -Prudenceand -Im!rudence-& &rudence, 3e will only @8E3 that all theorems come out true under the intended inter!retation if we mana e to P;EGE it& That is the cautious, thou htful way to !roceed& 3mprudence, En the contrary& It is E0GIEFS that all theorems will come out true& If you doubt me, look a ain at the rules of the system& Hou will find that each rule makes a symbol act e(actly as the word it re!resents ou ht to be used& 'or instance, the .oinin rule makes the symbol _1 act as 1and1 ou ht to act= the rule of detachment makes 1 1 act as it ou ht to, if it is to stand for 1im!lies1, or 1if-then1= and so on& Fnless you are like the Tortoise, you will reco ni<e in each rule a codification of a !attern you use in your own thou ht !atterns& So if you trust your own thou ht !atterns, then you ?)GE to believe that all theorems come out true2 That1s the way I see it& I don1t need any

further !roof& If you think that some theorem comes out false, then !resumably you think that some rule must be wron & Show me which one& &rudence, I1m not sure that there is any faulty rule, so I can1t !oint one out to you& Still, I can ima ine the followin kind of scenario& Hou, followin the rules, come u! with a theoremsay 9& /eanwhile I, also followin the rules, come u! with another theorem it ha!!ens to be h9& 7an1t you force yourself to conceive of that? 3mprudence, )ll ri ht= let1s su!!ose it ha!!ened& 3hy would it bother you? Er let me !ut it another way& Su!!ose that in !layin with the /IF-system, I came u! with a theorem (, and you came u! with 9& 7an you force yourself to conceive of that? &rudence, Ef coursein fact both MI and MI& are theorems& 3mprudence, :oesn1t that bother you? &rudence, Ef course not& Hour e(am!le is ridiculous, because MI and MI& are 8ET contradictory, whereas two strin s 9 and h9 in the Pro!ositional 7alculus );E contradictory& 3mprudence, 3ell, yes!rovided you wish to inter!ret 1h1 as 1not1& 0ut what would lead you to think that 1h1 should be inter!reted as 1not1? &rudence, The rules themselves& 3hen you look at them, you reali<e that the only conceivable inter!retation for 1h1 is 1not1and likewise, the only conceivable inter!retation for 11 is 1and1, etc& 3mprudence, In other words, you are convinced that the rules ca!ture the meanin s of those words? &rudence, Precisely& 3mprudence, )nd yet you are still willin to entertain the thou ht that both 9 and h9 could be theorems? 3hy not also entertain the notion that hed eho s are fro s, or that " equals B, or that the moon is made of reen cheese? I for one am not !re!ared even to consider whether such basic in redients of my thou ht !rocesses are wron because if I entertained that notion, then I would also have to consider whether my modes of analy<in the entire question are also wron , and I would wind u! in a total tan le& &rudence, Hour ar uments are forceful&&& Het I would still like to see a P;EE' that all theorems come out true, or that 9 and h9 can never both be theorems& 3mprudence, Hou want a !roof& I uess that means that you want to be more convinced that the Pro!ositional 7alculus is consistent than you are convinced of your own sanity& )ny !roof I could think of would involve mental o!erations of a reater com!le(ity than anythin in the Pro!ositional 7alculus itself& So what would it !rove? Hour desire for a !roof of consistency of the Pro!ositional 7alculus makes me think of someone who is learnin En lish and insists on bein iven a dictionary which defines all the sim!le words in terms of com!licated ones&&&

The Carroll

ialogue Again

This little debate shows the difficulty of tryin to use lo ic and reasonin to defend themselves& )t some !oint, you reach rock bottom, and there is no defense e(ce!t loudly

shoutin , -I know I1m ri ht2- Ence a ain, we are u! a ainst the issue which Aewis 7arroll so shar!ly set forth in his :ialo ue, you can1t o on defendin your !atterns of reasonin forever& There comes a !oint where faith takes over& ) system of reasonin can be com!ared to an e & )n e has a shell which !rotects its insides& If you want to shi! an e somewhere, thou h, you don1t rely on the shell& Hou !ack the e in some sort of container, chosen accordin to how rou h you e(!ect the e 1s voya e to be& To be e(tra careful, you may !ut the e inside several nested bo(es& ?owever, no matter how many layers of bo(es you !ack your e in, you can ima ine some cataclysm which could break the e & 0ut that doesn1t mean that you1ll never risk trans!ortin your e & Similarly, one can never ive an ultimate, absolute !roof that a !roof in some system is correct& Ef course, one can ive a !roof of a !roof, or a !roof of a !roof of a !roofbut the validity of the outermost system always remains an un!roven assum!tion, acce!ted on faith& Ene can always ima ine that some unsus!ected subtlety will invalidate every sin le level of !roof down to the bottom, and the -!rovenresult will be seen not to be correct after all& 0ut that doesn1t mean that mathematicians and lo icians are constantly worryin that the whole edifice of mathematics mi ht be wron & En the other hand, when unorthodo( !roofs are !ro!osed, or e(tremely len thy !roofs, or !roofs enerated by com!uters, then !eo!le do sto! to think a bit about what they really mean by that quasi-sacred word -!roven-& )n e(cellent e(ercise for you at this !oint would be to o back 7arroll :ialo ue, and code the various sta es of the debate into our notationbe innin with the ori inal bone of contention, )chilles, If you have ggABa2a, and you also have gABa, then surely you have 2& Tortoise, Eh2 Hou mean, ggggABa2agABaa2a, don1t you?

4?int, 3hatever )chilles considers a rule of inference, the Tortoise immediately flattens into a mere strin of the system& If you use or letters A, B, and 2, you will et a recursive !attern of lon er and strin s&5

Shortcuts and

eri"ed $ules

3hen carryin out derivations in the Pro!ositional 7alculus, one quickly invents various ty!es of shortcut, which are not strictly !art of the system 'or instance, if the strin g8h8a were needed at some !oint, and g,h,a had been derived earlier, many !eo!le would !roceed as if g8h8a had been derived, since they know that its derivation is an e(act !arallel to that of g,h,a& The derived theorem is treated as a -theorem schema- a mold for other theorems& This turns out to be a !erfect valid !rocedure, in that it always leads you to new theorems, but it is not a rule of the Pro!ositional 7alculus as we !resented it& It is, rather, a deri#ed rule, It is !art of the knowled e which we have a(out the system& That this rule kee!s you within the s!ace of theorems needs !roof, of course but such a !roof is not like a derivation inside the system& It is a !roof in the ordinary,

intuitive sensea chain of reasonin carried out in the I-mode& The theory about the Pro!ositional 7alculus is a -metatheory-, and results in it can be called -metatheorems-Theorems about theorems& 4Incidentally, note the !eculiar ca!itali<ation in the !hrase -Theorems about theorems-& It is a consequence of our convention, metatheorems are Theorems 4!roven results5 concernin theorems 4derivable strin s5&5 In the Pro!ositional 7alculus, one could discover many metatheorems, or derived rules of inference& 'or instance, there is a :e /or an1s ;ule, gh9h)a and hg9)a are interchan eable& If this were a rule of the system, it could s!eed u! many derivations considerably& 0ut if we pro#e that it is correct, isn1t that ood enou h? 7an1t we use it .ust like a rule of inference, from then on? There is no reason to doubt the correctness of this !articular derived rule& 0ut once you start admittin derived rules as !art of your !rocedure in the Pro!ositional 7alculus, you have lost the formality of the system, since derived rules are derived informallyoutside the system& 8ow formal systems were !ro!osed as a way to e(hibit every ste! of a !roof e(!licitly, within one sin le, ri id framework, so that any mathematician could check another1s work mechanically& 0ut if you are willin to ste! outside of that framework at the dro! of a hat, you mi ht as well never have created it at all& Therefore, there is a drawback to usin such shortcuts&

+ormali:ing /igher Le"els

En the other hand, there is an alternative way out& 3hy not formali<e the metatheory, too? That way, derived rules 4metatheorems5 would be theorems of a lar er formal system, and it would be le itimate to look for shortcuts and derive them as theorems-that is, theorems of the formali<ed metatheorywhich could then be used to s!eed u! the derivations of theorems of the Pro!ositional 7alculus& This is an interestin idea, but as soon as it is su ested, one .um!s ahead to think of metametatheories, and so on& It is clear that no matter how many levels you formali<e, someone will eventually want to make shortcuts in the to! level& It mi ht even be su ested that a theory of reasonin could be identical to its own metatheory, if it were worked out carefully& Then, it mi ht seem, all levels would colla!se into one, and thinkin a(out the system would be .ust one way of workin in the system2 0ut it is not that easy& Even if a system can -think about itself-, it still is not outside itself& Hou, outside the system, !erceive it differently from the way it !erceives itself& So there still is a metatheorya view from outsideeven for a theory which can -think about itself- inside itself& 3e will find that there are theories which can -think about themselves-& In fact, we will soon see a system in which this ha!!ens com!letely accidentally, without our even intendin it2 )nd we will see what kinds of effects this !roduces& 0ut for our study of the Pro!ositional 7alculus, we will stick with the sim!lest ideasno mi(in of levels&

'allacies can result if you fail to distin uish carefully between workin in the system 4the /-mode5 and thinkin about the system 4the I-mode5& 'or e(am!le, it mi ht seem !erfectly reasonable to assume that, since g,h,a 4whose semi-inter!retation is -either , or not ,-5 is a theorem, either , or h, must be a theorem& 0ut this is dead wron , neither one of the latter !air is a theorem& In eneral, it is a dan erous !ractice to assume that symbols can be sli!!ed back and forth between different levelshere, the lan ua e of the formal system and its metalan ua e 4En lish5&

$eflections on the Strengths and >ea.nesses of the System

Hou have now seen one e(am!le of a system with a !ur!oseto re!resent !art of the architecture of lo ical thou ht& The conce!ts which this handles are very few in number, and they are very sim!le, !recise conce!ts& 0ut the sim!licity and !recision of the Pro!ositional 7alculus are e(actly the kinds of features which make it a!!ealin to mathematicians& There are two reasons for this& 4"5 It can be studied for its own !ro!erties, e(actly as eometry studies sim!le, ri id sha!es& Gariants can be made on it, em!loyin different symbols, rules of inference, a(ioms or a(iom schemata and so on& 4Incidentally, the version of the Pro!ositional 7alculus here !resented is related to one invented by 6& 6ent<en in the early "#%+1s& The other versions in which only one rule of inference is useddetachment usuallyand in which there are several a(ioms, or a(iom schemata&5 The study of ways to carry out !ro!ositional reasonin in ele ant formal systems is an a!!ealin branch of !ure mathematics& 4B5 The Pro!ositional 7alculus can easily be e(tended to include other fundamental as!ects of reasonin & Some of this will be shown in the ne(t 7ha!ter, where the Pro!ositional 7alculus is incor!orated lock, stock and barrel into a much lar er and dee!er system in which so!histicated numbertheoretical reasonin can be done&

,roofs #s! eri"ations

The Pro!ositional 7alculus is very much like reasonin in some ways, but one should not equate its rules with the rules of human thou ht& ) proof is somethin informal, or in other words a !roduct of normal thou ht, written in a human lan ua e, for human consum!tion& )ll sorts of com!le( features of thou ht may be used in !roofs, and, thou h they may -feel ri ht-, one may wonder if they can be defended lo ically& That is really what formali<ation is for& ) deri#ation is an artificial counter!art of a !roof, and its !ur!ose is to reach the same oal but via a lo ical structure whose methods are not only all e(!licit, but also very sim!le& Ifand this is usually the caseit ha!!ens that a formal derivation is e(tremely len thy com!ared with the corres!ondin -natural- !roof, that is .ust too bad& It is the !rice one !ays for makin each ste! so sim!le& 3hat often ha!!ens is that a derivation and a !roof are -sim!le- in com!lementary senses of the word& The !roof is sim!le in that each ste! sounds ri ht-, even thou h one may not know .ust why= the derivation is sim!le in that each of its myriad ste!s is considered so trivial that it is beyond re!roach, and

since the whole derivation consists .ust of such trivial ste!s it is su!!osedly error-free& Each ty!e of sim!licity, however, brin s alon a characteristic ty!e of com!le(ity& In the case of !roofs, it is the com!le(ity of the underlyin system on which they restnamely, human lan ua eand in the case of derivations, it is their astronomical si<e, which makes them almost im!ossible to ras!& Thus, the Pro!ositional 7alculus should be thou ht of as !art of a eneral method for synthesi<in artificial !roof-like structures& It does not, however, have much fle(ibility or enerality& It is intended only for use in connection with mathematical conce!tswhich are themselves quite ri id& )s a rather interestin e(am!le of this, let us make a derivation in which a very !eculiar strin is taken as a !remise in a fantasy, g,h,a& )t least its semi-inter!retation is !eculiar& The Pro!ositional 7alculus, however, does not think about semi-inter!retations= it .ust mani!ulates strin s ty!o ra!hically and ty!o ra!hically, there is really nothin !eculiar about this strin & ?ere is a fantasy with this strin as its !remise, 4"5 4B5 4%5 4C5 4L5 4D5 4$5 4*5 4#5 4"+5 4""5 4"B5 4"%5 4"C5 U g,h,a , h, U h8 , hh, V gh8hh,a gh,8a 8 !ush !remise se!aration se!aration !ush !remise carry-over line % double-tilde !o! fantasy contra!ositive detachment 4Aines C,""5 !o! fantasy

V gg,h, a8a

8ow this theorem has a very stran e semi-inter!retation, , and not , to ether im!ly 8 Since 8 is inter!retable by any statement, we can loosely take the theorem to say that -'rom a contradiction, anythin follows-2 Thus, in systems based on the Pro!ositional 7alculus, contradictions cannot be contained= they infect the whole system like an instantaneous lobal cancer&

The /andling of Contradictions

This does not sound much like human thou ht& If you found a contradiction in your own thou hts, it1s very unlikely that your whole mentality would break down& Instead, you would !robably be in to question the beliefs or modes of reasonin which you felt had led to the contradictory thou hts& In other words, to the e(tent you could, you would ste! out of the systems inside you which you felt were res!onsible for the contradiction, and try to re!air them& Ene of the least likely thin s for you to do would be to throw u! your arms and cry, -3ell, I uess that shows that I believe everythin now2- )s a .oke, yes but not seriously& Indeed, contradiction is a ma.or source of clarification and !ro ress in all domains of lifeand mathematics is no e(ce!tion& 3hen in times !ast, a contradiction in mathematics was found, mathematicians would immediately seek to !in!oint the system res!onsible for it, to .um! out of it, to reason about it, and to amend it& ;ather than weakenin mathematics, the discovery and re!air of a contradiction would stren then it& This mi ht take time and a number of false starts, but in the end it would yield fruit& 'or instance, in the /iddle ) es, the value of the infinite series " Y " ^ " Y " ^ " -&&& was hotly dis!uted& It was -!roven- to equal +, ", i, and !erha!s other values& Eut of such controversial findin s came a fuller, dee!er theory about infinite series& ) more relevant e(am!le is the contradiction ri ht now confrontin usnamely the discre!ancy between the way we really think, and the way the Pro!ositional 7alculus imitates us& This has been a source of discomfort for many lo icians, and much creative effort has one into tryin to !atch u! the Pro!ositional 7alculus so that it would not act so stu!idly and infle(ibly& Ene attem!t, !ut forth in the book Entailment by )& ;& )nderson and 8& 0elna!,% involves -relevant im!lication-, which tries to make the symbol for -if-then- reflect enuine causality, or at least connect meanin s& 7onsider the followin theorems of the Pro!ositional 7alculus, g,g8,aa g,g8h,aa gg,h,a8a gg,8ag8,aa They, and many others like them, all show that there need be no relationshi! at all between the first and second clauses of an if-then statement for it to be !rovable within the Pro!ositional 7alculus& In !rotest, -relevant im!lication- !uts certain restrictions on the conte(ts in which the rules of inference can be a!!lied& Intuitively, it says that -somethin can only be derived from somethin else if they have to do with each other-& 'or e(am!le, line "+ in the derivation iven above would not be allowed in such a system, and that would block the derivation of the gg,h, a8a

/ore radical attem!ts abandon com!letely the quest for com!leteness or consistency, and try to mimic human reasonin with all its inconsistencies& Such research no lon er has as its oal to !rovide a solid under!innin for mathematics, but !urely to study human thou ht !rocesses& :es!ite its quirks, the Pro!ositional 7alculus has some features to recommend itself& If one embeds it into a lar er system 4as we will do ne(t 7ha!ter5, and if one is sure that the lar er system contains no contradictions 4and we will be5, then the Pro!ositional 7alculus does all that one could ho!e, it !rovides valid !ro!ositional inferencesall that can be made& So if ever an incom!leteness or an inconsistency is uncovered, one can be sure that it will be the fault of the lar er system, and not of its subsystem which is the Pro!ositional 7alculus&

73G5=E DF& -7rab 7anon-, () '! 8! Escher XBIJG/

Crab Canon
*chilles and the Tortoise happen upon each other in the park one da) while strolling! Tortoise, 6ood day, /r& )& *chilles, 3hy, same to you& Tortoise, So nice to run into you& *chilles, That echoes my thou hts& Tortoise, )nd it1s a !erfect day for a walk& I think I1ll be walkin home soon& *chilles, Eh, really? I uess there1s nothin better for you than walkin & Tortoise, Incidentally, you1re lookin in very fine fettle these days, I must say& *chilles, Thank you very much& Tortoise, 8ot at all& ?ere, care for one of my ci ars? *chilles, Eh, you are such a !hilistine& In this area, the :utch contributions are of markedly inferior taste, don1t you think? Tortoise, I disa ree, in this case& 0ut s!eakin of taste, I finally saw that 7rab 7anon by your favorite artist, /& 7& Escher, in a allery the other day, and I fully a!!reciate the beauty and in enuity with which he made one sin le theme mesh with itself oin both backwards and forwards& 0ut I am afraid I will always feel 0ach is su!erior to Escher& *chilles, I don1t know& 0ut one thin for certain is that I don1t worry about ar uments of taste& .e gusti(us non est disputandum& Tortoise, Tell me, what1s it like to be your a e? Is it true that one has no worries at all? *chilles, To be !recise, one has no frets& Tortoise, Eh, well, it1s all the same to me& *chilles, 'iddle& It makes a bi difference, you know& Tortoise, Say, don1t you !lay the uitar? *chilles, That1s my ood friend& ?e often !lays, the fool& 0ut I myself wouldn1t touch a uitar with a ten-foot !ole2 -uddenl), the 8ra(, appearing from out of nowhere, wanders up e9citedl), pointing to a rather prominent (lack e)e!/ 8ra(, ?allo2 ?ulloo2 3hat1s u!? 3hat1s new? Hou see this bum!, this lum!? 6iven to me by a rum!& ?o2 )nd on such a fine day& Hou see, I was .ust idly loafin about the !ark when u! lumbers this iant fellow from 3arsawa colossal bear of a man !layin a lute& ?e was three meters tall, if I1m a day& I mosey on u! to the cha!, reach skyward and mana e to ta! him on the knee, sayin , -Pardon me, sir, but you are Pole-lutin our !ark with your ma<urkas&- 0ut wow2 he had no sense of humornot a bit, not a witand PE32he lets loose and belts me one, smack in the eye2 3ere it in my nature, I would crab u! a storm, but in the time-honored tradition of my s!ecies, I backed off& )fter all, when we walk forwards, we move backwards& It1s in our enes, you know, turnin round and round& That reminds meI1ve always

wondered, -3hich came firstthe 7rab, or the 6ene?- That is to say, -3hich came last the 6ene, or the 7rab?- I1m always turnin thin s round and round, you know& It1s in our enes, after all& 3hen we walk backwards, we move forwards& )h me, oh my2 I must lo!e alon on my merry wayso off I o on such a fine day& Sin -ho2- for the life of a 7rab2 T)T)2 .Ele2 *nd he disappears as suddenl) as he arri#ed!/ Tortoise, That1s my ood friend& ?e often !lays the fool& 0ut I myself wouldn1t touch a ten-foot Pole with a uitar2 *chilles, Say, don1t you !lay the uitar? Tortoise, 'iddle& It makes a bi difference, you know& *chilles, Eh, well, it1s all the same to me& Tortoise, To be !recise, one has no frets& *chilles, Tell me, what1s it like to be your a e? Is it true that one has no worries at all? Tortoise, I don1t know& 0ut one thin for certain is that I don1t worry about ar uments of taste& .isputandum non est de gusti(us!

73G5=E DH! ,ere is a short section one of the 8ra('s Genes, turning round and round! >hen the two .N* strands are unra#eled and laid out side () side, the) read this wa): &&& TTTTTTTTT76)))))))))&&& &&& )))))))67TTTTTTTTTT&&& Notice that the) are the same, onl) one goes forwards while the other goes (ackwards This is the defining propert) of the form called "cra( canon" in music! 3t is reminiscent of, though a little different from, a palindrome, which is a sentence that reads the same (ackwards and forwards! 3n molecular (iolog), such segments of .N* are called "palindromes "Oa slight misnomer, since "cra( canon" would (e more accurate! Not onl) is this .N* segment cra(%canonicalO(ut moreo#er its (ase se"uence codes for the .ialogue's structure! +ook carefull)A

*chilles, I disa ree, in this case& 0ut s!eakin of taste, I finally heard that 8ra( 8anon by your favorite com!oser, J& S& 0ach, in a concert other day, and I fully a!!reciate the beauty and in enuity with which he made one sin le theme mesh with itself oin both backwards and forwards& 0ut I1m afraid I will always feel Escher is su!erior to 0ach& Tortoise, Eh, you are such a !hilistine& In this area, the :utch contributions are of markedly inferior taste, don1t you think? *chilles, 8ot at all& ?ere, care for one of my ci ars? Tortoise, Thank you very much& *chilles, Incidentally, you1re lookin in very fine fettle these days, I must say&

73G5=E DH! 7rab 7anon from the /usical Efferin () ?! -! Bach! U'usic printed () .onald B)rd's program "-'5T"Y

Tortoise, Eh, really? I uess there1s nothin better for you than walkin & *chilles, )nd it1s a !erfect day for a walk& I think I1ll be walkin home soon& Tortoise, That echoes my thou hts& *chilles, So nice to run into you& Tortoise, 3hy, same to you& *chilles, 6ood day, /r& T&

C/A,TE$ 4III

Ty!ogra!hical ;umber Theory

The 8ra( 8anon and Indirect Self-$eference
T?;EE ET)/PAES E' indirect self-reference are found in the 7rab 7anon& )chilles and the Tortoise both describe artistic creations they knowand, quite accidentally, those creations ha!!en to have the same structure as the :ialo ue they1re in& 4Ima ine my sur!rise, when I, the author, noticed this25 )lso, the 7rab describes a biolo ical structure and that, too, has the same !ro!erty& Ef course, one could read the :ialo ue and understand it and somehow fail to notice that it, too, has the form of a crab canon& This would be understandin it on one level, but not on another& To see the self-reference, one has to look at the form, as well as the content, of the :ialo ue& 6>del1s construction de!ends on describin the form, as well as the content, of strin s of the formal system we shall define in this 7ha!ter T)pographical Num(er Theor) 4T8T5& The une(!ected twist is that, because of the subtle ma!!in which 6>del discovered, the form of strin s can be described in the formal system itself& Aet us acquaint ourselves with this stran e system with the ca!acity for wra!!in around&

>hat >e >ant to Be Able to E-!ress in T;T

3e1ll be in by citin some ty!ical sentences belon in to number theory= then we will try to find a set of basic notions in terms of which all our sentences can be re!hrased& Those notions will then be iven individual symbols& Incidentally, it should be stated at the outset that the term -number theory- will refer only to !ro!erties of !ositive inte ers and <ero 4and sets of such inte ers5& These numbers are called the natural num(ers& 8e ative numbers !lay no role in this theory& Thus the word -number-, when used, will mean e(clusively a natural number& )nd it is im!ortantvitalfor you to kee! se!arate in your mind the formal system 4T8T5 and the rather ill-defined but comfortable old branch of mathematics that is number theory itself= this I shall call -8-& Some ty!ical sentences of 8number theoryare, 4"5 4B5 4%5 4C5 4L5 4D5 L is !rime& B is not a square& "$B# is a sum of two cubes& 8o sum of two !ositive cubes is itself a cube& There are infinitely many !rime numbers& D is even&

8ow it may seem that we will need a symbol for each notion such as -!rime- or -cubeor -!ositive-but those notions are really not !rimitive& Primeness, for instance, has to do with the factors which a number has, which in turn has to do with multi!lication& 7ubeness as well is defined in terms of multi!lication& Aet us re!hrase the sentences, then, in terms of what seem to be more elementary notions& 4"15 There do not e(ist numbers a and b, both reater than ", such that L equals a times b& 4B15 There does not e(ist a number b, such that b times b equals B& 4%15 There e(ist numbers b and c such that b times b times b, !lus c times c times c, equals "$B#& 4C15 'or all numbers b and c, reater than +, there is no number a such that a times a times a equals b times b times b !lus c times c times c& 4L15 'or each number a, there e(ists a number b, reater than a, with the !ro!erty that there do not e(ist numbers c and d, both reater than ", such that b equals c times d& 4D15 There e(ists a number e such that B times e equals D& This analysis has otten us a lon ways towards the basic elements of lan ua e of number theory& It is clear that a few !hrases rea!!ear over a over, for all numbers b there e(ists a number b, such that reater than equals times !lus +, ", B, & & /ost of these will be ranted individual symbols& )n e(ce!tion is - reater than-, which can be further reduced& In fact, the sentence -a is reater than b- becomes there e(ists a number c, not equal to +, such that a equals b !lus c&

;umerals
3e will not have a distinct symbol for each natural number& Instead, we will have a very sim!le, uniform way of ivin a com!ound symbol to each natural numbervery much as we did in the !q-system& ?ere is notation for natural numbers, <ero, one, two, three, etc& The symbol S has an inter!retation-the successor of-& ?ence, the inter!retation of SS? is literally -the successor of the successor of <ero-& Strin s of this form are called numerals! ? S? SS? SSS?

4ariables and Terms

7learly, we need a way of referrin to uns!ecified, or variable, numbers& 'or that, we will use the letters a, b, c, d, e& 0ut five will not be enou h& 3e need an unlimited su!!ly of them, .ust as we had of atoms in the Pro!ositional 7alculus& 3e will use a similar method for makin more variables, tackin on any number of !rimes& 48ote, Ef course the symbol - 1 -read -!rime-is not to be confused with !rime numbers25 'or instance, e d1 c11 b111 a1111 are all #aria(les& In a way it is a lu(ury to use the first five letters of the al!habet when we could et away with .ust a and the !rime& Aater on, I will actually dro! b, c, d, and e, which will result in a sort of -austere- version of T8Taustere in the sense that it is a little harder to deci!her com!le( formulas& 0ut for now we1ll be lu(urious& 8ow what about addition and multi!lication? Gery sim!le, we will use the ordinary symbols 1^1 and 1j1& ?owever, we will also introduce a !arenthesi<in requirement 4we are now slowly sli!!in into the rules which define well-formed strin s of T8T5& To write -b !lus c- and -b times c-, for instance, we use the strin s

4b^c5 4bjc5 There is no la(ness about such !arentheses= to violate the convention is to !roduce a nonwell-formed formula& 4-'ormula-? I use the term instead of -strin - because it is conventional to do so& ) formula is no more and no less than a strin of T8T&5 Incidentally, addition and multi!lication are always to be thou ht of as (inar) o!erations that is, they unite !recisely two numbers, never three or more& ?ence, if you wish to translate -" !lus B !lus %-, you have to decide which of the followin two e(!ressions you want, 4S?^4SS?^SSS?55 44S?^SS?5^SSS?5 The ne(t notion we1ll symboli<e is e"uals& That is very sim!le, we use 1 X 1& The advanta e of takin over the standard symbol used in 8nonformal number theoryis obvious, easy le ibility& The disadvanta e is very much like the disadvanta e of usin the words -!oint- and -line- in a formal treatment of eometry, unless one is very conscious and careful, one may blur the distinction between the familiar meanin and the strictly rule- overned behavior of the formal symbol& In discussin eometry, I distin uished between the everyday word and the formal term by ca!itali<in the formal term, thus, in elli!tical eometry, a PEI8T was the union of two ordinary !oints& ?ere, there is no such distinction= hence mental effort is needed not to confuse a symbol with all of the associations it is laden with& )s I said earlier, with reference to the !q-system, the strin --- is not the number %, but it acts isomor!hically to %, at least in the conte(t of additions& Similar remarks o for the strin SSS?&

Atoms and ,ro!ositional Symbols

)ll the symbols of the Pro!ositional 7alculus e(ce!t the letters used makin atoms 4 ,, 8, and $5 will be used in T8T, and they retain their inter!retations& The role of atoms will be !layed by strin s which, when inter!reted, are statements of equality, such as S?XSS? or 4S?jS?5 8ow, we have the equi!ment to do a fair amount of translation of sim!le sentences into the notation of T8T, B !lus % equals C, B !lus B is not equal to %, If " equals +, then + equals ", 4SS?^SSS?5XSSSS? h4SS?^SS?5XSSS? gS?X??XS?a

The first of these strin s is an atom= the rest are com!ound formulas 43arnin , The 1and1 in the !hrase -" and " make B- is .ust another word for 1!lus1, and must be re!resented by 1^1 4and the requisite !arentheses5&5

+ree 4ariables and 8uantifiers

)ll the well-formed formulas above have the !ro!erty that their inter!retations are sentences which are either true or false& There are, however, well-formed formulas which do not have that !ro!erty, such as this one 4b^S?5XSS? Its inter!retation is -b !lus " equals B-& Since b is uns!ecified, there is way to assi n a truth value to the statement& It is like an out-of-conte(t statement with a !ronoun, such as -she is clumsy-& It is neither true nor false= it is waitin for you to !ut it into a conte(t& 0ecause it is neither true nor false, such a formula is called open, and the variable b is called a free #aria(le& Ene way of chan in an o!en formula into a closed formula, or sentence, is by !refi(in it with a "uantifierOeither the !hrase -there e(ists a number b such that&&& -, or the !hrase -for all numbers b-& In the first instance, you et the sentence There e(ists a number b such that b !lus " equals B& 7learly this is true& In the second instance, you et the sentence 'or all numbers b, b !lus " equals B& 7learly this is false& 3e now introduce symbols for both of these "uantifiers& These sentences are translated into T8T-notation as follows, b,4b^S?5XSS? 411 stands for 1e(ists1&5 "b,4b^S?5XSS? 41"1 stands for 1all1&5 It is very im!ortant to note that these statements are no lon er about uns!ecified numbers= the first one is an assertion of e(istence, and the second one is a uni#ersal assertion& They would mean the same thin , even if written with c instead of b, c,4c^S?5XSS? "c,4c^S?5XSS? ) variable which is under the dominion of a quantifier is called a "uantified #aria(le& The followin two formulas illustrate the difference between free variables and quantified variables, 4bjb5XSS? ~b,4bjb5XSS? 4o!en5 4closed= a sentence of T8T5

The first one e(!resses a propert) which mi ht be !ossessed by some natural number& Ef course, no natural number has that !ro!erty& )nd that is !recisely what is e(!ressed by

the second one& It is very crucial to understand this difference between a strin with a free #aria(le, which e(!resses a propert), and a strin where the variable is "uantified, which e(!resses a truth or falsit)& The En lish translation of a formula with at least one free variablean o!en formulais called a predicate& It is a sentence without a sub.ect 4or a sentence whose sub.ect is an out-of-conte(t !ronoun5& 'or instance, -is a sentence without a sub.ect-would be an anomaly-runs backwards and forwards simultaneously-im!rovised a si(-!art fu ue on demandare nonarithmetical !redicates& They e(!ress properties which s!ecific entities mi ht or mi ht not !ossess& Ene could as well stick on a -dummy sub.ect-, such as -so-and-so-& ) strin with free variables is like a !redicate with -so-and-so- as its sub.ect& 'or instance, 4S?^S?5Xb is like sayin -" !lus " equals so-and-so-& This is a !redicate in the variable b& It e(!resses a !ro!erty which the number b mi ht have& If one were to substitute various numerals for b, one would et a succession of forms most of which would e(!ress falsehoods& ?ere is another e(am!le of difference between o!en formulas and sentences,
"b,"c,4b^c5X4c^b5

The above formula is a sentence re!resentin , of course, the commutativity of addition& En the other hand,
"c,4b^c5X4c^b5

is an o!en formula, since b is free& It e(!resses a !ro!erty which uns!ecified number b mi ht or mi ht not havenamely of commutin with all numbers c&

Translating 0ur Sam!le Sentences

This com!letes the vocabulary with which we will e(!ress all number-theoretical statements2 It takes considerable !ractice to et the han of e(!ressin com!licated statements of 8 in this notation, and conversely of fi urin out the meanin of wellformed formulas& 'or this reason we return to the si( sam!le sentences iven at the be innin , and work their translations into T8T& 0y the way, don1t think that the translations iven below are uniquefar from it& There are manyinfinitely many ways to e(!ress each one&

Aet us be in with the last one, -D is even-& This we re!hrased in to of more !rimitive notions as -There e(ists a number e such that B times e equals D-& This one is easy, e,4SS?je5XSSSSSS? 8ote the necessity of the quantifier= it sim!ly would not do to write 4SS?je5XSSSSSS? alone& This strin 1s inter!retation is of course neither true nor false= it e(!resses a !ro!erty which the number e mi ht have& It is curious that, since we know multi!lication is commutative mi ht easily have written e,4ejSS?5XSSSSSS? instead& Er, knowin that equality is a symmetrical relation, we mi ht have chosen to write the sides of the equation in the o!!osite order, e,SSSSSS?X4SS?je5 8ow these three translations of -D is even- are quite different strin s, and it is by no means obvious that theoremhood of any one of them is tied to theoremhood of any of the others& 4Similarly, the fact that --!-q--- was a theorem had very little to do with the fact that its -equivalent- strin -!--q--- was a theorem& The equivalence lies in our minds, since, as humans, we almost automatically think about inter!retations, not structural !ro!erties of formulas&5 3e can dis!ense with sentence B, -B is not a square-, almost immediately, ~b,4bjb5XSS? ?owever, once a ain, we find an ambi uity& 3hat if we had chosen to write it this way?
"b,h4bjb5 XSS?

The first way says, -It is not the case that there e(ists a number b with the !ro!erty that b1s square is B-, while the second way says, -'or all numbers b, it is not the case that b1s square is B&- Ence a ain, to us, they are conce!tually equivalentbut to T8T, they are distinct strin s& Aet us !roceed to sentence %, -"$B# is a sum of two cubes&- This one will involve two e(istential quantifiers, one after the other, as follows,

b,c,SSSSSSS&SSSS?X444bjb5jb5^44cjc5jc55

There are alternatives alore& ;everse the order of the quantifiers= switch the sides of the equation= chan e the variables to d and e= reverse the addition= write the multi!lications differently= etc&, etc& ?owever, I !refer the followin two translations of the sentence, b,c,444SSSSSSSSSS?jSSSSSSSSSS?5&SSSSSSSSSS?5^ 44SSSSSSSSS?jSSSSSSSSS?5jSSSSSSSSS?55X444bjb5jb5^44cjc5jc55 and b,c,444SSSSSSSSSSSS?&SSSSSSSSSSSS?5&SSSSSSSSSSSS?5^ 44S?jS?5jS?55X444bjb5jb5^44cjc5jc55 :o you see why?

8ow let us tackle the related sentence C, -8o sum of two !ositive cubes is itself a cube-& Su!!ose that we wished merely to state that $ is not a sum of two !ositive cubes& The easiest way to do this is by negating the formula which asserts that $ is a sum of two !ositive cubes& This will be .ust like the !recedin sentence involvin "$B#, e(ce!t that we have to add in the !roviso of the cubes bein !ositive& 3e can do this with a trick, !refi( variables with the symbol S, as follows, b,c,SSSSSSS?X444SbjSb5jSb5^44ScjSc5-Sc55 Hou see, we are cubin not b and c, but their successors, which must be !ositive, since the smallest value which either b or c can take on is <ero& ?ence the ri ht-hand side re!resents a sum of two !ositive cubes& Incidentally, notice that the !hrase -there e(ist numbers b and c such thatS&&-, when translated, does not involve the symbol 1 n1 which stands for _and1& That symbol is used for connectin entire well-formed strin s, not for .oinin two quantifiers& 8ow that we have translated -$ is a sum of two !ositive cubes-, we wish to ne ate it& That sim!ly involves !refi(in the whole thin by a sin le tilde& 48ote, you should not ne ate each quantifier, even thou h the desired !hrase runs -There do not e(ist numbers b and c such that&&&-&5 Thus we et, ~b,c,SSSSSSS?X444SbjSb@jSb5^44ScjSc5jSc55

b
BCFI of them

Tric.s of the Trade

8ow our ori inal oal was to assert this !ro!erty not of the number of all cubes& Therefore, let us re!lace the numeral SSSSSSS? by the 44aja5ja5, which is the translation of -a cubed-, b,c,44aja5ja5X444SbjSb5jSb5^44Sc-Sc5-Sc55 )t this sta e, we are in !ossession of an open formula, since a is still free& This formula e(!resses a !ro!erty which a number a mi ht or mi ht not haveand it is our !ur!ose to assert that all numbers do have that !ro!erty& That is sim!le.ust !refi( the whole thin with a universal quantifier
"a,hb,c,44aja5ja5X444SbjSb5jSb5 ^44Sc-Sc5-Sc55

)n equally ood translation would be this, ~~a,b,c,44aja5ja5X444SbjSb5jSb5^44ScjSc5jSc55 In austere T8T, we could use a1 instead of b, and a- instead of c, and the formula would become, ~~a, a1, a-,44aja5ja5 X444Sa1jSa15jSa15 ^44Sa%%jSa%%5jSa%%55 3hat about sentence ", -L is !rime-? 3e had reworded it in this way -There do not e(ist numbers a and b, both reater than ", such equals a times b-& 3e can sli htly modify it, as follows, -There do not e(ist numbers a and b such that L equals a !lus B, times b !lus B-& This is another tricksince a and b are restricted to natural number values, this is an adequate way to say the same thin & 8ow - b !lus B- could be translated into 4b ^ SS?5, but there is a shorter way to write itnamely, SSb& Aikewise, -c !lus B- can be written SSc& 8ow, our translation is e(tremely concise, b, c,SSSSS?X4SSbjSSc5 3ithout the initial tilde, it would be an assertion that two natural numbers do e(ist, which, when au mented by B, have a !roduct equal to L& 3ith the tilde in front, that whole statement is denied, resultin in an assertion that L is !rime& If we wanted to assert that d !lus e !lus ", rather than L, is !rime, the most economical way would be to re!lace the numeral for L by the strin 4d+Se5, b, c,4d^Se5X4SSbjSSc5

Ence a ain, an o!en formula, one whose inter!retation is neither a true nor a false sentence, but .ust an assertion about two uns!ecified numbers, d and e& 8otice that the number re!resented by the strin 4 d+Se5 is necessarily reater than d, since one has added to d an uns!ecified but definitely !ositive amount& Therefore, if we e(istentially quantify over the variable e, we will have a formula which asserts that, There e(ists a number which is reater than d and which is !rime& e,hb,c,4d^Se5X4SSbjSSc5 3ell, all we have left to do now is to assert that this !ro!erty actually obtains, no matter what d is& The way to do that is to universally quantify over the variable d,
"d,e,hb,c,4d^Se5X4SSbjSSc5

That1s the translation of sentence L2

Translation ,u::les for =ou

This com!letes the e(ercise of translatin all si( ty!ical number-theoretical sentences& ?owever, it does not necessarily make you an e(!ert in the notation of T8T& There are still some tricky issues to be mastered& The followin si( well-formed formulas will test your understandin of T8T notation& 3hat do they mean? 3hich ones are true 4under inter!retation, of course5, and which ones are false? 4?int, the way to tackle this e(ercise is to move leftwards& 'irst, translate the atom= ne(t, fi ure out what addin a sin le quantifier or a tilde does= then move leftwards, addin another quantifier or tilde= then move leftwards a ain, and do the same&5 h"c, b,4SS?jb5Xc "c,hb,4SS?jb5Xc "c, b,h4SS?jb5Xc hb,"c,4SS?jb5Xc b,h"c,4SS?jb5Xc b,h"c,h4SS?jb5Xc 4Second hint, Either four of them are true and two false, or four false and two true&5

/o' to

istinguish True from +alse7

)t this .uncture, it is worthwhile !ausin for breath and contem!latin what it would mean to have a formal system that could sift out the true ones from the false ones& This system would treat all these strin swhich look like statementsas desi ns havin form, but no content& )nd this system would be like a sieve throu h which could !ass only desi ns with a s!ecial stylethe -style of truth-& If you yourself have one throu h the si( formulas above, and have se!arated the true from the false by thinkin about meanin , you will a!!reciate the subtlety that any system would have to have, that could do the same thin but ty!o ra!hically2 The boundary se!aratin the set of true statements from the set of false statements 4as written in the T8T-notation5 is anythin but strai ht= it is a boundary with many treacherous curves 4recall 'i & "*5, a boundary of which mathematicians have delineated stretches, here and there, workin over hundreds of years& Just think what a cou! it would be to have a ty!o ra!hical method which was uaranteed to !lace any formula on the !ro!er side of the border2

The $ules of >ell-+ormedness

It is useful to have a table of ;ules of 'ormation for well-formed formulas This is !rovided below& There are some !reliminary sta es, definin numerals, #aria(les, and terms& Those three classes of strin s are in redients of well-formed formulas, but are not in themselves well-formed& The smallest well-formed formulas are the atoms= then there are ways of com!oundin atoms& /any of these rules are recursive len thenin rules, in that they take as in!ut an item of a iven class and !roduce a lon er item of the class& In this table, I use 191 and 1)1 to stand for well-formed formulas, and 1 s1, 1t', and 1u' to stand for other kinds of T8T-strin s& 8eedless to say, none of these five symbols is itself a symbol of T8T& 8F/E;)AS& ? is a numeral& ) numeral !receded by S is also a numeral& E9amples: ? S? SS? SSS? SSSS? SSSSS?

G);I)0AES& a is a variable& If we1re not bein austere, so are b, c, d and e& ) variable followed by a !rime is also a variable& E9amples, a b% c- d-1 e-TE;/S& )ll numerals and variables are terms& ) term !receded by S is also a term& If s and t are terms, then so are sPt/ and sZt/! E9amples, ? b SSa1 4S?j4SS?^c55 S4Saj4SbjSc55
TE;/S

may be divided into two cate ories, 4"5 :E'I8ITE terms& These contain no variables& E9amples, ? 4S?^S?5 SS44SS?jSS?5^4S?jS?55 4B5 I8:E'I8ITE terms& These contain variables& E9amples, b Sa 4b^S?5 444S?^S?5^S?5^e5

The above rules tell how to make parts of well-formed formulas= the remainin rules tell how to make complete well-formed formulas& )TE/S& If s and t are terms, then s Q t is an atom& E9amples, S?X? 4SS?^SS?5XSSSS? S4b^c5X44cjd5je5 If an atom contains a variable u, then u is free in it& Thus there are four free variables in the last e(am!le& 8E6)TIE8S& ) well-formed formula !receded by a tilde is well-formed& E9amples, hS?X? hb,4b^b5XS? hg?X?S?X?a hbXS? The "uantification status of a variable 4which says whether the variable is free or quantified5 does not chan e under ne ation& 7E/PEF8:S& If ( and y are well-formed formulas, and !rovided that no variable which is free in one is quantified in the other, then the followin are all well-formed formulas, g (ya, g (ya, g (ya& E9amples, g?X?h?X?a gbXbhc,cXba gS?X?"c,hb,4b^b5Xca The quantification status of a variable doesn1t chan e here& MF)8TI'I7)TIE8S& If u is a variable, and 9 is a well-formed formula in which u is free then the followin strin s are well-formed formulas, u,9 and "u,9& E9amples, "b,gbXbhc,cXba "c,hb,4b^b5Xc hc,ScXd

EPE8 'E;/FA)S

contain at least one free variable& E9amples, hcXc bXb g"b,bXbhcXca contain no free variables& E9amples, S?X? h"d,dX? c,g"b,bXbhcXca

7AESE: 'E;/FA)S 4SE8TE87ES5

This com!letes the table of ;ules of 'ormation for the well-formed formulas of T8T&

A +e' More Translation E-ercises

)nd now, a few !ractice e(ercises for you, to test your understandin of the notation of T8T& Try to translate the first four of the followin 8-sentences into T8T-sentences, and the last one into an o!en formed formula& )ll natural numbers are equal to C& There is no natural number which equals its own square& :ifferent natural numbers have different successors& If " equals +, then every number is odd& b is a !ower of B& The last one you may find a little tricky& 0ut it is nothin , com!ared to this one, b is a !ower of "+& Stran ely, this one takes reat cleverness to render in our notation& I would caution you to try it only if you are willin to s!end hours and hours on itand if you know quite a bit of number theory2

A ;onty!ogra!hical System
This concludes the e(!osition of the notation of T8T= however, we are still left with the !roblem of makin T8T into the ambitious system which we have described& Success would .ustify the inter!retations which we iven to the various symbols& Fntil we have done that, however, !articular inter!retations are no more .ustified than the -horse-a!!le ha!!y- inter!retations were for the !q-system1s symbols& Someone mi ht su est the followin way of constructin T8T, 4"5 :o not have any rules of inference= they are unnecessary, because 4B5 3e take as a(ioms all true statements of number theory 4as written in T8T-notation5& 3hat a sim!le !rescri!tion2 Fnfortunately it is as em!ty as instantaneous reaction says it is& Part 4B5 is, of course, not a ty!o ra!hical descri!tion of strin s& The whole !ur!ose of T8T is to fi ure out if and how it is !ossible tocharacteri<e the true strin s ty!o ra!hically&

The +i"e A-ioms and +irst $ules of T;T

Thus we will follow a more difficult route than the su estion above= we will have a(ioms and rules of inference& 'irstly, as was !romised, all of the rules of the &ropositional 8alculus are taken o#er into T8T& Therefore one theorem of T8T will be this one, gS?X?hS?X?a which can be derived in the same way as gP-Pa was derived& 0efore we ive more rules, let us ive the five a9ioms of T8T, )(iom ", "a,hSaX? )(iom B, "a,4a^?5Xa )(iom %, "a,"b,4a^Sb5XS4a^b5 )(iom C, "a,4aj?5X? )(iom L, "a,"b,4ajSb5X44ajb5^a5 4In the austere versions, use a1 instead of b&5 )ll of them are very sim!le to understand& )(iom " states a s!ecial fact about the number += )(ioms B and % are concerned with the nature of addition= )(ioms C and L are concerned with the nature of multi!lication, and in !articular with its relation to addition&

The +i"e ,eano ,ostulates

0y the way, the inter!retation of )(iom "-9ero is not the successor of any natural number-is one of five famous !ro!erties of natural numbers first e(!licitly reco ni<ed by the mathematician and lo ician 6iuse!!e Peano, in "**#& In settin out his !ostulates, Peano was followin the !ath of Euclid in this way, he made no attem!t to formali<e the !rinci!les of reasonin , but tried to ive a small set of !ro!erties of natural numbers from which everythin else could be derived by reasonin & Peano1s attem!t mi ht thus be considered -semiformal-& Peano1s work had a si nificant influence, and thus it would be ood to show Peano1s five !ostulates& Since the notion of -natural number- is the one which Peano was attem!tin to define, we will not use the familiar term -natural number-, which is laden with connotation& 3e will re!lace it with the undefined term dMinn, a word which comes fresh and free of connotations to our mind& Then Peano1s five !ostulates !lace five restrictions on d.inns& There are two other undefined terms, Genie,

and meta& I will let you fi ure out for yourself what usual conce!t each of them is su!!osed to re!resent& The five Peano !ostulates, 4"5 6enie is a d.inn& 4B5 Every d.inn has a meta 4which is also a d.inn5& 4%5 6enie is not the meta of any d.inn& 4C5 :ifferent d.inns have different metas& 4L5 If 6enie has T, and each d.inn relays T to its meta, then all d.inns et T& In li ht of the lam!s of the +ittle ,armonic +a()rinth, we should name the set of all d.inns -6E:-& This harks back to a celebrated statement by the 6erman mathematician and lo ician Aeo!old @ronecker, archenemy of 6eor 7antor, -6od made the natural numbers= all the rest is the work of man&Hou may reco ni<e Peano1s fifth !ostulate as the !rinci!le of mathematical inductionanother term for a hereditary ar ument& Peano ho!ed that his five restrictions on the conce!ts -6enie-, -d.inn-, and -meta- were so stron that if two different !eo!le formed ima es in their minds of the conce!ts, the two ima es would have com!letely isomor!hic structures& 'or e(am!le, everybody1s ima e would include an infinite number of distinct d.inns& )nd !resumably everybody would a ree that no d.inn coincides with its own meta, or its meta1s meta, etc& Peano ho!ed to have !inned down the essence of natural numbers in his five !ostulates& /athematicians enerally rant that he succeeded, but that does not lessen the im!ortance of the question, -?ow is a true statement about natural numbers to be distin uished from a false one?- )t answer this question, mathematicians turned to totally formal systems, as T8T& ?owever, you will see the influence of Peano in T8T, because all of his !ostulates are incor!orated in T8T in one way or another&

;e' $ules of T;T5 S!ecification and Generali:ation

8ow we come to the new rules of T8T& /any of these rules will allow us to reach in and chan e the internal structure of the atoms of T8T& In that sense they deal with more -microsco!ic- !ro!erties of strin s than the rules of the Pro!ositional 7alculus, which treat atoms as indivisible units& 'or e(am!le, it would be nice if we could e(tract the strin hS?X? from the first a(iom& To do this we would need a rule which !ermits us to dro! a universal quantifier, and at the same time to chan e the internal strut of the strin which remains, if we wish& ?ere is such a rule, ;FAE E' SPE7I'I7)TIE8, Su!!ose u is a variable which occurs inside strin 9& If the strin "u,9 is a theorem, then so is 9, and so are any strin s made from 9 by re!lacin u, wherever it occurs, by one and the same term& 4=estriction, The term which re!laces u must not contain any variable that is quantified in 9&5

The rule of s!ecification allows the desired strin to be e(tracted )(iom "& It is a oneste! derivation,
"a, hSaX? hS?X?

a(iom " s!ecification

8otice that the rule of s!ecification will allow some formulas which contain free variables 4i&e&, o!en formulas5 to become theorems& 'or e(am!le, the followin strin s could also be derived from )(iom ", by s!ecification, hSaX? hS4c^SS?5X? There is another rule, the rule of generali@ation, which allows us to !ut back the universal quantifier on theorems which contain variables that became free as a result of usa e of s!ecification& )ctin on the lower strin , for e(am!le, enerali<ation would ive,
"c,hS4c^SS?5X?

6enerali<ation undoes the action of s!ecification, and vice versa& Fsually, enerali<ation is a!!lied after several intermediate ste!s have transformed the o!en formula in various ways& ?ere is the e(act statement of the rule, ;FAE
E'

6E8E;)AI9)TIE8, Su!!ose 9 is a theorem in which u, a variable, occurs free& Then "u,9 is a theorem& 4=estriction, 8o enerali<ation is allowed in a fantasy on any variable which a!!eared free in the fantasy1s !remise&5

The need for restrictions on these two rules will shortly be demonstrated e(!licitly& Incidentally, this enerali<ation is the same enerali<ation as was mentioned in 7ha!ter II, in Euclid1s !roof about the infinitude of !rimes& )lready we can see how the symbolmani!ulatin rules are startin to a!!ro(imate the kind of reasonin which a mathematician uses&

The E-istential 8uantifier

These !ast two rules told how to take off universal quantifiers and !ut them back on= the ne(t two rules tell how to handle e(istential quantifiers& ;FAE
E'

I8TE;7?)86E, Su!!ose u is a variable& Then the strin s "u,h and ~u, are interchan eable anywhere inside any theorem&

'or e(am!le, let us a!!ly this rule to )(iom ",

"a,hSaX? ha,SaX?

a(iom " interchan e

0y the way, you mi ht notice that both these strin s are !erfectly natural renditions, in T8T, of the sentence -9ero is not the successor of any natural number-& Therefore it is ood that they can be turned into each other with ease& The ne(t rule is, if anythin , even more intuitive& It corres!onds to the very sim!le kind of inference we make when we o from -B is !rime- to -There e(ists a !rime-& The name of this rule is self-e(!lanatory, ;FAE
E'

ETISTE87E, Su!!ose a term 4which may contain variables as lon as they are free5 a!!ears once, or multi!ly, in a theorem& Then any 4or several, or all5 of the a!!earances of the term may be re!laced by a variable which otherwise does not occur in the theorem, and the corres!ondin e(istential quantifier must be !laced in front&

Aet us a!!ly the rule toas usual)(iom ",

"a,hSaX? a(iom " b,"a,hSaXb e(istence

Hou mi ht now try to shunt symbols, accordin to rules so far iven, to !roduce the theorem ~"b: a:Sa=b.

$ules of Equality and Successorshi!

3e have iven rules for mani!ulatin quantifiers, but so far none for symbols 1X1 and 1S1& 3e rectify that situation now& In what follows, r, s, t all stand for arbitrary terms& ;FAES E' EMF)AITH, SH//ET;H, If r X s is a theorem, then so is s X r& T;)8SITIGITH, If r X s and s X t are theorems, then so is r X t& ;FAES E' SF77ESSE;S?IP, ):: S, If r X t is a theorem, then Sr X St is a theorem& :;EP S, If Sr X St is a theorem, then r X t is a theorem& 8ow we are equi!!ed with rules that can ive us a fantastic variety of theorems& 'or e(am!le, the followin derivations yield theorems which are !retty fundamental, 4"5 4B5
"a, "b,4a^Sb5XS4a^b5 "b,4S?^Sb5XS4S?^b5

a(iom % s!ecification 4S? for a5

4%5 4C5 4L5 4D5 4$5

4S?^S?5XS4S?^?5 "a,4a^?5Xa 4S?^?5XS? S4S?^?5XSS? 4S?^S?5XSS? [ [ [ [

s!ecification 4? for b5 a(iom B s!ecification 4S? for a5 add S transitivity 4lines %,D5 [ a(iom L s!ecification 4S? for a5 s!ecification 4? for b5 a(iom % s!ecification 44S?j?5 for a5 s!ecification 4? for b5 a(iom B s!ecification 44S?j?5 for a5 a(iom C s!ecification 4S? for a5 transitivity 4lines *,"+5 add S transitivity 4lines D,"B5 transitivity 4lines %,"%5

4"5 4B5 4%5 4C5 4L5 4D5 4$5 4*5 4#5 4"+5 4""5 4"B5 4"% 4"C5

"a,"b,4ajSb5X44ajb5^a5 "b,4S?jSb5X44S?jb5^S?5 4S?jS?5X44S?j?5^S?5 "a,"b,4a^Sb5XS4a^b5 "b,44S?j?5^Sb5XS44S?j?5^b5 44S?j?5^S?5XS44S?j?5^?5 "a,4a^?5Xa 44S?j?5^?5X4S?j?5 "a,4aj?5X? 4S?j?5X? 44S?j?5^?5X? S44S?j?5^?5XS? 44S?j?5^S?5XS? 4S?jS?5XS?

Illegal Shortcuts
8ow here is an interestin question, -?ow can we make a derivation for the strin ?X??It seems that the obvious route to o would be first to derive the strin "a,aXa, and then to use s!ecification& So, what about the followin -derivation- of "a,aXa&&& 3hat is wron with it? 7an you fi( it u!? 4"5 4B5 4%5
"a,4a^?5Xa "a,aX4a^?5 "a,aXa

a(iom B symmetry transitivity 4lines B,"5

I ave this mini-e(ercise to !oint out one sim!le fact, that one should not .um! too fast in mani!ulatin symbols 4such as 1X15 which are familiar& Ene must follow the rules, and not one1s knowled e of the !assive meanin s of the symbols& Ef course, this latter ty!e of knowled e is invaluable in uidin the route of a derivation&

>hy S!ecification and Generali:ation Are $estricted

8ow let us see why there are restrictions necessary on both s!ecification and enerali<ation& ?ere are two derivations& In each of them, one of the restrictions is violated& Aook at the disastrous results they !roduce, 4"5 4B5 4%5 4C5 4L5 4D5 4$5 4*5 4#5 4"+5 U aX? "a,aX? SaX? V gaX?SaX?a "a,gaX?SaX?a g?X?S?X?a ?X? S?X? !ush !remise enerali<ation 4>rong25 s!ecification !o! fantasy rule enerali<ation s!ecification !revious theorem detachment 4lines #,*5

This is the first disaster& The other one is via faulty s!ecification& 4"5 4B5 4%5 4C5 4L5
"a,aXa SaXSa b,bXSa "a, b,bXSa b,bXSb

!revious theorem s!ecification e(istence enerali<ation s!ecification 4>rong25

So now you can see why those restrictions are needed& ?ere is a sim!le !u<<le, translate 4if you have not already done so5 Peano1s fourth !ostulate into T8T-notation, and then derive that strin as a theorem&

Something Is Missing
8ow if you e(!eriment around for a while with the rules and a(ioms of T8T so far !resented, you will find that you can !roduce the followin p)ramidal famil) of theorems 4a set of strin s all cast from an identical mold, differin from one another only in that the numerals ?, S?, SS?, and s have been stuffed in5, 4?^?5X? 4?^S?5XS? 4?^SS?5XSS? 4?^SSS?5XSSS? 4?^SSSS?5XSSSS?

etc& )s a matter of fact, each of the theorems in this family can be derived from the one directly above it, in only a cou!le of lines& Thus it is a sort of -cascade- of theorems, each one tri erin the ne(t& 4These theorems are very reminiscent of the !q-theorems, where the middle and ri ht-hand rou!s of hy!hens rew simultaneously&5 8ow there is one strin which we can easily write down, and which summari<es the !assive meanin of them all, taken to ether& That universally quantified summari@ing string is this,
"a,4?^a5Xa

Het with the rules so far iven, this strin eludes !roduction& Try to !roduce it yourself if you don1t believe me& Hou may think that we should immediately remedy the situation with the followin 4P;EPESE:5 ;FAE E' )AA, If all the strin s in a !yramidal family are theorems, then so is the universally quantified strin which summari<es them& The !roblem with this rule is that it cannot be used in the / -mode& Enly !eo!le who are thinkin a(out the system can ever know that an infinite set of strin s are all theorems& Thus this is not a rule that can be stuck inside any formal system&

-Incom!lete Systems and &ndecidable Strings

So we find ourselves in a stran e situation, in which we can ty!o ra!hically !roduce theorems about the addition of any specific numbers, but even a sim!le strin as the one above, which e(!resses a !ro!erty of addition in general, is not a theorem& Hou mi ht think that is not all that stran e, since we were in !recisely that situation with the !qsystem& ?owever, the !q-system had no !retensions about what it ou ht to be able to do= and in fact there was no way to e9press eneral statements about addition in its symbolism, let alone !rove them& The equi!ment sim!ly was not there, and it did not even occur to us to think that the system was defective& ?ere, however, the e(!ressive ca!ability is far stron er, and we have corres!ondin ly hi her e(!ectations of T8T than of the !q-system& If the strin above is not a theorem, then we will have ood reason to consider T8T to be defective& )s a matter of fact, there is a name for systems with this kind of defect-they are called -incomplete& 4The !refi( 11-1ome a1- comes from the fact that the totality of natural numbers is sometimes denoted by 11&5 ?ere is the e(act definition, ) system is -incom!lete if all the strin s in a !yramidal family are theorems, but the universally quantified summari<in strin is not a theorem&

Incidentally, the ne ation of the above summari<in strin ~"a,4?^a@Xa is also a nontheorem of T8T& This means that the ori inal strin is undecida(le within the s)stem& If one or the other were a theorem, then we would say that it was decidable& )lthou h it may sound like a mystical term, there is nothin mystical about undecidability within a iven system& It is only a si n that the system could be e(tended& 'or e(am!le, within absolute eometry, Euclid1s fifth !ostulate is undecidable& It has to be added as an e(tra !ostulate of eometry, to yield Euclidean eometry= or conversely, its ne ation can be added, to yield non-Euclidean eometry& If you think back to eometry, you will remember why this curious thin ha!!ens& It is because the four !ostulates of absolute eometry sim!ly do not !in down the meanin s of the terms -!oint- and -line-, and there is room for different e9tensions of the notions& The !oints and lines of Euclidean eometry !rovide one kind of e(tension of the notions of -!ointand -line-= the PEI8Ts and AI8Es of non-Euclidean eometry, another& ?owever, usin the !re-flavored words -!oint- and -line- tended, for two millennia, to make !eo!le believe that those words were necessarily univalent, ca!able of only one meanin &

;on-Euclidean T;T
3e are now faced with a similar situation, involvin T8T& 3e have ado!ted a notation which !re.udices us in certain ways& 'or instance, usa e of the symbol 1^1 tends to make us think that every theorem with a !lus si n in it ou ht to say somethin known and familiar and -sensible- about the known and familiar o!eration we call -addition-& Therefore it would run a ainst the rain to !ro!ose addin the followin -si(th a(iom-, ~"a,4?^a5Xa It doesn1t .ibe with what we believe about addition& 0ut it is one !ossible e(tension of T8T, as we have so far formulated T8T& The system which uses this as its si(th a(iom is a consistent system, in the sense of not havin two theorems of the form 9 and -9& ?owever, when you .u(ta!ose this -si(th a(iom- with the !yramidal family of theorems shown above, you will !robably be bothered by a seemin inconsistency between the family and the new a(iom& 0ut this kind of inconsistency is not so dama in as the other kind 4where 9 and -9 are both theorems5& In fact, it is not a true inconsistency, because there is a way of inter!retin the symbols so that everythin comes out all ri ht&

-Inconsistency Is ;ot the Same as Inconsistency

This kind of inconsistency, created by the o!!osition of 4"5 a !yramidal family of theorems which collectively assert that all natural numbers have some !ro!erty, and 4B5 a sin le theorem which seems to assert that not all numbers have it, is iven the name of -

inconsistenc)& )n -inconsistent system is more like the at-the-outset-distasteful-but-inthe-end-acce!table non-Euclidean eometry& In order to form a mental model of what is oin on, you have to ima ine that there are some -e(tra-, unsus!ected numberslet us not call them -natural-, but supernatural numberswhich have no numerals& Therefore, facts about them cannot be re!resented in the !yramidal family& 4This is a little bit like )chilles1 conce!tion of 6E:as a sort of -su!erd.inn-, a bein reater than any of the d.inn& This was scoffed at by the 6enie, but it is a reasonable ima e, and may hel! you to ima ine su!ernatural numbers&5 3hat this tells us is that the a(ioms and rules of T8T, as so far !resented, do not fully !in down the inter!retations for the symbols of T8T& There is still room for variation in one1s mental model of the notions they stand for& Each of the various !ossible e(tensions would !in down some of the notions further= but in different ways& 3hich symbols would be in to take on -distasteful- !assive meanin s, if we added the -si(th a(iom- iven above? 3ould all of the symbols become tainted, or would some of them still mean what we want them to mean? I will let you think about that& 3e will encounter a similar question in 7ha!ter TIG, and discuss the matter then& In any case, we will not follow this e(tension now, but instead o on to try to re!air the -incom!leteness of T8T&

The Last $ule

The !roblem with the -;ule of )ll- was that it required knowin that all lines of an infinite !yramidal family are theoremstoo much for a finite bein & 0ut su!!ose that each line of the !yramid can be derived from its !redecessor in a patterned way& Then there would be a finite reason accountin for the fact that all the strin s in the !yramid are theorems& The trick then, is to find the pattern that causes the cascade, and show that !attern is a theorem in itself& That is like !rovin that each d.inn !asses a messa e to its meta, as in the children1s ame of -Tele!hone-& The other thin left to show is that 6enie starts the cascadin messa ethat is, to establish that the first line of the !yramid is a theorem& Then you know that 6E: will et the messa e2 In the !articular !yramid we were lookin at, there is a !attern, ca!tured by lines C-# of the derivation below& 4"5 4B5 4%5 4C5 4L5 4D5 4$5 4*5 4#5
"a,"b,4a^Sb5XS4a^b5 "b,4?^Sb5XS4?^b5 4?^Sb5XS4?^b5 U 4?^b5Xb S4?^b5XSb 4?^Sb5XS4?^b5 4?^Sb5XSb V

a(iom % s!ecification s!ecification !ush !remise add S carry over line % transitivity !o!

The !remise is 4?^b5Xb= the outcome is 4?^Sb5XSb&

The first line of the !yramid is also a theorem= it follows directly from )(iom B& )ll we need now is a rule which lets us deduce that the strin which summari<es the entire !yramid is itself a theorem& Such a rule will he a formali<ed statement of the fifth Peano !ostulate& To e(!ress that rule, we need a little notation& Aet us abbreviate a well-formed formula in which the variable a is free by the followin notation, Skab 4There may be other free variables, too, but that is irrelevant&5 Then the notation SkSaIab will stand for that strin but with every occurrence of a re!laced by Sa& Aikewise, Sk?Iab would stand for the same strin , with each a!!earance of a re!laced by ?& ) s!ecific e(am!le would be to let Skab stand for the strin in question, 4 ?^a5Xa& Then SkSaIab would re!resent the strin 4?^Sa5XSa, and Sk?Iab would re!resent 4?^?5X?& 43arnin , This notation is not !art of T8T= it is for our convenience in talkin a(out T8T&5 3ith this new notation, we can state the last rule of T8T quite !recisely, ;FAE E' I8:F7TIE8, Su!!ose u is a variable, and Skub is a well-formed formula in which u occurs free& If both "u,g Skub SkSuIuba and Sk?Iub are theorems, then "u, Skub is also a theorem& This is about as close as we can come to !uttin Peano1s fifth !ostulate into T8T& 8ow let us use it to show that "a,4?^a5Xa is indeed a theorem in T8T& Emer in from the fantasy in our derivation above, we can a!!ly the fantasy rule, to ive us 4"+5 4""5 g4?^b5Xb4?^Sb5XSba "b,g4?^b5Xb4?^Sb5XSba fantasy rule enerali<ation

This is the first of the two in!ut theorems required by the induction The other requirement is the first line of the !yramid, which we have& Therefore, we can a!!ly the rule of induction, to deduce what we wanted&
"b,4?^b5Xb

S!ecification and enerali<ation will allow us to chan e the variable from b to a= thus "a,4?^a5Xa is no lon er an undecidable strin of T8T&

A Long

eri"ation

8ow I wish to !resent one lon er derivation in T8T, so that you ca what one is like, and also because it !roves a si nificant, if sim!le, fact of number theory&

4"5 "a,"b,4a^Sb5XS4a^b5 4B5 "b,4d^Sb5XS4d^b5 4%5 4d^SSc5XS4d^Sc5 4C5 "b,4Sd^Sb5XS4Sd^b5 4L5 4Sd^Sc5XS4Sd^c5 D5 S4Sd^c5X4Sd^Sc5 4$5 U 4*5 "d,4d^Sc5X4Sd^c5 4#5 4d^Sc5X4Sd^c5 4"+5 S4d^Sc5XS4Sd^c5 4""5 4d^SSc5XS4d^Sc5 4"B5 4d^SSc5XS4Sd^c5 4"%5 S4Sd^c5X4Sd^Sc5 4"C5 4d^SSc5X4Sd^Sc5 4"L5 "d,4d^SSc5X4Sd^Sc5 4"D5 V 4"$5 g"d,4d^Sc5X4Sd^c5"d,4d^SSc5X4Sd^Sc5a 4"*5 "c,g"d,4d^Sc5X4Sd^c5 "d,4d^SSc5X4Sd^Sc5a [ 4"#5 4d^S?5XS4d^?5 4B+5 "a,4a^?5Xa 4B"5 4d^?5Xd 4BB5 S4d^?5XSd 4B%5 4d^S?5XSd 4BC5 4Sd^?5XSd 4BL5 SdX4Sd^?5 4BD5 4d^S?5X4Sd^?5 4B$5 "d,4d^S?5X4Sd^?5 [ 4B*5 "c,"d,4d^Sc5X4Sd^c5 [ [ [ [ [ [ [ [

a(iom % s!ecification s!ecification s!ecification 4line "5 s!ecification symmetry !ush !remise s!ecification add S carry over % transitivity carry over D transitivity enerali<ation !o! fantasy rule enerali<ation

s!ecification 4line B5 a(iom " s!ecification add S transitivity 4lines "#,B5 s!ecification 4line B+5 symmetry transitivity 4lines B%,BL5 enerali<ation

induction 4lines "*,B$5

US can be sli!!ed back and forth in an additionV [ [ [ [ [ 4B#5 "b,4c^Sb5XS4c^b5 4%+5 4c^Sd5XS4c^d5 4%"5 "b,4d^Sb5XS4d^b5 s!ecification 4line "5 s!ecification s!ecification 4line "5

4%B5 4d^Sc5XS4d^c5 4%%5 S4d^c5X4d^Sc5 4%C5 "d,4d^Sc5X4Sd^c5 4%L5 4d^Sc5X4Sd^c5 4%D5 U 4%$5 "c,4c^d5X4d^c5 4%*5 4c^d5X4d^c5 4%#5 S4c^d5XS4d^c5 4C+5 4c^Sd5XS4c^d5 4C"5 4c^Sd5XS4d^c5 4CB5 S4d^c5X4d^Sc5 4C%5 4c^Sd5X4d^Sc5 4CC5 4d^Sc5X4Sd^c5 4CL5 4c^Sd5X4Sd^c5 4CD5 "c,4c^Sd5X4Sd^c5 4C$5 V 4C*5 g"e,4c^d5X4d^c5 "c,4c^Sd5X4Sd^c5a 4C#5 "d,g"c,4c^d5X4d^c5 "c,4c^Sd5X4Sd^c5a

s!ecification symmetry s!ecification 4line B*5 s!ecification !ush !remise s!ecification add S carry over %+ transitivity carry over %% transitivity carry over %L transitivity enerali<ation !o! fantasy rule enerali<ation

UIf d commutes with every c, then Sd does too&V [ 4L+5 4c^?5Xc 4L"5 "a,4?^a5Xa 4LB5 4?^c5Xc 4L%5 cX4?^c5 4LC5 4c^?5X4?^c5 4LL5 "c,4c^?5X4?^c5 U? commutes with every c&V 4LD5 "d,"c,4c^d5X4d^c5 induction 4lines C#,LL5 [ [ [ [ s!ecification 4line B+5 !revious theorem s!ecification symmetry transitivity 4lines L+,L%5 enerali<ation

UTherefore, every d commutes with every c&V

Tension and $esolution in T;T

T8T has !roven the commutativity of addition& Even if you do not follow this derivation in detail, it is im!ortant to reali<e that, like a !iece of music, it has its own natural -rhythm-& It is not .ust a random walk that ha!!ens to have landed on the desired last

line& I have inserted -breathin marks- to show some of the -!hrasin - of this derivation& Aine B* in !articular is a turnin !oint in the derivation, somethin like the halfway !oint in an **BB ty!e of !iece, where you resolve momentarily, even if not in the tonic key& Such im!ortant intermediate sta es are often called -lemmas-& It is easy to ima ine a reader startin at line " of this derivation i norant of where it is to end u!, and ettin a sense of where it is oin as he sees each new line& This would set u! an inner tension, very much like the tension in a !iece of music caused by chord !ro ressions that let know what the tonality is, without resolvin & )rrival at line B* would confirm the reader1s intuition and ive him a momentary feelin of satisfaction while at the same time stren thenin his drive to !ro ress towards what he !resumes is the true oal& 8ow line C# is a critically im!ortant tension-increaser, because of -almost-therefeelin which it induces& It would be e(tremely unsatisfactory to leave off there2 'rom there on, it is almost !redictable how thin s must o& 0ut you wouldn1t want a !iece of music to quit on you .ust when it had made the mode of resolution a!!arent& Hou don1t want to imagine the endin you want to hear the endin & Aikewise here, we have to carry thin s throu h& Aine LL is inevitable, and sets u! all the final tensions, which are resolved by Aine LD& This is ty!ical of the structure not only of formal derivations, but also of informal !roofs& The mathematician1s sense of tension is intimately related to his sense of beauty, and is what makes mathematics worthwhile doin & 8otice, however, that in T8T itself, there seems to be no reflection of these tensions& In other words, T8T doesn1t formali<e the notions of tension and resolution, oal and sub oal, -naturalness- and -inevitabilityany more than a !iece of music is a book about harmony and rhythm& 7ould one devise a much fancier ty!o ra!hical system which is aware of the tensions and oals inside derivations?

+ormal $easoning "s1 Informal $easoning

I would have !referred to show how to derive Euclid1s Theorem 4the infinitude of !rimes5 in T8T, but it would !robably have doubled the len th of the book& 8ow after this theorem, the natural direction to o would be to !rove the associativity of addition, the commutativity and associativity of multi!lication, and the distributivity of multi!lication over addition& These would ive a !owerful base to work from& )s it is now formulated, T8T has reached -critical mass- 4!erha!s a stran e meta!hor to a!!ly to somethin called -T8T-5& It is of the same stren th as the system of &rincipia 'athematica= in T8T one can now !rove every theorem which you would find in a standard treatise on number theory& Ef course, no one would claim that derivin theorems in T8T is the best way to do number theory& )nybody who felt that way would fall in the same class of !eo!le as those who think that the best way to know what "+++ ( "+++ is, is to draw a "+++ by "+++ rid, and count all the squares in it&&& 8o= after total formali<ation, the only way to o is towards rela(ation of the formal system& Etherwise, it is so enormously unwieldy as to be, for all !ractical !ur!oses, useless& Thus, it is im!ortant to embed T8T within a wider conte(t, a conte(t which enables new rules of

inference to be derived, so that derivations can be s!eeded u!& This would require formali<ation of the lan ua e in which rules of inference are e(!ressedthat is, the metalan ua e& )nd one could o considerably further& ?owever, none of these s!eedin u! tricks would make T8T any more powerful= they would sim!ly make it more usa(le& The sim!le fact is that we have !ut into T8T every mode of thou ht that number theorists rely on& Embeddin it in ever lar er conte(ts will not enlar e the s!ace of theorems= it will .ust make workin in T8Tor in each -new, im!roved version-look more like doin conventional number theory&

;umber Theorists Go out of Business

Su!!ose that you didn1t have advance knowled e that T8T will turn out to be incom!lete, but rather, e(!ected that it is com!letethat is, that every true statement e(!ressible in the T8Tnotation is a theorem& In that case, you could make a decision !rocedure for all of number theory& The method would be easy, if you want to know if 8statement S is true or false, code it into T8T-sentence 9& 8ow if S is true, com!leteness says that 9 is a theorem= and conversely, if not-T is true, then com!leteness says that X9 is a theorem& So either 9 or X9 must be a theorem, since either S or not-S is true& 8ow be in systematically enumeratin all the theorems of T8T, in the way we did for the /IFsystem and !q-system& Hou must come to 9 or X9 after a while= and whichever one you hit tells you which of S and not-S is true& 4:id you follow this ar ument? It crucially de!ends on your bein able to hold se!arate in your mind the formal system T8T and its informal counter!art 8& /ake sure you understand it&5 Thus, in !rinci!le, if T8T were com!lete, number theorists would be !ut out of business any question in their field could be resolved, with sufficient time, in a !urely mechanical way& )s it turns out, this is im!ossible, which, de!endin on your !oint of view, is a cause either for re.oicin , or for mournin &

/ilbert%s ,rogram
The final question which we will take u! in this 7ha!ter is whether should have as much faith in the consistency of T8T as we did consistency of the Pro!ositional 7alculus= and, if we don1t, whether !ossible to increase our faith in T8T, by pro#ing it to be consistent could make the same o!enin statement on the -obviousness- of T8T1s consistency as Im!rudence did in re ard to the Pro!ositional 7alculus namely, that each rule embodies a reasonin !rinci!le which we believe in, and therefore to question the consistency of T8T is to question our own sanity& To some e(tent, this ar ument still carries wei htbut not quite so much wei ht as before& There are .ust too many rules of inference and some of them .ust mi ht be sli htly -off -& 'urthermore, how do we know that this mental model we have of some abstract entities called -natural numbers- is actually a coherent construct? Perha!s our own thou ht !rocesses, those informal !rocesses which we have tried to ca!ture in the formal rules of the system, are themselves inconsistent2 It is of course not the kind of thin we e(!ect, but it ets more and more conceivable that our thou hts mi ht lead us astray, the more com!le( the sub.ect matter

etsand natural numbers are by no means a trivial sub.ect matter& Prudence1s cry for a proof of consistency has to be taken more seriously in this case& It1s not that we seriously doubt that T8T could be inconsistent but there is a little doubt, a flicker, a limmer of a doubt in our minds, and a !roof would hel! to dis!el that doubt& 0ut what means of !roof would we like to see used? Ence a ain, faced with the recurrent question of circularity& If we use all the equi!ment in a !roof a(out our system as we have inserted into it, what will we have accom!lished? If we could mana e to convince ourselves consistency of T8T, but by usin a weaker system of reasonin than we will have beaten the circularity ob.ection2 Think of the way a heavy ro!e is !assed between shi!s 4or so I read when I was a kid5, first a li ht arrow is fired across the a!, !ullin behind it a thin ro!e& Ence a connection has been established between the two shi!s this way, then the heavy ro!e !ulled across the a!& If we can use a -li ht- system to show that a system is consistent, then we shall have really accom!lished somethin & 8ow on first si ht one mi ht think there is a thin ro!e& Eur oal is to !rove that T8T has a certain ty!o ra!hical !ro!erty 4consistency5, that no theorems of the form 9 and X9 ever occur& This is similar to tryin to show that /F is not a theorem of the /IFsystem& 0oth are statements about t)pographical !ro!erties of symbol-mani!ulation systems& The visions of a thin ro!e are based on the !resum!tion that facts a(out num(er theor) won't (e needed in !rovin that such a ty!o ra!hical !ro!erty holds& In other words, if !ro!erties of inte ers are not usedor if only a few e(tremely sim!le ones are usedthen we could achieve the oal of !rovin T8T consistent, by usin means which are weaker than its own internal modes of reasonin & This is the ho!e which was held by an im!ortant school of mathematicians and lo icians in the early !art of this century, led by :avid ?ilbert& The oal was to !rove the consistency of formali<ations of number theory similar to T8T by em!loyin a very restricted set of !rinci!les of reasonin called -finitistic- methods of reasonin & These would be the thin ro!e& Included amon finitistic methods are all of !ro!ositional reasonin , as embodied in the Pro!ositional 7alculus, and additionally some kinds of numerical reasonin & 0ut 6>del1s work showed that any effort to !ull the heavy ro!e of T8T1s consistency across the a! by usin the thin ro!e of finitistic methods is doomed to failure& 6>del showed that in order to !ull the heavy ro!e across the a!, you can1t use a li hter ro!e= there .ust isn1t a stron enou h one& Aess meta!horically, we can say, *n) s)stem that is strong enough to pro#e TNT's consistenc) is at least as strong as TNT itself! )nd so circularity is inevitable&

A $u %fferin!"
The Tortoise and *chilles ha#e Must (een to hear a lecture on the origins of the Genetic 8ode, and are now drinking some tea at *chilles' home! *chilles, I have somethin terrible to confess, /r& T& Tortoise, 3hat is it, )chilles? *chilles, :es!ite the fascinatin sub.ect matter of that lecture, I drifted off to slee! a time or two& 0ut in my drowsy state, I still was semi-aware of the words comin into my ears& Ene stran e ima e that floated u! from my lower levels was that 1)1 and 1T1, instead of standin for -adenine- and -thymine-, stood for my name and yoursand double-strands of :8) had tiny co!ies of me and you alon their backbones, always !aired u!, .ust as adenine and thymine always are& Isn1t that a stran e symbolic ima e? Tortoise, Phooey2 3ho believes in that silly kind of stuff? )nyway, about 171 and 161? *chilles, 3ell, I su!!ose 171 could stand for /r& 7rab, instead of for cytosine& I1m not sure about 161, but I1m sure one could think of somethin & )nyway, it was amusin to ima ine my :8) bein with minuscule co!ies of youas well as tiny co!ies of myself, for that matter& Just think of the infinite re ress T?)T leads to2 Tortoise, I can see you were not !ayin too much attention to the lecture& *chilles, 8o, you1re wron & I was doin my best, only I had a hard time kee!in fancy se!arated from fact& )fter all, it is such a stran e netherworld that those molecular biolo ists are e(!lorin & Tortoise, ?ow do you mean? *chilles, /olecular biolo y is filled with !eculiar convoluted loo!s which I can1t quite understand, such as the way that folded !roteins, which are coded for in :8), can loo! back and mani!ulate the :8) which they came from, !ossibly even destroyin it& Such stran e loo!s always confuse the dayli hts out of me& They1re eerie, in a way& Tortoise, I find them quite a!!ealin & *chilles, Hou would, of coursethey1re .ust down your alley& 0ut me, sometimes I like to retreat from all this analytic thou ht any meditate a little, as an antidote& It clears my mind of all those confusin loo!s and incredible com!le(ities which we were hearin about toni ht& Tortoise, 'ancy that& I wouldn1t have uessed that you were a meditator& *chilles, :id I never tell you that I am studyin 9en 0uddhism? Tortoise, ?eavens, how did you come u!on that? *chilles, I have always had a yen for the yin and yan , you knowthe whole Eriental mysticism tri!, with the 3 8hing, urus, and whatnot& So one day I1m thinkin to myself, -3hy not 9en too?- )nd that1s how it all be an& Tortoise, Eh, s!lendid& Then !erha!s I can finally become enli htened& *chilles, 3hoa, now& Enli htenment is not the first ste! on the road to 9en= if anythin , it1s the last one2 Enli htenment is not for novices like you, /r& T2

Tortoise, I see we have had a misunderstandin & 0y -enli htenment-, I hardly meant somethin so wei hty as is meant in 9en& )ll I meant is that I can !erha!s become enli htened as to what 9en is all about& *chilles, 'or Pete1s sake, why didn1t you say so? 3ell, I1d be only too ha!!y to tell you what I know of 9en& Perha!s you mi ht even be tem!ted to become a student of it, like me& Tortoise, 3ell, nothin 1s im!ossible& *chilles, Hou could study with me under my master, Ekanisamathe seventh !atriarch& Tortoise, 8ow what in the world does that mean? *chilles, Hou have to know the history of 9en to understand that& Tortoise, 3ould you tell me a little of the history of 9en, then? *chilles, )n e(cellent idea& 9en is a kind of 0uddhism which was founded by a monk named 0odhidharma, who left India and went to 7hina around the si(th century& 0odhidharma was the first !atriarch& The si(th one was Eno& 4I1ve finally ot it strai ht now25 Tortoise, The si(th !atriarch was 9eno, eh? I find it stran e that he, of all !eo!le, would et mi(ed u! in this business& *chilles, I daresay you underestimate the value of 9en& Aisten .ust a little more, and maybe you1ll come to a!!reciate it& )s I was sayin , about five hundred years later, 9en was brou ht to Ja!an, and it took hold very well there& Since that time it has been one of the !rinci!al reli ions in Ja!an& Tortoise, 3ho is this Ekanisama, the -seventh !atriarch-? *chilles, ?e is my master, and his teachin s descend directly from those of the si(th !atriarch& ?e has tau ht me that reality is one, immutable, and unchan in = all !lurality, chan e, and motion are mere illusions of the senses& Tortoise, Sure enou h, that1s 9eno, a mile away& 0ut how ever did he come to be tan led u! in 9en? Poor fellow2 *chilles, 3haaat? I wouldn1t !ut it that way& If )8HE8E is tan led u!, it1s&&& 0ut that1s another matter& )nyway, I don1t know the answer to your question& Instead, let me tell you somethin of the teachin s of my master& I have learned that in 9en, one seeks enli htenment, or S)TE;Ithe state of -8o-mind-& In this state, one does not think about the worldone .ust is& I have also learned that a student of 9en is not su!!osed to -attach- to any ob.ect or thou ht or !ersonwhich is to say, he must not believe in, or de!end on, any absolutenot even this !hiloso!hy of nonattachment& Tortoise, ?mm&&& 8ow T?E;E1S somethin I could like about 9en& *chilles, I had a hunch you1d et attached to it& Tortoise, 0ut tell me, if 9en re.ects intellectual activity, does it make sense to intellectuali<e about 9en, to study it ri orously? *chilles, That matter has troubled me quite a bit& 0ut I think I have finally worked out an answer& It seems to me that you may be in a!!roachin 9en throu h any !ath you knoweven if it is com!letely antithetical to 9en& )s you a!!roach it, you radually learn to stray from that !ath& The more you stray from the !ath, the closer you et to 9en& Tortoise, Eh, it all be ins to sound so clear now&

*chilles, /y favorite !ath to 9en is throu h the short, fascinatin and weird 9en !arables called -kRans-& Tortoise, 3hat is a kRan? *chilles, ) kRan is a story about 9en masters and their students& Sometimes it is like a riddle= other times like a fable= and other times like nothin you1ve ever heard before& Tortoise, Sounds rather intri uin & 3ould you say that to read and en.oy kRans is to !ractice 9en? *chilles, I doubt it& ?owever, in my o!inion, a deli ht in kRans comes a million times closer to real 9en than readin volume after about 9en, written in heavy !hiloso!hical .ar on& Tortoise, I would like to hear a kRan or two& *chilles, )nd I would like to tell you oneor a few& Perha!s be in with the most famous one of all& /any centuries a o, there was a 9en master named JRshl, who lived to be ""# years old& Tortoise, ) mere youn ster2 *chilles, 0y your standards, yes& 8ow one day while JRshl and another monk were standin to ether in the monastery, a do wandered by& The monk asked JRshl, -:oes a do have 0uddha-nature, or not?Tortoise, 3hatever that is& So tell mewhat did JRshl re!ly? *chilles, 1/F1& Tortoise, 1/F1? 3hat is this 1/F1? 3hat about the do ? 3hat about 0uddha-nature? 3hat1s the answer? *chilles, Eh, but 1/F1 is JRshl1s answer& 0y sayin 1/F1, JRshl let the other monk know that only by not askin such questions can one know the answer to them& Tortoise, JRshl -unasked- the question& *chilles, E(actly2 Tortoise, 1/F1 sounds like a handy thin to have around& I1d like to unask a question or two, sometimes& I uess I1m be innin to et the han of 9en& :o you know any other kRans, )chilles? I would like to hear some more& *chilles, /y !leasure& I can tell you a !air of kRans which o to ether Enly&&& Tortoise, 3hat1s the matter? *chilles, 3ell, there is one !roblem& )lthou h both are widely told kRans, my master has cautioned me that only one of them is enuine& )nd what is more, he does not know which one is enuine, and which one is a fraud& Tortoise, 7ra<y2 3hy don1t you tell them both to me and we can s!eculate to our hearts1 content2 *chilles, )ll ri ht& Ene of the alle ed kRans oes like this,
) monk asked 0aso, -3hat is 0uddha?0aso said, -This mind is 0uddha&-

Tortoise, ?mm&&& -This mind is 0uddha-? Sometimes I don1t quite understand what these 9en !eo!le are ettin at& *chilles, Hou mi ht !refer the other alle ed kRan then&

Tortoise, ?ow does it run? *chilles, Aike this,

) monk asked 0aso, -3hat is 0uddha?0aso said, -This mind is not 0uddha&-

Tortoise, /y, my2 If my shell isn1t reen and not reen2 I like that2 *chilles, 8ow, /r& Tyou1re not su!!osed to .ust -like- kRans& Tortoise, Gery well, thenI don1t like it& *chilles, That1s better& 8ow as I was sayin , my master believes only one of the two is enuine& Tortoise, I can1t ima ine what led him to such a belief& 0ut anyway, I su!!ose it1s all academic, since there1s no way to know if a kRan is enuine or !hony& *chilles, Eh, but there you are mistaken& /y master has shown us how to do it& Tortoise, Is that so? ) decision !rocedure for enuineness of kRans? I should very much like to hear about T?)T& *chilles, It is a fairly com!le( ritual, involvin two sta es& In the first sta e, you must T;)8SA)TE the kRan in question into a !iece of strin , folded all around in three dimensions& Tortoise, That1s a curious thin to do& )nd what is the second sta e?

73G5=E DG! Aa /e<quita, () '! 8! Escher (lack and white chalk, BIHJ/!

*chilles: Eh, that1s easyall you need to do is determine whether the strin has 0uddhanature, or not2 If it does, then the kRan is enuineif not, the kRan is a fraud& Tortoise: ?mm&&& It sounds as if all you1ve done is transfer the need for a decision !rocedure to another domain& 8ow it1s a decision !rocedure for 0uddha-nature that you need& 3hat ne(t? )fter all, if you can1t even tell whether a :E6 has 0uddhanature or not, how can you e(!ect to do so for every !ossible folded strin ? *chilles, 3ell, my master e(!lained to me that shiftin between domains can hel!& It1s like switchin your !oint of view& Thin s sometimes look com!licated from one an le, but sim!le from another& ?e ave the e(am!le of an orchard, in which from one direction no order is a!!arent, but from s!ecial an les, beautiful re ularity emer es& Hou1ve reordered the same information by chan in your way of lookin at it& Tortoise, I see& So !erha!s the enuineness of a kRan is concealed how very dee!ly inside it, but if you translate it into a strin it mana es in some way to float to the surface? *chilles, That1s what my master has discovered& Tortoise, Then I would very much like to learn about the technique& 0ut first, tell me, how can you turn a kRan 4a sequence of words5 into a folded strin 4a threedimensional ob.ect5? They are rather different kinds of entities& *chilles, That is one of the most mysterious thin s I have learned in 9en& There are two ste!s, -transcri!tion- and -translation-& T;)8S7;I0I86 a kRan involves writin it in a !honetic al!habet, which contains only four eometric symbols& This !honetic rendition of the kRan is called the /ESSE86E;& Tortoise, 3hat do the eometric symbols look like? *chilles, They are made of he(a ons and !enta ons& ?ere is what they look like 4 picks up a near() napkin, and draws for the Tortoise these four figures5,

Tortoise, They are mysterious-lookin & *chilles, Enly to the uninitiated& 8ow once you have made the messen er, you rub your hands in some ribo, and Tortoise, Some ribo? Is that a kind of ritual anointment? *chilles, 8ot e(actly& It is a s!ecial sticky !re!aration which makes the strin hold its sha!e, when folded u!& Tortoise, 3hat is it made of? *chilles, I don1t know, e(actly& 0ut it feels sort of luey, and it works e(ceedin ly well& )nyway, once you have some ribo on your hands, you can T;)8SA)TE the sequence of symbols in the messen er into certain kinds of folds in the strin & It1s as sim!le as that& Tortoise, ?old on2 8ot so fast2 ?ow do you do that? *chilles, Hou be in with the strin entirely strai ht& Then you o to one end and start makin folds of various ty!es, accordin to the eometric symbols in the messen er&

Tortoise, So each of those eometric symbols stands for a different way to curl the strin u!? *chilles, 8ot in isolation& Hou take them three at a time, instead of one at a time& Hou be in at one end of the strin , and one end of the messen er& 3hat to do with the first inch of the strin is determined by the first three eometric symbols& The ne(t three symbols tell you how to fold the second inch of strin & )nd so you inch your way alon the strin and simultaneously alon the messen er, foldin each little se ment of strin until you have e(hausted the messen er& If you have !ro!erly a!!lied some ribo, the strin will kee! its folded sha!e, and what you thereby !roduce is the translation of the kRan into a strin & Tortoise, The !rocedure has a certain ele ance to it& Hou must et some wild-lookin strin s that way& *chilles, That1s for sure& The lon er kRans translate into quite bi<arre sha!es& Tortoise, I can ima ine& 0ut in order to carry out the translation of the messen er into the strin , you need to know what kind of fold each tri!let of eometric symbols in the messen er stands for& ?ow do you know this? :o you have a dictionary? *chilles, Hesthere is a venerated book which lists the -6eometric 7ode-& If you don1t have a co!y of this book, of course, you can1t translate a kRan into a strin & Tortoise, Evidently not& 3hat is the ori in of the 6eometric 7ode *chilles, It came from an ancient master known as -6reat Tutor- who my master says is the only one ever to attain the Enli htenment 1Hond Enli htenment& Tortoise, 6ood ravy2 )s if one level of the stuff weren1t enou h& 0ut then there are luttons of every sortwhy not luttons for enli htenment? *chilles, :o you su!!ose that -Enli htenment 1Hond Enli htenment stands for -EHE-? Tortoise, In my o!inion, it1s rather doubtful that it stands for you, )chilles& /ore likely, it stands for -/eta-Enli htenment--/E-, that is *chilles, 'or you? 3hy would it stand for you? Hou haven1t even reached the 'I;ST sta e of enli htenment, let alone the Tortoise, Hou never know, )chilles& Perha!s those who have learned the lowdown on enli htenment return to their state before enli htenment& I1ve always held that -twice enli htened is unenli htened&- 0ut let1s et back to the 6rand Tortue-uh, I mean the 6reat Tutor& *chilles, Aittle is known of him, e(ce!t that he also invented the )rt of 9en Strin s& Tortoise, 3hat is that? *chilles, It is an art on which the decision !rocedure for 0uddha-nature is based& I shall tell you about it& Tortoise, I would be fascinated& There is so much for novices like me to absorb2 *chilles, There is even re!uted to be a kRan which tells how the )rt of 9en Strin s be an& 0ut unfortunately, all this has lon since been lost in the sands of time, and is no doubt one forever& 3hich may be .ust as well, for otherwise there would be imitators who would take on the master1s name, and co!y him in other ways& Tortoise, 0ut wouldn1t it be a ood thin if all students of 9en co!ied that most enli htened master of all, the 6reat Tutor? *chilles, Aet me tell you a kRan about an imitator&

9en master 6utei raised his fin er whenever he was asked a question about 9en& ) youn novice be an to imitate him in this way& 3hen 6utei was told about the novice1s imitation, he sent for him and asked him if were true& The novice admitted it was so& 6utei asked him if he understood& In re!ly the novice held u! his inde( fin er& 6utei !rom!tly cut it off& The novice ran from the room, howlin in !ain& )s he reached the threshold, 6utei called, -0oy2- 3hen the novice turned, 6utei raised his inde( fin er& )t that instant the novice was enli htened&

Tortoise, 3ell, what do you know2 Just when I thou ht 9en was all about JRshl and his shenani ans, now I find out that 6utei is in on the merriment too& ?e seems to have quite a sense of humor& *chilles, That kRan is very serious& I don1t know how you ot the idea that it is humorous& Tortoise, Perha!s 9en is instructive because it is humorous& I would uess that if you took all such stories entirely seriously, you would miss the !oint as often as you would et it& *chilles, /aybe there1s somethin to your Tortoise-9en& Tortoise, 7an you answer .ust one question for me? I would like to know this, 3hy did 0odhidharma come from India into 7hina? *chilles, Eho2 Shall I tell you what JRshl said when he was asked that very question? Tortoise, Please do& *chilles, ?e re!lied, -That oak tree in the arden&Tortoise, Ef course= that1s .ust what I would have said& E(ce!t that I would have said it in answer to a different questionnamely, -3here can I find some shade from the midday sun?*chilles, 3ithout knowin it, you have inadvertently hit u!on one of the basic questions of all 9en& That question, innocent thou h it sounds, actually means, -3hat is the basic !rinci!le of 9en?Tortoise, ?ow e(traordinary& I hadn1t the sli htest idea that the central aim of 9en was to find some shade& *chilles, Eh, noyou1ve misunderstood me entirely& I wasn1t referrin to T?)T question& I meant your question about why 0odhidharma came from India into 7hina& Tortoise, I see& 3ell, I had no idea that I was ettin into such dee! waters& 0ut let1s come back to this curious ma!!in & I ather that any kRan can be turned into a folded strin by followin the method you outlined& 8ow what about the reverse !rocess? 7an any folded strin be read in such a way as to yield a kRan? *chilles, 3ell, in a way& ?owever&&& Tortoise, 3hat1s wron ? *chilles, Hou1re .ust not su!!osed to do it that way 1round& It would violate the 7entral :o ma of 9en strin s, you see, which is contained in this !icture 4 picks up a napkin and draws5, kRan
transcription

messen er

translation

folded strin

Hou1re not su!!osed to o a ainst the arrowses!ecially not the second one&

Tortoise, Tell me, does this :o ma have 0uddha-nature, or not? 7ome to think of it, I think I1ll unask the question& Is that all ri ht? *chilles, I am lad you unasked the question& 0utI1ll let you in on a secret& Promise you won1t tell anyone? Tortoise, Tortoise1s honor& *chilles, 3ell, once in a while, I actually do o a ainst the arrows& I et sort of an illicit thrill out of it, I uess& Tortoise, 3hy, )chilles2 I had no idea you would do somethin so irreverent2 *chilles, I1ve never confessed it to anyone beforenot even Ekanisama& Tortoise, So tell me, what ha!!ens when you o a ainst the arrows in the 7entral :o ma? :oes that mean you be in with a strin and make a kRan? *chilles, Sometimesbut some weirder thin s can ha!!en& Tortoise, 3eirder than !roducin kRans? *chilles, Hes&&& 3hen you untranslate and untranscribe, you et SE/ET?I86, but not always a kRan& Some strin s, when read out this way, only ive nonsense& Tortoise, Isn1t that .ust another name for kRans? *chilles, Hou clearly don1t have the true s!irit of 9en yet& Tortoise, :o you always et stories, at least? *chilles, 8ot alwayssometimes you et nonsense syllables, other times you et un rammatical sentences& 0ut once in a while you et what seems to be a kRan& Tortoise, It only SEE/S to be one? *chilles, 3ell, it mi ht be fraudulent, you see& Tortoise, Eh, of course& *chilles, I call those strin s which yield a!!arent kRans -well-formed- strin s& Tortoise, 3hy don1t you tell me about the decision !rocedure which allows you to distin uish !hony kRans from the enuine article? *chilles, That1s what I was headin towards& 6iven the kRan, or non-kRan, as the case may be, the first thin is to translate it into the dimensional strin & )ll that1s left is to find out if the strin has 0uddha-nature or not& Tortoise, 0ut how do you do T?)T? *chilles, 3ell, my master has said that the 6reat Tutor was able, by .ust lancin at a strin , to tell if it had 0uddha-nature or not& Tortoise, 0ut what if you have not reached the sta e of the Enli htenment, 1Hond Enli htenment? Is there no other way to tell if a strin has 0uddha-nature? *chilles, Hes, there is& )nd this is where the )rt of 9en Strin s comes in& It is a technique for makin innumerably many strin s, all of which have 0uddha-nature& Tortoise, Hou don1t say2 )nd is there a corres!ondin way of makin strin s which :E81T have 0uddha-nature? *chilles, 3hy would you want to do that? Tortoise, Eh, I .ust thou ht it mi ht be useful& *chilles, Hou have the stran est taste& Ima ine2 0ein more interested in thin s that :E81T have 0uddha-nature than thin s that :E2 Tortoise, Just chalk it u! to my unenli htened state& 0ut o on& Tell me how to make a strin which :EES have 0uddha-nature&

*chilles, 3ell, you must be in by dra!in a loo! of strin over your hands in one of five le al startin !ositions, such as this one&&& 4 &icks up a string and drapes it in a simple loop (etween a finger on each hand!/ Tortoise, 3hat are the other four le al startin !ositions? *chilles, Each one is a !osition considered to be a SEA'-EGI:E8T manner of !ickin u! a strin & Even novices often !ick u! strin s in those !ositions& )nd these five strin s all have 0uddha-nature& Tortoise, Ef course& *chilles, Then there are some Strin /ani!ulation ;ules, by which you can make more com!le( strin fi ures& In !articular, you are allowed to modify your strin by doin certain basic motions of the hands& 'or instance, you can reach across like thisand !ull like thisand twist like this& 3ith each o!eration you are chan in the overall confi uration of the strin dra!ed over your hands& Tortoise, 3hy, it looks .ust like makin cat1s-cradles and such strin fi ures2 *chilles, That1s ri ht& 8ow as you watch, you1ll see that some of these rules make the strin more com!le(= some sim!lify it& 0ut whichever way you o, as lon as you follow the Strin /ani!ulation ;ules, every strin you !roduce will have 0uddhanature& Tortoise, That is truly marvelous& 8ow what about the kRan concealed inside this strin you1ve .ust made? 3ould it be enuine? *chilles, 3hy, accordin to what I1ve learned, it must& Since I made it accordin to the ;ules, and be an in one of the five self-evident !ositions, the strin must have 0uddha-nature, and consequently it must corres!ond to a enuine kRan& Tortoise, :o you know what the kRan is? *chilles, )re you askin me to violate the 7entral :o ma? Eh, you nau hty fellow2 *nd with furrowed (row and code (ook in hand, *chilles points along the string inch () inch, recording each fold () a triplet of geometric s)m(ols of the strange phonetic alpha(et for k[an, until he has nearl) a napkinful!/ :one2 Tortoise, Terrific& 8ow let1s hear it& *chilles, )ll ri ht&
) travelin monk asked an old woman the road to Tai<an, a !o!ular tem!le su!!osed to ive wisdom to the one who worshi!s there& The old woman said, -6o strai ht ahead&- )fter the monk had !roceeded a few ste!s, she said to herself, -?e also is a common church- oer&Someone told this incident to JRshl, who said, -3ait until I investi ate&- The ne(t day he went and asked the same question, and the old woman ave the same answer& JRshl remarked, -I have investi ated that old woman&-

Tortoise, 3hy, with his flair for investi ations, it1s a shame that JRshl never was hired by the '0I& 8ow tell mewhat you did, I could also do, if I followed the ;ules from the )rt of 9en Strin s, ri ht?

*chilles, ;i ht& Tortoise, 8ow would I have to !erform the o!erations in .ust the same E;:E; as you did? *chilles, 8o, any old order will do& Tortoise, Ef course, then I would et a different strin , and consequently a different kRan& 8ow would I have to !erform the same 8F/0E; of ste!s as you did? *chilles, 0y no means& )ny number of ste!s is fine& Tortoise, 3ell, then there are an infinite number of strin s with 0uddha natureand consequently an infinite number of enuine kRans ?ow do you know there is any strin which 7)81T be made by your ;ules? *chilles, Eh, yesback to thin s which lack 0uddha-nature& It .ust so ha!!ens that once you know how to make strin s 3IT? 0uddha nature, you can also make strin s 3IT?EFT 0uddha-nature& That is somethin which my master drilled into me ri ht at the be innin & Tortoise, 3onderful2 ?ow does it work? *chilles, Easy& ?ere, for e(am!leI1ll make a strin which lacks 0uddha-nature && & ,e picks up the string out of which the preceding k[an was "pulled", ties a little teen) knot at one end of it, pulling it tight with his thum( forefinger!/ This is itno 0uddha-nature here& Tortoise, Gery illuminatin & )ll it takes is addin a knot? ?ow do you know that the new strin lacks 0uddha-nature? *chilles, 0ecause of this fundamental !ro!erty of 0uddha-nature= when two well-formed strin s are identical but for a knot at one end, then only E8E of them can have 0uddha-nature& It1s a rule of thumb which my master tau ht me& Tortoise, I1m .ust wonderin about somethin & )re there some strin s with 0uddha-nature which you 7)81T reach by followin the ;ules of 9en Strin s, no matter in what order? *chilles, I hate to admit it, but I am a little confused on this !oint myself& )t first my master ave the stron est im!ression that 0uddha-nature in a strin was :E'I8E: by startin in one of the five le al !ositions, and then develo!in the strin accordin to the allowed ;ules& 0ut then later, he said somethin about somebody-or-other1s -Theorem-& I never ot it strai ht& /aybe I even misheard what he said& 0ut whatever he said, it !ut some doubt in my mind as to whether this method hits )AA strin s with 0uddha-nature& To the best of my knowled e, at least, it does& 0ut 0uddha-nature is a !retty elusive thin , you know& Tortoise, I athered as much, from JRshl1s 1/F1& I wonder&&& *chilles, 3hat is it? Tortoise, I was .ust wonderin about those two kRansI mean the kRan and its un-kRan the ones which say -This mind is 0uddha- and -This mind is not 0uddha-what do they look like, when turned into strin s via the 6eometric 7ode? *chilles, I1d be lad to show you&

,e writes down the phonetic transcriptions, and then pulls from his pocket a couple of pieces of string, which he carefull) folds inch () inch, following the triplets of s)m(ols written in the strange alpha(et! Then he places the finished strings side () side!/ Hou see, here is the difference& Tortoise, They are very similar, indeed& 3hy, I do believe there is only one difference between them, it1s that one of them has a little knot on its end2 *chilles, 0y JRshl, you1re ri ht& Tortoise, )ha2 8ow I understand why your master is sus!icious& *chilles, Hou do? Tortoise, )ccordin to your rule of thumb, )T /EST E8E of such a !air can have 0uddhanature, so you know ri ht away that one of the kRans must be !hony& *chilles, 0ut that doesn1t tell which one is !hony& I1ve worked, and so has my master, at tryin to !roduce these two strin s by followin the Strin /ani!ulation ;ules, but to no avail& 8either one ever turns u!& It1s quite frustratin & Sometimes you be in to wonder&&& Tortoise, Hou mean, to wonder if either one has 0uddha-nature? Perha!s neither of them has 0uddha-natureand neither kRan is enuine2 *chilles, I never carried my thou hts as far as thatbut you1re ri htit1s !ossible, I uess& 0ut I think you should not ask so many questions about 0uddha-nature& The 9en master /umon always warned his !u!ils of the dan er of too many questions& Tortoise, )ll ri htno more questions& Instead, I have a sort of hankerin to make a strin myself& It would be amusin to see if what I come u! with is well-formed or not& *chilles, That could be interestin & ?ere1s a !iece of strin & 4 ,e passes one to the Tortoise!/ Tortoise, 8ow you reali<e that I don1t have the sli htest idea what to do& 3e1ll .ust have to take !otluck with my awkward !roduction, which will follow no rules and will !robably wind u! bein com!letely undeci!herable& 4Grasps the loop (etween his feet and, with a few simple manipulations, creates a comple9 string which he proffers wordlessl) to *chilles! *t that moment, *chilles' face lights up!/ *chilles, Jee!ers cree!ers2 I1ll have to try out your method myself& I have never seen a strin like this2 Tortoise, I ho!e it is well-formed& *chilles, I see it1s ot a knot at one end& Tortoise, Eh .ust a moment2 /ay I have it back? I want to do one thin to it& *chilles, 3hy, certainly& ?ere you are& ,ands it (ack to the Tortoise, who ties another knot at the same end! Then the Tortoise gi#es a sharp tug, and suddenl) (oth knots disappearA/ *chilles, 3hat ha!!ened?

Tortoise, I wanted to et rid of that knot& *chilles, 0ut instead of untyin it, you tied another one, and then 0ET? disa!!eared2 3here did they o? Tortoise, Tumbolia, of course& That1s the Aaw of :ouble 8odulation& -uddenl), the two knots reappear from out of nowhereOthat is to sa), Tum(olia!/ *chilles, )ma<in & They must lie in a fairly accessible layer of Tumbolia if they can !o! into it and out of it so easily& Er is all of Tumbolia equally inaccessible? Tortoise, I couldn1t say& ?owever, it does occur to me that burnin the strin would make it quite im!robable for the knots to come back& In such a case, you could think of them as bein tra!!ed in a dee!er layer of Tumbolia& Perha!s there are layers and layers of Tumbolia& 0ut that1s neither here nor there& 3hat I would like to know is how my strin sounds, if you turn it back into !honetic symbols& *s he hands it (ack, once again, the knots pop into o(li#ion!/ *chilles, I always feel so uilty about violatin the 7entral :o ma&&& 4 Takes out his pen and code (ook, and carefull) Mots down the man) s)m(ol%triplets which correspond to the cur#) in#olutions of the Tortoise's stringR and when he is finished, he clears his #oice&5 )hem& )re you ready to hear what you have wrou ht? Tortoise, I1m willin if you1re willin & *chilles, )ll ri ht& It oes like this,
) certain monk had a habit of !esterin the 6rand Tortue 4the only one who had ever reached the Enli htenment 1Hond Enli htenment5, by askin whether various ob.ects had 0uddha-nature or not& To such questions Tortue invariably sat silent& The monk had already asked about a bean, a lake, and a moonlit ni ht& Ene day, he brou ht to Tortue a !iece of strin , and asked the same question& In re!ly, the 6rand Tortue ras!ed the loo! between his feet and

Tortoise, 0etween his feet? ?ow odd2 *chilles, 3hy should HEF find that odd? Tortoise, 3ell, ah&&& you1ve ot a !oint there& 0ut !lease o on2 *chilles, )ll ri ht&
The 6rand Tortue ras!ed the loo! between his feet and, with a few sim!le mani!ulations, created a com!le( strin which he !roffered wordlessly to the monk& )t that moment, the monk was enli htened&

Tortoise, I1d rather be twice-enli htened, !ersonally& *chilles, Then it tells how to make the 6rand Tortue1s strin , if you be in with a strin dra!ed over your feet& I1ll ski! those borin details& It concludes this way,
'rom then on, the monk did not bother Tortue& Instead, he made strin after strin by Tortue1s method= and he !assed the method on to his own disci!les, who !assed it on to theirs&

Tortoise, Muite a yarn& It1s hard to believe it was really hidden inside my strin &

*chilles, Het it was& )stonishin ly, you seem to have created a well-formed strin ri ht off the bat& Tortoise, 0ut what did the 6rand Tortue1s strin look like? That1s the main !oint of this kRan, I1d su!!ose& *chilles, I doubt it& Ene shouldn1t -attach- to small details like that inside kRans& It1s the s!irit of the whole kRan that counts, not little !arts of it& Say, do you know what I .ust reali<ed? I think, cra<y thou h it sounds, that you may have hit u!on that lon -lost kRan which describes the very ori in of the )rt of 9en Strin s2 Tortoise, Eh, that would almost be too ood to have 0uddha-nature& *chilles, 0ut that means that the reat masterthe only one who ever reached the mystical state of the Enli htenment 1Hond Enli htenmentwas named -Tortue-, not -Tutor-& 3hat a droll name2 Tortoise, I don1t a ree& I think it1s a handsome name& I still want to know how Tortue1s strin looked& 7an you !ossibly recreate it from the descri!tion iven in the kRan? *chilles, I could try&&& Ef course, I1ll have to use my feet, too, since it1s described in terms of foot motions& That1s !retty unusual& 0ut I think I can mana e it& Aet me ive it a o& 4,e picks up the k[an and a piece of string, and for a few minutes twists and (ends the string in arcane wa)s until he has the finished product &5 3ell, here it is& Edd, how familiar it looks& Tortoise, Hes, isn1t that so? I wonder where I saw it before? *chilles, I know2 3hy, this is HEF; strin , /r& T2 Er is it? Tortoise, 7ertainly not& *chilles, Ef course notit1s the strin which you first handed to me, before you took it back to tie an e(tra knot in it& Tortoise, Eh, yesindeed it is& 'ancy that& I wonder what that im!lies& *chilles, It1s stran e, to say the least& Tortoise, :o you su!!ose my kRan is enuine? *chilles, 3ait .ust a moment&&& Tortoise, Er that my strin has 0uddha-nature? *chilles, Somethin about your strin is be innin & to trouble me, /r& Tortoise& Tortoise 4looking most pleased with himself and pa)ing no attention to *chilles 5, )nd what about Tortue1s strin ? :oes it have 0uddha nature? There are a host of questions to ask2 *chilles, I would be scared to ask such questions, /r& T& There is somethin mi hty funny oin on here, and I1m not sure I like it& Tortoise, I1m sorry to hear it& I can1t ima ine what1s troublin you& *chilles, 3ell, the best way I know to e(!lain it is to quote the words of another old 9en master, @yo en& @yo en said,
9en is like a man han in in a tree by his teeth over a !reci!ice& ?is hands ras! no branch, his feet rest on no limb, and under the tree another !erson asks him, -3hy did 0odhidharma come to 7hina from India?- If the man in the tree does not answer, he fails= and if he does answer, falls and loses his life& 8ow what shall he do?

Tortoise, That1s clear= he should ive u! 9en, and take u! molecular biolo y&

C/A,TE$ I<

Mumon and G#del

>hat Is 2en7
I1/ 8ET SF;E I know what 9en is& In a way, I think I understand it very well= but in a way, I also think I can never understand it at all& Ever since my freshman En lish teacher in colle e read JRshl1s /F out loud to our class, I have stru led with 9en as!ects of life, and !robably I will never cease doin so& To me, 9en is intellectual quicksandanarchy, darkness, meanin lessness, chaos& It is tantali<in and infuriatin & )nd yet it is humorous, refreshin , enticin & 9en has its own s!ecial kind of meanin , bri htness, and clarity& I ho!e that in this 7ha!ter, I can et some of this cluster of reactions across to you& )nd then, stran e thou h it may seem, that will lead us directly to 6>delian matters& Ene of the basic tenets of 9en 0uddhism is that there is no way to characteri<e what 9en is& 8o matter what verbal s!ace you try to enclose 9en in, it resists, and s!ills over& It mi ht seem, then, that all efforts to e(!lain 9en are com!lete wastes of time& 0ut that is not the attitude of 9en masters and students& 'or instance, 9en kRans are a central !art of 9en study, verbal thou h they are& kRans are su!!osed to be -tri ers- which, thou h they do not contain enou h information in themselves to im!art enli htenment, may !ossibly be sufficient to unlock the mechanisms inside one1s mind that lead to enli htenment& 0ut in eneral, the 9en attitude is that words and truth are incom!atible, or at least that no words can ca!ture truth&

2en Master Mumon

Possibly in order to !oint this out in an e(treme way, the monk /umon 4-8o- ate-5, in the thirteenth century, com!iled forty-ei ht kRans, followin each with a commentary and a small -!oem-& This work is called -The 6ateless 6ate- or the /umonkan 4-8o- ate barrier-5& It is interestin to note that the lives of /umon and 'ibonacci coincided almost e(actly, /umon livin from ""*% to "BD+ in 7hina, 'ibonacci from ""*+ to "BL+ in Italy& To those who would look to the /umonkan in ho!es of makin sense of, or -understandin -, the kRans, the /umonkan may come as a rude shock, for the comments and !oems are entirely as o!aque as the kRans which they are su!!osed to clarify& Take this, for e(am!le,"

73G5=E DJ! Three 3orlds, () '! 8! Escher lithograph, BIGG/

;[an:
?R en of Seiryo monastery was about to lecture before dinner when he noticed that the bamboo screen, lowered for meditation, had not been rolled u!& ?e !ointed to it& Two monks arose wordlessly from the audience and rolled it u!& ?R en, observin the !hysical moment, said, -The state of the first monk is ood, not that of the second&-

'umon's 8ommentar):
I want to ask you, which of those two monks ained and which lost? If any of you has one eye, he will see the failure on the teacher1s !art& ?owever, I am not discussin ain and loss&

'umon's &oem:
3hen the screen is rolled u! the reat sky o!ens, Het the sky is not attuned to 9en& It is best to for et the reat sky )nd to retire from every wind&

Er then a ain, there is this one,B ;[an:

6oso said, -3hen a buffalo oes out of his enclosure to the ed e of the abyss, his horns and his head and his hoofs all !ass throu h, but why can1t the tail also !ass?-

'umon's 8ommentar): If anyone can o!en one eye at this !oint and say a word of 9en, he is qualified to re!ay the four ratifications, and, not only that, he can save all sentient bein s under him& 0ut if he cannot say such a word of 9en, he should turn back to his tail& 'umon's &oem:
If the buffalo runs, he will fall into the trench= If he returns, he will be butchered& That little tail Is a very stran e thin &

I think you will have to admit that /umon does not e(actly clear everythin u!& Ene mi ht say that the metalan ua e 4in which /umon writes5 is not very different from the ob.ect lan ua e 4the lan ua e of the kRan5& )ccordin to some, /umon1s comments are intentionally idiotic, !erha!s meant to show how useless it is to s!end one1s time in chatterin about 9en& ?owever, /umon1s comments can be taken on more than one level& 'or instance, consider this,% ;[an:
) monk asked 8ansen, -Is there a teachin no master ever tau ht before?8ansen said, -Hes, there is&-3hat is it?- asked the monk& 8ansen re!lied, -It is not mind, it is not 0uddha, it is not thin s&-

73G5=E DC! :ewdro!, () '! 8! Escher me@@otint, BIDE/!

'umon's 8ommentar): Eld 8ansen ave away his treasure-words& ?e must have been reatly u!set& 'umon's &oem:
8ansen was too kind and lost his treasure& Truly, words have no !ower& Even thou h the mountain becomes the sea, 3ords cannot o!en another1s mind&

In this !oem /umon seems to be sayin somethin very central to 9en, not makin idiotic statements& 7uriously, however, the !oem is referential, and thus it is a comment not only on 8ansen1s words, but on its own ineffectiveness& This ty!e of !arado( is quite characteristic of 9en& It is an attem!t to -break the mind of lo ic-& Hou see this !arado( quality in the kRan, as well& 7oncernin /umon1s commentary, do think that 8ansen was really so sure of his answer? Er did the -correctness- of his answer matter at all? Er does correctness !lay any role in 9en? 3hat is the difference between correctness and truth, or is there any? 3hat if 8ansen had said, -8o, there is not any such teachin -? 3ould it have made any difference? 3ould his remark have been immortali<ed in a kRan?

73G5=E DE! )nother 3orld, () '! 8! Escher wood%engra#ing, BIDC/!

?ere is another kRan which aims to break the mind of lo ic,C

The student :oko came to a 9en master, and said, -I am seekin the truth& In what state of mind should I train myself, so as to find it?Said the master, -There is no mind, so you cannot !ut it in any state& There is no truth, so you cannot train yourself for it&-If there is no mind to train, and no truth to find, why do you have these monks ather before you every day to study 9en and train themselves for this study?-0ut I haven1t an inch of room here,- said the master, -so how could the monks ather? I have no ton ue, so how could I call them to ether or teach them?-Eh, how can you lie like this?- asked :oko& -0ut if I have no ton ue to talk to others, how can I lie to you?- asked the master& Then :oko said sadly, -I cannot follow you& I cannot understand you&-I cannot understand myself,- said the master&

If any kRan serves to bewilder, this one does& )nd most likely, causin bewilderment is its !recise !ur!ose, for when one is in a bewildered state, one1s mind does be in to o!erate nonlo ically, to some e(tent& Enly by ste!!in outside of lo ic, so the theory oes, can one make the lea! to enli htenment& 0ut what is so bad about lo ic? 3hy does it !revent the lea! to enli htenment?

2en%s Struggle Against

ualism

To answer that, one needs to understand somethin about what enli htenment is& Perha!s the most concise summary of enli htenment would be, transcendin dualism& 8ow what is dualism? :ualism is the conce!tual division of the world into cate ories& Is it !ossible to transcend this natural tendency? 0y !refi(in the word -division- by the word -conce!tual-, I may have made it seem that this is an intellectual or conscious effort, and !erha!s thereby iven the im!ression that dualism could be overcome sim!ly by su!!ressin thou ht 4as if to su!!ress thinkin actually were sim!le25& 0ut the breakin of the world into cate ories takes !lace far below the u!!er strata of thou ht= in fact, dualism is .ust as much a !erce!tual division of the world into cate ories as it is a conce!tual division& In other words, human !erce!tion is by nature a dualistic !henomenonwhich makes the quest for enli htenment an u!hill stru le, to say the least& )t the core of dualism, accordin to 9en, are words.ust !lain words& The use of words is inherently dualistic, since each word re!resents, obviously, a conce!tual cate ory& Therefore, a ma.or !art of 9en is the fi ht a ainst reliance on words& To combat the use of words, one of the devices is the kRan, where words are so dee!ly abused that one1s mind is !ractically left reelin , if one takes the kRans seriously& Therefore !erha!s it is wron to say that the enemy of enli htenment is lo ic= rather it is dualistic, verbal thinkin & In fact, it is even more basic than that, it is !erce!tion& )s soon as you !erceive an ob.ect, you draw a line between it and the rest of the world= you divide the world, artificially, into !arts, and you thereby miss the 3ay& ?ere is a kRan which demonstrates the stru ;[an:
Shu<an held out his short staff and said, -If you call this a short staff, you o!!ose its reality& If you do not call it a short staff, you i nore the fact& 8ow, what do you wish to call this?-

le a ainst words,L

73G5=E DI! :ay and 8i ht, () '! 8! Escher woodcut, BIHE/!

'umon's 8ommentar): If you call this a short staff, you o!!ose its reality& If you do not call it a short staff, you i nore the fact& It cannot be e(!ressed with words and it cannot be e(!ressed without words& 8ow say quickly what it is& 'umon's &oem:
?oldin out the short staff, ?e ave an order of life or death& Positive and ne ative interwoven, Even 0uddhas and !atriarchs cannot esca!e this attack&

4-Patriarchs- refers to si( venerated founders of 9en 0uddhism, of whom 0odhidharma is the first, and EnR is the si(th&5 3hy is callin it a short staff o!!osin its reality? Probably because such a cate ori<ation ives the a!!earance of ca!turin reality, whereas the surface has not even been scratched by such a statement& It could be com!ared to sayin -L is a !rime number-& There is so much morean infinity of factsthat has been omitted& En the other hand, not to call it a staff is, indeed, to i nore that !articular fact, minuscule as it may be& Thus words lead to some truthsome falsehood, !erha!s, as wellbut certainly not to all truth& ;elyin on words to lead you to the truth is like relyin on an incom!lete formal system to lead you to the truth& ) formal system will ive you some truths, but as we shall soon see, a formal systemno matter how !owerfulcannot lead to all truths& The dilemma of mathematicians is, what else is there to rely on, but formal systems? )nd

the dilemma of 9en !eo!le is, what else is there to rely on, but words? /umon states the dilemma very clearly, -It cannot be e(!ressed with words and it cannot be e(!ressed without words&?ere is 8ansen, once a ain,D
JRshl asked the teacher 8ansen, -3hat is the true 3ay?8ansen answered, -Everyday way is the true 3ay&JRshl asked, -7an I study it?8ansen answered, -The more you study, the further from the 3ay&JRshl asked, -If I don1t study it, how can I know it?8ansen answered, -The 3ay does not belon to thin s seen, nor to thin s unseen& It does not belon to thin s known, nor to thin s unknown& :o not seek it, study it, or name it& To find yourself on it, o!en yourself wide as the sky&- USee 'i & L+&V 73G5=E GK! ;ind, () '! 8! Escher wood%engra#ing, BIGG/!

This curious statement seems to abound with !arado(& It is a little reminiscent of this surefire cure for hiccu!s, -;un around the house three times without thinkin of the word 1wolf1&- 9en is a !hiloso!hy which seems to have embraced the notion that the road to ultimate truth, like the only surefire cure for hiccu!s, may bristle with !arado(es&

Ism) The &n-Mode) and &nmon

If words are bad, and thinkin is bad, what is ood? Ef course, to ask this is already horribly dualistic, but we are makin no !retense of bein faithful to 9en in discussin 9enso we can try to answer the question seriously& I have a name for what 9en strives for, ism& Ism is an anti!hiloso!hy, a way of bein without thinkin & The masters of ism are rocks, trees, clams= but it is the fate of hi her animal s!ecies to have to strive for ism, without ever bein able to attain it fully& Still, one is occasionally ranted lim!ses of ism& Perha!s the followin kRan offers such a lim!se,$
?yaku.o wished to send a monk to o!en a new monastery& ?e told his !u!ils that whoever answered a question most ably would be a!!ointed& Placin a water vase on the round, he asked, -3ho can say what this is without callin its name?The chief monk said, -8o one can call it a wooden shoe&Isan, the cookin monk, ti!!ed over the vase with his foot and went out& ?yaku.o smiled and said, -The chief monk loses&- )nd Isan became the master of the new monastery&

To su!!ress !erce!tion, to su!!ress lo ical, verbal, dualistic thinkin this is the essence of 9en, the essence of ism& This is the Fnmodenot Intelli ent, not /echanical, .ust -Fn-& JRshl was in the Fnmode, and that is why his 1/F1 unasks the question& The Fnmode came naturally to 9en /aster Fnmon,*
Ene day Fnmon said to his disci!les, -This staff of mine has transformed itself into a dra on and has swallowed u! the universe2 Eh, where are the rivers and mountains and the reat earth?-

9en is holism, carried to its lo ical e(treme& If holism claims that thin s can only be understood as wholes, not as sums of their !arts, 9en oes one further, in maintainin that the world cannot be broken into !arts at all& To divide the world into !arts is to be deluded, and to miss enli htenment&
) master was asked the question, -3hat is the 3ay?- by a curious monk& -It is ri ht before your eyes,- said the master& -3hy do I not see it for myself?-0ecause you are thinkin of yourself&-3hat about you, do you see it?-So lon as you see double, sayin 1I don1t1, and 1you do1, and so on, your eyes are clouded,- said the master& -3hen there is neither 1I1 nor 1Hou1, can one see it?-3hen there is neither 1I1 nor 1Hou1, who is the one that wants to see it?- #

)!!arently the master wants to et across the idea that an enli htened state is one where the borderlines between the self and the rest of universe are dissolved& This would truly be the end of dualism, for as he says, there is no system left which has any desire for !erce!tion& 0ut what is that state, if not death? ?ow can a live human bein dissolve the borderlines between himself and the outside world?

2en and Tumbolia

The 9en monk 0assui wrote a letter to one of his disci!les who was about to die, and in it he said, -Hour end which is endless is as a snowflake dissolvin in the !ure air&- The snowflake, which was once very much a discernible subsystem of the universe, now dissolves into the lar er system which once held it& Thou h it is no lon er !resent as a distinct subsystem, its essence somehow still !resent, and will remain so& It floats in Tumbolia, alon hiccu!s that are not bein hiccu!!ed and characters in stories that are bein read&&& That is how I understand 0assui1s messa e& 9en reco ni<es its own limitations, .ust as mathematicians have learned to reco ni<e the limitations of the a(iomatic method as a method for attainin truth& This does not mean that 9en has an answer to what lies beyond 9en any more than mathematicians have a clear understandin the forms of valid reasonin which lie outside of formali<ation& Ene of the clearest 9en statements about the borderlines of 9en is iven in the followin stran e kRan, very much in the s!irit of 8ansen,"+
TR<an said to his monks, -Hou monks should know there is an even hi her understandin in 0uddhism&- ) monk ste!!ed forward and asked, -3hat the hi her 0uddhism?- TR<an answered, -It is not 0uddha&-

There is always further to o= enli htenment is not the end-all of 9en& )nd there is no reci!e which tells how to transcend 9en= the only thin one can rely on for sure is that 0uddha is not the way& 9en is a system and cannot be its own metasystem= there is always somethin outside of 9en, which cannot be fully understood or described within 9en&

Escher and 2en

In questionin !erce!tion and !osin absurd answerless riddles, 9en has com!any, in the !erson of /& 7& Escher& 7onsider .a) and Night 4'i & C#5, a master!iece of -!ositive and ne ative interwoven- 4in the words of /umon5& Ene mi ht ask, -)re those really birds, or are they really fields? Is it really ni ht, or day?- Het we all know there is no !oint to such questions& The !icture, like a 9en kRan, is tryin to break the mind of lo ic& Escher also deli hts in settin u! contradictory !ictures, such as *nother >orld 4'i & C*5 !ictures that !lay with reality and unreality the same way as 9en !lays with reality and unreality& Should one take Escher seriously? Should one take 9en seriously?

73G5=E GB! Puddle, () '! 8! Escher woodcut, BIGF/&

There is a delicate, haiku-like study of reflections in .ewdrop 4'i & C$5= and then there are two tranquil ima es of the moon reflected in still waters, &uddle 4'i & L"5, and =ippled -urface 4'i & LB5& The reflected moon is a theme which recurs in various kRans& ?ere is an e(am!le,""
7hiyono studied 9en for many years under 0ukko of En aku& Still, she could not attain the fruits of meditation& )t last one moonlit ni ht she was carryin water in an old wooden !ail irded with bamboo& The bamboo broke, and the bottom fell out of the !ail& )t that moment, she was set free& 7hiyono said, -8o more water in the !ail, no more moon in the water&-

Three >orlds, an Escher !icture 4'i & CD5, and the sub.ect of a 9en kRan,"B
) monk asked 6antR, -3hen the three worlds threaten me, what shall I do?- 6antR answered, -Sit down&- -I do not understand,- said the monk& 7anto said, -Pick u! the mountain and brin it to me& Then I will tell you&-

/emiolia and Escher

In <er(um 4'i & "C#5, o!!ositions are made into unities on several levels& 6oin around we see radual transitions from black birds to white birds to black fish to white fish to black fro s to white fro s to black birds&&& )fter si( ste!s, back where we started2 Is this a reconciliation of the dichotomy of black and white? Er of the trichotomy of birds, fish, and fro s? Er si(fold unity made from the o!!osition of the evenness of B an oddness of %? In music, si( notes of equal time value create a rhythmic ambi uityare they B rou!s of %, or % rou!s of B? This ambi uity has a name, hemiolia& 7ho!in was a master of hemiolia, see his 3alt< o!& CB, or his Etude o!& BL, no& B& In 0ach, there is the Tempo di 'enuetto from the keyboard Partita no& L, or the incredible 7inale of the first Sonata for unaccom!anied violin, in 6 /inor& )s one lides inward toward the center of <er(um, the distinctions radually blur, so that in the end there remains not three, not two, but one sin le essence, -GE;0F/-, which lows with brilliancy!erha!s a symbol of enli htenment& Ironically, -verbumnot only is a word, but means -word-not e(actly the most com!atible notion with 9en& En the hand, -verbum- is the only word in the !icture& )nd 9en master TR<an once said, -The com!lete Tri!itaka can be e(!ressed in one character&- 4-Tri!itaka-, meanin -three baskets-, refers to the com!lete te(ts of the ori inal 0uddhist writin s&5 3hat kind of decodin -mechanism, I wonder, would it take to suck the three baskets out of one character? Perha!s one with two hemis!heres&
73G5=E GF! ;i!!led Surface, () '! 8! Escher lino%cut, BIGK/!

73G5=E GH! Three S!heres II, () '! 8! Escher lithograph, BIDJ/!

Indra%s ;et
'inally, consider Three -pheres 33 4'i & L%5, in which every !art of the world seems to contain, and be contained in, every other !art, the writin table reflects the s!heres on to! of it, the s!heres reflect each other, as well as the writin table, the drawin of them, and the artist drawin it& The endless connections which all thin s have to each other is only hinted at here, yet the hint is enou h& The 0uddhist alle ory of -Indra1s 8et- tells of an endless net of threads throu hout the universe, the hori<ontal threads runnin throu h s!ace, the vertical ones throu h time& )t every crossin of threads is an individual, and every individual is a crystal bead& The reat li ht of -)bsolute 0ein - illuminates and !enetrates every crystal bead= moreover, every crystal bead reflects not only the li ht from every other crystal in the netbut also every reflection of every reflection throu hout the universe& To my mind, this brin s forth an ima e of renormali<ed !articles, in every electron, there are virtual !hotons, !ositrons, neutrinos, muons&&& = in every !hoton, there are virtual electrons, !rotons, neutrons, !ions&&& = in every !ion, there are&&& 0ut then another ima e rises, that of !eo!le, each one reflected in the minds of many others, who in turn are mirrored in yet others, and so on& 0oth of these ima es could be re!resented in a concise, ele ant way by usin )u mented Transition 8etworks& In the case of !articles, there would be one network for each cate ory of !article= in the case of !eo!le, one for each !erson& Each one would contain calls to many others, thus creatin a virtual cloud of )T81s around each )T8&

7allin one would create calls on others, and this !rocess mi ht cascade arbitrarily far, until it bottomed out&

Mumon on M&
Aet us conclude this brief e(cursion into 9en by returnin comment on JRshl1s /F,"% to /umon& ?ere is his

To reali<e 9en one has to !ass throu h the barrier of the !atriarchs& Enli htenment always comes after the road of thinkin is blocked& If you do not !ass the barrier of the !atriarchs or if your thinkin road is not blocked, whatever you think, whatever you do, is like a tan lin host& Hou may ask -3hat is a barrier of a !atriarch?- This one word, 1/F1, is it& This is the barrier of 9en& If you !ass throu h it, you will see JRshl face to face& Then you can work hand in hand with the whole line of !atriarchs& Is this not a !leasant thin to do? If you want to !ass this barrier, you must work throu h every bone in your body, throu h every !ore of your skin, filled with this question, -3hat is 1/F1?- and carry it day and ni ht& :o not believe it is the common ne ative symbol meanin nothin & It is not nothin ness, the o!!osite of e(istence& If you really want to !ass this barrier, you should feel like drinkin a hot iron ball that you can neither swallow nor s!it out& Then your !revious lesser knowled e disa!!ears& )s a fruit ri!enin in season, your sub.ectivity and ob.ectivity naturally become one& It is like a dumb man who has had a dream& ?e knows about it but he cannot tell it& 3hen he enters this condition his e o-shell is crushed and he can shake the heaven and move the earth& ?e is like a reat warrior with a shar! sword& If a 0uddha stands in his way, he will cut him down= if a !atriarch offers him any obstacle, he will kill him= and he will be free in his way of birth and death& ?e can enter any world as if it were his own !lay round& I will tell you how to do this with this kRan, Just concentrate your whole ener y into this /F, and do not allow any discontinuation& 3hen you enter this /F and there is no discontinuationyour attainment will be as a candle burnin and illuminatin the whole universe&

+rom Mumon to the M&-!u::le

'rom the ethereal hei hts of JRshl1s /F, we now descend to the !rivate lowlinesses of ?ofstadter1s M&&&& I know that you have already concentrated your whole ener y into this M& 4when you read 7ha!ter "5& So now I wish to answer the question which was !osed there, ?as M& theorem-nature, or not? The answer to this question is not an evasive /F= rather, it is a resoundin show this, we will take advanta e of dualistic, lo ical thinkin &
8E&

In order to

3e made two crucial observations in 7ha!ter I, 4"5 that the /F-!u<<le has de!th lar ely because it involves the inter!lay of len thenin and shortenin rules= 4B5 that ho!e nevertheless e(ists for crackin the !roblem by em!loyin a tool which is in some sense of adequate de!th to handle matters of that com!le(ity, the theory of numbers& 3e did not analy<e the /F-!u<<le in those terms very carefully in 7ha!ter I= we shall do so now& )nd we will see how the second observation 4when enerali<ed beyond the insi nificant /IF-system5 is one of the most fruitful reali<ations of all mathematics, and how it chan ed mathematicians1 view of their own disci!line& 'or your ease of reference, here is a reca!itulation of the /IF-system, SH/0EAS, /, I, F )TIE/, /I ;FAES, I& If 9I is a theorem, so is 9I&& II& If M9 is a theorem, so is M99& III& In any theorem, III can be re!laced by && IG& && can be dro!!ed from any theorem&

Mumon Sho's &s /o' to Sol"e the M&-!u::le

)ccordin to the observations above, then, the /F-!u<<le is merely a !u<<le about natural numbers in ty!o ra!hical dis uise& If we could only find a way to transfer it to the domain of number theory, we mi ht be able to solve it& Aet us !onder the words of /umon, who said, -If any of you has one eye, he will see the failure on the teacher1s !art&- 0ut why should it matter to have one eye? If you try countin the number of I1s contained in theorems, you will soon notice that it seems never to be +& In other words, it seems that no matter how much len thenin and shortenin is involved, we can never work in such a way that all I1s are eliminated& Aet us call the number of I1s in any strin the I-count of that strin & 8ote that the I-count of the a(iom /I is "& 3e can do more than show that the I-count can1t be +we can show that the I-count can never be any multi!le of %& To be in with, notice that rules I and IG leave the I-count totally undisturbed& Therefore we need only think about rules II and III& )s far as rule III is concerned, it diminishes the I-count by e(actly %& )fter an a!!lication of this rule, the I-count of the out!ut mi ht conceivably be a multi!le of %but only if the I-count of the in!ut was also& ;ule III, in short, never creates a multi!le of % from scratch& It can only create one when it be an with one& The same holds for rule II, which doubles the I-count& The

reason is that if % divides Bn, thenbecause % does not divide Bit must divide n 4a sim!le fact from the theory of numbers5& 8either rule II nor rule III can create a multi!le of % from scratch& 0ut this is the key to the /F-!u<<le2 ?ere is what we know, 4"5 The I-count be ins at " 4not a multi!le of %5= 4B5 Two of the rules do not affect the I-count at all= 4%5 4%5 The two remainin rules which do affect the I-count do so in such a way as never to create a multi!le of % unless iven one initially& The conclusionand a ty!ically hereditary one it is, toois that I-count can never become any multi!le of %& In !articular, + is a forbidden value of the I-count& ?ence, M& is not a theorem of the '35%s)stem! 8otice that, even as a !u<<le about I-counts, this !roblem was !la ued by the crossfire of len thenin and shortenin rules& 9ero became the oal= I-counts could increase 4rule II5, could decrease 4rule III5& Fntil we analy<ed the situation, we mi ht have thou ht that, with enou h switchin back and forth between the rules, we mi ht eventually hit +& 8ow, thanks to a sim!le number-theoretical ar ument, we know that this is im!ossible&

G#del-;umbering the MI&-System

8ot all !roblems of the the ty!e which the /F-!u<<le symboli<es are so easy to solve as this one& 0ut we have seen that at least one such !u<<le could be embedded within, and solved within, number theory& 3e are oin to see that there is a way to embed all !roblems about any formal system, in number theory& This can ha!!en thanks to the discovery, by 6>del, of a s!ecial kind of isomor!hism& To illustrate it, I will use /IFsystem& 3e be in by considerin the notation of the /IF-system& 3e ma! each symbol onto a new symbol, M % I " & + The corres!ondence was chosen arbitrarily= the only rhyme or reason is that each symbol looks a little like the one it is ma!!ed onto& Each number is called the 6>del number of the corres!ondin letter& 8ow I am sure you can uess what the 6>del number of a multiletter strin will be,

M& %+ MII& %""+ etc& It is easy& 7learly this ma!!in between notations is an information-!reservin transformation= it is like !layin the same melody on two different instruments& Aet us now take a look at a ty!ical derivation in the /IF-system, written simultaneously in both notations, 4"5 MI 4B5 MII 4%5 MIIII 4C5 M&I 4L5 M&I& 4D5 M&I&&I& 4$5 M&II& a(iom rule B rule B rule % rule " rule B rule C %" %"" %"""" %+" %+"+ %+"++"+ %+""+

The left-hand column is obtained by a!!lyin our four familiar ty!o ra!hical rules& The ri ht-hand column, too, could be thou ht of as havin been enerated by a similar set of ty!o ra!hical rules& Het the ri ht-hand column has a dual nature& Aet me e(!lain what this means&

Seeing Things Both Ty!ogra!hically and Arithmetically

3e could say of the fifth strin 41%+"+15 that it was made from the fourth, by a!!endin a 1+1 on the ri ht= on the other hand we could equally well view the transition as caused by an arithmetical o!erationmulti!lication by "+, to be e(act& 3hen natural numbers are written in the decimal system, multi!lication by "+ and !uttin a 1+1 on the ri ht are indistin uishable o!erations& 3e can take advanta e of this to write an arithmetical rule which corres!onds to ty!o ra!hical rule I, );IT?/ETI7)A ;FAE Ia, ) number whose decimal e(!ansion ends on the ri ht in 1"1 can be multi!lied by "+& 3e can eliminate the reference to the symbols in the decimal e(!ansion by arithmetically describin the ri htmost di it, );IT?/ETI7)A ;FAE Ib, ) number whose remainder when divided by "+ is ", can be multi!lied by "+& 8ow we could have stuck with a !urely ty!o ra!hical rule, such as the followin one,

THPE6;)P?I7)A ;FAE I, 'rom any theorem whose ri htmost symbol is 1"1 a new theorem can be made, by a!!endin 1+1 to the ri ht of that 1"1& They would have the same effect& This is why the ri ht-hand column has a -dual nature-, it can be viewed either as a series of ty!o ra!hical o!erations chan in one !attern of symbols into another, or as a series of arithmetical o!erations chan in one ma nitude into another& 0ut there are !owerful reasons for bein more interested in the arithmetical version& Ste!!in out of one !urely ty!o ra!hical system into another isomor!hic ty!o ra!hical system is not a very e(citin thin to do= whereas ste!!in clear out of the ty!o ra!hical domain into an isomor!hic !art of number theory has some kind of une(!lored !otential& It is as if somebody had known musical scores all his life, but !urely visuallyand then, all of a sudden, someone introduced him to the ma!!in between sounds a musical scores& 3hat a rich, new world2 Then a ain, it is as if somebody had been familiar with strin fi ures all his life, but !urely as strin fi ures devoid of meanin and then, all of a sudden, someone introduced him the ma!!in between stories and strin s& 3hat a revelation2 The discovery of 6>del-numberin has been likened to the discovery, by :escartes, of the isomor!hism between curves in a !lane and equations in two variables= incredibly sim!le, once you see itand o!enin onto a vast new world 0efore we .um! to conclusions, thou h, !erha!s you would like to see a more com!lete renderin of this hi her level of the isomor!hism& It is a very ood e(ercise& The idea is to ive an arithmetical rule whose action is indistin uishable from that of each ty!o ra!hical rule of the /IF-system, ) solution is iven below& In the rules, m and k are arbitrary natural numbers, and n is any natural number which is less than "+m& ;FAE ", If we have made "+m ^ ", then we can make "+ W 4"+m ^ "5 E9ample, 6oin from line C to line L& ?ere, m X %+& ;FAE B, If we have made % W "+m ^ n, then we can make "+m W 4% W "+m ^ n5 ^ n& E9ample, 6oin from line " to line B, where both m and n equal "& ;FAE %, If we have made k W "+m^% ^ """ W "+m ^ n, then we can make k W "+m^" ^ n& E9ample, 6oin from line % to line C& ?ere, m and n are ", and k is %& ;FAE C, If we have made k W "+m^B ^ n then we can make k W "+m ^ n E(am!le, 6oin from line D to line $& ?ere, m X B, n X "+, and k X %+"& Aet us not for et our a(iom2 3ithout it we can o nowhere& Therefore, let us !ostulate that, 3e can make %"&

8ow the ri ht-hand column can be seen as a full-fled ed arithmetic !rocess, in a new arithmetical system which we mi ht call the HBK%s)stem, 4"5 4B5 4%5 4C5 4L5 4D5 4$5 %" %"" %"""" %+" %+"+ %+"++"+ %+""+ iven rule B 4mX", nX"5 rule B 4mXB, nX""5 rule % 4mX", nX", kX%5 rule " 4mX%+5 rule B 4mX%, nX"+5 rule C 4mXB, nX"+, kX%+"5

8otice once a ain that the len thenin and shortenin rules are ever with us in this -%"+system-= they have merely been trans!osed into the domain of numbers, so that the 6>del numbers o u! and down& If you look carefully at what is oin on, you will discover that the rules are based on nothin more !rofound than the idea that shiftin di its to left and ri ht in decimal re!resentations of inte ers is related to multi!lications and divisions by !owers of "+& This sim!le observation finds its enerali<ation in the followin 7E8T;)A P;EPESITIE8, If there is a ty!o ra!hical rule which tells how certain di its are to be shifted, chan ed, dro!!ed, or inserted in any number re!resented decimally, then this rule can be re!resented equally well by an arithmetical counter!art which involves arithmetical o!erations with !owers of "+ as well as additions, subtractions, and so forth& /ore briefly, Ty!o ra!hical rules for mani!ulatin numerals are actually arithmetical rules for o!eratin on num(ers& This sim!le observation is at the heart of 6>del1s method, and it will have an absolutely shatterin effect& It tells us that once we have a 6>del numberin for any formal system, we can strai htaway form a set of arithmetical rules which com!lete the 6>del isomor!hism& The u!shot is that we can transfer the study of any formal systemin fact the study of all formal systemsinto number theory&

MI&-,roducible ;umbers
Just as any set of ty!o ra!hical rules enerates a set of theorems, a corres!ondin set of natural numbers will be enerated by re!eated a!!lications of arithmetical rules& These produci(le num(ers !lay the same role inside number theory as theorems do inside any formal system& Ef course, different numbers will be !roducible, de!endin on which rules are ado!ted& -Producible numbers- are only !roducible relati#e to a s)stem of arithmetical rules& 'or e(am!le, such numbers as %", %+"++"+, %""", and so forth could be called /IF-produci(le numbersan un ainly name, which mi ht be shortened to

/IF-num(ers, symboli<in the fact that those numbers are the ones that result when you transcribe the /IF-system into number theory, via 6>del-numberin & If we were to 6>del-number the !q-system and then -arithmeti<e- its rules, we could call the !roducible numbers -!q-numbers-and so on& 8ote that the !roducible numbers 4in any iven system5 are defined by a recursive method, iven numbers which are known to be !roducible, we have rules tellin how to make more !roducible numbers& Thus, the class of numbers known to be !roducible is constantly e(tendin itself, in much the same way that the list of 'ibonacci numbers, or M-numbers, does& The set of !roducible numbers of any system is a recursi#el) enumera(le set& 3hat about its com!lementthe set of non!roducible numbers? Is that set always recursively enumerable? :o numbers which are non!roducible share some common arithmetical feature? This is the sort of issue which arises when you trans!ose the study of formal systems into number theory& 'or each system which is arithmeti<ed, one can ask, -7an we characteri<e !roducible numbers in a sim!le way?- -7an we characteri<e non!roducible numbers in a recursively enumerable way?- These are difficult questions of number theory& :e!endin on the system which has been arithmeti<ed, such questions mi ht !rove too hard for us to resolve& 0ut if there is any ho!e for solvin such !roblems, it would have to reside in the usual kind of ste!-by-ste! reasonin as it a!!lies to natural numbers& )nd that, of course, was !ut in its quintessential form in the !revious 7ha!ter& T8T seemed, to all a!!earances, to have ca!tured all valid mathematical thinkin !rocesses in one sin le, com!act system&

Ans'ering 8uestions about ,roducible ;umbers by Consulting T;T

7ould it be, therefore, that the means with which to answer any question about any formal system lies within .ust a sin le formal systemT8T? It seems !lausible& Take, for instance, this question, Is M& a theorem of the /IF-system? 'indin the answer is equivalent to determinin whether %+ is a /IF number or not& 0ecause it is a statement of number theory, we should e(!ect that, with some hard work, we could fi ure out how to translate the sentence -%+ is a /IF-number- into T8Tnotation, in somewhat the same way as we fi ured out how to translate other numbertheoretical sentences into T8T-notation& I should immediately caution the reader that such a translation, thou h it does e(ist, is immensely com!le(& If you recall, I !ointed out in 7ha!ter GIII that even such a sim!le arithmetical !redicate as - b is a !ower of "+- is very tricky to code into T8T-notationand the !redicate -b is a /IF-number- is a lot more com!licated than that2 Still, it can be found= and the numeral SSSSSSSSSSSSSSSSSSSSSSSSSSSSSS? can be substituted for every b& This will result in a /E8strous strin of T8T, a strin of T8T which s!eaks about the /F-!u<<le& Aet us

therefore call that strin -/F/E8-& Throu h /F/E8 and strin s like it, T8T is now ca!able of s!eakin -in code- about the /IF-system&

The

ual ;ature of M&M0;

In order to ain some benefit from this !eculiar transformation of the ori inal question, we would have to seek the answer to a new question, Is /F/E8 a theorem of T8T? )ll we have done is re!lace one relatively short strin 4 M&5 by another 4the monstrous /F/E85, and a sim!le formal system 4the /IF-system5 by a com!licated one 4T8T5& It isn1t likely that the answer will be any more forthcomin even thou h the question has been resha!ed& In fact, T8T has a full com!lement of both len thenin and shortenin rules, and the reformulation of the question is likely to be far harder than the ori inal& Ene mi ht even say that lookin at M& via /F/E8 is an intentionally idiotic way of doin thin s& ?owever, /F/E8 can be looked at on more than one level& In fact, this is an intri uin !oint, /F/E8 has two different !assive meanin s& 'irstly, it has the one which was iven before, %+ is a /IF-number& 0ut secondly, we know that this statement is tied 4via isomor!hism5 to the statement M& is a theorem of the /IF-system& So we can le itimately quote this latter as the second !assive meanin of /F/E8& It may seem very stran e because, after all, /F/E8 contains nothin but !lus si ns, !arentheses, and so forthsymbols of T8T& ?ow can it !ossibly e(!ress any statement with other than arithmetical content? The fact is, it can& Just as a sin le musical line may serve as both harmony and melody in a sin le !iece= .ust as -0)7?- may be inter!reted as both a name and a melody= .ust as a sin le sentence may be an accurate structural descri!tion of a !icture by Escher, of a section of :8), of a !iece by 0ach, and of the dialo ue in which the sentence is embedded, so /F/E8 can be taken in 4at least5 two entirely different ways& This state of affairs comes about because of two facts, 'act "& Statements such as -M& is a theorem- can be coded into number theory via 6>del1s isomor!hism& 'act B& Statements of number theory can be translated into T8T&

It could be said that /F/E8 is, by 'act ", a coded messa e, where the symbols of the code are, by 'act B, .ust symbols of T8T&

Codes and Im!licit Meaning

8ow it could be ob.ected here that a coded messa e, unlike an uncoded messa e, does not e(!ress anythin on its ownit requires knowled e of the code& 0ut in reality there is no such thin as an uncoded messa e& There are only messa es written in more familiar codes, and messa es written in less familiar codes& If the meanin of a messa e is to be revealed it must be !ulled out of the code by some sort of mechanism, or isomor!hism& It may be difficult to discover the method by which the decodin should be done= but once that method has been discovered, the messa e becomes trans!arent as water& 3hen a code is familiar enou h, it ceases a!!earin like a code= one for ets that there is a decodin mechanism& The messa e is identified with its meanin & ?ere we have a case where the identification of messa e and meanin is so stron that it is hard for us to conceive of an alternate meanin residin in the same symbols& 8amely, we are so !re.udiced by the symbols of T8T towards seein number-theoretical meanin 4and only number-theoretical meanin 5 in strin s of T8T, that to conceive of certain strin s of T8T as statements about the /IF-system is quite difficult& 0ut 6>del1s isomor!hism com!els us to reco ni<e this second level of meanin in certain strin s of T8T& :ecoded in the more familiar way, /F/E8 bears the messa e, %+ is a /IF-number& This is a statement of number theory, otten by inter!retin each si n the conventional way& 0ut in discoverin 6>del-numberin and the whole isomor!hism built u!on it, we have in a sense broken a code in which messa es about the /IF-system are written in strin s of T8T& 6>del1s isomor!hism is a new information-revealer, .ust as the deci!herments of ancient scri!ts were information-revealers& :ecoded by this new and less familiar mechanism, /F/E8 bears the messa e M& is a theorem of the /IF-system& The moral of the story is one we have heard before, that meanin is an automatic by!roduct of our reco nition of any isomor!hism= therefore there are at least two !assive meanin s of /F/E8maybe more2

The Boomerang5 G#del-;umbering T;T

Ef course thin s do not sto! here& 3e have only be un reali<in the, !otential of 6>del1s isomor!hism& The natural trick would be to turn T8T1s ca!ability of mirrorin other

formal systems back on itself, as the Tortoise turned the 7rab1s !hono ra!hs a ainst themselves, and as his 6oblet 6 turned a ainst itself, in destroyin itself& In order to do this, we will have to 6>del-number T8T itself, .ust as we did the /IF-system, and then -arithmeti<e- its rules of inference& The 6>del-numberin is easy to do& 'or instance, we could make the followin corres!ondence, -)m(ol &&&&& &&&&& &&&&& &&&&& . &&&&& 4 &&&&& 5 &&&&& g &&&&& a &&&&& U &&&&& V &&&&& a &&&&& ' &&&&& &&&&& &&&&& &&&&& ~ &&&&& &&&&& " &&&&& , &&&&& !unc& &&&&& ? S X ^ 8odon DDD "B% """ ""B B%D %DB %B% B"B B"% %"B %"% BDB "D% "D" D"D D%% BB% %%% DBD D%D D"" 'nemonic ?ustification 8umber of the 0east for the /ysterious 9ero successorshi!, ", B, %, visual resemblance, turned sideways "^"XB BW%XD ends in B ends in % ends in B these three pairs ends in % form a pattern ends in B ends in % o!!osite to " 4DBD5 "D% is !rime 11 is a - ra!h- of the sequence "-D-" 11 is a - ra!h- of the sequence D-"-D 1 D -im!lies- % and %, in some sense & B ^ B is not % 11 looks like 1%1 o!!osite to a= also a - ra!h- of D-B-D two dots, two si(es s!ecial number, as on 0ell system 4C"", #""5

Each symbol of T8T is matched u! with a tri!let com!osed of the di its ", B, %, and D, in a manner chosen for mnemonic value& I shall call each such tri!let of di its a Gdel codon, or codon for short& 8otice that I have iven no codon for b, c, d, or e= we are usin austere T8T& There is a hidden motivation for this, which you will find out about in 7ha!ter TGI& I will e(!lain the bottom entry, -!unctuation-, in 7ha!ter TIG& 8ow we can rewrite any strin or rule of T8T in the new arb& ?ere, for instance, is )(iom " in the two notations, the old below the new, DBD,BDB,D%D,BB%,"B%,BDB,""",DDD " a , h S a X + 7onveniently, the standard convention of !uttin in a comma every third di it ha!!ens to coincide with our colons, settin them off for -easy- le ibility& ?ere is the ;ule of :etachment, in the new notation,

;FAE, If 9 and B"B9D%%)B"% are both theorems, then ) is a theorem& 'inally, here is an entire derivation taken from last 7ha!ter, iven in austere T8T and also transcribed into the new notation,
DBD,BDB,D%D,DBD,BDB,"D%,D%D,%DB,BDB,""B,"B%,BDB,"D%,%B%,""","B%,%DB,BDB,""B,BDB,"D%,%B% a(iom % " a , " a 1 , 4 a ^ S a 1 5 X S 4 a ^ a 1 5 DBD,BDB&"D%,D%D,%DB,"B%,DDD,""B,"B%,BDB,"D%,%B%,""","B%,%DB,"B%,DDD,""B,BDB,"D%,%B% s!ecification " a 1 , 4 S ? ^ S a 1 5 X S 4 S ? ^ a 1 5 %DB,"B%,DDD,""B,"B%,DDD,%B%,""","B%,%DB,"B%,DDD,""B,DDD,%B% s!ecification 4 S ? ^ S ? 5 X S 4 S ? ^ ? 5 DBD,BDB,D%D,%DB BDB,""B,DDD, %B%,""",BDB a(iom B " a , 4 a ^ ? 5 X a %DB,"B%,DDD,""B,DDD,%B%,""","B%,DDD s!ecification 4 S ? ^ ? 5 X S ? "B%,%DB,"B%,DDD,""B,DDD,%B%,"" ","B%,"B%,DDD insert 1"B%1 S 4 S ? ^ ? 5 X S S ? %DB,"B%,DDD,""B,"B%,DDD,%B%,""","B%,"B%,DDD transitivity 4 S ? ^ S ? 5 X S S ?

8otice that I chan ed the name of the -)dd S- rule to -Insert 1"B%1 -, since that is the ty!o ra!hical o!eration which it now le itimi<es& This new notation has a !retty stran e feel to it& Hou lose all sense of meanin = but if you had been brou ht u! on it, you could read strin s it this notation as easily as you do T8T& Hou would be able to look and, at a lance, distin uish well-formed formulas from ill-formed ones& 8aturally, since it is so visual, you would think of this as a ty!o ra!hical o!erationbut at the same time, !ickin out well-formed formulas in this notation is !ickin out a s!ecial class of integers, which have an arithmetical definition too& 8ow what about -arithmeti<in - all the rules of inference? )s matters stand, they are all still ty!o ra!hical rules& 0ut wait2 )ccordin to the 7entral Pro!osition, a ty!o ra!hical rule is really equivalent to an arithmetical rule& Insertin and movin di its in decimally re!resented numbers is an arithmetical o!eration, which can be carried out ty!o ra!hically& Just as a!!endin a 1+1 on the end is e(actly the same as multi!lyin by "+, so each rule is a condensed way of describin a messy arithmetical o!eration& Therefore, in a sense, we do not even need to look for equivalent arithmetical rules, because all of the rules are alread) arithmetical2

T;T-;umbers5 A $ecursi"ely Enumerable Set of ;umbers

Aooked at this way, the !recedin derivation of the theorem -%DB,"B%,DDD,""B, "B%,DDD,%B%,""","B%,"B%,DDD- is a sequence of hi hly convoluted number-theoretical transformations, each of which acts on one or more in!ut numbers, and yields an out!ut number, which is, as before, called a produci(le number, or, to be more s!ecific, a T8Tnum(er& Some of the arithmetical rules take an old T8T-number and increase it in a !articular way, to yield a new T8T-number= some take an old T8T-number a and decrease it= other rules take two T8T-numbers, o!erate on each of them some odd way,

and then combine the results into a new T8T-number and so on and so forth& )nd instead of startin with .ust one known T8T-number, we have fi#e initial T8T-numbersone for each 4austere a(iom, of course& )rithmeti<ed T8T is actually e(tremely similar to the arithmeti<ed /IF-system, only there are more rules and a(ioms, and to write out arithmetical equivalents e(!licitly would be a bi botherand quite unenli htenin , incidentally& If you followed how it was done for the /IF-system, there ou ht to be no doubt on your !art that it is quite analo ous here& There is a new number-theoretical !redicate brou ht into bein by this -6>deli<ation- of T8T, the !redicate a is a T8T-number& 'or e(am!le, we know from the !recedin derivation that %DB,"B%,DDD,""B,"B%,DDD,%B%, ""","B%,"B%,DDD is a T8T-number, while on the other hand, !resumably "B%,DDD,""",DDD is not a T8T-number& 8ow it occurs to us that this new number-theoretical !redicate is e9pressi(le by some strin of T8T with one free variable, say a& 3e could !ut a tilde in front, and that strin would e(!ress the com!lementary notion a is not a T8T-number& 8ow if we re!laced all the occurrences of in this second strin by the T8T-numeral for "B%,DDD,""",DDDa numeral which would contain e(actly "B%,DDD,""",DDD S1s, much too lon to write outwe would have a T8T-strin which, .ust like /F/E8, is ca!able of bein inter!reted on two levels& In the first !lace, that strin would say "B%,DDD,""",DDD is not a T8T-number& 0ut because of the isomor!hism which links T8T-numbers to theorems of T8T, there would be a second-level meanin of this strin , which is, S?X? is not a theorem of T8T&

T;T Tries to S'allo' Itself

This une(!ected double-entendre demonstrates that T8T contains strin s which talk about other strin s of T8T& In other words, the metalan ua e in which we, on the outside, can s!eak about T8T, is at least !artially imitated inside T8T itself& )nd this is not an accidental feature of T8T= it ha!!ens because the architecture of any formal system can be mirrored inside 8 4number theory5& It is .ust as inevitable a feature of T8T as are the vibrations induced in a record !layer when it !lays a record& It seems as if vibrations should come from the outside worldfor instance, from .um!in children or bouncin balls= but a side effect of !roducin soundsand an unavoidable oneis that they wra!

around and shake the very mechanism which !roduces them& It is no accident= it is a side effect which cannot be hel!ed& It is in the nature of record !layers& )nd it is in the nature of any formali<ation of number theory that its metalan ua e is embedded within it& 3e can di nify this observation by callin it the 8entral .ogma of 'athematical +ogic, and de!ictin it in a two-ste! dia ram, T8T 8 meta-T8T In words, a strin of T8T has an inter!retation in 8= and a statement of 8 may have a second meanin as a statement about T8T&

G5 A String >hich Tal.s about Itself in Code

This much is intri uin yet it is only half the story& The rest of the story involves an intensification of the self-reference& 3e are now at the sta e where the Tortoise was when he reali<ed that a record could be made which would make the !hono ra!h !layin it breakbut now the quest is, -6iven a record !layer, how do you actually fi ure out what to !ut on the record?- That is a tricky matter& 3e want to find a strin of T8Twhich we1ll call 161which is about itself, in the sense that one of its !assive meanin s is a sentence about 6& !articular the !assive meanin will turn out to be -6 is not a theorem of T8T&I should quickly add that 6 also has a !assive meanin which is a sta tement of num(er theor)= .ust like /F/E8 it is susce!tible to bein construed in 4at least5 two different ways& The im!ortant thin is that each !assive meanin is valid and useful and doesn1t cast doubt on the other !assive meanin in any way& 4The fact that a !hono ra!h !layin a record can induce vibrations in itself and in the record does not diminish in any way the fact that those vibrations are musical sounds25

G%s E-istence Is >hat Causes T;T%s Incom!leteness

The in enious method of creatin 6, and some im!ortant conce!ts relatin to T8T, will be develo!ed in 7ha!ters TIII and TIG= for now it is .ust interestin to lance ahead, a bit su!erficially, at the consequences findin a self-referential !iece of T8T& 3ho knows? It mi ht blow u!2 In a sense it does& 3e focus down on the obvious question, Is 6 a theorem of T8T, or not? Aet us be sure to form our own o!inion on this matter, rather than rely on 61s o!inion about itself& )fter all, 6 may not understand itself any better than a 9en master understands himself& Aike /F/E8, 6 may e(!ress a falsity& Aike /F, 6 may be a

nontheorem& 3e don1t need to believe every !ossible strin of T8Tonly its theorems& 8ow let us use our !ower of reasonin to clarify the issue as best we can at this !oint& 3e will make our usual assum!tion, that T8T incor!orates valid methods of reasonin , and therefore that T8T never has falsities for theorems& In other words, anythin which is a theorem of T8T e(!resses a truth& So if 6 were a theorem, it would e(!ress a truth, namely, -6 is not a theorem-& The full force of its self-reference hits us& 0y bein a theorem, 6 would have to be a falsity& ;elyin on our assum!tion that T8T never has falsities for theorems, we1d be forced to conclude that 6 is not a theorem& This is all ri ht= it leaves us, however, with a lesser !roblem& @nowin that 6 is not a theorem, we1d have to concede that 6 e(!resses a truth& ?ere is a situation in which T8T doesn1t live u! to our e(!ectationswe have found a strin which e(!resses a true statement yet the strin is not a theorem& )nd in our ama<ement, we shouldn1t lose track of the fact that 6 has an arithmetical inter!retation, toowhich allows us to summari<e our findin s this way, ) strin of T8T has been found= it e(!resses, unambi uously, a statement about certain arithmetical !ro!erties of natural numbers= moreover, by reasonin outside the system we can determine not only that the statement is a true one, but also that the strin fails to be a theorem of T8T& )nd thus, if we ask T8T whether the statement is true, T8T says neither yes nor no& Is the Tortoise1s strin in the 'u :ffering the analo ue of 6? 8ot quite& The analo ue of the Tortoise1s strin is h6& 3hy is this so? 3ell, let us think a moment about what h6 says& It must say the o!!osite of what 6 says& 6 says, -6 is not a theorem of T8T-, so h6 must say -6 is a theorem-& 3e could re!hrase both 6 and h6 this way, 6, -I am not a theorem 4of T8T5&h6, -/y ne ation is a theorem 4of T8T5&It is h6 which is !arallel to the Tortoise1s strin , for that strin s!oke not about itself, but about the strin which the Tortoise first !roffered to )chilleswhich had an e(tra knot on it 4or one too few, however you want to look at it5&

Mumon /as the Last >ord

/umon !enetrated into the /ystery of the Fndecidable anyone, in his concise !oem on JRshl1s /F, ?as a do 0uddha-nature? This is the most serious question of all& If you say yes or no, Hou lose your own 0uddha-nature&

,A$T II

Prelude
*chilles and the Tortoise ha#e come to the residence of their friend the 8ra(, to make the ac"uaintance of one of his friends, the *nteater! The introductions ha#ing (een made, the four of them settle down to tea! Tortoise, 3e have brou ht alon a little somethin for you, /r& 7rab& 8ra(, That1s most kind of you& 0ut you shouldn1t have& Tortoise, Just a token of our esteem& )chilles, would you like to ive it to /r& 7? *chilles, Surely& 0est wishes, /r& 7rab& I ho!e you en.oy it& *chilles hands the 8ra( an elegantl) wrapped present, s"uare and #er) thin! The 8ra( (egins unwrapping it!/ *nteater, I wonder what it could be& 8ra(, 3e1ll soon find out& 8ompletes the unwrapping, and pulls out the gift/! Two records2 ?ow e(citin 2 0ut there1s no label& Fh-ohis this another of your -s!ecials-, /r& T? Tortoise, If you mean a !hono ra!h-breaker, not this time& 0ut it is in fact a customrecorded item, the only one of its kind in the entire world& In fact, it1s never even been heard beforee(ce!t, of course, when 0ach !layed it& 8ra(, 3hen 0ach !layed it? 3hat do you mean, e(actly? *chilles, Eh, you are oin to be fabulously e(cited, /r& 7rab, when /r& T tells you what these records in fact are& Tortoise, Eh, you o ahead and tell him, )chilles& *chilles, /ay I? Eh, boy2 I1d better consult my notes, then& &ulls out a small filing card, and clears his #oice&5 )hem& 3ould you be interested in hearin about the remarkable new result in mathematics, to which your records owe their e(istence? 8ra(, /y records derive from some !iece of mathematics? ?ow curious2 3ell, now that you1ve !rovoked my interest, I must hear about it& *chilles, Gery well, then& 4&auses for a moment to sip his tea, then resumes!/ ?ave you heard of 'ermat1s infamous -Aast Theorem-? *nteater, I1m not sure&&& It sounds stran ely familiar, and yet I can1t quite !lace it& *chilles, It1s a very sim!le idea& Pierre de 'ermat, a lawyer by vocation but a mathematician by avocation, had been readin in his co!y of the class te(t *rithmetica by :io!hantus, and came across a !a e containin the equation

aB^bBXcB

73G5=E GD! />bius Stri! II, () '! 8! Escher woodcut, BIJH/!

?e immediately reali<ed that this equation has infinitely many solutions a, b, c, and then wrote in the mar in the followin notorious comment,
The equation an^bnXcn has solutions in !ositive inte ers a, b, c, and n only when n X B 4and then there are infinitely many tri!lets a, b, c which satisfy the equation5= but there are no solutions for n a B& I have discovered a truly marvelous !roof of this statement, which, unfortunately, this mar in is too small to contain&

Ever since that day, some three hundred years a o, mathematicians have been vainly tryin to do one of two thin s, either to !rove 'ermat1s claim, and thereby vindicate 'ermat1s re!utation, which, althou h very hi h, has been somewhat tarnished by ske!tics who think he never really found the !roof he claimed to have foundor else to refute the claim, by findin a countere(am!le, a set of four inte ers a, b, c, and n, with n a B, which satisfy the equation& Fntil very recently, every attem!t in either direction had met with failure& To be sure, the Theorem has been !roven for many s!ecific values of nin !articular, all n u! to "BL,+++& *nteater, Shouldn1t it be called a -7on.ecture- rather than a -Theorem- it1s never been iven a !ro!er !roof? *chilles, Strictly s!eakin , you1re ri ht, but tradition has ke!t it this way& 8ra(, ?as someone at last mana ed to resolve this celebrated question? *chilles, Indeed2 In fact, /r& Tortoise has done so, and as usual, by a wi<ardly stroke& ?e has not only found a P;EE' of 'ermat1s Theorem 4thus .ustifyin its name as well as vindicatin 'ermat= also a 7EF8TE;ET)/PAE, thus showin that the ske!tics had ood intuition2 8ra(, Eh my racious2 That is a revolutionary discovery& *nteater, 0ut !lease don1t leave us in sus!ense& 3hat ma ical inte er they, that satisfy 'ermat1s equation? I1m es!ecially curious about the value of n& *chilles, Eh, horrors2 I1m most embarrassed2 7an you believe this? I left the values at home on a truly colossal !iece of !a!er& Fnfortunately it was too hu e to brin alon & I wish I had them here to show to you& If it1s of any hel! to you, I do remember one thin the value of n is the only !ositive inte er which does not occur anywhere in the continued fraction for m& 8ra(, Eh, what a shame that you don1t have them here& 0ut there reason to doubt what you have told us&

73G5=E GG! &ierre de 7ermat!

*nteater, )nyway, who needs to see n written out decimally? )chilles has .ust told us how to find it& 3ell, /r& T, !lease acce!t my hearty felicitations, on the occasion of your e!och-makin discovery2 Tortoise, Thank you& 0ut what I feel is more im!ortant than the result itself is the !ractical use to which my result immediately led& 8ra(, I am dyin to hear about it, since I always thou ht number theory was the Mueen of /athematicsthe !urest branch of mathematicsthe one branch of mathematics which has 8E a!!lications2 Tortoise, Hou1re not the only one with that belief, but in fact it is quite im!ossible to make a blanket statement about when or how some branchor even some individual Theoremof !ure mathematics will have im!ortant re!ercussions outside of mathematics& It is quite un!redictableand this case is a !erfect e(am!le of that !henomenon& *chilles, /r& Tortoise1s double-barreled result has created a breakthrou h in the field of acoustico-retrieval2 *nteater, 3hat is acoustico-retrieval? *chilles, The name tells it all, it is the retrieval of acoustic information from e(tremely com!le( sources& ) ty!ical task of acoustico-retrieval is to reconstruct the sound which a rock made on !lummetin into a lake from the ri!!les which s!read out over the lake1s surface& 8ra(, 3hy, that sounds ne(t to im!ossible2 *chilles, 8ot so& It is actually quite similar to what one1s brain does, when it reconstructs the sound made in the vocal cords of another !erson from the vibrations transmitted by the eardrum to the fibers in the cochlea& 8ra(, I see& 0ut I still don1t see where number theory enters the !icture, or what this all has to do with my new records&

)chilles, 3ell, in the mathematics of acoustico-retrieval, there arise many questions which have to do with the number of solutions of certain :io!hantine equations& 8ow /r& T has been for years tryin to find a way of reconstructin the sounds of 0ach !layin his har!sichord, which took !lace over two hundred years a o, from calculations involvin the motions of all the molecules in the atmos!here at the !resent time& )nteater, Surely that is im!ossible2 They are irretrievably one, one forever2 )chilles, Thus think the naPve&&& 0ut /r& T has devoted many years to this !roblem, and came to the reali<ation that the whole thin hin ed on the number of solutions to the equation

an^bnXcn
in !ositive inte ers, with n a B& Tortoise, I could e(!lain, of course, .ust how this equation arises, but I1m sure it would bore you& *chilles, It turned out that acoustico-retrieval theory !redicts that 0ach sounds can be retrieved from the motion of all the molecules in the atmos!here, !rovided that EIT?E; there e(ists at least one solution to the equation 8ra(, )ma<in 2 *nteater, 'antastic2 Tortoise, 3ho would have thou ht2 *chilles, I was about to say, -!rovided that there e(ists EIT?E; such a solution E; a !roof that there are 8E solutions2- )nd therefore, /r& T, in careful fashion, set about workin at both ends of the !roblem, simultaneously& )s it turns out, the discovery of the countere(am!le was the key in redient to findin the !roof, so the one led directly to the other& 8ra(, ?ow could that be? Tortoise, 3ell, you see, I had shown that the structural layout of any !roof of 'ermat1s Aast Theoremif one e(istedcould be described by an ele ant formula, which, it so ha!!ened, de!ended on the values of the solution to a certain equation& 3hen I found this second equation, to my sur!rise it turned out to be the 'ermat equation& )n amusin accidental relationshi! between form and content& So when I found the countere(am!le, all I needed to do was to use those numbers as a blue!rint for constructin my !roof that there were no solutions to the equation& ;emarkably sim!le, when you think about it& I can1t ima ine why no one had ever found the result before& *chilles, )s a result of this unantici!atedly rich mathematical success, /r& T was able to carry out the acoustico-retrieval which he had so lon dreamed of& )nd /r& 7rab1s !resent here re!resents a !al!able reali<ation of all this abstract work& 8ra(, :on1t tell me it1s a recordin of 0ach !layin his own works for har!sichord2 *chilles, I1m sorry, but I have to, for that is indeed .ust what it is2 This is a set of two records of Johann Sebastian 0ach !layin all of his >ell%Tempered 8la#ier& Each

record contains one of the two volumes of the >ell%Tempered 8la#ier= that is to say, each record contains BC !reludes and fu uesone in each ma.or and minor key& 8ra(, 3ell, we must absolutely !ut one of these !riceless records on, immediately2 )nd how can I ever thank the two of you? Tortoise, Hou have already thanked us !lentifully, with this delicious tea which you have !re!ared& The 8ra( slides one of the records out of its Macket, and puts it on! The sound of an incredi(l) masterful harpsichordist fills the room, in the highest imagina(le fidelit)! :ne e#en hearsOor is it one's imagination$Othe soft sounds of Bach singing to himself as he pla)s!!!/ 8ra(, 3ould any of you like to follow alon in the score? I ha!!en to have a unique edition of the >ell%Tempered 8la#ier, s!ecially illuminated by a teacher of mine who ha!!ens also to be an unusually fine calli ra!her& Tortoise, I would very much en.oy that& The 8ra( goes to his elegant glass%enclosed wooden (ookcase, opens the doors, and draws out two large #olumes!/ 8ra(, ?ere you are, /r& Tortoise& I1ve never really otten to know all the beautiful illustrations in this edition& Perha!s your ift will !rovide the needed im!etus for me to do so& Tortoise, I do ho!e so& *nteater, ?ave you ever noticed how in these !ieces the !relude always sets the mood !erfectly for the followin fu ue? 8ra(, Hes& )lthou h it may be hard to !ut it into words, there is always some subtle relation between the two& Even if the !relude and fu ue do not have a common melodic sub.ect, there is nevertheless always some intan ible abstract quality which underlies both of them, bindin them to ether very stron ly& Tortoise, )nd there is somethin very dramatic about the few moments of silent sus!ense han in between !relude and fu uethat moment where the the theme of the fu ue is about to rin out, in sin le tones, and then to .oin with itself in ever-increasin ly com!le( levels of weird, e(quisite harmony& *chilles, I know .ust what you mean& There are so many !reludes and fu ues which I haven1t yet otten to know, and for me that fleetin interlude of silence is very e(citin = it1s a time when I try to second- uess old 0ach& 'or e(am!le, I always wonder what the fu ue1s tem!o will be, alle ro, or ada io? 3ill it be in DI*, or CIC? 3ill it have three voices, or fiveor four? )nd then, the first voice starts&&& Such an e(quisite moment& 8ra(, )h, yes, well do I remember those lon - one days of my youth, days when I thrilled to each new !relude and fu ue, filled with e(citement of their novelty and beauty and the many une(!ected sur!rises which they conceal& *chilles, )nd now? Is that thrill all one?

8ra(, It1s been su!!lanted by familiarity, as thrills always will be& 0ut in that familiarity there is also a kind of de!th, which has its own com!ensations& 'or instance, I find that there are always new sur!rises which I hadn1t noticed before& *chilles, Eccurrences of the theme which you had overlooked? 8ra(, Perha!ses!ecially when it is inverted and hidden amon several other voices, or where it seems to come rushin u! from the de!ths, out of nowhere& 0ut there are also ama<in modulations which it is marvelous to listen to over and over a ain, and wonder how old 0ach dreamt them u!& *chilles, I am very lad to hear that there is somethin to look forward to, after I have been throu h the first flush of infatuation with the >ell Tempered 8la#ierOalthou h it also makes me sad that this sta e could not last forever and ever& 8ra(, Eh, you needn1t fear that your infatuation will totally die& Ene of the nice thin s about that sort of youthful thrill is that it can always resuscitated, .ust when you thou ht it was finally dead& It .ust takes the ri ht kind of tri erin from the outside& *chilles, Eh, really? Such as what? 8ra(, Such as hearin it throu h the ears, so to s!eak, of someone to whom it is a totally new e(!eriencesomeone such as you, )chilles& Somehow the e(citement transmits itself, and I can feel thrilled a ain& *chilles, That is intri uin & The thrill has remained dormant somewhere inside you, but by yourself, you aren1t able to fish it u! out of your subconscious& 8ra(, E(actly& The !otential of relivin the thrill is -coded-, in some unknown way, in the structure of my brain, but I don1t have the !ower to summon it u! at will= I have to wait for chance circumstance to tri er it& *chilles, I have a question about fu ues which I feel a little embarrass about askin , but as I am .ust a novice at fu ue-listenin , I was wonderin if !erha!s one of you seasoned fu ue-listeners mi ht hel! me in learnin &&& Tortoise, I1d certainly like to offer my own mea er knowled e, if it mi ht !rove of some assistance& *chilles, Eh, thank you& Aet me come at the question from an an le& )re you familiar with the !rint called 8u(e with 'agic =i((ons, by /& 7& Escher? Tortoise, In which there are circular bands havin bubble-like distortions which, as soon as you1ve decided that they are bum!s, seem to turn it dentsand vice versa? *chilles, E(actly& 8ra(, I remember that !icture& Those little bubbles always seem to fli! back and forth between bein concave and conve(, de!endin on the direction that you a!!roach them from& There1s no way to see them simultaneously as concave )8: conve( somehow one1s brain doesn1t allow that& There are two mutually e(clusive -modes- in which one can !erceive the bubbles& *chilles, Just so& 3ell, I seem to have discovered two somewhat analo ous modes in which I can listen to a fu ue& The modes are these, either to follow one individual voice at a time, or to listen to the total effect of all of them to ether, without tryin to disentan le one from another& I have tried out both of these modes, and, much to my frustration, each one of them shuts out the other& It1s sim!ly not in my !ower to follow

73G5=E GJ! 7ube with /a ic ;ibbons, () '! 8! Escher lithograph, BIGC5&

the !aths of individual voices and at the same time to hear the whole effect& I find that I fli! back and forth between one mode and the other, more or less s!ontaneously and involuntarily& *nteater, Just as when you look at the ma ic bands, eh? *chilles, Hes& I was .ust wonderin &&& does my descri!tion of these two modes of fu uelistenin brand me unmistakably as a naPve, ine(!erienced listener, who couldn1t even be in to ras! the dee!er modes of !erce!tion which e(ist beyond his ken? Tortoise, 8o, not at all, )chilles& I can only s!eak for myself, but I to find myself shiftin back and forth from one mode to the other without e(ertin any conscious control over which mode should he dominant& I don1t know if our other com!anions here have also e(!erienced anythin similar& 8ra(, /ost definitely& It1s quite a tantali<in !henomenon, since you feel that the essence of the fu ue is flittin about you, and you can1t quite ras! all of it, because you can1t quite make yourself function both ways at once& *nteater, 'u ues have that interestin !ro!erty, that each of their voices is a !iece of music in itself= and thus a fu ue mi ht be thou ht of as a collection of several distinct !ieces of music, all based on one sin le theme, and all !layed simultaneously& )nd it

is u! to the listener 4or his subconscious5 to decide whether it should be !erceived as a unit, or as a collection of inde!endent !arts, all of which harmoni<e& *chilles, Hou say that the !arts are -inde!endent-, yet that can1t be literally true& There has to be some coordination between them, otherwise when they were !ut to ether one would .ust have an unsystematic clashin of tonesand that is as far from the truth as could be& *nteater, ) better way to state it mi ht be this, if you listened to each on its own, you would find that it seemed to make sense all by itself& It could stand alone, and that is the sense in which I meant that it is inde!endent& 0ut you are quite ri ht in !ointin out that each of these individually meanin ful lines fuses with the others in a hi hly nonrandom way, to make a raceful totality& The art of writin a beautiful fu ue lies !recisely in this ability, to manufacture several different lines, each one of which ives the illusion of havin been written for its own beauty, and yet which when taken to ether forms a whole, which does not feel forced in any way& 8ow, this dichotomy between hearin a fu ue as a whole, and hearin its com!onent voices, is a !articular e(am!le of a very eneral dichotomy, which a!!lies to many kinds of structures built u! from lower levels& *chilles, Eh, really? Hou mean that my two -modes- may have some eneral ty!e of a!!licability, in situations other than fu ue-listenin ? *nteater, )bsolutely& *chilles, I wonder how that could be& I uess it has to do with alternatin between !erceivin somethin as a whole, and !erceivin it as a collection of !arts& 0ut the only !lace I have ever run into that dichotomy is in listenin to fu ues& Tortoise, Eh, my, look at this2 I .ust turned the !a e while followin the music, and came across this ma nificent illustration facin the first !a e of the fu ue& 8ra(, I have never seen that illustration before& 3hy don1t you !ass it 1round? The Tortoise passes the (ook around! Each of the foursome looks at it in a characteristic wa)Othis one from afar, that one from close up, e#er)one tipping his head this wa) and that in pu@@lement! 7inall) it has made the rounds, and returns to the Tortoise, who peers at it rather intentl)!/ *chilles, 3ell, I uess the !relude is .ust about over& I wonder if, as I listen to this fu ue, I will ain any more insi ht into the question, -3hat is the ri ht way to listen to a fu ue, as a whole, or as the sum of its !arts?Tortoise, Aisten carefully, and you will2 The prelude ends! There is a moment of silenceR and!!! U*TT*88*V

C/A,TE$ <

Le"els of escri!tion) and Com!uter Systems

Le"els of escri!tion

6n:EA1S ST;I86 6, and a 0ach fu ue, they both have the !ro!erty that they can be understood on different levels& 3e are all familiar with this kind of thin = and yet in some cases it confuses us, while in others we handle it without any difficulty at all& 'or e(am!le, we all know that we human bein s are com!osed of an enormous number of cells 4around twenty-five trillion5, and therefore that everythin we do could in !rinci!le be described in terms of cells& Er it could even be described on the level of molecules& /ost of us acce!t this in a rather matter-of-fact way= we o to the doctor, who looks at us on lower levels than we think of ourselves& 3e read about :8) and - enetic en ineerin - and si! our coffee& 3e seem to have reconciled these two inconceivably different !ictures of ourselves sim!ly by disconnectin them from each other& 3e have almost no way to relate a microsco!ic descri!tion of ourselves to that which we feel ourselves to be, and hence it is !ossible to store se!arate re!resentations of ourselves in quite se!arate -com!artments- of our minds& Seldom do we have to fli! back and forth between these two conce!ts of ourselves, wonderin -?ow can these two totally different thin s be the same me?Er take a sequence of ima es on a television screen which shows Shirley /acAaine lau hin & 3hen we watch that sequence, we know that we are actually lookin not at a woman, but at sets of flickerin dots on a flat surface& 3e know it, but it is the furthest thin from our mind& 3e have these two wildly o!!osin re!resentations of what is on the screen, but that does not confuse us& 3e can .ust shut one out, and !ay attention to the otherwhich is what all of us do& 3hich one is -more real-? It de!ends on whether you1re a human, a do , a com!uter, or a television set&

Chun.ing and Chess S.ill

Ene of the ma.or !roblems of )rtificial Intelli ence research is to fi ure out how to brid e the a! between these two descri!tions= how to construe a system which can acce!t one level of descri!tion, and !roduce the other& Ene way in which this a! enters )rtificial Intelli ence is well illustrated by the !ro ress in knowled e about how to !ro ram a com!uter to !lay ood chess& It used to be thou htin the "#L+1s and on into the "#D+1sthat the trick to makin a machine !lay well was to make the machine look further ahead into the branchin network of !ossible sequences of !lay than any chess

master can& ?owever, as this oal radually became attained, the level of com!uter chess did not have any sudden s!urt, and sur!ass human e(!erts& In fact, a human e(!ert can quite soundly and confidently trounce the best chess !ro rams of this day& The reason for this had actually been in !rint for many years& In the "#C+1s, the :utch !sycholo ist )driaan de 6root made studies of how chess novices and chess masters !erceive a chess situation& Put in their starkest terms, his results im!ly that chess masters !erceive the distribution of !ieces in chunks& There is a hi her-level descri!tion of the board than the strai htforward -white !awn on @L, black rook on MD- ty!e of descri!tion, and the master somehow !roduces such a mental ima e of the board& This was !roven by the hi h s!eed with which a master could re!roduce an actual !osition taken from a ame, com!ared with the novice1s !loddin reconstruction of the !osition, after both of them had had five-second lances at the board& ?i hly revealin was the fact that masters1 mistakes involved !lacin whole groups of !ieces in the wron !lace, which left the ame strate ically almost the same, but to a novice1s eyes, not at all the same& The clincher was to do the same e(!eriment but with !ieces randomly assi ned to the squares on the board, instead of co!ied from actual ames& The masters were found to be sim!ly no better than the novices in reconstructin such random boards& The conclusion is that in normal chess !lay, certain ty!es of situation recur certain !atternsand it is to those hi h-level !atterns that the master is sensitive& ?e thinks on a different le#el from the novice= his set of conce!ts is different& 8early everyone is sur!rised to find out that in actual !lay, a master rarely looks ahead any further than a novice doesand moreover, a master usually e(amines only a handful of !ossible moves2 The trick is that his mode of !erceivin the board is like a filter, he literally does not see (ad mo#es when he looks at a chess situationno more than chess amateurs see ille al moves when they look at a chess situation& )nyone who has !layed even a little chess has or ani<ed his !erce!tion so that dia onal rook-moves, forward ca!tures by !awns, and so forth, are never brou ht to mind& Similarly, master-level !layers have built u! hi her levels of or ani<ation in the way they see the board= consequently, to them, bad moves are as unlikely to come to mind as ille al moves are, to most !eo!le& This mi ht be called implicit pruning of the iant branchin tree of !ossibilities& 0y contrast, e9plicit pruning would involve thinkin of a move, and after su!erficial e(amination, decidin not to !ursue e(aminin it any further& The distinction can a!!ly .ust as well to other intellectual activitiesfor instance, doin mathematics& ) ifted mathematician doesn1t usually think u! and try out all sorts of false !athways to the desired theorem, as less ifted !eo!le mi ht do= rather, he .ust -smells- the !romisin !aths, and takes them immediately& 7om!uter chess !ro rams which rely on lookin ahead have not been tau ht to think on a hi her level= the strate y has .ust been to use brute force look-ahead, ho!in to crush all ty!es of o!!osition& 0ut it has not worked& Perha!s someday, a look-ahead !ro ram with enou h brute force will indeed overcome the best human !layersbut that will be a small intellectual ain, com!ared to the revelation that intelli ence de!ends crucially on the ability to create hi h-level descri!tions of com!le( such as chess boards, television screens, !rinted !a es, or !aintin s&

Similar Le"els
Fsually, we are not required to hold more than one level of understandin of a situation in our minds at once& /oreover, the different descri!tions a sin le system are usually so conce!tually distant from each other that, as was mentioned earlier, there is no !roblem in maintainin them both= they are .ust maintained in se!arate mental com!artments& 3hat is confusin , thou h, is when a sin le system admits of two or more descri!tions on different levels which nevertheless resem(le each other in some way& Then we find it hard to avoid mi(in levels when we think about the system, and can easily et totally lost& Fndoubtedly this ha!!ens when we think about our !sycholo yfor instance, when we try to understand !eo!le1s motivations for various actions& There are many levels in the human mental structurecertainly it is a system which we do not understand very well yet& 0ut there are hundreds of rival theories which tell why !eo!le act the way they do, each theory based on some underlyin assum!tions about how far down in this set of levels various kinds of !sycholo ical -forces- are found& Since at this time we use !retty much the same kind of lan ua e for all mental levels, this makes for much levelmi(in and most certainly for hundreds of wron theories& 'or instance, we talk of -drives-for se(, for !ower, for fame, for love, etc&, etc&without knowin where these drives come from in the human mental structure& 3ithout belaborin the !oint, I sim!ly wish to say that our confusion about who we are is certainly related to the fact that we consist of a lar e set of levels, and we use overla!!in lan ua e to describe ourselves on all of those levels&

Com!uter Systems
There is another !lace where many levels of descri!tion coe(ist for a system, and where all the levels are conce!tually quite close to one another& I am referrin to com!uter systems& 3hen a com!uter !ro ram is runnin , it can be viewed on a number of levels& En each level, the descri!tion is iven in the lan ua e of com!uter science, which makes all the descri!tions similar in some ways to each otheryet there are e(tremely im!ortant differences between the views one ets on the different levels& )t the lowest level, the descri!tion can be so com!licated that it is like the dot-descri!tion of a television !icture& 'or some !ur!oses, however, this is by far the im!ortant view& )t the hi hest level, the descri!tion is reatly chunked and takes on a com!letely different feel, des!ite the fact that many of the same conce!ts a!!ear on the lowest and hi hest levels& The chunks on the hi h-level descri!tion are like the chess e(!ert1s chunks, and like the chunked descri!tion of the ima e on the screen, they summari<e in ca!sule form a number of thin s which on lower levels are seen as se!arate& 4See 'i & L$&5 8ow before thin s become too abstract, let us !ass on to the concrete facts about com!uters, be innin with a very quick skim of what a com!uter system is like on the lowest level& The lowest level? 3ell, not really, for I am not oin to talk about elementary !articles but it is the lowest level which we wish to think about&

73G5=E GC! The idea of "chunking": a group of items is repercei#ed as a single "chunk"! The chunk's (oundar) is a little like a cell mem(rane or a national (order: it esta(lishes a separate identit) for the cluster within! *ccording to conte9t, one ma) wish to ignore the chunk's internal structure or to take it into account!

)t the conce!tual rock-bottom of a com!uter, we find a memory, a central !rocessin unit 47PF5, and some in!ut-out!ut 4IIE5 devices& Aet us first describe the memory& It is divided u! into distinct !hysical !ieces, called words& 'or the sake of concreteness, let us say there are DL,L%D words of memory 4a ty!ical number, bein B to the "Dth !ower5& ) word is further divided into what we shall consider the atoms of com!uter science-bits& The number of bits in a ty!ical word mi ht be around thirty-si(& Physically, a bit is .ust a ma netic -switch- that can be in either of two !ositions&

%%% a word of HJ (its %%% Hou could call the two !ositions -u!- and -down-, or -T- and -E-, or -"- and -+-&&& The third is the usual convention& It is !erfectly fine, but it has the !ossibly misleadin effect of makin !eo!le think that a com!uter, dee! down, is storin numbers& This is not true& ) set of thirty-si( bits does not have to be thou ht of as a number any more than two bits has to be thou ht of as the !rice of an ice cream cone& Just as money can do various thin s de!endin on how you use it, so a word in memory can serve many functions& Sometimes, to be sure, those thirty-si( bits will indeed re!resent a number in binary notation& Ether times, they may re!resent thirty-si( dots on a television screen& )nd other times, they may re!resent a few letters of te(t& ?ow a word in memory is to be thou ht of de!ends entirely on the role that this word !lays in the !ro ram which uses it& It may, of course, !lay more than one rolelike a note in a canon&

Instructions and

ata

There is one inter!retation of a word which I haven1t yet mentioned, that is as an instruction& The words of memory contain not only data to be acted on, but also the !ro ram to act on the data& There e(ists a limited re!ertoire of o!erations which can be carried out by the central !rocessin unitthe 7PFand !art of a word, usually its first several bitsis inter!retable as the name of the instruction-ty!e which is to be carried 3hat do the rest of the bits in a word-inter!reted-as-instruction stand for? /ost often, they tell which other words in memory are to be acted u!on& In other words, the remainin bits constitute a pointer to some other word 4or words5 in memory& Every word in memory has a distinct location, like a house on a street= and its location is called its address& /emory may have one -street-, or many -streets-they are called -!a es-& So a iven word is addressed by its !a e number 4if memory is !a ed5 to ether with its !osition within the !a e& ?ence the -!ointer- !art of an instruction is the numerical address of some word4s5 in memory& There are no restrictions on the !ointer, so an instruction may even -!oint- at itself, so that when it is e(ecuted, it causes a chan e in itself to be made& ?ow does the com!uter know what instruction to e(ecute at any iven time? This is ke!t track of in the 7PF& The 7PF has a s!ecial !ointer which !oints at 4i&e&, stores the address of5 the ne(t word which is to be inter!reted as an instruction& The 7PF fetches that word from memory, and co!ies it electronically into a s!ecial word belon in to the 7PF itself& 43ords in the 7PF are usually not called -words-, but rather, registers&5 Then the 7PF e(ecutes that instruction& 8ow the instruction may call for any of a number of ty!es of o!erations to be carried out& Ty!ical ones include, A the word !ointed to in the instruction, to a re ister& 4In this case, the word !ointed to is obviously inter!reted as number&5

,$I;T the word !ointed to in the instruction, as letters& 4In this case, the word is obviously inter!reted not as a number, but as a strin of letters&5 3&M, to the word !ointed to in the instruction& 4In this case, the 7PF is bein told to inter!ret that !articular word as its ne(t instruction&5 Fnless the instruction e(!licitly dictates otherwise, the 7PF will !ick u! the very ne(t word and inter!ret it as an instruction& In other words, the 7PF assumes that it should move down the -street- sequentially, like a mailman, inter!retin word after word as an instruction& 0ut this sequential order can be broken by such instructions as the 3&M, instruction, and others&

Machine Language #s. Assembly Language This is a very brief sketch of machine language& In this lan ua e, the ty!es of o!erations which e(ist constitute a finite re!ertoire which cannot be e(tended& Thus all !ro rams, no matter how lar e and com!le(, must be made out of com!ounds of these ty!es of instructions& Aookin at a !ro ram written in machine lan ua e is va uely com!arable to lookin at a :8) molecule atom by atom& If you lance back to 'i & C", showin the nucleotide sequence of a :8) moleculeand then if you consider that each nucleotide contains two do<en atoms or soand if you ima ine tryin to write the :8), atom by atom, for a small virus 4not to mention a human bein 25then you will et a feelin for what it is like to write a com!le( !ro ram in machine lan ua e, and what it is like to try to ras! what is oin on in a !ro ram if you have access only to its machine lan ua e descri!tion& It must be mentioned, however, that com!uter !ro rammin was ori inally done on an even lower level, if !ossible, than that of machine lan ua enamely, connectin wires to each other, so that the !ro!er o!erations were -hard-wired- in& This is so ama<in ly !rimitive by modern standards that it is !ainful even to ima ine& Het undoubtedly the !eo!le who first did it e(!erienced as much e(hilaration as the !ioneers of modern com!uters ever do&&& 3e now wish to move to a hi her level of the hierarchy of levels of descri!tion of !ro rams& This is the assem(l) language level& There is not a i antic s!read between assembly lan ua e and machine lan ua e= indeed, the ste! is rather entle& In essence, there is a one-to-one corres!ondence between assembly lan ua e instructions and machine lan ua e instructions& The idea of assembly lan ua e is to -chunk- the individual machine lan ua e instructions, so that instead of writin the sequence of bits -+"+"""+++- when you want an instruction which adds one number to another, you sim!ly write A , and then instead of ivin the address in binary re!resentation, you can refer to the word in memory by a name& Therefore, a !ro ram in assembly lan ua e is very much like a machine lan ua e !ro ram made le ible to humans& Hou mi ht com!are the machine lan ua e version of a !ro ram to a T8T-derivation done in the obscure 6>del-numbered notation, and the assembly lan ua e version to the isomor!hic T8T-derivation, done in the ori inal T8Tnotation, which is much easier to understand& Er, oin back to the :8) ima e, we can liken the difference between machine lan ua e and assembly lan ua e to the difference between !ainfully s!ecifyin each nucleotide, atom by atom, and s!ecifyin a nucleotide by sim!ly ivin its name 4i&e&, 1)1, 161, 171, or 1T15& There is a tremendous savin of labor in this very sim!le -chunkin - o!eration, althou h conce!tually not much has been chan ed&

,rograms That Translate ,rograms

Perha!s the central !oint about assembly lan ua e is not its differences from machine lan ua e, which are not that enormous, but .ust the key idea that !ro rams could be written on a different level at all2 Just think about it, the hardware is built to -understand-

machine lan ua e !ro ramssequences of bitsbut not letters and decimal numbers& 3hat ha!!ens when hardware is fed a !ro ram in assembly lan ua e? It is as if you tried to et a cell to acce!t a !iece of !a!er with the nucleotide sequence written out in letters of the al!habet, instead of in chemicals& 3hat can a cell do with a !iece of !a!er? 3hat can a com!uter do with an assembly lan ua e !ro ram? )nd here is the vital !oint, someone can write, in machine lan ua e, a translation program& This !ro ram, called an assem(ler, acce!ts mnemonic instruction names, decimal numbers, and other convenient abbreviations which a !ro rammer can remember easily, and carries out the conversion into the monotonous but critical bit-sequences& )fter the assembly lan ua e !ro ram has been assem(led 4i&e&, translated5, it is runOor rather, its machine lan ua e equivalent is run& 0ut this is a matter of terminolo y& 3hich level !ro ram is runnin ? Hou can never o wron if you say that the machine lan ua e !ro ram is runnin , for hardware is always involved when any !ro ram runsbut it is also quite reasonable to think of the runnin !ro ram in terms of assembly lan ua e& 'or instance, you mi ht very well say, -;i ht now, the 7PF is e(ecutin a 3&M, instruction-, instead of sayin , -;i ht now, the 7PF is e(ecutin a 1"""+"++++1 instruction-& ) !ianist who !lays the notes 6-E-0 E-6-0 is also !layin an ar!e io in the chord of E minor& There is no reason to be reluctant about describin thin s from a hi her-level !oint of view& So one can think of the assembly lan ua e !ro ram runnin concurrently with the machine lan ua e !ro ram& 3e have two modes of describin what the 7PF is doin &

/igher-Le"el Languages) Com!ilers) and Inter!reters

The ne(t level of the hierarchy carries much further the e(tremely !owerful idea of usin the com!uter itself to translate !ro rams from a hi h level into lower levels& )fter !eo!le had !ro rammed in assembly lan ua e for a number of years, in the early "#L+1s, they reali<ed that there were a number of characteristic structures which ke!t rea!!earin in !ro ram after !ro ram& There seemed to be, .ust as in chess, certain fundamental !atterns which cro!!ed u! naturally when human bein s tried to formulate al orithmse(act descri!tions of !rocesses they wanted carried out& In other words, al orithms seemed to have certain hi her-level com!onents, in terms of which they could be much more easily and esthetically s!ecified than in the very restricted machine lan ua e, or assembly lan ua e& Ty!ically, a hi h-level al orithm com!onent consists not of one or two machine lan ua e instructions, but of a whole collection of them, not necessarily all conti uous in memory& Such a com!onent could be re!resented in a hi her-level lan ua e by a sin le itema chunk& )side from standard chunksthe newly discovered com!onents out of which all al orithms can be built!eo!le reali<ed that almost all !ro rams contain even lar er chunkssu!erchunks, so to s!eak& These su!erchunks differ from !ro ram to !ro ram, de!endin on the kinds of hi h-level tasks the !ro ram is su!!osed to carry out& 3e discussed su!erchunks in 7ha!ter G, callin them by their usual names, -subroutines- and -!rocedures-& It was clear that a most !owerful addition to any !ro rammin lan ua e would be the ability to define new hi her-level entities in terms of !reviously known ones, and then to call them by name& This would build the chunkin o!eration ri ht into

the lan ua e& Instead of there bein a determinate re!ertoire of instructions out of which all !ro rams had to be e(!licitly assembled, the !ro rammer could construct his own modules, each with its own name, each usable anywhere inside the !ro ram, .ust as if it had been a built-in feature of the lan ua e& Ef course, there is no ettin away from the fact that down below, on a machine lan ua e level, everythin would still be com!osed of the same old machine lan ua e instructions, but that would not be e(!licitly visible to the hi h-level !ro rammer= it would be im!licit& The new lan ua es based on these ideas were called compiler languages& Ene of the earliest and most ele ant was called -)l ol-, for -)l orithmic Aan ua e-& Fnlike the case with assembly lan ua e, there is no strai htforward one-to-one corres!ondence between statements in )l ol and machine lan ua e instructions& To be sure, there is still a ty!e of ma!!in from )l ol into machine lan ua e, but it is far more -scrambled- than that between assembly lan ua e and machine lan ua e& ;ou hly s!eakin , an )l ol !ro ram is to its machine lan ua e translation as a word !roblem in an elementary al ebra te(t is to the equation it translates into& 4)ctually, ettin from a word !roblem to an equation is far more com!le(, but it ives some inklin of the ty!es of -unscramblin that have to be carried out in translatin from a hi h-level lan ua e to a lower-level lan ua e&5 In the mid-"#L+1s, successful !ro rams called compilers were written whose function was to carry out the translation from com!iler lan ua es to machine lan ua e& )lso, interpreters were invented& Aike com!ilers, inter!reters translate from hi hlevel lan ua es into machine lan ua e, but instead of translatin all the statements first and then e(ecutin the machine code, they read one line and1 e(ecute it immediately& This has the advanta e that a user need not have written a com!lete !ro ram to use an inter!reter& ?e may invent his !ro ram line by line, and test it out as he oes alon & Thus, an inter!reter is to a com!iler as a simultaneous inter!reter is to a translator of a written s!eech& Ene of the most im!ortant and fascinatin of all com!uter lan ua es is AISP 4standin for -Aist Processin -5, which was invented by John /c7arthy around the time )l ol was invented& Subsequently, AISP has en.oyed reat !o!ularity with workers in )rtificial Intelli ence& There is one interestin difference between the way inter!reters work and com!ilers work& ) com!iler takes in!ut 4a finished )l ol !ro ram, for instance5 and !roduces out!ut 4a lon sequence of machine lan ua e instructions5& )t this !oint, the com!iler has done its duty& The out!ut is then iven to the com!uter to run& 0y contrast, the inter!reter is constantly runnin while the !ro rammer ty!es in one AISP statement after another, and each one ets e(ecuted then and there& 0ut this doesn1t mean that each statement ets first translated, then e(ecuted, for then an inter!reter would be nothin but a line-by-line com!iler& Instead, in an inter!reter, the o!erations of readin a new line, -understandin - it, and e(ecutin it are intertwined, they occur simultaneously& ?ere is the idea, e(!anded a little more& Each time a new line of AISP is ty!ed in, the inter!reter tries to !rocess it& This means that the inter!reter .olts into action, and certain 4machine lan ua e5 instructions inside it et e(ecuted& Precisely which ones et e(ecuted de!ends on the AISP statement itself, of course& There are many 3&M, instructions inside the inter!reter, so that the new line of AISP may cause control to move around in a com!le( wayforwards, backwards, then forwards a ain, etc& Thus, each

AISP statement ets converted into a -!athway- inside the inter!reter, and the act of followin that !athway achieves the desired effect& Sometimes it is hel!ful to think of the AISP statements as mere !ieces of data which are fed sequentially to a constantly runnin machine lan ua e !ro ram 4the AISP inter!reter5& 3hen you think of thin s this way, you et a different ima e of the relation between a !ro ram written in a hi her-level lan ua e and the machine which is e(ecutin it&

Bootstra!!ing
Ef course a com!iler, bein itself a !ro ram, has to be written in some lan ua e& The first com!ilers were written in assembly lan ua e, rather than machine lan ua e, thus takin full advanta e of the already accom!lished first ste! u! from machine lan ua e& ) summary of these rather tricky conce!ts is !resented in 'i ure L*&

73G5=E GE! *ssem(lers and compilers are (oth translators into machine language! This is indicated () the solid lines! 'oreo#er, since the) are themsel#es programs, the) are originall) written in a language also! The wa#) lines indicate that a compiler can (e written in assem(l) language, and an assem(ler in machine language!

8ow as so!histication increased, !eo!le reali<ed that a !artially written com!iler could be used to com!ile e(tensions of itself& In other words, once a certain minimal core of a com!iler had been written, then that minimal com!iler could translate bi er com!ilers into machine lan ua ewhich in turn could translate yet bi er com!ilers, until the final, full-blown com!iler had been com!iled& This !rocess is affectionately known as -bootstra!!in -for obvious reasons 4at least if your native lan ua e is En lish it is obvious5& It is not so different from the attainment by a child of a critical level of fluency in his native lan ua e, from which !oint on his vocabulary and fluency can row by lea!s and bounds, since he can use lan ua e to acquire new lan ua e&

Le"els on >hich to

escribe $unning ,rograms

7om!iler lan ua es ty!ically do not reflect the structure of the machines which will run !ro rams written in them& This is one of their chief advanta es over the hi hly s!eciali<ed assembly and machine lan ua es& Ef course, when a com!iler lan ua e !ro ram is translated into machine lan ua e, the resultin !ro ram is machine-

de!endent& Therefore one can describe a !ro ram which is bein e(ecuted in a machineinde!endent way or a machine-de!endent way& It is like referrin to a !ara ra!h in a book by its sub.ect matter 4!ublisher-inde!endent5, or its !a e number and !osition on the !a e 4!ublisher-de!endent5& )s lon as a !ro ram is runnin correctly, it hardly matters how you describe it or think of its functionin & It is when somethin oes wron that it is im!ortant to be able to think on different levels& If, for instance, the machine is instructed to divide by <ero at some sta e, it will come to a halt and let the user know of this !roblem, by tellin where in the !ro ram the questionable event occurred& ?owever, the s!ecification is often iven on a lower level than that in which the !ro rammer wrote the !ro ram& ?ere are three !arallel descri!tions of a !ro ram rindin to a halt, /achine Aan ua e Aevel, -E(ecution of the !ro ram sto!!ed in location """++"+"+"""+""")ssembly Aan ua e Aevel, -E(ecution of the !ro ram sto!!ed when the :IG 4divide5 instruction was hit7om!iler Aan ua e Aevel, -E(ecution of the !ro ram sto!!ed durin e(!ression 14) ^ 05I91 evaluation of the al ebraic

Ene of the reatest !roblems for systems !ro rammers 4the !eo!le who write com!ilers, inter!reters, assemblers, and other !ro rams to be used by many !eo!le5 is to fi ure out how to write error-detectin routines in such a way that the messa es which they feed to the user whose !ro ram has a -bu - !rovide hi h-level, rather than low-level, descri!tions of the !roblem& It is an interestin reversal that when somethin oes wron in a enetic -!ro ram- 4e& &, a mutation5, the -bu - is manifest only to !eo!le on a hi h levelnamely on the !henoty!e level, not the enoty!e level& )ctually, modern biolo y uses mutations as one of its !rinci!al windows onto enetic !rocesses, because of their multilevel traceability&

Micro!rogramming and 0!erating Systems

In modern com!uter systems, there are several other levels of the hierarchy& 'or instance, some systemsoften the so-called -microcom!uters-come with machine lan ua e instructions which are even more rudimentary than the instruction to add a number in memory to a number in a re ister& It is u! to the user to decide what kinds of ordinary machine-level instructions he would like to be able to !ro ram in= he -micro!ro ramsthese instructions in terms of the -micro-instructions- which are available& Then the -hi her-level machine lan ua e- instructions which he has desi ned may be burned into the circuitry and become hard-wired, althou h they need not be& Thus micro!ro rammin allows the user to ste! a little below the conventional machine lan ua e level& Ene of the consequences is that a com!uter of one manufacturer can be hard-wired 4via

micro!ro rammin 5 so as to have the same machine lan ua e instruction set as a com!uter of the same, or even another, manufacturer& The micro!ro rammed com!uter is said to be -emulatin - the other com!uter& Then there is the level of the operating s)stem, which fits between the machine lan ua e !ro ram and whatever hi her level the user is !ro rammin in& The o!eratin system is itself a !ro ram which has the functions of shieldin the bare machine from access by users 4thus !rotectin the system5, and also of insulatin the !ro rammer from the many e(tremely intricate and messy !roblems of readin the !ro ram, callin a translator, runnin the translated !ro ram, directin the out!ut to the !ro!er channels at the !ro!er time, and !assin control to the ne(t user& If there are several users -talkin - to the same 7PF at once, then the o!eratin system is the !ro ram that shifts attention from one to the other in some orderly fashion& The com!le(ities of o!eratin systems are formidable indeed, and I shall only hint at them by the followin analo y& 7onsider the first tele!hone system& )le(ander 6raham 0ell could !hone his assistant in the ne(t room, electronic transmission of a voice2 8ow that is like a bare com!uter minus o!eratin system, electronic com!utation2 7onsider now a modern tele!hone system& Hou have a choice of other tele!hones to connect to& 8ot only that, but many different calls can be handled simultaneously& Hou can add a !refi( and dial into different areas& Hou can call direct, throu h the o!erator, collect, by credit card, !ersonto-!erson, on a conference call& Hou can have a call rerouted or traced& Hou can et a busy si nal& Hou can et a siren-like si nal that says that the number you dialed isn1t -wellformed-, or that you have taken too in lon in dialin & Hou can install a local switchboard so that a rou! of !hones are all locally connectedetc&, etc& The list is ama<in , when you think of how much fle(ibility there is, !articularly in com!arison to the erstwhile miracle of a -bare- tele!hone& 8ow so!histicated o!eratin systems carry out similar traffic-handlin and level-switchin o!erations with res!ect to users and their !ro rams& It is virtually certain that there are somewhat !arallel thin s which take !lace in the brain, handlin of many stimuli at the same time= decisions of what should have !riority over what and for how lon = instantaneous -interru!ts- caused by emer encies or other une(!ected occurrences= and so on&

Cushioning the &ser and ,rotecting the System

The many levels in a com!le( com!uter system have the combined effect of -cushionin the user, !reventin him from havin to think about the many lower-level oin s-on which are most likely totally irrelevant to him anyway& ) !assen er in an air!lane does not usually want to be aware of the levels of fuel in the tanks, or the wind s!eeds, or how many chicken dinners are to be served, or the status of the rest of the air traffic around the destinationthis is all left to em!loyees on different levels of the airline1s hierarchy, and the !assen er sim!ly ets from one !lace to another& ?ere a ain, it is when somethin oes wron such as his ba a e not arrivin that the !assen er is made aware of the confusin system of levels underneath him&

Are Com!uters Su!er-+le-ible or Su!er-$igid7

Ene of the ma.or oals of the drive to hi her levels has always been to make as natural as !ossible the task of communicatin to the com!uter what you want it to do& 7ertainly, the hi h-level constructs in com!iler lan ua es are closer to the conce!ts which humans naturally think in, than are lower-level constructs such as those in machine lan ua e& 0ut in this drive towards ease of communication, one as!ect of -naturalness- has been quite ne lected& That is the fact that interhuman communication is far less ri idly constrained than human-machine communication& 'or instance, we often !roduce meanin less sentence fra ments as we search for the best way to e(!ress somethin , we cou h in the middle of sentences, we interru!t each other, we use ambi uous descri!tions and -im!ro!er- synta(, we coin !hrases and distort meanin sbut our messa e still ets throu h, mostly& 3ith !ro rammin lan ua es, it has enerally been the rule that there is a very strict synta( which has to be obeyed one hundred !er cent of the time= there are no ambi uous words or constructions& Interestin ly, the !rinted equivalent of cou hin 4i&e&, a nonessential or irrelevant comment5 is allowed, but only !rovided it is si naled in advance by a key word 4e& &, C0MME;T5, and then terminated by another key word 4e& &, a semicolon5& This small esture towards fle(ibility has its own little !itfall, ironically, if a semicolon 4or whatever key word is used for terminatin a comment5 is used inside a comment, the translatin !ro ram will inter!ret that semicolon as si nalin the end of the comment, and havoc will ensue& If a !rocedure named I;SIG/T has been defined and then called seventeen times in the !ro ram, and the ei hteenth time it is miss!elled as I;SI/GT, woe to the !ro rammer& The com!iler will balk and !rint a ri idly unsym!athetic error messa e, sayin that it has never heard of I;SI/GT& Eften, when such an error is detected by a com!iler, the com!iler tries to continue, but because of its lack of insi ht, it has not understood what the !ro rammer meant& In fact, it may very well su!!ose that somethin entirely different was meant, and !roceed under that erroneous assum!tion& Then a lon series of error messa es will !e!!er the rest of the !ro ram, because the com!ilernot the !ro rammer ot confused& Ima ine the chaos that would result if a simultaneous En lish-;ussian inter!reter, u!on hearin one !hrase of 'rench in the En lish, be an tryin to inter!ret all the remainin En lish as 'rench& 7om!ilers often et lost in such !athetic ways& 8'est la #ie! Perha!s this sounds condemnatory of com!uters, but it is not meant to be& In some sense, thin s had to be that way& 3hen you sto! to think what most !eo!le use com!uters for, you reali<e that it is to carry out very definite and !recise tasks, which are too com!le( for !eo!le to do& If the com!uter is to be reliable, then it is necessary that it should understand, without the sli htest chance of ambi uity, what it is su!!osed to do& It is also necessary that it should do neither more nor less than it is e(!licitly instructed to do& If there is, in the cushion underneath the !ro rammer, a !ro ram whose !ur!ose is to - uess- what the !ro rammer wants or means, then it is quite conceivable that the !ro rammer could try to communicate his task and be totally misunderstood& So it is im!ortant that the hi h-level !ro ram, while comfortable for the human, still should be unambi uous and !recise&

Second-Guessing the ,rogrammer

8ow it is !ossible to devise a !ro rammin lan ua eand a !ro ram which translates it into the lower levelswhich allows some sorts of im!recision& Ene way of !uttin it would be to say that a translator for such a !ro rammin lan ua e tries to make sense of thin s which are done -outside of the rules of the lan ua e-& 0ut if a lan ua e allows certain -trans ressions-, then trans ressions of that ty!e are no lon er true trans ressions, because they have been included inside the rules2 If a !ro rammer is aware that he may make certain ty!es of miss!ellin , then he may use this feature of the lan ua e deliberately, knowin that he is actually o!eratin within the ri id rules of the lan ua e, des!ite a!!earances& In other words, if the user is aware of all the fle(ibilities !ro rammed into the translator for his convenience, then he knows the bounds which he cannot overste!, and therefore, to him, the translator still a!!ears ri id and infle(ible, althou h it may allow him much more freedom than early versions of the lan ua e, which did not incor!orate -automatic com!ensation for human error-& 3ith -rubbery- lan ua es of that ty!e, there would seem to be two alternatives, 4"5 the user is aware of the built-in fle(ibilities of the lan ua e and its translator= 4B5 the user is unaware of them& In the first case, the lan ua e is still usable for communicatin !ro rams !recisely, because the !ro rammer can !redict how the com!uter will inter!ret the !ro rams he writes in the lan ua e& In the second case, the -cushion- has hidden features which may do thin s that are un!redictable 4from the vanta e !oint of a user who doesn1t know the inner workin s of the translator5& This may result in ross misinter!retations of !ro rams, so such a lan ua e is unsuitable for !ur!oses where com!uters are used mainly for their s!eed and reliability& 8ow there is actually a third alternative, 4%5 the user is aware of the built-in fle(ibilities of the lan ua e and its translator, but there are so many of them and they interact with each other in such a com!le( way that he cannot tell how his !ro rams will be inter!reted& This may well a!!ly to the !erson who wrote the translatin !ro ram= he certainly knows its insides as well as anyone couldbut he still may not be able to antici!ate how it will react to a iven ty!e of unusual construction& Ene of the ma.or areas of research in )rtificial Intelli ence today is called automatic programming, which is concerned with the develo!ment of yet hi her-level lan ua eslan ua es whose translators are so!histicated& in that they can do at least some of the followin im!ressive thin s, enerali<e from e(am!les, correct some mis!rints or rammatical errors, try to make sense of ambi uous descri!tions, try to second- uess the user by havin a !rimitive user model, ask questions when thin s are unclear, use En lish itself, etc& The ho!e is that one can walk the ti htro!e between reliability and fle(ibility&

AI Ad"ances Are Language Ad"ances

It is strikin how ti ht the connection is between !ro ress in com!uter science 4!articularly )rtificial Intelli ence5 and the develo!ment of new lan ua es& ) clear trend has emer ed in the last decade, the trend to consolidate new ty!es of discoveries in new

lan ua es& Ene key for the understandin and creation of intelli ence lies in the constant develo!ment and refinement of the lan ua es in terms of which !rocesses for symbol mani!ulation are describable& Today, there are !robably three or four do<en e(!erimental lan ua es which have been develo!ed e(clusively for )rtificial Intelli ence research& It is im!ortant to reali<e that any !ro ram which can be written in one of these lan ua es is in !rinci!le !ro rammable in lower-level lan ua es, but it would require a su!reme effort for a human= and the resultin !ro ram would be so lon that it would e(ceed the ras! of humans& It is not that each hi her level e(tends the !otential of the com!uter= the full !otential of the com!uter already e(ists in its machine lan ua e instruction set& It is that the new conce!ts in a hi h-level lan ua e su est directions and !ers!ectives by their very nature& The -s!ace- of all !ossible !ro rams is so hu e that no one can have a sense of what is !ossible& Each hi her-level lan ua e is naturally suited for e(!lorin certain re ions of -!ro ram s!ace-= thus the !ro rammer, by usin that lan ua e, is channeled into those areas of !ro ram s!ace& ?e is not forced by the lan ua e into writin !ro rams of any !articular ty!e, but the lan ua e makes it easy for him to do certain kinds of thin s& Pro(imity to a conce!t, and a entle shove, are often all that is needed for a ma.or discoveryand that is the reason for the drive towards lan ua es of ever hi her levels& Pro rammin in different lan ua es is like com!osin !ieces in different keys, !articularly if you work at the keyboard& If you have learned or written !ieces in many keys, each key will have its own s!ecial emotional aura& )lso, certain kinds of fi urations -lie in the hand- in one key but are awkward in another& So you are channeled by your choice of key& In some ways, even enharmonic keys, such as 7-shar! and :-flat, are quite distinct in feelin & This shows how a notational system can !lay a si nificant role in sha!in the final !roduct& ) -stratified- !icture of )I is shown in 'i ure L#, with machine com!onents such as transistors on the bottom, and -intelli ent !ro rams- on the to!& The !icture is taken from the book *rtificial 3ntelligence by Patrick ?enry 3inston, and it re!resents a vision of )I shared by nearly all )I workers& )lthou h I a ree with the idea that )I must be stratified in some such way, I do not think that, with so few layers, intelli ent !ro rams can he reached& 0etween the machine lan ua e level and the level where true intelli ence will be reached, I am convinced there will lie !erha!s another do<en 4or even several do<en25 layers, each new layer buildin on and e(tendin the fle(ibilities of the layer below& 3hat they will be like we can hardly dream of now&&&

73G5=E GI! To create intelligent programs, one needs to (uild up a series of le#els of hardware and software, so that one is spared the agon) of seeing e#er)thing onl) on the lowest le#el! .escriptions of a single process on different le#els will sound #er( different from each other, onl) the top one (eing sufficientl) chunked that it is comprehensi(le to us! U*dapted from &! ,! >inston, )rtificial Intelli ence =eading, 'ass!: *ddison%>esle), BICC/V

The ,aranoid and the 0!erating System

The similarity of all levels in a com!uter system can lead to some stran e level-mi(in e(!eriences& I once watched a cou!le of friendsboth com!uter novices!layin with the !ro ram -P);;H- on a terminal& P);;H is a rather infamous !ro ram which simulates a !aranoid in an e(tremely rudimentary way, by s!ittin out canned !hrases in En lish chosen from a wide re!ertoire= its !lausibility is due to its ability to tell which of its stock !hrases mi ht sound reasonable in res!onse to En lish sentences ty!ed to it by a human& )t one !oint, the res!onse time ot very slu ishP);;H was takin very lon to re!lyand I e(!lained to my friends that this was !robably because of the heavy load on the time-sharin system& I told them they could find out how many users were lo ed on, by ty!in a s!ecial -control- character which would o directly to the o!eratin system, and would be unseen by P);;H& Ene of my friends !ushed the control character& In a flash, some internal data about the o!eratin system1s status overwrote some of P);;H1s words on the screen& P);;H knew nothin of this, it is a !ro ram with -knowled e- only of horse racin and bookiesnot o!eratin systems and terminals and s!ecial control characters& 0ut to my friends, both P);;H and the o!eratin system were .ust -the com!uter-a mysterious, remote, amor!hous entity that res!onded to them when they ty!ed& )nd so it made !erfect sense when one of them blithely ty!ed, in En lish, -3hy are you overty!in what1s on the screen?- The idea that P);;H could know nothin about the o!eratin system it was runnin under was not clear to my friends& The idea that -you- know all about -yourself- is so familiar from interaction with !eo!le that it was natural to e(tend it to the com!uterafter all, it was intelli ent enou h that it could -talk- to them in En lish2 Their question was not unlike askin a !erson, -3hy are you makin so few red blood cells today?- Peo!le do not know about that level the -o!eratin system level-of their bodies& The main cause of this level-confusion was that communication with all levels of the com!uter system was takin !lace on a sin le screen, on a sin le terminal& )lthou h my friends1 naPvetK mi ht seem rather e(treme, even e(!erienced com!uter !eo!le often

make similar errors when several levels of a com!le( system are all !resent at once on the same screen& They for et -who- they are talkin to, and ty!e somethin which makes no sense at that level, althou h it would have made !erfect sense on another level& It mi ht seem desirable, therefore, to have the system itself sort out the levelsto inter!ret commands accordin to what -makes sense-& Fnfortunately, such inter!retation would require the system to have a lot of common sense, as well as !erfect knowled e of the !ro rammer1s overall intentboth of which would require more artificial intelli ence than e(ists at the !resent time&

The Border bet'een Soft'are and /ard'are

Ene can also be confused by the fle(ibility of some levels and the ri idity of others& 'or instance, on some com!uters there are marvelous te(t-editin systems which allow !ieces of te(t to be -!oured- from one format into another, !ractically as liquids can be !oured from one vessel into another& ) thin !a e can turn into a wide !a e, or vice versa& 3ith such !ower, you mi ht e(!ect that it would be equally trivial to chan e from one font to anothersay from roman to italics& Het there may be only a sin le font available on the screen, so that such chan es are im!ossible& Er it may be feasible on the screen but not !rintable by the !rinteror the other way around& )fter dealin with com!uters for a lon time, one ets s!oiled, and thinks that everythin should be !ro rammable, no !rinter should be so ri id as to have only one character set, or even a finite re!ertoire of themty!efaces should be user-s!ecifiable2 0ut once that de ree of fle(ibility has been attained, then one may be annoyed that the !rinter cannot !rint in different colors of ink, or that it cannot acce!t !a!er of all sha!es and si<es, or that it does not fi( itself when it breaks&&& The trouble is that somewhere, all this fle(ibility has to -bottom out-, to use the !hrase from 7ha!ter G& There must be a hardware level which underlies it all, and which is infle(ible& It may lie dee!ly hidden, and there may be so much fle(ibility on levels above it that few users feel the hardware limitationsbut it is inevitably there& 3hat is this !roverbial distinction between software and hardware? It is the distinction between !ro rams and machinesbetween lon com!licated sequences of instructions, and the !hysical machines which carry them out& I like to think of software as -anythin which you could send over the tele!hone lines-, and hardware as -anythin else-& ) !iano is hardware, but !rinted music is software& ) tele!hone set is hardware, but a tele!hone number is software& 1The distinction is a useful one, but not always so clearcut& 3e humans also have -software- and -hardware- as!ects, and the difference is second nature to us& 3e are used to the ri idity of our !hysiolo y, the fact that we cannot, at will, cure ourselves of diseases, or row hair of any colorto mention .ust a cou!le of sim!le e(am!les& 3e can, however, -re!ro ram- our minds so that we o!erate in new conce!tual frameworks& The ama<in fle(ibility of our minds seems nearly irreconcilable with the notion that our brains must be made out of fi(ed-rule hardware, which cannot be re!ro rammed& 3e cannot make our neurons fire faster or slower, we cannot rewire our

brains, we cannot redesi n the interior of a neuron, we cannot make any choices about the hardwareand yet, we can control how we think& 0ut there are clearly as!ects of thou ht which are beyond our control& 3e cannot make ourselves smarter by an act of will= we cannot learn a new lan ua e as fast as we want= we cannot make ourselves think faster than we do= we cannot make ourselves think about several thin s at once= and so on& This is a kind of !rimordial self-knowled e which is so obvious that it is hard to see it at all= it is like bein conscious that the air is there& 3e never really bother to think about what mi ht cause these -defects- of our minds, namely, the or ani<ation of our brains& To su est ways of reconcilin the software of mind with the hardware of brain is a main oal of this book&

Intermediate Le"els and the >eather

3e have seen that in com!uter systems, there are a number of rather shar!ly defined strata, in terms of any one of which the o!eration of a runnin !ro ram can be described& Thus there is not merely a sin le low level and a sin le hi h levelthere are all de rees of lowness and hi hness& Is the e(istence of intermediate levels a eneral feature of systems which have low and hi h levels? 7onsider, for e(am!le, the system whose 1hardware- is the earth1s atmos!here 4not very hard, but no matter5, and whose -softwareis the weather& @ee!in track of the motions of all of the molecules simultaneously would be a very low-level way of -understandin - he weather, rather like lookin at a hu e, com!licated !ro ram on the machine lan ua e level& Ebviously it is way beyond human com!rehension& 0ut we still have our own !eculiarly human ways of lookin at, and describin , weather !henomena& Eur chunked view of the weather is based on very hi hlevel !henomena, such as, rain, fo , snow, hurricanes, cold fronts, seasons, !ressures, trade winds, the .et stream, cumulo-nimbus clouds, thunderstorms, inversion layers, and so on& )ll of these !henomena involve astronomical numbers of molecules, somehow behavin in concert so that lar e-scale trends emer e& This is a little like lookin at the weather in a com!iler lan ua e& Is there somethin analo ous to lookin at the weather in an intermediate-level lan ua e, such as assembly lan ua e? 'or instance, are there very small local -ministorms-, somethin like the small whirlwinds which one occasionally sees, whi!!in u! some dust in a swirlin column a few feet wide, at most? Is a local ust of wind an intermediate-level chunk which !lays a role in creatin hi her-level weather !henomena? Er is there .ust no !ractical way of combinin knowled e of such kinds of !henomena to create a more com!rehensive e(!lanation of the weather? Two other questions come to my mind& The first is, -7ould it be that the weather !henomena which we !erceive on our scalea tornado, a drou htare .ust intermediatelevel !henomena, !arts of vaster, slower !henomena?- If so, then true hi h-level weather !henomena would be lobal, and their time scale would be eolo ical& The Ice ) e would be a hi h-level weather event& The second question is, -)re there intermediate level weather !henomena which have so far esca!ed human !erce!tion, but which, if !erceived, could ive reater insi ht into why the weather is as it is?-

+rom Tornadoes to 8uar.s

This last su estion may sound fanciful, but it is not all that far-fetched& 3e need only look to the hardest of the hard sciences!hysicsto find !eculiar e(am!les of systems which are e(!lained in terms of interactin -!arts- which are themselves invisible& In !hysics, as in any other disci!line, a system is a rou! of interactin !arts& In most systems that we know, the !arts retain their identities durin the interaction, so that we still see the !arts inside the system& 'or e(am!le, when a team of football !layers assembles, the individual !layers retain their se!aratenessthey do not melt into some com!osite entity, in which their individuality is lost& Stilland this is im!ortantsome !rocesses are oin on in their brains which are evoked by the team-conte(t, and which would not o on otherwise, so that in a minor way, the !layers chan e identity when they become !art of the lar er system, the team& This kind of system is called a nearl) decomposa(le s)stem 4the term comes from ?& )& Simon1s article -The )rchitecture of 7om!le(ity-= see the 0iblio ra!hy5& Such a system consists of weakly interactin modules, each of which maintains its own !rivate identity throu hout the interaction but by becomin sli htly different from how it is when outside of the system, contributes to the cohesive behavior of the whole system& The systems studied in !hysics are usually of this ty!e& 'or instance, an atom is seen as made of a nucleus whose !ositive char e ca!tures a number of electrons in -orbits-, or bound states& The bound electrons are very much like free electrons, des!ite their bein internal to a com!osite ob.ect& Some systems studied in !hysics offer a contrast to the relatively strai htforward atom& Such systems involve e(tremely stron interactions, as a result of which the !arts are swallowed u! into the lar er system, and lose some or all of their individuality& )n e(am!le of this is the nucleus of an atom, which is usually described as bein -a collection of !rotons and neutrons-& 0ut the forces which !ull the com!onent !articles to ether are so stron that the com!onent !articles do not survive in anythin like their -free- form 4the form they have when outside a nucleus5& )nd in fact a nucleus acts in many ways as a sin le !article, rather than as a collection of interactin !articles& 3hen a nucleus is s!lit, !rotons and neutrons are often released, but also other !articles, such as !i-mesons and amma rays, are commonly !roduced& )re all those different !articles !hysically !resent inside a nucleus before it is s!lit, or are they .ust -s!arks- which fly off when the nucleus is s!lit? It is !erha!s not meanin ful to try to ive an answer to such a question& En the level of !article !hysics, the difference between storin the !otential to make -s!arks- and storin actual sub!articles is not so clear& ) nucleus is thus one systems whose -!arts-, even thou h they are not visible while on the inside, can be !ulled out and made visible& ?owever, there are more !atholo ical cases& such as the !roton and neutron seen as systems themselves& Each of them has been hy!othesi<ed to be constituted from a trio of -quarks-hy!othetical !articles which can be combined in twos or threes to make many known fundamental !articles& ?owever, the interaction between quarks is so stron that not only can they not be seen inside the !roton and neutron, but they cannot even be !ulled out at all2 Thus,

althou h quarks hel! to ive a theoretical understandin of certain !ro!erties of !rotons and neutrons, their own e(istence may !erha!s never be inde!endently established& ?ere we have the antithesis of a -nearly decom!osable system-it is a system which, if anythin , is -nearly indecom!osable-& Het what is curious is that a quark-based theory of !rotons and neutrons 4and other !articles5 has considerable e(!lanatory !ower, in that many e(!erimental results concernin the !articles which quarks su!!osedly com!ose can be accounted for quite well, quantitatively, by usin the -quark model-&

Su!erconducti"ity5 A (,arado-( of $enormali:ation

In 7ha!ter G we discussed how renormali<ed !articles emer e from their bare cores, by recursively com!ounded interactions with virtual !articles& ) renormali<ed !article can be seen either as this com!le( mathematical construct, or as the sin le lum! which it is, !hysically& Ene of the stran est and most dramatic consequences of this way of describin !articles is the e(!lanation it !rovides for the famous !henomenon of superconducti#it), resistance-free flow of electrons in certain solids, at e(tremely low tem!eratures& It turns out that electrons in solids are renormali<ed by their interactions with stran e quanta of vibration called phonons 4themselves renormali<ed as well25& These renormali<ed electrons are called polarons& 7alculation shows that at very low tem!eratures, two o!!ositely s!innin !olarons will be in to attract each other, and can actually become bound to ether in in a certain way& Fnder the !ro!er conditions& all the current-carryin !olarons will et !aired u!, formin 8ooper pairs& Ironically, this !airin comes about !recisely because electronsthe bare cores of the !aired !olarons re!el each other electrically& In contrast to the electrons, each 7oo!er !air feels neither attracted to nor re!elled by another 7oo!er !air, and consequently it can sli! freely throu h a metal as if the metal were a vacuum& If you convert the mathematical descri!tion of such a metal from one whose !rimitive units are !olarons into one whose !rimitive units are 7oo!er !airs& you et a considerably sim!lified set of equations& This mathematical sim!licity is the !hysicist1s way of knowin that -chunkin - into 7oo!er !airs is the natural way to look at su!erconductivity& ?ere we have several levels of !article, the 7oo!er !air itself, the two o!!ositelys!innin !olarons which com!ose it, the electrons and !honons which make u! the !olarons, and then, within the electrons, the virtual !hotons and !ositrons, etc& etc& 3e can look at each level and !erceive !henomena there, which are e(!lained by an understandin of the levels below&

(Sealing-off(
Similarly, and fortunately& one does not have to know all about quarks to understand many thin s about the !articles which they may com!ose& Thus, a nuclear !hysicist can !roceed with theories of nuclei that are based on !rotons and neutrons, and i nore quark theories and their rivals& The nuclear !hysicist has a chunked !icture of !rotons and neutronsa descri!tion derived from lower-level theories but which does not require

understandin the lower-level theories& Aikewise, an atomic !hysicist has a chunked !icture of an atomic nucleus derived from nuclear theory& Then a chemist has a chunked !icture of the electrons and their orbits, and builds theories of small molecules, theories which can be taken over in a chunked way by the molecular biolo ist, who has an intuition for how small molecules han to ether, but whose technical e(!ertise is in the field of e(tremely lar e molecules and how they interact& Then the cell biolo ist has a chunked !icture of the units which the molecular biolo ist !ores over, and tries to use them to account for the ways that cells interact& The !oint is clear& Each level is, in some sense, -sealed off- from the levels below it& This is another of Simon1s vivid terms, recallin the way in which a submarine is built in com!artments, so that if one !art is dama ed, and water be ins !ourin in, the trouble can be !revented from s!readin , by closin the doors, thereby sealin off the dama ed com!artment from nei hborin com!artments& )lthou h there is always some -leaka e- between the hierarchical levels of science, so that a chemist cannot afford to i nore lower-level !hysics totally, or a biolo ist to i nore chemistry totally, there is almost no leaka e from one level to a distant level& That is why !eo!le can have intuitive understandin s of other !eo!le without necessarily understandin the quark model, the structure of nuclei, the nature of electron orbits, the chemical bond, the structure of !roteins, the or anelles in a cell, the methods of intercellular communication, the !hysiolo y of the various or ans of the human body, or the com!le( interactions amon or ans& )ll that a !erson needs is a chunked model of how the hi hest level acts= and as we all know, such models are very realistic and successful&

The Trade-off bet'een Chun.ing and

eterminism

There is, however, !erha!s one si nificant ne ative feature of a chunked model, it usually does not have e(act !redictive !ower& That is, we save ourselves from the im!ossible task of seein !eo!le as collections of quarks 4or whatever is at the lowest level5 by usin chunked models, but of course such models only ive us !robabilistic estimates of how other !eo!le feel, will react to what we say or do, and so on& In short, in usin chunked hi h-level models, we sacrifice determinism for sim!licity& :es!ite not bein sure how !eo!le will react to a .oke, we tell it with the e(!ectation at they will do somethin such as lau h, or not lau hrather than, say, climb the nearest fla !ole& 49en masters mi ht well do the latter25 ) chunked model defines a -s!ace- within which behavior is e(!ected to fall, and s!ecifies !robabilities of its fallin in different !arts of that s!ace&

(Com!uters Can 0nly

o >hat =ou Tell Them to

o(

8ow these ideas can be a!!lied as well to com!uter !ro rams as to com!osite !hysical systems& There is an old saw which says, -7om!uters can only do what you tell them to do&- This is ri ht in one sense, but it misses the !oint, you don1t know in advance the consequences of what you tell a com!uter to do= therefore its behavior can be as bafflin and sur!risin and un!redictable to you as that of a !erson& Hou enerally know in

advance the s!ace in which the out!ut will fall, but you don1t know details of where it will fall& 'or instance, you mi ht write a !ro ram to calculate the first million di its of m& Hour !ro ram will s!ew forth di its of m much faster than you canbut there is no !arado( in the fact that the com!uter outracin its !ro rammer& Hou know in advance the s!ace in which the out!ut will lienamely the s!ace of di its between + and #which is to say, you have a chunked model of the !ro ram1s behavior= but if you1d known the rest, you wouldn1t have written the !ro ram& There is another sense in which this old saw is rusty& This involves the fact that as you !ro ram in ever hi her-level lan ua es, you know less and less !recisely what you1ve told the com!uter to do2 Aayers and layers of translation may se!arate the -front end- of a com!le( !ro ram from the actual machine lan ua e instructions& )t the level you think and !ro ram, your statements may resemble declaratives and su estions more than they resemble im!eratives or commands& )nd all the internal rumblin !rovoked by the in!ut of a hi h-level statement is invisible to you, enerally, .ust as when you eat a sandwich, you are s!ared conscious awareness of the di estive !rocesses it tri ers& In any case, this notion that -com!uters can only do what they are told to do,- first !ro!ounded by Aady Aovelace in her famous memoir, is so !revalent and so connected with the notion that -com!uters cannot think- that we shall return to it in later 7ha!ters when our level of so!histication is reater&

T'o Ty!es of System

There is an im!ortant division between two ty!es of system built u! from many !arts& There are those systems in which the behavior of some !arts tends to cancel out the behavior of other !arts, with the result that it does not matter too much what ha!!ens on the low level, because most anythin will yield similar hi h-level behavior& )n e(am!le of this kind of system is a container of as, where all the molecules bum! and ban a ainst each other in very com!le( microsco!ic ways= but the total outcome, from a macrosco!ic !oint of view, is a very calm, stable system with a certain tem!erature, !ressure, and volume& Then there are systems where the effect of a sin le low-level event may et magnified into an enormous hi h-level consequence& Such a system is a !inball machine, where the e(act an le with which a ball strikes each !ost is crucial in determinin the rest of its descendin !athway& ) com!uter is an elaborate combination of these two ty!es of system& It contains subunits such as wires, which behave in a hi hly !redictable fashion, they conduct electricity accordin to Ehm1s law, a very !recise, chunked law which resembles the laws overnin ases in containers, since it de!ends on statistical effects in which billions of random effects cancel each other out, yieldin a !redictable overall behavior& ) com!uter also contains macrosco!ic subunits, such as a !rinter, whose behavior is com!letely determined by delicate !atterns of currents& 3hat the !rinter !rints is not by any means created by a myriad of cancelin microsco!ic effects& In fact, in the case of most com!uter !ro rams, the value of every sin le bit in the !ro ram !lays a critical role in the out!ut that ets !rinted& If any bit were chan ed, the out!ut would also chan e drastically&

Systems which are made u! of -reliable- subsystems onlythat is, subsystems whose behavior can be reliably !redicted from chunked descri!tions!lay inestimably im!ortant roles in our daily lives, because they are !illars of stability& 3e can rely on walls not to fall down, on sidewalks to o where they went yesterday, on the sun to shine, on clocks to tell the time correctly, and so on& 7hunked models of such systems are virtually entirely deterministic& Ef course, the other kind of system which !lays a very lar e role in our lives is a system that has variable behavior which de!ends on some internal microsco!ic !arametersoften a very lar e number of them, moreoverwhich we cannot directly observe& Eur chunked model of such a system is necessarily in terms of the -s!ace- of o!eration, and involves !robabilistic estimates of landin in different re ions of that s!ace& ) container of as, which, as I already !ointed out, is a reliable system because of many cancelin effects, obeys !recise, deterministic laws of !hysics& Such laws are chunked laws, in that they deal with the as as a whole, and i nore its constituents& 'urthermore, the microsco!ic and macrosco!ic descri!tions of a as use entirely different terms& The former requires the s!ecification of the !osition and velocity of every sin le com!onent molecule= the latter requires only the s!ecification of three new quantities, tem!erature, !ressure, and volume, the first two of which do not even have microsco!ic counter!arts& The sim!le mathematical relationshi! which elates these three !arameters p< X cT, where c is a constantis a law which de!ends on, yet is inde!endent of, the lower-level !henomena& Aess !arado(ically, this law can be derived from the laws overnin the molecular level= in that sense it de!ends on the lower level& En the other hand, it is law which allows you to i nore the lower level com!letely, if you wish= in that sense it is inde!endent of the lower level& It is im!ortant to reali<e that the hi h-level law cannot be stated in the vocabulary of the low-level descri!tion& -Pressure- and -tem!erature- are new terms which e(!erience with the low level alone cannot convey& 3e humans !erceive tem!erature and !ressure directly= that is how we are built, so that it is not ama<in that we should have found this law& 0ut creatures which knew ases only as theoretical mathematical constructs would have to have an ability to synthesi<e new conce!ts, if they were to discover this law&

E!i!henomena
In drawin this 7ha!ter to a close, I would like to relate a story about a com!le( system& I was talkin one day with two systems !ro rammers for the com!uter I was usin & They mentioned that the o!eratin system seemed to be able to handle u! to about thirty-five users with reat comfort, but at about thirty-five users or so, the res!onse time all of a sudden shot u!, ettin so slow that you mi ht as well lo off and o home and wait until later& Jokin ly I said, -3ell, that1s sim!le to fi(.ust find the !lace in the o!eratin system where the number 1%L1 is stored, and chan e it to 1D+12- Everyone lau hed& The !oint is, of course, that there is no such !lace& where, then, does the critical number%L userscome from? The answer is, 3t is a #isi(le conse"uence of the o#erall s)stem organi@ationOan -epiphenomenon"&

Similarly, you mi ht ask about a s!rinter, -3here is the 1#&%1 stored, that makes him be able to run "++ yards in #&% seconds?- Ebviously, it is not stored anywhere& ?is time is a result of how he is built, what his reaction time is, a million factors all interactin when he runs& The time is quite re!roducible, but it is not stored in his body anywhere& It is s!read around amon all the cells of his body and only manifests itself in the act of the s!rint itself& E!i!henomena abound& In the ame of -6o-, there is the feature that -two eyes live-& It is not built into the rules, but it is a consequence of the rules& In the human brain, there is ullibility& ?ow ullible are you? Is your ullibility located in some - ullibility center- in your brain? 7ould a neurosur eon reach in and !erform some delicate o!eration to lower your ullibility, otherwise leavin you alone? If you believe this, you are !retty ullible, and should !erha!s consider such an o!eration&

Mind #s& Brain

In comin 7ha!ters, where we discuss the brain, we shall e(amine whether the brain1s to! levelthe mindcan be understood without understandin the lower levels on which it both de!ends and does not de!end& )re there laws of thinkin which are -sealed offfrom the lower laws that overn the microsco!ic activity in the cells of the brain? 7an mind be -skimmed- off of brain and trans!lanted into other systems? Er is it im!ossible to unravel thinkin !rocesses into neat and modular subsystems? Is the brain more like an atom, a renormali<ed electron, a nucleus, a neutron, or a quark? Is consciousness an e!i!henomenon? To understand the mind, must one o all the way down to the level of nerve cells?

& Ant "u!ue

& then, one () one, the four #oices of the fugue chime in!/
*chilles, I know the rest of you won1t believe this, but the answer to the question is starin us all in the face, hidden in the !icture& It is sim!ly one wordbut what an im!ortant one, -/F-2 8ra(, I know the rest of you won1t believe this, but the answer to the question is starin us all in the face, hidden in the !icture& It is sim!ly one wordbut what an im!ortant one, -?EAIS/-2 *chilles, 8ow hold on a minute& Hou must be seein thin s& It1s !lain as day that the messa e of this !icture is -/F-, not -?EAIS/-2 8ra(, I be your !ardon, but my eyesi ht is e(tremely ood& Please look a ain, and then tell me if the the !icture doesn1t say what I said it says2 *nteater, I know the rest of you won1t believe this, but the answer to the question is starin us all in the face, hidden in the !icture& It is sim!ly one wordbut what an im!ortant one, -;E:F7TIE8IS/-2 8ra(, 8ow hold on a minute& Hou must be seein thin s& It1s !lain as day that the messa e of this !icture is -?EAIS/-, not -;E:F7TIE8IS/-2 *chilles, )nother deluded one2 8ot -?EAIS/-, not -;E:F7TIE8IS/-, but -/F- is the messa e of this !icture, and that much is certain& *nteater, I be your !ardon, but my eyesi ht is e(tremely clear& Please look a ain, and then see if the !icture doesn1t say what I said it says& *chilles, :on1t you see that the !icture is com!osed of two !ieces, and that each of them is a sin le letter? 8ra(, Hou are ri ht about the two !ieces, but you are wron in your identification of what they are& The !iece on the left is entirely com!osed of three co!ies of one word, -?EAIS/-= and the !iece on the ri ht is com!osed of many co!ies, in smaller letters, of the same word& 3hy the letters are of different si<es in the two !arts, I don1t know, but I know what I see, and what I see is - ?EAIS/-, !lain as day& ?ow you see anythin else is beyond me& *nteater, Hou are ri ht about the two !ieces, but you are wron in your identification of what they are& The !iece on the left is entirely com!osed of many co!ies of one word, -;E:F7TIE8IS/-= and the !iece on the ri ht is com!osed of one sin le co!y, in lar er letters, of the same word& 3hy the letters are of different si<es in the two !arts, I don1t know, but I know what I see, and what I see is - ;E:F7TIE8IS/-, !lain as day& ?ow you see anythin else is beyond me& *chilles, I know what is oin on here& Each of you has seen letters which com!ose, or are com!osed of, other letters& In the left-hand !iece, there are indeed three -?EAIS/-1s, but each one of them is com!osed out of smaller co!ies of the word -;E:F7TIE8IS/-& )nd in com!lementary fashion, in the ri ht-hand !iece, there is indeed one -;E:F7TIE8IS/-, but it is com!osed out of smaller co!ies of the word -?EAIS/-& 8ow this is all fine and ood, but in your silly squabble, the two of you have actually missed the forest for the trees& Hou see, what ood is it to ar ue about

whether -?EAIS/- or -;E:F7TIE8IS/- is ri ht, when the !ro!er way to understand the matter is to transcend the question, by answerin -/F-, 8ra(, I now see the !icture as you have described it, )chilles, but I have no idea of what you mean by the stran e e(!ression -transcendin the question-& *nteater, I now see the !icture as you have described it, )chilles, but I have no idea of what you mean by the stran e e(!ression -/F-& *chilles, I will be lad to indul e both of you, if you will first obli e me, by tellin me the meanin of these stran e e(!ressions, -?EAIS/- and -;E:F7TIE8IS/-& 8ra(, ?EAIS/ is the most natural thin in the world to ras!& It1s sim!ly the belief that -the whole is reater than the sum of its !arts-& 8o one in his ri ht mind could re.ect holism& *nteater, ;E:F7TIE8IS/ is the most natural thin in the world to ras!& It1s sim!ly the belief that -a whole can be understood com!letely if you understand its !arts, and the nature of their 1sum1-& 8o one in her left brain could re.ect reductionism& 8ra(, I re.ect reductionism& I challen e you to tell me, for instance, how to understand a brain reductionistically& )ny reductionistic e(!lanation of a brain will inevitably fall far short of e(!lainin where the consciousness e(!erienced by a brain arises from& *nteater, I re.ect holism& I challen e you to tell me, for instance, how a holistic descri!tion of an ant colony sheds any more li ht on it than is shed by a descri!tion of the ants inside it, and their roles, and their interrelationshi!s& )ny holistic e(!lanation of an ant colony will inevitably fall far short of e(!lainin where the consciousness e(!erienced by an ant colony arises from& *chilles, Eh, no2 The last thin which I wanted to do was to !rovoke another ar ument& )nyway, now that I understand the controversy, I believe that my e(!lanation of -/F- will hel! reatly& Hou see, -/F- is an ancient 9en answer which, when iven to a question, F8)S@S the question& ?ere, the question seems to be, -Should the world be understood via holism, or via reductionism?- )nd the answer of -/F- here re.ects the !remises of the question, which are that one or the other must be chosen& 0y unaskin the question, it reveals a wider truth, that there is a lar er conte(t into which both holistic and reductionistic e(!lanations fit& *nteater, )bsurd2 Hour -/F- is as silly as a cow1s moo& I1ll have none of this 9en wishywashiness& 8ra(, ;idiculous2 Hour -/F- is as silly as a kitten1s mew& I1ll have none of this 9en washy-wishiness& *chilles, Eh, dear2 3e1re ettin nowhere fast& 3hy have you stayed so stran ely silent, /r& Tortoise? It makes me very uneasy& Surely you must somehow be ca!able of hel!in strai hten out this mess? Tortoise, I know the rest of you won1t believe this, but the answer to the question is starin us all in the face, hidden in the !icture& It is sim!ly one wordbut what an im!ortant one, -/F-2 ?ust as he sa)s this, the fourth #oice in the fugue (eing pla)ed enters, e9actl) one octa#e (elow the first entr)!/

*chilles, Eh, /r& T, for once you have let me down& I was sure that you, who always see the most dee!ly into thin s, would be able to resolve this dilemmabut a!!arently, you have seen no further than I myself saw& Eh, well, I uess I should feel !leased to have seen as far as /r& Tortoise, for once& Tortoise, I be your !ardon, but my eyesi ht is e(tremely fine& Please look a ain, and then tell me if the !icture doesn1t say what I said it says& *chilles, 0ut of course it does2 Hou have merely re!eated my own ori inal observation& Tortoise, Perha!s -/F- e(ists in this !icture on a dee!er level than you ima ine, )chilles an octave lower 4fi uratively s!eakin 5& 0ut for now I doubt that we can settle the dis!ute in the abstract& I would like to see both the holistic and reductionistic !oints of view laid out more e(!licitly= then there may be more of a basis for a decision& I would very much like to hear a reductionistic descri!tion of an ant colony, for instance& 8ra(, Perha!s :r& )nteater will tell you somethin of his e(!eriences in that re ard& )fter all, he is by !rofession somethin of an e(!ert on that sub.ect& Tortoise, I am sure that we have much to learn from you, :r& )nteater& 7ould you tell us more about ant colonies, from a reductionistic !oint of view? *nteater, 6ladly& )s /r& 7rab mentioned to you, my !rofession has led me quite a lon way into the understandin of ant colonies& *chilles, I can ima ine2 The !rofession of anteater would seem to be synonymous with bein an e(!ert on ant colonies2 *nteater, I be your !ardon& -)nteater- is not my !rofession= it is my s!ecies& 0y !rofession, I am a colony sur eon& I s!eciali<e in correctin nervous disorders of the colony by the technique of sur ical removal& *chilles, Eh, I see& 0ut what do you mean by -nervous disorders- of an ant colony? *nteater, /ost of my clients suffer from some sort of s!eech im!airment& Hou know, colonies which have to ro!e for words in everyday situations& It can be quite tra ic& I attem!t to remedy the situation by, uhhremovin the defective !art of the colony& These o!erations are sometimes quite involved, and of course years of study are required before one can !erform them& *chilles, 0utisn1t it true that, before one can suffer from s!eech im!airment, one must have the faculty of s!eech? *nteater, ;i ht& *chilles, Since ant colonies don1t have that faculty, I am a little confused& 8ra(, It1s too bad, )chilles, that you weren1t here last week, when :r& )nteater and )unt ?illary were my house uests& I should have thou ht of havin you over then& *chilles, Is )unt ?illary your aunt, /r& 7rab? 8ra(, Eh, no, she1s not really anybody1s aunt& *nteater, 0ut the !oor dear insists that everybody should call her that, even stran ers& It1s .ust one of her many endearin quirks& 8ra(, Hes, )unt ?illary is quite eccentric, but such a merry old soul& It1s a shame I didn1t have you over to meet her last week&

*nteater, She1s certainly one of the best-educated ant colonies I have ever had the ood fortune to know& The two of us have s!ent many a lon evenin in conversation on the widest ran e of to!ics& *chilles, I thou ht anteaters were devourers of ants, not !atrons of antintellectualism2 *nteater, 3ell, of course the two are not mutually inconsistent& I am on the best of terms with ant colonies& It1s .ust )8TS that I eat, not coloniesand that is ood for both !arties, me, and the colony& *chilles, ?ow is it !ossible that Tortoise, ?ow is it !ossible that *chilles, havin its ants eaten can do an ant colony any ood? 8ra(, ?ow is it !ossible that Tortoise, havin a forest fire can do a forest any ood? *nteater, ?ow is it !ossible that 8ra(, havin its branches !runed can do a tree any ood? *nteater, havin a haircut can do )chilles any ood? Tortoise, Probably the rest of you were too en rossed in the discussion to notice the lovely stretto which .ust occurred in this 0ach fu ue& *chilles, 3hat is a stretto? Tortoise, Eh, I1m sorry= I thou ht you knew the term& It is where one theme re!eatedly enters in one voice after another, with very little delay between entries& *chilles, If I listen to enou h fu ues, soon I1ll know all of these thin s and will be able to !ick them out myself, without their havin to be !ointed out& Tortoise, Pardon me, my friends& I am sorry to have interru!ted& :r& )nteater was tryin to e(!lain how eatin ants is !erfectly consistent with bein a friend of an ant colony& *chilles, 3ell, I can va uely see how it mi ht be !ossible for a limited and re ulated amount of ant consum!tion to im!rove the overall health of a colonybut what is far more !er!le(in is all this talk about havin conversations with ant colonies& That1s im!ossible& )n ant colony is sim!ly a bunch of individual ants runnin around at random lookin for food and makin a nest& *nteater, Hou could !ut it that way if you want to insist on seein the trees but missin the forest, )chilles& In fact, ant colonies, seen as wholes, are quite well-defined units, with their own qualities, at times includin the mastery of lan ua e& *chilles, I find it hard to ima ine myself shoutin somethin out loud in the middle of the forest, and hearin an ant colony answer back& *nteater, Silly fellow2 That1s not the way it ha!!ens& )nt colonies don1t converse out loud, but in writin & Hou know how ants form trails leadin them hither and thither? *chilles, Eh, yesusually strai ht throu h the kitchen sink and into my !each .am& *nteater, )ctually, some trails contain information in coded form& If you know the system, you can read what they1re sayin .ust like a book& *chilles, ;emarkable& )nd can you communicate back to them? *nteater, 3ithout any trouble at all& That1s how )unt ?illary and I have conversations for hours& I take a stick and draw trails in the moist round, and watch the ants follow my trails& Presently, a new trail starts ettin formed somewhere& I reatly en.oy watchin trails develo!& )s they are formin , I antici!ate how they will continue 4and

more often I am wron than ri ht5& 3hen the trail is com!leted, I know what )unt ?illary is thinkin , and I in turn make my re!ly& *chilles, There must be some ama<in ly smart ants in that colony, I1ll say that& *nteater, I think you are still havin some difficulty reali<in the difference in levels here& Just as you would never confuse an individual tree with a forest, so here you must not take an ant for the colony& Hou see, all the ants in )unt ?illary are as dumb as can be& They couldn1t converse to save their little thora(es2 *chilles, 3ell then, where does the ability to converse come from? It must reside somewhere inside the colony2 I don1t understand how the ants can all be unintelli ent, if )unt ?illary can entertain you for hours with witty banter& Tortoise, It seems to me that the situation is not unlike the com!osition of a human brain out of neurons& 7ertainly no one would insist that individual brain cells have to be intelli ent bein s on their own, in order to e(!lain the fact that a !erson can have an intelli ent conversation& *chilles, Eh, no, clearly not& 3ith brain cells, I see your !oint com!letely& Enly&&& ants are a horse of another color& I mean, ants .ust roam about at will, com!letely randomly, chancin now and then u!on a morsel of food&&& They are free to do what they want to do, and with that freedom, I don1t see at all how their behavior, seen as a whole, can amount to anythin coherentes!ecially somethin so coherent as the brain behavior necessary for conversin & 8ra(, It seems to me that the ants are free only within certain constraints& 'or e(am!le, they are free to wander, to brush a ainst each other, to !ick u! small items, to work on trails, and so on& 0ut they never ste! out of that small world, that ant-system, which they are in& It would never occur to them, for they don1t have the mentality to ima ine anythin of the kind& Thus the ants are very reliable com!onents, in the sense that you can de!end on them to !erform certain kinds of tasks in certain ways& *chilles, 0ut even so, within those limits they are still free, and they .ust act at random, runnin about incoherently without any re ard for the thou ht mechanisms of a hi her-level bein which :r& )nteater asserts they are merely com!onents of& *nteater, )h, but you fail to reco ni<e one thin & )chillesthe re ularity of statistics& *chilles, ?ow is that? *nteater, 'or e(am!le, even thou h ants as individuals wander about in what seems a random way, there are nevertheless overall trends, involvin lar e numbers of ants, which can emer e from that chaos& *chilles, Eh, I know what you mean& In fact, ant trails are a !erfect e(am!le of such a !henomenon& There, you have really quite un!redictable motion on the !art of any sin le antand yet, the trail itself seems to remain well-defined and stable& 7ertainly that must mean that the individual ants are not .ust runnin about totally at random& *nteater, E(actly, )chilles& There is some de ree of communication amon the ants, .ust enou h to kee! them from wanderin off com!letely at random& 0y this minimal communication they can remind each other that they are not alone but are coo!eratin with teammates& It takes a lar e number of ants, all reinforcin each other this way, to sustain any activitysuch as trail-buildin for any len th of time& 8ow my very ha<y understandin of the o!eration of brains leads me to believe that

somethin similar !ertains to the firin of neurons& Isn1t it true, /r& 7rab, that it takes a rou! of neurons firin in order to make another neuron fire? 8ra(, :efinitely& Take the neurons in )chilles1 brain, for e(am!le& Each neuron receives si nals from neurons attached to its in!ut lines, and if the sum total of in!uts at any moment e(ceeds a critical threshold, then that neuron will fire and send its own out!ut !ulse rushin off to other neurons, which may in turn fireand on down the line it oes& The neural flash swoo!s relentlessly in its )chillean !ath, in sha!es stran er then the dash of a nat-hun ry swallow= every twist, every turn foreordained by the neural structure in )chilles1 brain, until sensory in!ut messa es interfere& *chilles, 8ormally, I think that I1/ in control of what I thinkbut the way you !ut it turns it all inside out, so that it sounds as thou h -I- am .ust what comes out of all this neural structure, and natural law& It makes what I consider my SEA' sound at best like a by-!roduct of an or anism overned by natural law, and at worst, an artificial notion !roduced by my distorted !ers!ective& In other words, you make me feel like I don1t know who or whatI am, if anythin & Tortoise, Hou1ll come to understand much better as we o alon & 0ut :r& )nteaterwhat do you make of this similarity? *nteater, I knew there was somethin !arallel oin on in the two very different systems& 8ow I understand it much better& It seems that rou! !henomena which have coherencetrail-buildin , for e(am!lewill take !lace only when a certain threshold number of ants et involved& If an effort is initiated, !erha!s at random, by a few ants in some locale, one of two thin s can ha!!en, either it will fi<<le out after a brief s!utterin start *chilles, 3hen there aren1t enou h ants to kee! the thin rollin ? *nteater, E(actly& The other thin that can ha!!en is that a critical mass of ants is !resent, and the thin will snowball, brin in more and more ants into the !icture& In the latter case, a whole -team- is brou ht into bein which works on a sin le !ro.ect& That !ro.ect mi ht be trailmakin , or food- atherin , or it mi ht involve nestkee!in & :es!ite the e(treme sim!licity of this scheme on a small scale, it can ive rise to very com!le( consequences on a lar er scale& *chilles, I can ras! the eneral idea of order emer in from chaos, as you sketch it, but that still is a lon way from the ability to converse& )fter all, order also emer es from chaos when molecules of a as bounce a ainst each other randomlyyet all that results there is an amor!hous mass with but three !arameters to characteri<e it, volume, !ressure, and tem!erature& 8ow that1s a far cry from the ability to understand the world, or to talk about it2 *nteater, That hi hli hts a very interestin difference between the e(!lanation of the behavior of an ant colony and the e(!lanation of the behavior of as inside a container& Ene can e(!lain the behavior of the as sim!ly by calculatin the statistical !ro!erties of the motions of its molecules& There is no need to discuss any hi her elements of structure than molecules, e(ce!t the full as itself& En the other hand, in an ant colony, you can1t even be in to understand the activities of the colony unless you o throu h several layers of structure&

*chilles, I see what you mean& In a as, one .um! takes you from the lowest level moleculesto the hi hest levelthe full as& There are no intermediate levels of or ani<ation& 8ow how do intermediate levels of or ani<ed activity arise in an ant colony? *nteater, It has to do with the e(istence of several different varieties of ants inside any colony& *chilles, Eh, yes& I think I have heard about that& They are called -castes-, aren1t they? *nteater, That1s correct& )side from the queen, there are males, who do !ractically nothin towards, the u!kee! of the nest, and then *chilles, )nd of course there are soldiers6lorious 'i hters ) ainst 7ommunism2 8ra(, ?mm&&& I hardly think that could be ri ht, )chilles& )n ant colony is quite communistic internally, so why would its soldiers fi ht a ainst communism? Er am I ri ht, :r& )nteater? *nteater, Hes, about colonies you are ri ht, /r& 7rab= they are indeed based on somewhat communistic !rinci!les& 0ut about soldiers )chilles is somewhat naPve& In fact, the so-called -soldiers- are hardly ade!t at fi htin at all& They are slow, un ainly ants with iant heads, who can sna! with their stron .aws, but are hardly to be lorified& )s in a true communistic state, it is rather the workers who are to be lorified& It is they who do most of the chores, such as food- atherin , huntin , and nursin of the youn & It is even they who do most of the fi htin & *chilles, 0ah& That is an absurd state of affairs& Soldiers who won1t fi ht2 *nteater, 3ell, as I .ust said, they really aren1t soldiers at all& It1s the workers who are soldiers= the soldiers are .ust la<y fatheads& *chilles, Eh, how dis raceful2 3hy, if I were an ant, I1d !ut some disci!line in their ranks2 I1d knock some sense into those fatheads2 Tortoise, If you were an ant? ?ow could you be an ant? There is no way to ma! your brain onto an ant brain, so it seems to me to be a !retty fruitless question to worry over& /ore reasonable would be the !ro!osition of ma!!in your brain onto an ant colony&&& 0ut let us not et sidetracked& Aet :r& )nteater continue with his most illuminatin descri!tion of castes and their role in the hi her levels of or ani<ation& *nteater, Gery well& There are all sorts of tasks which must be accom!lished in a colony, and individual ants develo! s!eciali<ations& Fsually an ant1s s!eciali<ation chan es as the ant a es& )nd of course it is also de!endent on the ant1s caste& )t any one moment, in any small area of a colony, there are ants of all ty!es !resent& Ef course, one caste may be be very s!arse in some !laces and very dense in others& 8ra(, Is the density of a iven caste, or s!eciali<ation, .ust a random thin ? Er is there a reason why ants of one ty!e mi ht be more heavily concentrated in certain areas, and less heavily in others? *nteater, I1m lad you brou ht that u!, since it is of crucial im!ortance in understandin how a colony thinks& In fact, there evolves, over a lon !eriod of time, a very delicate distribution of castes inside a colony& )nd it is this distribution which allows the colony to have the com!le(ity which underlies the ability to converse with me& *chilles, It would seem to me that the constant motion of ants to and fro would com!letely !revent the !ossibility of a very delicate distribution& )ny delicate

distribution would be quickly destroyed by all the random motions of ants, .ust as any delicate !attern amon molecules in a as would not survive for an instant, due to the random bombardment from all sides& *nteater, In an ant colony& the situation is quite the contrary& In fact, it is .ust e(actly the to-in and fro-in of ants inside the colony which ada!ts the caste distribution to varyin situations, and thereby !reserves the delicate caste distribution& Hou see, the caste distribution cannot remain as one sin le ri id !attern= rather, it must constantly be chan in so as to reflect, in some manner, the real-world situation with which the colony is dealin , and it is !recisely the motion inside the colony which u!dates the caste distribution, so as to kee! it in line with the !resent circumstances facin the colony& Tortoise, 7ould you ive an e(am!le? *nteater, 6ladly& 3hen I, an anteater, arrive to !ay a visit to )unt ?illary, all the foolish ants, u!on sniffin my odor, o into a !anicwhich means, of course, that they be in runnin around com!letely differently from the way they were before I arrived& *chilles, 0ut that1s understandable, since you1re a dreaded enemy of the colony& *nteater, Eh, no& I must reiterate that, far from bein an enemy of the colony, I am )unt ?illary1s favorite com!anion& )nd )unt ?illary is my favorite aunt& I rant you, I1m quite feared by all the individual ants in the colonybut that1s another matter entirely& In any case, you see that the ants1 action in res!onse to my arrival com!letely chan es the internal distribution of ants& *chilles, That1s clear& *nteater, )nd that sort of thin is the u!datin which I s!oke of& The new distribution reflects my !resence& Ene can describe the chan e from old state to new as havin added a -!iece of knowled e- to the colony& *chilles, ?ow can you refer to the distribution of different ty!es of ants inside a colony as a -!iece of knowled e-? *nteater, 8ow there1s a vital !oint& It requires some elaboration& Hou see, what it comes down to is how you choose to describe the caste distribution& If you continue to think in terms of the lower levelsindividual antsthen you miss the forest for the trees& That1s .ust too microsco!ic a level, and when you think microsco!ically, you1re bound to miss some lar e-scale features& Hou1ve ot to find the !ro!er hi h-level framework in which to describe the caste distributiononly then will it make sense how the caste distribution can encode many !ieces of knowled e& *chilles, 3ell, how :E you find the !ro!er-si<ed units in which to describe the !resent state of the colony, then? *nteater, )ll ri ht& Aet1s be in at the bottom& 3hen ants need to et somethin done, they form little -teams-, which stick to ether to !erform a chore& )s I mentioned earlier, small rou!s of ants are constantly formin and unformin & Those which actually e(ist for a while are the teams, and the reason they don1t fall a!art is that there really is somethin for them to do& *chilles, Earlier you said that a rou! will stick to ether if its si<e e(ceeds a certain threshold& 8ow you1re sayin that a rou! will stick to ether if there is somethin for it to do&

*nteater, They are equivalent statements& 'or instance, in food- atherin , if there is an inconsequential amount of food somewhere which ets discovered by some wanderin ant who then attem!ts to communicate its enthusiasm to other ants, the number of ants who res!ond will be !ro!ortional to the si<e of the food sam!leand an inconsequential amount will not attract enou h ants to sur!ass the threshold& 3hich is e(actly what I meant by sayin there is nothin to dotoo little food ou ht to be i nored& *chilles, I see& I assume that these -teams- are one of the levels of structure fallin somewhere in between the sin le-ant level and the colony level& *nteater, Precisely& There e(ists a s!ecial kind of team, which I call a -si nal-and all the hi her levels of structure are based on si nals& In fact, all the hi her entities are collections of si nals actin in concert& There are teams on hi her levels whose members are not ants, but teams on lower levels& Eventually you reach the lowestlevel teamswhich is to say, si nalsand below them, ants& *chilles, 3hy do si nals deserve their su estive name? *nteater, It comes from their function& The effect of si nals is to trans!ort ants of various s!eciali<ations to a!!ro!riate !arts of the colony& So the ty!ical story of a si nal is thus, it comes into e(istence by e(ceedin the threshold needed for survival, then it mi rates for some distance throu h the colony, and at some !oint it more or less disinte rates into its individual members, leavin them on their own& *chilles, It sounds like a wave, carryin sand dollars and seaweed from afar, and leavin them strewn, hi h and dry, on the shore& *nteater, In a way that1s analo ous, since the team does indeed de!osit somethin which it has carried from a distance, but whereas the water in the wave rolls back to the sea, there is no analo ous carrier substance in the case of a si nal, since the ants themselves com!ose it& Tortoise, )nd I su!!ose that a si nal loses its coherency .ust at some s!ot in the colony where ants of that ty!e were needed in the first !lace& *nteater, 8aturally& *chilles, 8aturally? It1s not so obvious to /E that a si nal should always o .ust where it is needed& )nd even if it oes in the ri ht direction, how does it fi ure out where to decom!ose? ?ow does it know it has arrived? *nteater, Those are e(tremely im!ortant matters, since they involve e(!lainin the e(istence of !ur!oseful behavioror what seems to be !ur!oseful behavioron the !art of si nals& 'rom the descri!tion, one would be inclined to characteri<e the si nals1 behavior as bein oriented towards fillin a need, and to call it -!ur!oseful-& 0ut you can look at it otherwise& *chilles, Eh, wait& Either the behavior IS !ur!oseful, or it is 8ET& I don1t see how you can have it both ways& *nteater, Aet me e(!lain my way of seein thin s, and then see if you a ree& Ence a si nal is formed, there is no awareness on its !art that it should head off in any !articular direction& 0ut here, the delicate caste distribution !lays a crucial role& It is what determines the motion of si nals throu h the colony, and also how lon a si nal will remain stable, and where it will -dissolve-&

*chilles, So everythin de!ends on the caste distribution, eh? *nteater, ;i ht& Aet1s say a si nal is movin alon & )s it oes, the ants which com!ose it interact, either by direct contact or by e(chan e of scents, with ants of the local nei hborhoods which it !asses throu h& The contacts and scents !rovide information about local matters of ur ency, such as nest-buildin , or nursin , or whatever& The si nal will remain lued to ether as lon as the local needs are different from what it can su!!ly= but if it 7)8 contribute, it disinte rates, s!illin a fresh team of usable ants onto the scene& :o you see now how the caste distribution acts as an overall uide of the teams inside the colony? *chilles, I do see that& *nteater, )nd do you see how this way of lookin at thin s requires attributin no sense of !ur!ose to the si nal? *chilles, I think so& )ctually, I1m be innin to see thin s from two different vanta e !oints& 'rom an ant1s-eye !oint of view, a si nal has 8E !ur!ose& The ty!ical ant in a si nal is .ust meanderin around the colony, in search of nothin in !articular, until it finds that it feels like sto!!in & Its teammates usually a ree, and at that moment the team unloads itself by crumblin a!art, leavin .ust its members but none of its coherency& 8o !lannin is required, no lookin ahead= nor is any search required, to determine the !ro!er direction& 0ut from the 7EAE8H1S !oint of view, the team has .ust res!onded to a messa e which was written in the lan ua e of the caste distribution& 8ow from this !ers!ective, it looks very much like !ur!oseful activity& 8ra(, 3hat would ha!!en if the caste distribution were entirely random? 3ould si nals still band and disband? *nteater, 7ertainly& 0ut the colony would not last lon , due to the meanin lessness of the caste distribution& 8ra(, Precisely the !oint I wanted to make& 7olonies survive because their caste distribution has meanin , and that meanin is a holistic as!ect, invisible on lower levels& Hou lose e(!lanatory !ower unless you take that hi her level into account& *nteater, I see your side= but I believe you see thin s too narrowly& 8ra(, ?ow so? *nteater, )nt colonies have been sub.ected to the ri ors of evolution for billions of years& ) few mechanisms were selected for, and most were selected a ainst& The end result was a set of mechanisms which make ant colonies work as we have been describin & If you could watch the whole !rocess in a movierunnin a billion or so times faster than life, of coursethe emer ence of various mechanisms would be seen as natural res!onses to e(ternal !ressures, .ust as bubbles in boilin water are natural res!onses to an e(ternal heat source& I don1t su!!ose you see -meanin - and -!ur!ose-, in the bubbles in boilin wateror do you? 8ra(, 8o, but *nteater, 8ow that1s /H !oint& 8o matter how bi a bubble is, it owes its e(istence to !rocesses on the molecular level, and you can for et about any -hi her-level laws-& The same oes for ant colonies and their teams& 0y lookin at thin s from the vast !ers!ective of evolution, you can drain the whole colony of meanin and !ur!ose& They become su!erfluous notions&

*chilles, 3hy, then, :r& )nteater, did you tell me that you talked with )unt ?illary? It now seems that you would deny that she can talk or think at all& *nteater, I am not bein inconsistent, )chilles& Hou see, I have as much difficulty as anyone else in seein thin s on such a randiose time scale, so I find it much easier to chan e !oints of view& 3hen I do so, for ettin about evolution and seein thin s in the here and now, the vocabulary of teleolo y comes back, the /E)8I86 of the caste distribution and the PF;PESE'FA8ESS of si nals& This not only ha!!ens when I think of ant colonies, but also when I think about my own brain and other brains& ?owever, with some effort I can always remember the other !oint of view if necessary, and drain all these systems of meanin , too& 8ra(, Evolution certainly works some miracles& Hou never know the ne(t trick it will !ull out of its sleeve& 'or instance, it wouldn1t sur!rise me one bit if it were theoretically !ossible for two or more -si nals- to !ass throu h each other, each one unaware that the other one is also a si nal= each one treatin the other as if it were .ust !art of the back round !o!ulation& *nteater, It is better than theoretically !ossible= in fact it ha!!ens routinely2 *chilles, ?mm&&& 3hat a stran e ima e that con.ures u! in my mind& I can .ust ima ine ants movin in four different directions, some black, some white, criss-crossin , to ether formin an orderly !attern, almost likelike Tortoise, ) fu ue, !erha!s? *chilles, Hesthat1s it2 )n ant fu ue2 8ra(, )n interestin ima e, )chilles& 0y the way, all that talk of boilin water made me think of tea& 3ho would like some more? *chilles, I could do with another cu!, /r& 7& 8ra(, Gery ood& *chilles, :o you su!!ose one could se!arate out the different visual -voices- of such an -ant fu ue-? I know how hard it is for me Tortoise, 8ot for me, thank you& *chilles, to track a sin le voice *nteater, I1d like some, too, /r& 7rab *chilles, in a musical fu ue *nteater, if it isn1t too much trouble& *chilles, when all of them 8ra(, 8ot at all& 'our cu!s of tea Tortoise, Three? *chilles, are oin at once& 8ra(, comin ri ht u!2 *nteater, That1s an interestin thou ht, )chilles& 0ut its unlikely that anyone could draw such a !icture in a convincin way *chilles, That1s too bad& Tortoise, Perha!s you could answer this, :r& )nteater& :oes a si nal, from its creation until its dissolution, always consist of the same set of ants? *nteater, )s a matter of fact, the individuals in a si nal sometimes break off and et re!laced by others of the same caste, if there are a few in the area& /ost often, si nals

73G5=E JB! ")nt 'u ue-, () '! 8! Escher woodcut, BIGH/!

arrive at their disinte ration !oints with nary an ant in common with their startin lineu!& 8ra(, I can see that the si nals are constantly affectin the caste distribution throu hout the colony, and are doin so in res!onse to the internal needs of the colonywhich in turn reflect the e(ternal situation which the colony is faced with& Therefore the caste distribution, as you said, :r& )nteater, ets continually u!dated in a way which ultimately reflects the outer world& *chilles, 0ut what about those intermediate levels of structure? Hou were sayin that the caste distribution should best be !ictured not in terms of ants or si nals, but in terms of teams whose members were other teams, whose members were other teams, and so on until you come down to the ant level& )nd you said that that was the key to understandin how it was !ossible to describe the caste distribution as encodin !ieces of information about the world& *nteater, Hes, we are comin to all that& I !refer to ive teams of a sufficiently hi h level the name of -symbols-& /ind you, this sense of the word has some si nificant differences from the usual sense& /y -symbols- are )7TIGE SF0SHSTE/S of a com!le( system, and they are com!osed of lower-level active subsystems&&& They are therefore

quite different from P)SSIGE symbols, e(ternal to the system, such as letters of the al!habet or musical notes, which sit there immobile, waitin for an active system to !rocess them& *chilles, Eh, this is rather com!licated, isn1t it? I .ust had no idea that ant colonies had such an abstract structure& *nteater, Hes, it1s quite remarkable& 0ut all these layers of structure are necessary for the stora e of the kinds of knowled e which enable an or anism to be -intelli ent- in any reasonable sense of the word& )ny system which has a mastery of lan ua e has essentially the same underlyin sets of levels& *chilles, 8ow .ust a cotton-!ickin minute& )re you insinuatin that my brain consists of, at bottom, .ust a bunch of ants runnin around? *nteater, Eh, hardly& Hou took me a little too literally& The lowest level may be utterly different& Indeed, the brains of anteaters, for instance, are not com!osed of ants& 0ut when you o u! a level or two in a brain, you reach a level whose elements have e(act counter!arts in other systems of equal intellectual stren thsuch as ant colonies& Tortoise, That is why it would be reasonable to think of ma!!in your brain, )chilles, onto an ant colony, but not onto the brain of a mere ant& *chilles, I a!!reciate the com!liment& 0ut how would such a ma!!in be carried out? 'or instance, what in my brain corres!onds to the low level teams which you call si nals? *nteater, Eh, I but dabble in brains, and therefore couldn1t set u! the ma! in its lorious detail& 0utand correct me if I1m wron , /r& 7rabI would surmise that the brain counter!art to an ant colony1s si nal is the firin of a neuron= or !erha!s it is a lar erscale event, such as a !attern of neural firin s& 8ra(, I would tend to a ree& 0ut don1t you think that, for the !ur!oses of our discussion, delineatin the e(act counter!art is not in itself crucial, desirable thou h it mi ht be? It seems to me that the main idea is that such a corres!ondence does e(ist, even if we don1t know e(actly how to define it ri ht now& I would only question one !oint, :r& )nteater, which you raised, and that concerns the level at which one can have faith that the corres!ondence be ins& Hou seemed to think that a SI68)A mi ht have a direct counter!art in a brain= whereas I feel that it is only at the level of your )7TIGE SH/0EAS and above that it is likely that a corres!ondence must e(ist& *nteater, Hour inter!retation may very well be more accurate than mine, /r& 7rab& Thank you for brin in out that subtle !oint& *chilles, 3hat does a symbol do that a si nal couldn1t do? *nteater, It is somethin like the difference between words and letters& 3ords, which are meanin -carryin entities, are com!osed of letters, which in themselves carry no meanin & This ives a ood idea of the difference between symbols and si nals& In fact it is a useful analo y, as lon as you kee! in mind the fact that words and letters are P)SSIGE, symbols and si nals are )7TIGE& *chilles, I1ll do so, but I1m not sure I understand why it is so vital to stress the difference between active and !assive entities& *nteater, The reason is that the meanin which you attribute to any !assive symbol, such as a word on a !a e, actually derives from the meanin which is carried by

corres!ondin active symbols in your brain& So that the meanin of !assive symbols can only be !ro!erly understood when it is related to the meanin of active symbols& *chilles, )ll ri ht& 0ut what is it that endows a SH/0EAan active one, to be surewith meanin , when you say that a SI68)A, which is a !erfectly ood entity in its own ri ht, has none? *nteater, It all has to do with the way that symbols can cause other symbols to be tri ered& 3hen one symbol becomes active, it does not do so in isolation& It is floatin about, indeed, in a medium, which is characteri<ed by its caste distribution& 8ra(, Ef course, in a brain there is no such thin as a caste distribution, but the counter!art is the -brain state-& There, you describe the states of all the neurons, and all the interconnections, and the threshold for firin of each neuron& *nteater, Gery well= let1s lum! -caste distribution- and -brain state- under a common headin , and call them .ust the -state-& 8ow the state can be described on a low level or on a hi h level& ) low-level descri!tion of the state of an ant colony would involve !ainfully s!ecifyin the location of each ant, its a e and caste, and other similar items& ) very detailed descri!tion, yieldin !ractically no lobal insi ht as to 3?H it is in that state& En the other hand, a descri!tion on a hi h level would involve s!ecifyin which symbols could be tri ered by which combinations of other symbols, under what conditions, and so forth& *chilles, 3hat about a descri!tion on the level of si nals, or teams? *nteater, ) descri!tion on that level would fall somewhere in between the low-level and symbol-level descri!tions& It would contain a reat deal of information about what is actually oin on in s!ecific locations throu hout the colony, althou h certainly less than an ant-by-ant descri!tion, since teams consist of clum!s of ants& ) team-by-team descri!tion is like a summary of an ant-by-ant descri!tion& ?owever, you have to add e(tra thin s which were not !resent in the ant-by-ant descri!tionsuch as the relationshi!s between teams, and the su!!ly of various castes here and there& This e(tra com!lication is the !rice you !ay for the ri ht to summari<e& *chilles, It is interestin to me to com!are the merits of the descri!tions at various levels& The hi hest-level descri!tion seems to carry the most e(!lanatory !ower, in that it ives you the most intuitive !icture of the ant colony, althou h stran ely enou h, it leaves out seemin ly the most im!ortant featurethe ants& *nteater, 0ut you see, des!ite a!!earances, the ants are not the most im!ortant feature& )dmittedly, were it not for them, the colony wouldn1t e(ist, but somethin equivalent a braincan e(ist, ant-free& So, at least from a hi h-level !oint of view, the ants are dis!ensable& *chilles, I1m sure no ant would embrace your theory with ea erness& *nteater, 3ell, I never met an ant with a hi h-level !oint of view& 8ra(, 3hat a counterintuitive !icture you !aint, :r& )nteater& It seems that, if what you say is true, in order to ras! the whole structure, you have to describe it omittin any mention of its fundamental buildin blocks& *nteater, Perha!s I can make it a little clearer by an analo y& Ima ine you have before you a 7harles :ickens novel& *chilles, The &ickwick &aperswill that do?

*nteater, E(cellently2 )nd now ima ine tryin the followin ame, you must find a way of ma!!in letters onto ideas, so that the entire &ickwick &apers makes sense when you read it letter by letter& *chilles, ?mm&&& Hou mean that every time I hit a word such as -the-, I have to think of three definite conce!ts, one after another, with no room for variation? *nteater, E(actly& They are the 1t1-conce!t, the 1h1-conce!t, and the 1e1-conce!tand every time, those conce!ts are as they were the !recedin time& *chilles, 3ell, it sounds like that would turn the e(!erience of -readin - The &ickwick &apers into an indescribably borin ni htmare& It would be an e(ercise in meanin lessness, no matter what conce!t I associated with each letter& *nteater, E(actly& There is no natural ma!!in from the individual letters into the real world& The natural ma!!in occurs on a hi her level between words, and !arts of the real world& If you wanted to describe the book, therefore, you would make no mention of the letter level& *chilles, Ef course not2 I1d describe the !lot and the characters, and so forth& *nteater, So there you are& Hou would omit all mention of the buildin blocks, even thou h the book e(ists thanks to them& They are the medium, but not the messa e& *chilles, )ll ri htbut what about ant colonies? *nteater, ?ere, there are active si nals instead of !assive letters, and active symbols instead of !assive wordsbut the idea carries over& *chilles, :o you mean I couldn1t establish a ma!!in between si nals and thin s in the real world? *nteater, Hou would find that you could not do it in such a way that the tri erin of new si nals would make any sense& 8or could you succeed on any lower levelfor e(am!le the ant level& Enly on the symbol level do the tri erin !atterns make sense& Ima ine, for instance, that one day you were watchin )unt ?illary when I arrived to !ay a call& Hou could watch as carefully as you wanted, and yet you would !robably !erceive nothin more than a rearran ement of ants& *chilles, I1m sure that1s accurate& *nteater, )nd yet, as I watched, readin the hi her level instead of the lower level, I would see several dormant symbols bein awakened, those which translate into the thou ht, -Eh, here1s that charmin :r& )nteater a ainhow !leasant2-or words to that effect& *chilles, That sounds like what ha!!ened when the four of us all found different levels to read in the /F-!ictureor at least T?;EE of us did && & Tortoise, 3hat an astonishin coincidence that there should be such a resemblance between that stran e !icture which I chanced u!on in the >ell%Tempered 8la#ier, and the trend of our conversation& *chilles, :o you think it1s .ust coincidence? Tortoise, Ef course& *nteater, 3ell, I ho!e you can ras! now how the thou hts in )unt ?illary emer e from the mani!ulation of symbols com!osed of si nals com!osed of teams com!osed of lower-level teams, all the way down to ants&

*chilles, 3hy do you call it -symbol mani!ulation-? 3ho does the mani!ulatin , if the symbols are themselves active? 3ho is the a ent? *nteater, This ets back to the question which you earlier raised about !ur!ose& Hou1re ri ht that symbols themselves are active, but the activities which they follow are nevertheless not absolutely free& The activities of all symbols are strictly determined by the state of the full system in which they reside& Therefore, the full system is res!onsible for how its symbols tri er each other, and so it is quite reasonable to s!eak of the full system as the -a ent-& )s the symbols o!erate, the state of the system ets slowly transformed, or u!dated& 0ut there are many features which remain over time& It is this !artially constant, !artially varyin system which is the a ent& Ene can ive a name to the full system& 'or e(am!le, )unt ?illary is the -who- who can be said to mani!ulate her symbols= and you are similar, )chilles& *chilles, That1s quite a stran e characteri<ation of the notion of who I am& I1m not sure I can fully understand it, but I will ive it some thou ht& Tortoise, It would be quite interestin to follow the symbols in your brain as you do that thinkin about the symbols in your brain& *chilles, That1s too com!licated for me& I have trouble enou h .ust tryin to !icture how it is !ossible to look at an ant colony and read it on the symbol level& I can certainly ima ine !erceivin it at the ant level= and with a little trouble, I can ima ine what it must be like to !erceive it at the si nal level= but what in the world can it be like to !erceive an ant colony at the symbol level? *nteater, Ene only learns throu h lon !ractice& 0ut when one is at my sta e, one reads the to! level of an ant colony as easily as you yourself read the - /F- in the /F!icture& *chilles, ;eally? That must be an ama<in e(!erience& *nteater, In a waybut it is also one which is quite familiar to you, )chilles& *chilles, 'amiliar to me? 3hat do you mean? I have never looked at an ant colony on anythin but the ant level& *nteater, /aybe not= but ant colonies are no different from brains in many res!ects& *chilles, I have never seen nor read any brain either, however& *nteater, 3hat about your E38 brain? )ren1t you aware of your own thou hts? Isn1t that the essence of consciousness? 3hat else are you doin but readin your own brain directly at the symbol level? *chilles, I never thou ht of it that way& Hou mean that I by!ass all the lower levels, and only see the to!most level? *nteater, That1s the way it is, with conscious systems& They !erceive themselves on the symbol level only, and have no awareness of the lower levels, such as the si nal levels& *chilles, :oes it follow that in a brain, there are active symbols which are constantly u!datin themselves so that they reflect the overall state of the brain itself, always on the symbol level? *nteater, 7ertainly& In any conscious system there are symbols which re!resent the brain state, and they are themselves !art of the very brain state which they symboli<e& 'or consciousness requires a lar e de ree of self-consciousness&

*chilles, That is a weird notion& It means that althou h there is frantic activity occurrin in my brain at all times, I am only ca!able of re isterin that activity in one wayon the symbol level= and I am com!letely insensitive to the lower levels& It is like bein able to read a :ickens novel by direct visual !erce!tion, without ever havin learned the letters of the al!habet& I can1t ima ine anythin as weird as that really ha!!enin & 8ra(, 0ut !recisely that sort of thin :I: ha!!en when you read -/F-, without !erceivin the lower levels -?EAIS/- and -;E:F7TIE8IS/-& *chilles, Hou1re ri htI by!assed the lower levels, and saw only the to!& I wonder if I1m missin all sorts of meanin on lower levels of my brain as well, by readin only the symbol level& It1s too bad that the to! level doesn1t contain all the information about the bottom level, so that by readin the to!, one also learns what the bottom level says& 0ut I uess it would be naPve to ho!e that the to! level encodes anythin from the bottom levelit !robably doesn1t !ercolate u!& The /F-!icture is the most strikin !ossible e(am!le of that, there, the to!most level says only - /F-, which bears no relation whatever to the lower levels2 8ra(, That1s absolutely true& 4&icks up the '5%picture, to inspect it more closel)&5 ?mm&&& There1s somethin stran e about the smallest letters in this !icture= they1re very wi ly&&& *nteater, Aet me take a look& 4&eers closel) at the '5%picture&5 I think there1s yet another level, which all of us missed2 Tortoise, S!eak for yourself, :r& )nteater& *chilles, Eh, nothat can1t be2 Aet me see& 4 +ooks #er) carefull)&5 I know the rest of you won1t believe this, but the messa e of this !icture is starin us all in the face, hidden in its de!ths& It is sim!ly one word, re!eated over and over a ain, like a mantrabut what an im!ortant one, -/F-2 3hat do you know2 It is the same as the to! level2 )nd none of us sus!ected it in the least& 8ra(, 3e would never have noticed it if it hadn1t been for you, )chilles& *nteater, I wonder if the coincidence of the hi hest and lowest levels ha!!ened by chance? Er was it a !ur!oseful act carried out by some creator? 8ra(, ?ow could one ever decide that? Tortoise, I don1t see any way to do so, since we have no idea why that !articular !icture is in the 7rab1s edition of the >ell%Tempered 8la#ier& *nteater, )lthou h we have been havin a lively discussion, I have still mana ed to listen with a ood fraction of an ear to this very lon and com!le( four-voice fu ue& It is e(traordinarily beautiful& Tortoise, It certainly is& )nd now, in .ust a moment, comes an or an !oint& *chilles, Isn1t an or an !oint what ha!!ens when a !iece of music slows down sli htly, settles for a moment or two on a sin le note or chord, and then resumes at normal s!eed after a short silence? Tortoise, 8o, you1re thinkin of a -fermata-a sort of musical semicolon& :id you notice there was one of those in the !relude? *chilles, I uess I must have missed it& Tortoise, 3ell, you have another chance comin u! to hear a fermatain fact, there are a cou!le of them comin u!, towards the end of this fu ue&

*chilles, Eh, ood& Hou1ll !oint them out in advance, won1t you? Tortoise, If you like& *chilles, 0ut do tell me, what is an or an !oint? Tortoise, )n or an !oint is the sustainin of a sin le note by one of the voices in a !oly!honic !iece 4often the lowest voice5, while the other voices continue their own inde!endent lines& This or an !oint is on the note of 6& Aisten carefully, and you1ll hear it& *nteater, There occurred an incident one day when I visited with )unt ?illary which reminds me of your su estion of observin the symbols in )chilles1 brain as they create thou hts which are about themselves& 8ra(, :o tell us about it& *nteater, )unt ?illary had been feelin very lonely, and was very ha!!y to have someone to talk to that day& So she ratefully told me to hel! myself to the .uiciest ants I could find& 4She1s always been most enerous with her ants&5 *chilles, 6ee2 *nteater, It .ust ha!!ened that I had been watchin the symbols which were carryin out her thou hts, because in them were some !articularly .uicy-lookin ants& *chilles, 6ee2 *nteater, So I hel!ed myself to a few of the fattest ants which had been !arts of the hi her-level symbols which I had been readin & S!ecifically, the symbols which they were !art of were the ones which had e(!ressed the thou ht, -?el! yourself to any of the ants which look a!!eti<in &*chilles, 6ee2 *nteater, Fnfortunately for them, but fortunately for me, the little bu s didn1t have the sli htest inklin of what they were collectively tellin me, on the symbol level& *chilles, 6ee2 That is an ama<in wra!around& They were com!letely unconscious of what they were !artici!atin in& Their acts could be seen as !art of a !attern on a hi her level, but of course they were com!letely unaware of that& )h, what a !itya su!reme irony, in factthat they missed it& 8ra(, Hou are ri ht, /r& Tthat was a lovely or an !oint& *nteater, I had never heard one before, but that one was so cons!icuous that no one could miss it& Gery effective& *chilles, 3hat? ?as the or an !oint already occurred? ?ow can I not have noticed it, if it was so blatant? Tortoise, Perha!s you were so wra!!ed u! in what you were sayin that you were com!letely unaware of it& )h, what a !itya su!reme irony, in factthat you missed it& 8ra(, Tell me, does )unt ?illary live in an anthill? *nteater, 3ell, she owns a rather lar e !iece of !ro!erty& It used to belon to someone else, but that is rather a sad story& In any case, her estate is quite e(!ansive& She lives rather sum!tuously, com!ared to many other colonies& *chilles, ?ow does that .ibe with the communistic nature of ant colonies which you earlier described to us? It sounds quite inconsistent, to me, to !reach communism and to live in a fancy estate&

*nteater, The communism is on the ant level& In an ant colony all ants work for the common ood, even to their own individual detriment at times& 8ow this is sim!ly a built-in as!ect of )unt ?illary1s structure, but for all I know, she may not even be aware of this internal communism& /ost human bein s are not aware of anythin about their neurons= in fact they !robably are quite content not to know anythin about their brains, bein somewhat squeamish creatures& )unt ?illary is also somewhat squeamish= she ets rather antsy whenever she starts to think about ants at all& So she avoids thinkin about them whenever !ossible& I truly doubt that she knows anythin about the communistic society which is built into her very structure& She herself is a staunch believer in libertarianismyou know, laisse<-faire and all that& So it makes !erfect sense, to me at least, that she should live in a rather sum!tuous manor&

73G5=E JF! U.rawing () the author!V

Tortoise, )s I turned the !a e .ust now, while followin alon in this lovely edition of the 3ell-Tem!ered 7lavier, I noticed that the first of the two fermatas is comin u! soon so you mi ht listen for it, )chilles& *chilles, I will, I will& Tortoise, )lso, there1s a most curious !icture facin this !a e& 8ra(, )nother one? 3hat ne(t? Tortoise, See for yourself& &asses the score o#er to the 8ra(!/ 8ra(, )ha2 It1s .ust a few bunches of letters& Aet1s seethere are various numbers of the letters 1J1, 1S1, 101, 1m1, 1a1, and 1t1& It1s stran e, how the first three letters row, and then the last three letters shrink a ain& *nteater, /ay I see it? 8ra(, 3hy, certainly& *nteater, Eh, by concentratin on details, you have utterly missed the bi !icture& In reality, this rou! of letters is 1f1, 1e1, 1r1, 1)1, 171, 1?1, without any re!etitions& 'irst they et smaller, then they et bi er& ?ere, )chilleswhat do you make of it? *chilles, Aet me see& ?mm& 3ell, I see it as a set of u!!er-case letters which row as you move to the ri ht& Tortoise, :o they s!ell anythin ? *chilles, )h&&& -J& S& 0)7?-& Eh2 I understand now& It1s 0ach1s name2

Tortoise, Stran e that you should see it that way& I see it as a set of lower-case letters, shrinkin as they move to the ri ht, and&&& s!ellin out&&& the name of&&& -lows down slightl), especiall) drawing out the last few words! Then there is a (rief silence! -uddenl) he resumes as if nothing unusual had happened!/ -fermat-& *chilles, Eh, you1ve ot 'ermat on the brain, I do believe& Hou see 'ermat1s Aast Theorem everywhere& *nteater, Hou were ri ht, /r& TortoiseI .ust heard a charmin little fermata in the fu ue& 8ra(, So did I& *chilles, :o you mean everybody heard it but me? I1m be innin to feel stu!id& Tortoise, There, there, )chillesdon1t feel bad& I1m sure you won1t miss 'u ue1s Aast 'ermata 4which is comin u! quite soon5& 0ut, to return to our !revious to!ic, :r& )nteater, what is the very sad story which you alluded to, concernin the former owner of )unt ?illary1s !ro!erty? *nteater, The former owner was an e(traordinary individual, one of the most creative ant colonies who ever lived& ?is name was Johant Sebastiant 'ermant, and he was a mathematiciant by vocation, but a musiciant by avocation& *chilles, ?ow very versantile of him2 *nteater, )t the hei ht of his creative !owers, he met with a most untimely demise& Ene day, a very hot summer day, he was out soakin u! the warmth, when a freak thundershowerthe kind that hits only once every hundred years or soa!!eared from out of the blue, and thorou hly drenched J& S '& Since the storm came utterly without warnin , the ants ot com!letely disoriented and confused& The intricate or ani<ation which had been so finely built u! over decades, all went down the drain in a matter of minutes& It was tra ic& *chilles, :o you mean that all the ants drowned, which obviously would s!ell the end of !oor J& S& '& *nteater, )ctually, no& The ants mana ed to survive, every last one of them, by crawlin onto various sticks and lo s which floated above the ra in torrents& 0ut when the waters receded and left the ants back on their home rounds, there was no or ani<ation left& The caste distribution was utterly destroyed, and the ants themselves had no ability to reconstruct what had once before been such a finely tuned or ani<ation& They were as hel!less as the !ieces of ?um!ty :um!ty in !uttin themselves back to ether a ain& I myself tried, like all the kin 1s horses and all the kin 1s men, to !ut !oor 'ermant to ether a ain& I faithfully !ut out su ar and cheese, ho!in a ainst ho!e that somehow 'ermant would rea!!ear&&& 4 &ulls out a handkerchief and wipes his e)es!/ *chilles, ?ow valiant of you2 I never knew )nteaters had such bi hearts& *nteater, 0ut it was all to no avail& ?e was one, beyond reconstitution& ?owever, somethin very stran e then be an to take !lace, over the ne(t few months, the ants which had been com!onents of J& S& '& slowly re rou!ed, and built u! a new or ani<ation& )nd thus was )unt ?illary born& 8ra(, ;emarkable2 )unt ?illary is com!osed of the very same ants as 'ermant was?

*nteater, 3ell, ori inally she was, yes& 0y now, some of the older ants have died, and been re!laced& 0ut there are still many holdovers from the J& S& '&-days& 8ra(, )nd can1t you reco ni<e some of J& S& '&1s old traits comin to the fore, from time to time, in )unt ?illary? *nteater, 8ot a one& They have nothin in common& )nd there is no reason they should, as I see it& There are, after all, often several distinct ways to rearran e a rou! of !arts to form a -sum-& )nd )unt ?illary was .ust a new -sum- of the old !arts& 8ot /E;E than the sum, mind you .ust that !articular @I8: of sum& Tortoise, S!eakin of sums, I am reminded of number theory, where occasionally one will be able to take a!art a theorem into its com!onent symbols, rearran e them in a new order, and come u! with a new theorem& *nteater, I1ve never heard of such a !henomenon, althou h I confess to bein a total i noramus in the field& *chilles, 8or have I heard of itand I am rather well versed in the field, If I don1t say so myself& I sus!ect /r& T is .ust settin u! one of his elaborate s!oofs& I know him !retty well by now& *nteater, S!eakin of number theory, I am reminded of J& S& '& a ain, for number theory is one of the domains in which he e(celled& In fact, he made some rather remarkable contributions to number theory& )unt ?illary, on the other hand, is remarkably dullwitted in anythin that has even the remotest connection with mathematics& )lso, she has only a rather banal taste in music, whereas Sebastiant was e(tremely ifted in music& *chilles, I am very fond of number theory& 7ould you !ossibly relate to us somethin of the nature of Sebastiant1s contributions, *nteater, Gery well, then& &auses for a moment to sip his tea, then resumes!/ ?ave you heard of 'ourmi1s infamous -3ell-Tested 7on.ecture-? *chilles& I1m not sure&&& It sounds stran ely familiar, and yet I can1t quite !lace it& *nteater, It1s a very sim!le idea& Aierre de 'ourmi, a mathematiciant by vocation but lawyer by avocation, had been readin in his co!y of the classic te(t *rithmetica by :i of )ntus, and came across a !a e containin the equation

Ba^BbXBc
?e immediately reali<ed that this equation has infinitely many solutions a& b, c, and then wrote in the mar in the followin notorious comment,
The equation na^nbXnc has solutions in !ositive inte ers a, b, c, and n only when n X B 4and then there are infinitely many tri!lets a, b, c which satisfy the equation5= but there are no solutions for n a B& I have discovered a truly marvelous !roof of this statement, which, unfortunately, is so small that it would be well-ni h invisible if written in the mar in&

'I6F;E D%& .uring emigrations arm) ants sometimes create li#ing (ridges of their own (odies! 3n this photograph of such a (ridge de 7ourmi +ierre/, the workers of an Eciton burchelli colon) can (e seen linking their legs and, along the top of the (ridge, hooking their tarsal claws together to form irregular s)stems of chains! * s)m(iotic sil#erfish, Trichatelura manni, is seen crossing the (ridge in the center! U7rom E! :! >ilson, The Insect Societies 8am(ridge, 'ass!: ,ar#ard 5ni#ersit) &ress, BICB/, p! JFV

Ever since that year, some three hundred days a o, mathematiciants have been vainly tryin to do one of two thin s, either to !rove 'ourmi1s claim, and thereby vindicate 'ourmi1s re!utation, which, althou h very hi h, has been somewhat tarnished by ske!tics who think he never really found the !roof he claimed to have foundor else to refute the claim, by findin a countere(am!le, a set of four inte ers a, b, c, and n, with n a B, which satisfy the equation& Fntil very recently, every attem!t in either direction had met with failure& To be sure, the 7on.ecture has been verified for many s!ecific values of nin !articular, all n u! to "BL,+++& 0ut no one had succeeded in !rovin it for )AA nno one, that is, until Johant Sebastiant 'ermant came u!on the scene& It was he who found the !roof that cleared 'ourmi1s name& It now oes under the name -Johant Sebastiant1s 3ell-Tested 7on.ecture-& *chilles, Shouldn1t it be called a -Theorem- rather than a -7on.ecture-, if it1s finally been iven a !ro!er !roof? *nteater, Strictly s!eakin , you1re ri ht, but tradition has ke!t it this way& Tortoise, 3hat sort of music did Sebastiant do? *nteater, ?e had reat ifts for com!osition& Fnfortunately, his reatest work is shrouded in mystery, for he never reached the !oint of !ublishin it& Some believe that he had it all in his mind= others are more unkind, sayin that he !robably never worked it out at all, but merely blustered about it&

*chilles, 3hat was the nature of this ma num o!us? *nteater, It was to be a iant !relude and fu ue= the fu ue was to have twenty-four voices, and to involve twenty-four distinct sub.ects, one in each of the ma.or and minor keys& *chilles, It would certainly be hard to listen to a twenty-four-voice fu ue as a whole2 8ra(, 8ot to mention com!osin one2 *nteater, 0ut all that we know of it is Sebastiant1s descri!tion of it, which he wrote in the mar in of his co!y of 0u(tehude1s Preludes and 'u ues for Er an& The last words which he wrote before his tra ic demise were,
I have com!osed a truly marvelous fu ue& In it, I have added to ether the !ower of BC keys, and the !ower of BC themes= I came u! with a fu ue with the !ower of BC voices& Fnfortunately, this mar in is too narrow to contain it&

)nd the unreali<ed master!iece sim!ly oes by the name, -'ermant1s Aast 'u ue-& *chilles, Eh, that is unbearably tra ic& Tortoise, S!eakin of fu ues, this fu ue which we have been listenin to is nearly over& Towards the end, there occurs a stran e new twist on its theme& 7lips the page in the >ell%Tempered 8la#ier!/ 3ell, what have we here? ) new illustration-how a!!ealin 2 -hows it to the 8ra(&5
73G5=E JD! U.rawing () the authorV

8ra(, 3ell, what have we here? Eh, I see, It1s -?EAIS/IE8IS/-, written in lar e letters that first shrink and then row back to their ori inal si<e& 0ut that doesn1t make any sense, because it1s not a word& Eh me, oh my2 &asses it to the *nteater!/ *nteater, 3ell, what have we here? Eh, I see, it1s - ;E:F7T?EAIS/-, written in small letters that first row and then shrink back to their ori inal si<e& 0ut that doesn1t make any sense, because it1s not a word& Eh my, oh me2 &asses it to *chilles!/ *chilles, I know the rest of you won1t believe this, but in fact this !icture consists of the word -?EAIS/- written twice, with the letters continually shrinkin as they !roceed from left to ri ht& =eturns it to the Tortoise!/

Tortoise, I know the rest of you won1t believe this, but in fact this !icture consists of the word -;E:F7TIE8IS/- written once, with the letters continually rowin as they !roceed from left to ri ht& *chilles, )t lastI heard the new twist on the theme this time2 I am so lad that you !ointed it out to me, /r& Tortoise& 'inally, I think I am be innin to ras! the art of listenin to fu ues&

Cha!ter <I

Brains and Thoughts

;e' ,ers!ecti"es on Thought
IT 3)S E8AH with the advent of com!uters that !eo!le actually tried to create -thinkin machines, and witnessed bi<arre variations on the theme, of thou ht& Pro rams were devised whose -thinkin - was to human thinkin as a slinky fli!!in end over end down a staircase is to human locomotion& )ll of a sudden the idiosyncrasies, the weaknesses and !owers, the va aries and vicissitudes of human thou ht were hinted at by the newfound ability to e(!eriment with alien, yet hand-tailored forms of thou htor a!!ro(imations of thou ht& )s a result, we have acquired, in the last twenty years or so, a new kind of !ers!ective on what thou ht is, and what it is not& /eanwhile, brain researchers have found out much about the small-scale and lar e-scale hardware of the brain& This a!!roach has not yet been able to shed much li ht on how the brain mani!ulates conce!ts, but it ives us some ideas about the biolo ical mechanisms on which thou ht mani!ulation rests& In the comin two 7ha!ters, then, we will try to unite some insi hts leaned from attem!ts at com!uter intelli ence with some of the facts learned from in enious e(!eriments on livin animal brains, as well as with results from research on human thou ht !rocesses done by co nitive !sycholo ists& The sta e has been set by the &relude, *nt 7ugue= now we develo! the ideas more dee!ly&

Intensionality and E-tensionality

Thou ht must de!end on representing realit) in the hardware of the (rain& In the !recedin 7ha!ters, we have develo!ed formal systems which re!resent domains of mathematical reality in their symbolisms& To what e(tent is it reasonable to use such formal systems as models for how the brain mi ht mani!ulate ideas? 3e saw, in the !q-system and then in other more com!licated systems, how meanin , in a limited sense of the term, arose as a result of an isomor!hism which ma!s ty!o ra!hical symbols onto numbers, o!erations, and relations= and strin s of ty!o ra!hical symbols onto statements& 8ow in the brain we don1t have ty!o ra!hical symbols, but we have somethin even better, active elements which can store information and transmit it and receive it from other active elements& Thus we have acti#e symbols, rather than !assive ty!o ra!hical symbols& In the brain, the rules are mi(ed ri ht in with the symbols themselves, whereas on !a!er, the symbols are static entities, and the rules are in our heads& It is im!ortant not to et the idea, from the rather strict nature of all the formal systems we have seen, that the isomor!hism between symbols and real thin s is a ri id,

one-to-one ma!!in , like the strin s which link a marionette and the hand uidin it& In T8T, the notion -fifty- can be e(!ressed in different symbolic ways= for e(am!le, 44SSSSSSS?jSSSSSSS?5^4S?jS?55 44SSSSS?jSSSSS?5^4SSSSS?jSSSSS?55 That these both re!resent the same number is not a !riori clear& Hou can mani!ulate each e(!ression inde!endently, and at some !oint stumble cross a theorem which makes you e(claim, -Ehit1s that number2In your mind, you can also have different mental descri!tions for a sin le !erson= for e(am!le, The !erson whose book I sent to a friend in Poland a while back& The stran er who started talkin with me and my friends toni ht in this coffee house& That they both re!resent the same !erson is not a !riori clear& 0oth descri!tions may sit in your mind, unconnected& )t some !oint durin the evenin you may stumble across a to!ic of conversation which leads to the revelation that they desi nate the same !erson, makin you e(claim, -Ehyou1re that !erson28ot all descri!tions of a !erson need be attached to some central symbol for that !erson, which stores the !erson1s name& :escri!tions can be manufactured and mani!ulated in themselves& 3e can invent none(istent !eo!le by makin descri!tions of them= we can mer e two descri!tions when we find they re!resent a sin le entity= we can s!lit one descri!tion into two when we find it re!resents two thin s, not oneand so on& This -calculus of descri!tions- is at the heart of thinkin & It is said to be intentional and not e(tensional, which means that descri!tions can -float- without bein anchored down to s!ecific, known ob.ects& The intensionality of thou ht is connected to its fle(ibility= it ives us the ability to ima ine hy!othetical worlds, to amal amate different descri!tions or cho! one descri!tion into se!arate !ieces, and so on& Su!!ose a friend who has borrowed your car tele!hones you to say that your car skidded off a wet mountain road, careened a ainst a bank, and overturned, and she narrowly esca!ed death& Hou con.ure u! a series of ima es in your mind, which et !ro ressively more vivid as she adds details, and in the end you -see it all in your mind1s eye-& Then she tells you that it1s all been an )!ril 'ool1s .oke, and both she and the car are fine2 In many ways that is irrelevant& The story and the ima es lose nothin of their vividness, and the memory will stay with you for a lon , lon time& Aater, you may even think of her as an unsafe driver because of the stren th of the first im!ression, which should have been wi!ed out when you learned it was all untrue& 'antasy and fact intermin le very closely in our minds, and this is because thinkin involves the manufacture and mani!ulation of com!le( descri!tions, which need in no way be tied down to real events or thin s&

) fle(ible, intensional re!resentation of the world is what thinkin is all about& 8ow how can a !hysiolo ical system such as the brain su!!ort such a system?

The Brain%s (Ants(

The most im!ortant cells in the brain are nerve cells, or neurons 4see 'i & DL5, of which there are about ten billion& 47uriously, outnumberin the neurons by about ten to one are the lial cells, or lia& 6lia are believed to !lay more of a su!!ortin role to the neurons1 starrin role, and therefore we will not discuss them&5 Each neuron !ossesses a number of syna!ses 4-entry !orts-5 and one a(on 4-out!ut channel-5& The in!ut and out!ut are electrochemical flows, that is, movin ions& In between the entry !orts of a neuron and its out!ut channel is its cell body, where -decisions- are made&

73G5=E JG! -chematic drawing of a neuron! U*dapted 7rom .! >ooldridge, The /achinery of the 0rain New York: 'cGraw%,ill, BIJH/, p! J!V

The ty!e of decision which a neuron facesand this can take !lace u! to a thousand times !er secondis this, whether or not to firethat is, to release ions down its a(on, which eventually will cross over into the entry !orts of one or more other neurons, thus causin them to make the same sort of decision& The decision is made in a very sim!le manner, if the sum all in!uts e(ceeds a certain threshold, yes= otherwise, no& Some of the in!uts can be ne ative in!uts, which cancel out !ositive in!uts comin from somewhere else& In any case, it is sim!le addition which rules the lowest level of the mind& To !ara!hrase :escartes1 famous remark, -I think, therefore I sum- 4from the Aatin 8ogito, ergo am5& 8ow althou h the manner of makin the decision sounds very sim!le, here is one fact which com!licates the issue, there may be as many as B++,+++ se!arate entry !orts to a neuron, which means that u! to B++,+++ se!arate summands may be involved in determinin the neuron1s ne(t ion& Ence the decision has been made, a !ulse of ions streaks down the on towards its terminal end& 0efore the ions reach the end, however, they may encounter a bifurcationor several& In such cases, the sin le out!ut !ulse s!lits u! as it moves down the bifurcatin a(on, and by the time it has reached the end, -it- has become -they-and they may reach their destinations at se!arate times, since the a(on branches alon which they travel may be of different len ths and have different resistivities& The im!ortant thin , thou h, is that they all be an as one sin le !ulse, movin away from the cell body& )fter a neuron fires, it needs a short recovery time before firin a ain= characteristically this is measured in milliseconds, so at a neuron may fire u! to about a thousand times !er second&

Larger Structures in the Brain

8ow we have described the brain1s -ants-& 3hat about -teams-, or -si nals-? 3hat about -symbols-? 3e make the followin observation, des!ite the com!le(ity of its in!ut, a sin le neuron can res!ond only in a very !rimitive wayby firin , or not firin & This is a very small amount of information& 7ertainly for lar e amounts of information to be carried or !rocessed, many neurons must be involved& )nd therefore one mi ht uess that lar er structures, com!osed from many neurons, would e(ist, which handle conce!ts on a hi her level& This is undoubtedly true, but the most naPve assum!tionthat there is a fi(ed rou! of neurons for each different conce!tis almost certainly false& There are many anatomical !ortions of the brain which can be distin uished from each other, such as the cerebrum, the cerebellum, the hy!othalamus 4see 'i & DD5& The cerebrum is the lar est !art of the human brain, and is divided into a left hemis!here and a ri ht hemis!here& The outer few millimeters of each cerebral hemis!here are coated with a layered -bark-, or cerebral corte(& The amount of cerebral corte( is the ma.or distin uishin feature, in terms of anatomy, between human brains and brains of less intelli ent s!ecies& 3e will not describe any of the brain1s subor ans in detail because, as it turns out, only the rou hest ma!!in can at this time be made between such lar e-scale subor ans and the activities, mental or !hysical, which they are res!onsible for& 'or instance, it is known that lan ua e is !rimarily handled in one of the two cerebral

73G5=E JJ! The human (rain, seen from the left side! 3t is strange that the #isual area is in the (ack of the head! U7rom -te#en =ose, The 7onscious 0rain, updated ed! New York: <intage, BIJJ/, p! GK! V

hemis!heresin fact, usually the left hemis!here& )lso, the cerebellum is the !lace where trains of im!ulses are sent off to muscles to control motor activity& 0ut how these areas carry out their functions is still lar ely a mystery&

Ma!!ings bet'een Brains

8ow an e(tremely im!ortant question comes u! here& If thinkin does take !lace in the brain, then how are two brains different from each other? ?ow is my brain different from yours? 7ertainly you do not think e(actly as I do, nor as anyone else does& 0ut we all have the same anatomical divisions in our brains& ?ow far does this identity of brains e(tend? :oes it o to the neural level? Hes, if you look at animals on a low enou h level of the thinkin -hierarchythe lowly earthworm, for instance& The followin quote is from the neuro!hysiolo ist, :avid ?ubel, s!eakin at a conference on communication with e(traterrestrial intelli ence,
The number of nerve cells in an animal like a worm would be measured, I su!!ose, in the thousands& Ene very interestin thin is that we may !oint to a !articular individual cell in a !articular earthworm, and then identify the same cell, the corres!ondin cell in another earthworm of the same s!ecies&"

Earthworms have isomor!hic brains2 Ene could say, -There is only one earthworm&0ut such one-to-one ma!!ability between individuals1 brains disa!!ears very soon as you ascend in the thinkin -hierarchy and the number of neurons increasesconfirmin one1s sus!icions that there is not .ust one human2 Het considerable !hysical similarity can be detected between different human brains when they are com!ared on a scale lar er than a sin le neuron but smaller than the ma.or subor ans of the brain& 3hat s this im!ly about how individual mental differences are re!resented in !hysical brain? If we looked at my neurons1 interconnections, could we find various structures that could be identified as codin for s!ecific thin s I know, s!ecific beliefs I have, s!ecific ho!es, fears, likes and dislikes I harbor? If mental e(!eriences can be attributed to the brain, can knowled e and other as!ects of mental life likewise be traced to s!ecific locations inside the brain, or to s!ecific !hysical subsystems of the brain? This will be a central question to which we will often return in this 7ha!ter and the ne(t&

Locali:ation of Brain ,rocesses5 An Enigma

In an attem!t to answer this question, the neurolo ist @arl Aashley, in a series of e(!eriments be innin around "#B+ and runnin for many years, tried to discover where in its brain a rat stores its knowled e about ma<e runnin & In his book The 8onscious Brain, Steven ;ose describes Aashley1s trials and tribulations this way,
Aashley was attem!tin to identify the locus of memory within the corte(, and, to do so, first trained rats to run ma<es, and then removed various cortical re ions& ?e allowed the animals to recover and tested the retention of the ma<e-runnin skills& To his sur!rise it was not !ossible to find a !articular re ion corres!ondin to the ability to remember the way throu h a ma<e& Instead all the rats which had had corte( re ions removed suffered some kind of im!airment, and the e(tent of the im!airment was rou hly !ro!ortional to the amount of corte( taken off& ;emovin corte( dama ed the motor and sensory ca!acities of the animals, and they would lim!, ho!, roll, or sta er, but somehow they always mana ed to traverse the ma<e& So far as memory was concerned, the corte( a!!eared to be equi!otential, that is, with all re ions of equal !ossible utility& Indeed, Aashley concluded rather loomily in is last !a!er -In Search of the En ram-, which a!!eared in "#L+, that the only conclusion was that memory was not !ossible at all& B

7uriously, evidence for the o!!osite !oint of view was bein develo!ed in 7anada at rou hly the same time that Aashley was doin his last work, in the late "#C+1s& The neurosur eon 3ilder Penfield was e(aminin the reactions of !atients whose brains had been o!erated on, by insertin electrodes into various !arts of their e(!osed brains, and then usin small electrical !ulses to stimulate the neuron or neurons to which the electrodes been attached& These !ulses were similar to the !ulses which come other neurons& 3hat Penfield found was that stimulation of certain neurons would reliably create s!ecific ima es or sensations in the !atient& These artificially !rovoked im!ressions ran ed from stran e but indefinable fears to bu<<es and colors, and, most im!ressively of all, to entire successions of events recalled from some earlier time of life, such as a childhood birthday !arty& The set of locations which could tri er such s!ecific

events was e(tremely smallbasically centered u!on a sin le neuron& 8ow these results of Penfield dramatically o!!ose the conclusions of Aashley, since they seem to im!ly that local areas are res!onsible for s!ecific memories, after all& 3hat can one make of this? Ene !ossible e(!lanation could be that memories are coded locally, but over and over a ain in different areas of the corte(a strate y !erha!s develo!ed in evolution as security a ainst !ossible loss of corte( in fi hts, or in e(!eriments conducted by neuro!hysiolo ists& )nother e(!lanation would be that memories can be reconstructed from dynamic !rocesses s!read over the whole brain, but can be tri ered from local s!ots& This theory is based on the notion of modern tele!hone networks, where the routin of a lon -distance call is not !redictable in advance, for it is selected at the time the call is !laced, and de!ends on the situation all over the whole country& :estroyin any local !art of the network would not block calls= it would .ust cause them to be routed around the dama ed area& In this sense any call is !otentially nonlocali<able& Het any call .ust connects u! two s!ecific !oints= in this sense any call is locali<able&

S!ecificity in 4isual ,rocessing

Some of the most interestin and si nificant work on locali<ation of brain !rocesses has been done in the last fifteen years by :avid ?ubel and Torsten 3iesel, at ?arvard& They have ma!!ed out visual !athways in the brains of cats, startin with the neurons in the retina, followin their connections towards the rear of the head, !assin throu h the -relay station- of the lateral eniculate, and endin u! in the visual corte(, at the very back of the brain& 'irst of all, it is remarkable that there e(ist well defined neural !athways, in li ht of Aashley1s results& 0ut more remarkable are the !ro!erties of the neurons located at different sta es alon the !athway& It turns out that retinal neurons are !rimarily contrast sensors& /ore s!ecifically, the way they act is this& Each retinal neuron is normally firin at a -cruisin s!eed-& 3hen its !ortion of the retina is struck by li ht, it may either fire faster or slow down and even sto! firin & ?owever, it will do so only !rovided that the surroundin !art of the retina is less illuminated& So this means that there are two ty!es of neuron, -on-center-, and -off-center-& The on%center neurons are those whose firin rate increases whenever, in the small circular retinal area to which they are sensitive, the center is bri ht but the outskirts are dark= the off%center neurons are those which fire faster when there is darkness in the center and bri htness in the outer rin & If an on-center !attern is shown to an off-center neuron, the neuron will slow down in firin 4and vice versa5& Fniform illumination will leave both ty!es of retinal neuron unaffected= they will continue to fire at cruisin s!eed& 'rom the retina, si nals from these neurons !roceed via the o!tic nerve to the lateral eniculate, located somewhere towards the middle of the brain& There, one can find a direct ma!!in of the retinal surface in the sense that there are lateral- eniculate neurons which are tri ered only by s!ecific stimuli fallin on s!ecific areas of the retina& In that sense, the eneral eniculate is disa!!ointin = it seems to be only a -relay station-, and not a further !rocessor 4althou h to ive it its due, the contrast sensitivity seems to be

enhanced in the lateral eniculate5& The retinal ima e is coded a strai htforward way in the firin !atterns of the neurons in the lateral eniculate, des!ite the fact that the neurons there are not arran ed on a two-dimensional surface in the form of the retina, but in a three-dimensional block& So two dimensions et ma!!ed onto three, yet the formation is !reserved, an isomor!hism& There is !robably some dee! meanin to the chan e in the dimensionality of the re!resentation, which is not yet fully a!!reciated& In any case, there are so many further une(!lained sta es of vision that we should not be disa!!ointed but !leased by the fact thatto some e(tentwe have fi ured out this one sta e2 'rom the lateral eniculate, the si nals !roceed back to the visual corte(& ?ere, some new ty!es of !rocessin occur& The cells of the visual corte( are divided into three cate ories, sim!le, com!le(, and hy!er-com!le(& Sim!le cells act very much like retinal cells or lateral eniculate cells, they res!ond to !oint-like li ht or dark s!ots with contrastin surrounds, in !articular re ions of the retina& 8omple9 cells, by contrast, usually receive in!ut from a hundred or more other cells, and they detect li ht or dark bars oriented at s!ecific an les on the retina 4see 'i & D$5& ,)per%comple9 cells res!ond to corners, bars, or even -ton ues- movin in s!ecific directions 4a ain see 'i & D$5& These latter cells are so hi hly s!eciali<ed at they are sometimes called -hi her-order hy!er-com!le( cells-&

A (Grandmother Cell(7
0ecause of the discovery of cells in the visual corte( which can be tri ered by stimuli of ever-increasin com!le(ity, some !eo!le have wondered if thin s are not leadin in the direction of -one cell, one conce!t-for e(am!le, you would have a - randmother cell- which would fire if, and only if, your randmother came into view& This somewhat humorous e(am!le of a -su!erhy!ercom!le( cell- is not taken very seriously& ?owever, it is not obvious what alternative theory seems reasonable& Ene !ossibility is that lar er neural networks are e(cited collectively by sufficiently com!le( visual stimuli& Ef course, the tri erin of these lar er multineuron units would somehow have to come from inte ration of si nals emanatin from the many hy!er com!le( cells& ?ow this mi ht be done, nobody knows& Just when we seem to be a!!roachin the threshold where -symbol- mi ht emer e from -si nal-, the trail ets losta tantali<in ly unfinished story& 3e will return to this story shortly, however, and try to fill in some of it& Earlier I mentioned the coarse- rained isomor!hism between all human brains which e(ists on a lar e anatomical scale, and the very fine- rained, neural-level isomor!hism which e(ists between earthworm brains& It is quite interestin that there is also an isomor!hism between the visual !rocessin a!!aratus of cat, monkey, and human, the - rain- of which is somewhere between coarse and fine& ?ere is how that isomor!hism works& 'irst of all, all three s!ecies have -dedicated- areas of corte( at the back of their brains where visual !rocessin is done, the visual corte(& Secondly, in each of them, the visual corte( breaks u! into three subre ions, called areas "*, and "# of the corte(& These areas are still universal, in the sense that they can be located in the brain of any normal individual in any of the three ties& 3ithin each area you can o still further, reachin the -columnar- or ani<ation of the visual corte(& Per!endicular to the surface of

73G5=E JC! =esponses to patterns () certain sample neurons! a/ This edge%detecting neuron looks for #ertical edges with light on the left and dark on the right! The first column shows how the orientation of an edge is rele#ant to this neuron! The second column shows how the position of the edge within the field is irrele#ant, for this particular neuron! (/ -howing how a h)per comple9 cell responds more selecti#el): here, onl) when the descending tongue is in the middle of the field! c/ The responses of a h)pothetical "grandmother cell" to #arious random stimuliR the reader ma) enMo) pondering how an "octopus cell" would respond to the same stimuli!

the corte(, movin radially inwards towards the inner brain, visual neurons are arran ed in -columns-that is, almost all connections move alon the radial, columnar direction, and not between columns& )nd each column ma!s onto a small, s!ecific retinal re ion& The number of columns is not same in each individual, so that one can1t find -the same column-& 'inally, within a column, there are layers in which sim!le neurons tend to be found, and other layers in which com!le( neurons tend to be found& 4The hy!ercom!le( neurons tend to be found in areas "* and "# !redominately, while the sim!le and com!le( ones are found mostly in area "$&5 It a!!ears that we run out of isomor!hisms at this level of detail& 'rom here down to the individual neuron level, each individual cat, monkey, or man has a com!letely unique !atternsomewhat like a fin er!rint or a si nature& Ene minor but !erha!s tellin difference between visual !rocessin in cats1 brains and monkeys1 brains has to do with the sta e at which information from the two eyes is inte rated to yield a sin le combined hi her-level si nal& It turns out that it takes !lace

sli htly later in the monkey than in the cat, which ives each se!arate eye1s si nal a sli htly lon er time to et !rocessed by itself& This is not too sur!risin , since one would e(!ect that hi her a s!ecies lies in the intelli ence hierarchy, the more com!le( will the !roblems which its visual system will be called u!on to handle= and therefore si nals ou ht to !ass throu h more and more early !rocessin before receivin a final -label-& This is quite dramatically confirmed by observations of the visual abilities of a newborn calf, which seems to be born with as much !ower of visual discrimination as it will ever have& It will shy away from !eo!le or do s, but not from other cattle& Probably its entire visual system is -hard-wired- before birth, and involves relatively little o!tical !rocessin & En the other hand, a human1s visual system, so dee!ly reliant on the corte(, takes several years to reach maturity&

+unneling into ;eural Modules

) !u<<lin thin about the discoveries so far made about the or ani<ation the brain is that few direct corres!ondences have been found between lar e-scale hardware and hi h-level software& The visual corte(, for instance, is a lar e-scale !iece of hardware, which is entirely dedicated to a clear software !ur!osethe !rocessin of visual informationyet all of the !rocessin so far discovered is still quite low-level& 8othin a!!roachin reco nition of ob.ects has been locali<ed in the visual corte(& This means that no one knows where or how the out!ut from com!le( and hy!ercom!le( cells ets transformed into conscious reco nition of sha!es, rooms, !ictures, faces, and so on& Peo!le have looked for evidence of the -funnelin - of many low-level neural res!onses into fewer and fewer hi her-level ones, culminatin in somethin such as the !roverbial randmother cell, or some kind of multineuron network, as mentioned above& It is evident that this will not be found in some ross anatomical division of the brain, but rather in a more microsco!ic analysis& Ene !ossible alternative to the the randmother cell mi ht be a fi(ed set of neurons, say a few do<en, at the thin end of the -funnel-, all of which fire when 6ranny comes into view& )nd for each different reco ni<able ob.ect, there would be a unique network and a funnelin !rocess that would focus down onto that network& There are more com!licated alternatives alon similar lines, involvin networks which can be e(cited in different manners, instead of in a fi(ed manner& Such networks would be the -symbols- in our brains& 0ut is such funnelin necessary? Perha!s an ob.ect bein looked at is im!licitly identified by its -si nature- in the visual corte(that is, the collected res!onses of sim!le, com!le(, and hy!ercom!le( cells& Perha!s the brain does not need any further reco ni<er for a !articular form& This theory, however, !oses the followin !roblem& Su!!ose you are lookin at a scene& It re isters its si nature on your visual corte(= but then how do you et from that si nature to a verbal descri!tion of the scene? 'or instance, the !aintin s of Edouard Guillard, a 'rench !ost-im!ressionist, often take a few seconds of scrutiny, and then suddenly a human fi ure will .um! out at you& Presumably the si nature ets im!rinted on the visual corte( in the first fraction of a secondbut the !icture is only understood after a few seconds& This is but one e(am!le of what is actually

a common !henomenona sensation of somethin -crystalli<in - in your mind at the moment of reco nition, which takes !lace not when the li ht rays hit your retina, but sometime later, after some !art of your intelli ence has had a chance to act on the retinal si nals& The crystalli<ation meta!hor yields a !retty ima e derived from statistical mechanics, of a myriad microsco!ic and uncorrelated activities in a medium, slowly !roducin local re ions of coherence which s!read and enlar e= in the end, the myriad small events will have !erformed a com!lete structural revam!in of their medium from the bottom u!, chan in it from a chaotic assembly of inde!endent elements into one lar e, coherent, fully linked structure& If one thinks of the early neural activities as inde!endent, and of the end result of their many inde!endent firin s as the tri erin of a well-defined lar e -module- of neurons, then the word -crystalli<ation- seems quite a!t& )nother ar ument for funnelin is based on the fact that there are a myriad distinct scenes which can cause you to feel you have !erceived the same ob.ectfor e(am!le, your randmother, who may be smilin or frownin , wearin a hat or not, in a bri ht arden or a dark train station, seen from near or far, from side or front, and so on& )ll these scenes !roduce e(tremely different si natures on the visual corte(= yet all of them could !rom!t you to say -?ello, 6ranny&- So a funnelin !rocess must take !lace at some !oint after the rece!tion of the visual si nature and before the words are uttered& Ene could claim that this funnelin is not !art of the !erce!tion of 6ranny, but .ust !art of verbali<ation& 0ut it seems quite unnatural to !artition the !rocess that way, for you could internally use the formation that it is 6ranny without verbali<in it& It would be very unwieldy to handle all of the information in the entire visual corte(, when much of it could be thrown away, since you don1t care about where shadows fall or how many buttons there are on her blouse, etc& )nother difficulty with a non-funnelin theory is to e(!lain how there can be different inter!retations for a sin le si naturefor e(am!le, the Escher !icture 8on#e9 and 8onca#e 4'i & B%5& Just as it seems obvious to us that we do not merely !erceive dots on a television screen, but chunks, likewise it seems ridiculous to !ostulate that !erce!tion has taken !lace when a iant dot-like -si nature- has been created on the visual corte(& There must be some funnelin , whose end result is to tri er some s!ecific modules of neurons, each of which is associated with the conce!tsthe chunksin the scene&

Modules >hich Mediate Thought ,rocesses

Thus we are led to the conclusion that for each conce!t there is a fairly well-defined module which can be tri ereda module that consists of a small rou! of neuronsa -neural com!le(- of the ty!e su ested earlier& ) !roblem with this theoryat least if it is taken naPvelyis that it would su est that one should be able to locate such modules somewhere within the brain& This has not yet been done, and some evidence, such as the e(!eriments by Aashley, !oints a ainst locali<ation& ?owever, it is still too early to tell& There may be many co!ies of each module s!read around, or modules may overla! !hysically= both of these effects would tend to obscure any division of neurons into

-!ackets-& Perha!s the com!le(es are like very thin !ancakes !acked in layers which occasionally !ass throu h each other= !erha!s they are like lon snakes which curl around each other, here and there flattenin out, like cobras1 heads= !erha!s they are like s!iderwebs= or !erha!s they are circuits in which si nals travel round and round in sha!es stran er than the dash of a nat-hun ry swallow& There is no tellin & It is even !ossible that these modules are software, rather than hardware, !henomenabut this is somethin which we will discuss later There are many questions that come to mind concernin these hy!othesi<ed neural com!le(es& 'or instance, :o they e(tend into the lower re ions of the brain, such as the midbrain, the hy!othalamus, etc&? 7an a sin le neuron belon to more than one such com!le(? To how many such com!le(es can a sin le neuron belon ? 0y how many neurons can such com!le(es overla!? )re these com!le(es !retty much the same for everybody? )re corres!ondin ones found in corres!ondin !laces in different !eo!le1s brains? :o they overla! in the same way in everybody1s brain? Philoso!hically, the most im!ortant question of all is this, 3hat would the e(istence of modulesfor instance, a randmother moduletell us? 3ould this ive us any insi ht into the !henomenon of our own consciousness? Er would it still leave us as much in the dark about what consciousness is, as does knowled e that a brain is built out of neurons and lia? )s you mi ht uess from readin the )nt 'u ue, my feelin is that it would o a lon way towards ivin us an understandin of the !henomenon of consciousness& The crucial ste! that needs to be taken is from a low-level-neuron-by-neuron-descri!tion of the state of a brain, to a hi h-level-module-by-module-descri!tion of the same state of the same brain& Er, to revert to the su estive terminolo y of the *nt 7ugue, we want to shift the descri!tion of the brain state from the si nal level to the symbol level&

Acti"e Symbols
Aet us from now on refer to these hy!othetical neural com!le(es, neural modules, neural !ackets, neural networks, multineuron unitscall them what you will, whether they come in the form of !ancakes, arden rakes, rattlesnakes, snowflakes, or even ri!!les on lakesas symbols& ) descri!tion of a brain state in terms of symbols was alluded to in the :ialo ue& 3hat would such a descri!tion be like? 3hat kinds of conce!ts is it reasonable to think actually mi ht be -symboli<ed-? 3hat kinds of interrelations would symbols have? )nd what insi hts would this whole !icture !rovide into consciousness? The first thin to em!hasi<e is that symbols can be either dormant, or awake 4activated5& )n active symbol is one which has been tri eredthat is, one in which a threshold number of neurons have been caused to fire by stimuli comin from outside& Since a symbol can be tri ered in many different ways, it can act in many different ways

when awakened& This su ests that we should think of a symbol not as a fi(ed entity, but as a variable entity& Therefore it would not suffice to describe a brain state by sayin -Symbols ), 0,&&&, 8 are all active-= rather, we would have to su!!ly in addition a set of !arameters for each active symbol, characteri<in some as!ects of the symbol1s internal workin s& It is an interestin question whether in each symbol there are certain core neurons, which invariably fire when the symbol is activated& If such a core set of neurons e(ists, we mi ht refer to it as the -invariant core- of the symbol& It is tem!tin to assume that each time you think of, say, a waterfall, some fi(ed neural !rocess is re!eated, without doubt embellished in different ways de!endin on the conte(t, but reliably occurrin & ?owever, it is not clear that this must be so& 8ow what does a symbol do, when awakened? ) low-level descri!tion would say, -/any of its neurons fire&- 0ut this no lon er interests us& The hi h-level descri!tion should eliminate all reference to neurons, and concentrate e(clusively on symbols& So a hi h-level descri!tion of what makes a symbol active, as distin uished from dormant, would be, -It sends out messa es, or si nals, whose !ur!ose is to try to awaken, or tri er, other symbols&- Ef course these messa es would be carried as streams of nerve im!ulses, by neuronsbut to the e(tent that we can avoid such !hraseolo y, we should, for it re!resents a low-level way of lookin at thin s, and we ho!e that we can et alon on !urely a hi h level& In other words, we ho!e that thou ht !rocesses can be thou ht of as bein sealed off from neural events in the same way that the behavior of a clock is sealed off from the laws of quantum mechanics, or the biolo y of cells is sealed off from the laws of quarks& 0ut what is the advanta e of this hi h-level !icture? 3hy is it better to say, -Symbols ) and 0 tri ered symbol 7- than to say, -8eurons "*% throu h D"B e(cited neuron $L and caused it to fire-? This question was answered in the *nt 7ugue, It is better because symbols s)m(oli@e thin s, and neurons don1t& Symbols are the hardware reali<ations of conce!ts& 3hereas a rou! of neurons tri erin another neuron corres!onds to no outer event, the tri erin of some symbol by other symbols bears a relation to events in the real worldor in an ima inary world& Symbols are related to each other by the messa es which they can send back and forth, in such a way that their tri erin !atterns are very much like the lar e-scale events which do ha!!en in our world, or could ha!!en in a world similar to ours& In essence, meanin arises here for the same reason as it did in the !q-systemisomor!hism= only here, the isomor!hism is infinitely more com!le(, subtle, delicate, versatile, and intensional& Incidentally, the requirement that symbols should be able to !ass so!histicated messa es to and fro is !robably sufficient to e(clude neurons themselves from !layin the role of symbols& Since a neuron has only a sin le way of sendin information out of itself, and has no way of selectively selectin a si nal now in one direction, now in another, it sim!ly does not have the kind of selective tri erin !ower which a symbol must have to act like an ob.ect in the real world& In his book The 3nsect -ocieties, E& E& 3ilson makes a similar !oint about how messa es !ro!a ate around inside ant colonies,
U/ass communicationV is defined as the transfer, amon individual could not !ass to another&% rou!s, of information that a sin le

It is not such a bad ima e, the brain as an ant colony2 The ne(t questionand an e(tremely im!ortant one it is, tooconcerns the nature and -si<e- of the conce!ts which are re!resented in the brain by sin le symbols& )bout the nature of symbols there are questions like this, 3ould there be a symbol for the eneral notion of waterfalls, or would there be different symbols for various s!ecific waterfalls? Er would both of these alternatives be reali<ed? )bout the -si<e- of symbols, there are questions like this, 3ould there be a symbol for an entire story? Er for a melody? Er a .oke? Er is it more likely that there would only be symbols for conce!ts rou hly the si<e of words, and that lar er ideas, such as !hrases or sentences, would be re!resented by concurrent or sequential activation of various symbols? Aet us consider the issue of the si<e of conce!ts re!resented by symbols& /ost thou hts e(!ressed in sentences are made u! out of basic, quasi-atomic com!onents which we do not usually analy<e further& These are of word si<e, rou hlysometimes a little lon er, sometimes a little shorter& 'or instance, the noun -waterfall-, the !ro!er noun -8ia ara 'alls-, the !ast-tense suffi( --ed-, the verb -to catch u! with-, and lon er idiomatic !hrases are all close to atomic& These are ty!ical elementary brush strokes which we use in !aintin !ortraits of more com!le( conce!ts, such as the !lot of a movie, the flavor of a city, the nature of consciousness, etc& Such com!le( ideas are not sin le brush strokes& It seems reasonable to think that the brush strokes of lan ua e are also brush strokes of thou ht, and therefore that symbols re!resent conce!ts of about this si<e& Thus a symbol would be rou hly somethin for which you know a word or stock !hrase, or with which you associate a !ro!er name& )nd the re!resentation in the brain of a more com!le( idea, such as a !roblem in a love affair, would be a very com!licated sequence of activations of various symbols by other symbols&

Classes and Instances

There is a eneral distinction concernin thinkin , that between cate ories and individuals, or classes and instances& 4Two other terms sometimes used are -ty!es- and -tokens-&5 It mi ht seem at first si ht that a iven symbol would inherently be either a symbol for a class or a symbol for an instancebut that is an oversim!lification& )ctually, most symbols may !lay either role, de!endin on the conte(t of their activation& 'or e(am!le, look at the list below, 4"5 a !ublication 4B5 a news!a!er 4%5 The San 'rancisco 7hronicle 4C5 the /ay "* edition of the 7hronicle 4L5 my co!y of the /ay "* edition of the 7hronicle 4D5 my co!y of the /ay "* edition of the 7hronicle as it was when I first !icked it u! 4as contrasted with my co!y as it was a few days later, in my fire!lace, burnin 5

?ere, lines B to L all !lay both roles& Thus, line C is an instance of of the eneral class of line %, and line L is an instance of line C& Aine D is a s!ecial kind of instance of a class, a manifestation& The successive sta es of an ob.ect durin its life history are its manifestations& It is interestin to wonder if the cows on a farm !erceive the invariant individual underneath all the manifestations of the .olly farmer who feeds them hay&

The ,rototy!e ,rinci!le

The list above seems to be a hierarchy of eneralitythe to! bein a very broad conce!tual cate ory, the bottom some very humble !articular thin located in s!ace and time& ?owever, the idea that a -class- must always be enormously broad and abstract is far too limited& The reason is that our thou ht makes use of an in enious !rinci!le, which mi ht be called the protot)pe principle, The most s!ecific event can serve as a eneral e(am!le of a class of events& Everyone knows that s!ecific events have a vividness which im!rints them so stron ly on the memory that they can later be used as models for other events which are like them in some way& Thus in each s!ecific event, there is the erm of a whole class of similar events& This idea that there is enerality in the s!ecific is of far-reachin im!ortance& 8ow it is natural to ask, :o the symbols in the brain re!resent classes, or instances? )re there certain symbols which re!resent only classes, while other symbols re!resent only instances? Er can a sin le symbol serve duty either as a class symbol or instance symbol, de!endin which !arts of it are activated? The latter theory seems a!!ealin = one mi ht think that a -li ht- activation of a symbol mi ht re!resent a class, and that a dee!er, or more com!le(, activation would contain more detailed internal neural firin !atterns, and hence would re!resent an instance& 0ut on second thou ht, this is cra<y, it would im!ly, for e(am!le, that by activatin the symbol for -!ublication- in a sufficiently com!le( way, you would et the very com!le( symbol which re!resents a s!ecific news!a!er burnin in my fire!lace& )nd very other !ossible manifestation of every other !iece of !rinted matter would be re!resented internally by some manner of activatin the sin le symbol for -!ublication-& That seems much too heavy a burden to !lace on to sin le symbol -!ublication-& Ene must conclude, therefore, that finance symbols can e(ist side by side with class symbols, and are not .ust modes of activation of the latter&

The S!litting-off of Instances from Classes

En the other hand, instance symbols often inherit many of their !ro!erties from the classes to which those instances belon & If I tell you I went to see a movie, you will be in -mintin - a fresh new instance symbol for that !articular movie= but in the absence of more information, the new instance symbol will have to lean rather heavily on your !ree(istin class symbol for -movie-& Fnconsciously, you will rely on a host of !resu!!ositions about at moviefor e(am!le, that it lasted between one and three hours,

that it was shown in a local theater, that it told a story about some !eo!le, and so on& These are built into the class symbol as e(!ected links to other symbols 4i&e&, !otential tri erin relations5, and are called default options& In any freshly minted instance symbol, the default o!tions can easily be overridden, but unless this is e(!licitly done, they will remain in the instance symbol, inherited from its class symbol& Fntil they are overridden, they !rovide some !reliminary basis for you to think about the new instance for e(am!le, the movie I went to seeby usin the reasonable uesses which are su!!lied by the -stereoty!e-, or class symbol& ) fresh and sim!le instance is like a child without its own ideas or e(!eriencesit relies entirely on its !arents1 e(!eriences and o!inions and .ust !arrots them& 0ut radually, as it interacts more and more with the rest of the world, the child acquires its own idiosyncratic e(!eriences and inevitably be ins to s!lit away from the !arents& Eventually, the child becomes a full-fled ed adult& In the same way, a fresh instance can s!lit off from its !arent class over a !eriod of time, and become a class, or !rototy!e, in its own ri ht& 'or a ra!hic illustration of such a s!littin -off !rocess, su!!ose that some Saturday afternoon you turn on your car radio, and ha!!en to tune in on a football ame between two -random- teams& )t first you do not know the names of the !layers on either team& )ll you re ister, when the announcer says, -Palindromi made the sto! on the twentyseven yard line, and that brin s u! fourth down and si( to o,- is that some !layer sto!!ed some other !layer& Thus it is a case of activation of the class symbol -football !layer-, with some sort of coordinated activation of the symbol for tacklin & 0ut then as Palindromi fi ures in a few more key !lays, you be in buildin u! a fresh instance symbol for him in !articular, usin his name, !erha!s, as a focal !oint& This symbol is de!endent, like a child, on the class symbol for -football !layer-, most of your ima e of Palindromi is su!!lied by your stereoty!e of a football !layer as contained in the -football !layer- symbol& 0ut radually, as more information comes to you, the -Palindromi- symbol becomes more autonomous, and relies less and less on concurrent activation of its !arent class symbol& This may ha!!en in a few minutes, as Palindromi makes a few ood !lays and stands out& ?is teammates may still all be re!resented by activations of the class symbol, however& Eventually, !erha!s after a few days, when you have read some articles in the s!orts section of your !a!er, the umbilical cord is broken, and Palindromi can stand on his own two feet& 8ow you know such thin s as his home town and his ma.or in colle e= you reco ni<e his face= and so on& )t this !oint, Palindromi is no lon er conceived of merely as a football !layer, but as a human bein who ha!!ens also to be a football !layer& -Palindromi- is an instance symbol which can become active while its !arent class symbol 4football !layer5 remains dormant& Ence, the Palindromi symbol was a satellite orbitin around its mother symbol, like an artificial satellite circlin the Earth, which is so much bi er and more massive& Then there came an intermediate sta e, where one symbol was more im!ortant than the other, but they could be seen as orbitin around each othersomethin like the Earth and the /oon& 'inally, the new symbol becomes quite autonomous= now it mi ht easily serve as a class symbol around which could start rotatin new satellitessymbols for other !eo!le who are less familiar but who have somethin in common with Palindromi, and for

whom he can serve as a tem!orary stereoty!e, until you acquire more information, enablin the new symbols so to become autonomous&

The

ifficulty of

isentangling Symbols from Each 0ther

These sta es of rowth and eventual detachment of an instance from a class will be distin uishable from each other by the way in which the symbols involved are linked& Sometimes it will no doubt be very difficult to tell .ust where one symbol leaves off and the other one be ins& ?ow -active- the one symbol, com!ared to the other? If one can be activated inde!endently of the other, then it would be quite sensible to call them autonomous& 3e have used an astronomy meta!hor above, and it is interestin that the !roblem of the motion of !lanets is an e(tremely com!le( onein fact the eneral !roblem of three ravitationally interactin bodies 4such as the Earth, /oon, and Sun5 is far from solved, even after several centuries of work& Ene situation in which it is !ossible to obtain ood a!!ro(imate solutions, however, is when one body is much more massive than the other two 4here, the Sun5= then it makes sense to consider that body as stationary, with the other two rotatin about it, on to! of this can finally be added the interaction between the two satellites& 0ut this a!!ro(imation de!ends on breakin u! the system into the Sun, and a -cluster-, the Earth-/oon system& This is an a!!ro(imation, but it enables the system to be understood quite dee!ly& So to what e(tent is this cluster a !art of reality, and to what e(tent is it a mental fabrication, a human im!osition of structure on the universe? This !roblem of the -reality- of boundaries drawn between what are !erceived to be autonomous or semi-autonomous clusters will create endless trouble when we relate it to symbols in the brain& Ene reatly !u<<lin question is the sim!le issue of !lurals& ?ow do we visuali<e, say, three do s in a teacu!? Er several !eo!le in an elevator? :o we be in with the class symbol for -do - and then rub three -co!ies- off of it? That is, do we manufacture three fresh instance symbols usin the class symbol -do - as tem!late? Er do we .ointly activate the symbols -three- and -do -? 0y addin more or less detail to the scene bein ima ined, either theory becomes hard to maintain& 'or instance, we certainly do not have a se!arate instance symbol for each nose, mustache, rain of salt, etc&, that we have ever seen& 3e let class symbols take care of such numerous items, and when we !ass !eo!le on the street who have mustaches, we somehow .ust activate the -mustache- class symbol, without mintin fresh instance symbols, unless we scrutini<e them carefully& En the other hand, once we be in to distin uish individuals, we cannot rely on a sin le class symbol 4e& &, -!erson-5 to timeshare itself amon all the different !eo!le& 7learly there must come into e(istence se!arate stance symbols for individual !eo!le& It would be ridiculous to ima ine that this feat could be accom!lished by 1.u lin -that is, by the sin le class symbol flittin back and forth between several different modes of activation 4one for each !erson5& 0etween the e(tremes, there must be room for many sorts of intermediate cases& There may be a whole hierarchy of ways of creatin the class-instance distinction in the brain, ivin rise to symbolsand symbol or ani<ationsof varyin de rees of

s!ecificity& The followin different kinds of individual and .oint activation of symbols mi ht be res!onsible for mental ima es of various de rees of s!ecificity, 4"5 various different modes or de!ths of activation of a sin le class symbol, 4B5 simultaneous activation of several class symbols in some in some coordinated manner, 4%5 activation of a sin le instance symbol, 4C5 activation of a sin le instance symbol in con.unction with activation of several class symbols, 4L5 simultaneous activation of several instance symbols and several class symbols in some coordinated manner& This brin s us ri ht hack to the question, -3hen is a symbol a distin uishable subsystem of the brain?- 'or instance, consider the second e(am!lesimultaneous activation of several class symbols in some coordinated manner& This could easily be what ha!!ens when -!iano sonata- is the conce!t under consideration 4the symbols for -!iano- and -sonata- bein at least two of the activated symbols5& 0ut if this !air of symbols ets activated in con.unction often enou h, it is reasonable to assume that the link between them will become stron enou h that they will act as a unit, when activated to ether in the !ro!er way& So two or more symbols can act as one, under the !ro!er conditions, which means that the !roblem of enumeratin the number of symbols in the brain is trickier than one mi ht uess& Sometimes conditions can arise where two !reviously unlinked symbols et activated simultaneously and in a coordinated fashion& They may fit to ether so well that it seems like an inevitable union, and a sin le new symbol is formed by the ti ht interaction of the two old symbols& If this ha!!ens, would it be fair to say that the new symbol -always had been there but never had been activated-or should one say that it has been -created-? In case this sounds too abstract, let us take a concrete e(am!le, the :ialo ue 8ra( 8anon& In the invention of this :ialo ue, two e(istin symbolsthat for -musical crab canon-, and that for -verbal dialo ue- had to be activated simultaneously and in some way forced to interact& Ence this was done, the rest was quite inevitable, a new symbol a class symbolwas born from the interaction of these two, and from then on it was able to be activated on its own& 8ow had it always been a dormant symbol in my brain? If so, then it must have also been a dormant symbol in the brain of every human who ever had its com!onent symbols, even if it never was awakened in them& This would mean that to enumerate the symbols in anyone1s brain, one would have to count all dormant symbols all !ossible combinations and !ermutations of all ty!es of activations of all known symbols& This would even include those fantastic creatures of software that one1s brain invents when one is aslee!the stran e mi(tures of ideas which wake u! when their host oes to slee!&&& The e(istence of these -!otential symbols- shows that it is really a hu e oversim!lification to ima ine that the brain is a well-defined collection of symbols in well-defined states of activation& It is much harder than that to !in down a brain state on the symbol level&

Symbols Soft'are or /ard'are7

3ith the enormous and ever- rowin re!ertoire of symbols that e(ist in each brain, you mi ht wonder whether there eventually comes a !oint when the brain is saturatedwhen there is .ust no more room for a new symbol& This would come about, !resumably, if symbols never overla!!ed each otherif a iven neuron never served a double function, so that symbols would be like !eo!le ettin into an elevator& -3arnin , This brain has a ma(imum ca!acity of %L+,B$L symbols2This is not a necessary feature of the symbol model of brain function, however& In fact, overla!!in and com!letely tan led symbols are !robably the rule, so that each neuron, far from bein a member of a unique symbol, is !robably a functionin !art of hundreds of symbols& This ets a little disturbin , because if it is true, then mi ht it not .ust as easily be the case that each neuron is !art of every sin le symbol? If that were so, then there would be no locali<ability whatsoever of symbolsevery symbol would be identified with the whole of the brain& This would account for results like Aashley1s corte( removal in ratsbut it would also mean abandonment of our ori inal idea of breakin the brain u! into !hysically distinct subsystems& Eur earlier characteri<ation of symbols as -hardware reali<ations of conce!ts- could at best be a reat oversim!lification& In fact, if every symbol were made u! of the same com!onent neurons as every other symbol, then what sense would it make to s!eak of distinct symbols at all? 3hat would be the si nature of a iven symbol1s activationthat is, how could the activation of symbol ) be distin uished from the activation of symbol 0? 3ouldn1t our whole theory o down the drain? )nd even if there is not a total overla! of symbols, is our theory not more and more difficult to maintain, the more that symbols do overla!? 4Ene !ossible way of !ortrayin overla!!in symbols is shown in 'i ure D*&5 There is a way to kee! a theory based on symbols even if !hysically, they overla! considerably or totally& 7onsider the surface of a !ond, which can su!!ort many different ty!es of waves or ri!!les& The hardware namely the water itselfis the same in all cases, but it !ossesses different !ossible modes of e(citation& Such software e(citations of the same hardware can all be distin uished from each other& 0y this analo y, I do not mean to o so far as to su est that all the different symbols are .ust different kinds of -waves!ro!a atin throu h a uniform neural medium which admits of no meanin ful division into !hysically distinct symbols& 0ut it may be that in order to distin uish one symbol1s activation from that of another symbol, a !rocess must be carried out which involves not only locatin the neurons which are firin , but also identifyin very !recise details of the timin of the firin of those neurons& That is, which neuron !receded which other neuron, and by how much? ?ow many times a second was a !articular neuron firin ? Thus !erha!s several symbols can coe(ist in the same set of neurons by havin different characteristic neural firin !atterns& The difference between a theory havin !hysically distinct symbols, and a theory havin overla!!in symbols which are distin uished from each other by modes of e(citation, is that the former ives hardware reali<ations of conce!ts, while the latter ives !artly hardware, !artly software reali<ations of conce!ts&

73G5=E JE! 3n this schematic diagram, neurons are imagined as laid out as dots in one plane! Two o#erlapping neural pathwa)s are shown in different shades of gra)! 3t ma) happen that two independent "neural flashes" simultaneousl) race down these two pathwa)s, passing through one another like two ripples on a pond's surface as in 7ig! GF/! This is illustrati#e of the idea of two "acti#e s)m(ols" which share neurons and which ma) e#en (e simultaneousl) acti#ated! U7rom ?ohn 8! Eccles, 'acin ;eality New York: -pringer <erlag, BICK/, p!FB!V

Liftability of Intelligence
Thus we are left with two basic !roblems in the unravelin of thou ht !rocesses, as they take !lace in the brain& Ene is to e(!lain how the low-level traffic of neuron firin s ives rise to the hi h-level traffic of symbol activations& The other is to e(!lain the hi h-level traffic of symbol activation in its own termsto make a theory which does not talk about the low-level neural events& If this latter is !ossibleand it is a key assum!tion at the basis of all !resent research into )rtificial Intelli encethen intelli ence can be reali<ed in other ty!es of hardware than brains& Then intelli ence will have been shown to be a !ro!erty that can be -lifted- ri ht out of the hardware in which it residesor in other words, intelli ence will be a software !ro!erty& This will mean that the !henomena of consciousness and intelli ence are indeed hi h-level in the same sense as most other com!le( !henomena of nature, they have their own hi h-level laws which de!end on, yet are -liftable- out of, the lower levels& If, on the other hand, there is absolutely no way to

73G5=E JI! The construction of an arch () workers of the termite /acrotermes belosus! Each column is (uilt up () the addition of pellets of soil and e9crement! :n the outer part of the left column a worker is seen depositing a round fecal pellet! :ther workers, ha#ing carried pellets in their mandi(les up the columns, are now placing them at the growing ends of the columns! >hen a column reaches a certain height the termites, e#identl) guided () odor, (egin to e9tend it at an angle in the direction of a neigh(oring column! * completed arch is shown in the (ackground! U.rawing () Turid ,olldo(lerR from E! K! >ilson, The Insect Societies 8am(ridge, 'ass!: ,ar#ard 5ni#ersit) &ress, BICB/, p! FHKV

reali<e symbol-tri erin !atterns without havin all the hardware of neurons 4or simulated neurons5, this will im!ly that intelli ence is a brain-bound !henomenon, and much more difficult to unravel than one which owes its e(istence to a hierarchy of laws on several different levels& ?ere we come back to the mysterious collective behavior of ant colonies, which can build hu e and intricate nests, des!ite the fact that the rou hly "++,+++ neurons of an ant brain almost certainly do not carry any information about nest structure& ?ow, then, does the nest et created? 3here does the information reside? In !articular, !onder where the information describin an arch such as is shown in 'i ure D# can be found& Somehow, it must be s!read about in the colony, in the caste distribution, the a e distributionand !robably lar ely in the !hysical !ro!erties of the ant-body itself& That is, the interaction between ants is determined .ust as much by their si(-le edness and their si<e and so on, as by the information stored in their brain& 7ould there be an )rtificial )nt 7olony?

Can 0ne Symbol Be Isolated7

Is it !ossible that one sin le symbol could be awakened in isolation from all others? Probably not& Just as ob.ects in the world always e(ist in a conte(t of other ob.ects, so symbols are always connected to a constellation of other symbols& This does not necessarily mean that symbols can never be disentan led from each other& To make a rather sim!le analo y, males and females always arise in a s!ecies to ether, their roles are com!letely intertwined, and yet this does not mean that a male cannot be distin uished from a female& Each is reflected in the other, as the beads in Indra1s net reflect each other& The recursive intertwinin of the functions '4n5 and /4n5 in 7ha!ter G does not !revent each function from havin its own characteristics& The intertwinin of ' and / could be mirrored in a !air of ;T81s which call each other& 'rom this we can .um! to a whole network of )T81s intertwined with each othera heterarchy of interactin recursive !rocedures& ?ere, the meshin is so inherent that no one )T8 could be activated in isolation= yet its activation may be com!letely distinctive, not confusable with that of any other of the )T81s& It is not such a bad ima e, the brain as an )T8colony2 Aikewise, symbols, with all their multi!le links to each other, are meshed to ether and yet ou ht to be able to be teased a!art& This mi ht involve identifyin a neural network, a network !lus a mode of e(citationor !ossibly somethin of a com!letely different kind& In any case, if symbols are !art of reality, !resumably there e(ists a natural way to chart them out in a real brain& ?owever, if some symbols were finally identified in a brain, this would not mean that any one of them could be awakened in isolation& The fact that a symbol cannot be awakened in isolation does not diminish the se!arate identity of the symbol= in fact, quite to the contrary, a symbol1s identity lies !recisely in its ways of bein connected 4via !otential tri erin links5 to other symbols& The network by which symbols can !otentially tri er each other constitutes the brain1s workin model of the real universe, as well as of the alternate universes which it considers 4and which are every bit as im!ortant for the individual1s survival in the real world as the real world is5&

The Symbols of Insects

Eur facility for makin instances out of classes and classes out of instances lies at the basis of our intelli ence, and it is one of the reat differences between human thou ht and the thou ht !rocesses of other animals& 8ot that I have ever belon ed to another s!ecies and e(!erienced at first hand how it feels to think their waybut from the outside it is a!!arent that no other s!ecies forms eneral conce!ts as we do, or ima ines hy!othetical worldsvariants on the world as it is, which aid in fi urin out which future !athway to choose& 'or instance, consider the celebrated -lan ua e of the bees-information-laden dances which are !erformed by worker bees returnin to the hive, to inform other bees of the location of nectar& 3hile there may be in each bee a set of rudimentary symbols which are activated by such a dance, there is no reason to believe that a bee has an e(!andable vocabulary of symbols& 0ees and other insects do not seem to have the !ower

to enerali<ethat is, to develo! new class symbols from instances which we would !erceive as nearly identical& ) classic e(!eriment with solitary was!s is re!orted in :ean 3ooldrid e1s book, /echanical /an, from which I quote,
3hen the time comes for e layin , the was! S!he( builds a burrow for the !ur!ose and seeks out a cricket which she stin s in such a way as to !araly<e but not kill it& She dra s the cricket into the burrow, lays her e s alon side, closes the burrow, then flies away, never to return& In due course, the e s hatch and the was! rubs feed off the !araly<ed cricket, which has not decayed, havin been ke!t in the was! equivalent of a dee!free<e& To the human mind, such an elaborately or ani<ed and seemin ly !ur!oseful routine conveys a convincin flavor of lo ic and thou htfulnessuntil more details are e(amined& 'or e(am!le, the was!1s routine is to brin the !araly<ed cricket to the burrow, leave it on the threshold, o inside to see that all is well, emer e, and then dra the cricket in& If the cricket is moved a few inches away while the was! is inside makin her !reliminary ins!ection, the was!, on emer in from the burrow, will brin the cricket back to the threshold, but not inside, and will then re!eat the !re!aratory !rocedure of enterin the burrow to see that everythin is all ri ht& If a ain the cricket is removed a few inches while the was! is inside, once a ain she will move the cricket u! to the threshold and reenter the burrow for a final check& The was! never thinks of !ullin the cricket strai ht in& En one occasion this !rocedure was re!eated forty times, always with the same result& C

This seems to be com!letely hard-wired behavior& 8ow in the was! brain, there may be rudimentary symbols, ca!able of tri erin each other= but there is nothin like the human ca!acity to see several instances as instances of an as-yet-unformed class, and then to make the class symbol= nor is there anythin like the human ability to wonder, -3hat if I did thiswhat would ensue in that hy!othetical world?- This ty!e of thou ht !rocess requires an ability to manufacture instances and to mani!ulate them as if they were symbols standin for ob.ects in a real situation, althou h that situation may not be the case, and may never be the case&

Class Symbols and Imaginary >orlds

Aet us reconsider the )!ril 'ool1s .oke about the borrowed car, and the ima es con.ured u! in your mind durin the tele!hone call& To be in with, you need to activate symbols which re!resent a road, a car, a !erson in a car& 8ow the conce!t -road- is a very eneral one, with !erha!s several stock sam!les which you can unconsciously !ull out of dormant memory when the occasion arises& -;oad- is a class, rather than an instance& )s you listen to the tale, you quickly activate symbols which are instances with radually increasin s!ecificity& 'or instance, when you learn that the road was wet, this con.ures u! a more s!ecific ima e, thou h you reali<e that it is most likely quite different from the actual road where the incident took !lace& 0ut that is not im!ortant= what matters is whether your symbol is sufficiently well suited for the storythat is, whether the symbols which it can tri er are the ri ht kind& )s the story !ro resses, you fill in more as!ects of this road, there is a hi h bank a ainst which a car could smash& 8ow does this mean that you are activatin the symbol for -bank-, or does it mean that you are settin some !arameters in your symbol for -road-& Fndoubtedly both& That is, the network of neurons which re!resents -road- has

many different ways of firin , and you are selectin which subnetwork actually shall fire& )t the same time, you are activatin the symbol for -bank-, and this is !robably instrumental in the !rocess of selectin the !arameters for& -road-, in that its neurons may send si nals to some of those in -road-and vice versa& 4In case this seems a little confusin , it is because I am somewhat straddlin levels of descri!tionI am tryin to set u! an ima e of the symbols, as well as of their com!onent neurons&5 8o less im!ortant than the nouns are the verbs, !re!ositions, etc, They, too, activate symbols, which send messa es back and forth to each other& There are characteristic differences between the kinds of tri erin !atterns of symbols for verbs and symbols for nouns, of course, which means that they may be !hysically somewhat differently or ani<ed& 'or instance, nouns mi ht have fairly locali<ed symbols, while verbs and !re!ositions mi ht have symbols with many -tentacles- reachin all around the corte(= or any number of other !ossibilities& )fter the story is all over, you learn it was all untrue& The !ower of -rubbin offinstances from classes, in the way that one makes rubbin s from brasses in churches, has enabled you to re!resent the situation, and has freed you from the need to remain faithful to the real world& The fact that symbols can act as tem!lates for other symbols ives you some mental inde!endence of reality, you can create artificial universes, in which there can ha!!en nonreal events with any amount of detail that you care to imbue them with& 0ut the class symbols themselves, from which all of this richness s!rin s, are dee!ly rounded in reality& Fsually symbols !lay isomor!hic roles to events which seem like they could ha!!en, althou h sometimes symbols are activated which re!resent situations which could not ha!!enfor e(am!le, watches si<<lin , tubas layin e s, etc& The borderline between what could and what could not ha!!en is an e(tremely fu<<y one& )s we ima ine a hy!othetical event, we brin certain symbols into active statesand de!endin on how well they interact 4which is !resumably reflected in our comfort in continuin the train of thou ht5, we say the event -could- or -could not- ha!!en& Thus the terms -could- and -could not- are e(tremely sub.ective& )ctually, there is a ood deal of a reement amon !eo!le about which events could or could not ha!!en& This reflects the reat amount of mental structure which we all sharebut there is a borderline area where the sub.ective as!ect of what kinds of hy!othetical worlds we are willin to entertain is a!!arent& ) careful study of the kinds of ima inary events that !eo!le consider could and could not ha!!en would yield much insi ht into the tri erin !atterns of the symbols by which !eo!le think&

Intuiti"e La's of ,hysics

3hen the story has been com!letely told, you have built u! quite an elaborate mental model of a scene, and in this model all the ob.ects obey !hysical law& This means that !hysical law itself must be im!licitly !resent in the tri erin !atterns of the symbols& Ef course, the !hrase -!hysical law- here does not mean -the laws of !hysics as e(!ounded by a !hysicist-, but rather the intuitive, chunked laws which all of us have to have in our minds in order to survive&

) curious sideli ht is that one can voluntarily manufacture mental sequences of events which violate !hysical law, if one so desires& 'or instance, if I but su est that you ima ine a scene with two cars a!!roachin each other and then !assin ri ht throu h each other, you won1t have any trouble doin so& The intuitive !hysical laws can be overridden by ima inary laws of !hysics= but how this overridin is done, how such sequences of ima es are manufacturedindeed what any one visual ima e isall of these are dee!ly cloaked mysteriesinaccessible !ieces of knowled e& 8eedless to say, we have in our brains chunked laws not only of how inanimate ob.ects act, but also of how !lants, animals, !eo!le and societies actin other words, chunked laws of biolo y, !sycholo y, sociolo y, and so on& )ll of the internal re!resentations of such entities involve the inevitable feature of chunked models, determinism is sacrificed for sim!licity& Eur re!resentation of reality ends u! bein able only to !redict !robabilities of endin u! in certain !arts of abstract s!aces of behavior not to !redict anythin with the !recision of !hysics&

,rocedural and

eclarati"e Kno'ledge

) distinction which is made in )rtificial Intelli ence is that between !rocedural and declarative ty!es of knowled e& ) !iece of knowled e is said to be declarative if it is stored e(!licitly, so that not only the !ro rammer but also the !ro ram can -read- it as if it were in an encyclo!edia or an almanac& This usually means that it is encoded locally, not s!read around& 0y contrast, !rocedural knowled e is not encoded as factsonly as !ro rams& ) !ro rammer may be able to !eer in and say, -I see that because of these !rocedures here, the !ro ram 1knows1 how to write En lish sentences-but the !ro ram itself may have no e(!licit awareness of how it writes those sentences& 'or instance, its vocabulary may include none of the words -En lish-, -sentence-, and -write- at all2 Thus !rocedural knowled e is usually s!read around in !ieces, and you can1t retrieve it, or -key- on it& It is a lobal consequence of how the !ro ram works, not a local detail& In other words, a !iece of !urely !rocedural knowled e is an e!i!henomenon& In most !eo!le there coe(ists, alon with a !owerful !rocedural re!resentation of the rammar of their native lan ua e, a weaker declarative re!resentation of it& The two may easily be in conflict, so that a native s!eaker will often instruct a forei ner to say thin s he himself would never say, but which a ree with the declarative -book learnin he acquired in school sometime& The intuitive or chunked laws of !hysics and other disci!lines mentioned earlier fall mainly on the !rocedural side= the knowled e that an octo!us has ei ht tentacles falls mainly on the declarative side& In between the declarative and !rocedural e(tremes, there are all !ossible shades& 7onsider the recall of a melody& Is the melody stored in your brain, note by note? 7ould a sur eon e(tract a windin neural filament from your brain, then stretch it strai ht, and finally !roceed to !in!oint alon it the successively stored notes, almost as if it were a !iece of ma netic ta!e? If so, then melodies are stored declaratively& Er is the recall of a melody mediated by the interaction of a lar e number of symbols, some of which re!resent tonal relationshi!s, others of which re!resent emotional qualities, others of which re!resent rhythmic devices, and so on? If so, then melodies are stored

!rocedurally& In reality, there is !robably a mi(ture of these e(tremes in the way a melody is stored and recalled& It is interestin that, in !ullin a melody out of memory, most !eo!le do not discriminate as to key, so that they are as likely to sin -?a!!y 0irthday- in the key of 'shar! as in the key of 7& This indicates that tone relationshi!s, rather than absolute tones, are stored& 0ut there is no reason that tone relationshi!s could not be stored quite declaratively& En the other hand, some melodies are very easy to memori<e, whereas others are e(tremely elusive& If it were .ust a matter of storin successive notes, any melody could be stored as easily as any other& The fact that some melodies are catchy and others are not seems to indicate that the brain has a certain re!ertoire of familiar !atterns which are activated as the melody is heard& So, to -!lay back- the melody, those !atterns would have to be activated in the same order& This returns us to the conce!t of symbols tri erin one another, rather than a sim!le linear sequence of declaratively stored notes or tone relationshi!s& ?ow does the brain know whether a !iece of knowled e is stored declaratively? 'or instance, su!!ose you are asked, -3hat is the !o!ulation of 7hica o?- Somehow the number five million s!rin s to mind, without your wonderin , -6ee, how would I o about countin them all?- 8ow su!!ose I ask you, -?ow many chairs are there in your livin room?- ?ere, the o!!osite ha!!ensinstead of tryin to dred e the answer out of a mental almanac, you immediately either o to the room and count the chairs, or you manufacture the room in your head and count the chairs in the ima e of the room& The questions were of a sin le ty!e-how many?-yet one of them caused a !iece of declarative knowled e to be fetched, while the other one caused a !rocedural method of findin the answer to be invoked& This is one e(am!le where it is clear that you have knowled e about how you classify your own knowled e= and what is more, some of that metaknowledge may itself be stored !rocedurally, so that it is used without your even bein aware of how it is done&

4isual Imagery
Ene of the most remarkable and difficult-to-describe qualities of consciousness is visual ima ery& ?ow do we create a visual ima e of our livin room? Ef a roarin mountain brook? Ef an oran e? Even more mysterious, how do we manufacture ima es unconsciously, ima es which uide our thou hts, ivin them !ower and color and de!th? 'rom what store are they fetched? 3hat ma ic allows us to mesh two or three ima es, hardly ivin a thou ht as to how we should do it? @nowled e of how to do this is amon the most !rocedural of all, for we have almost no insi ht into what mental ima ery is& It may be that ima ery is based on our ability to su!!ress motor activity& 0y this, I mean the followin & If you ima ine an oran e, there may occur in your corte( a set of commands to !ick it u!, to smell it, to ins!ect it, and so on& 7learly these commands cannot be carried out, because the oran e is not there& 0ut they can be sent alon the usual channels towards the cerebellum or other subor ans of the brain, until, at some critical !oint, a -mental faucet- is closed, !reventin them from actually bein carried out& :e!endin on how far down the line this -faucet- is situated, the ima es may be more or less vivid and realseemin & )n er can cause us to ima ine quite vividly !ickin u! some ob.ect and throwin it,

or kickin somethin = yet we don1t actually do so& En the other hand, we feel so -near- to actually doin so& Probably the faucet catches the nerve im!ulses -at the last moment-& ?ere is another way in which visuali<ation !oints out the distinction between accessible and inaccessible knowled e& 7onsider how you visuali<ed the scene of the car skiddin on the mountain road& Fndoubtedly you ima ined the mountain as bein much lar er than the car& 8ow did this ha!!en because sometime lon a o you had occasion to note that -cars are not as bi as mountains-= then you committed this statement to rote memory, and in ima inin the story, you retrieved this fact, and made use of it in constructin your ima e? ) most unlikely theory& Er did it ha!!en instead as a consequence of some intros!ectively inaccessible interactions of the symbols which were activated in your brain? Ebviously the latter seems far more likely& This knowled e that cars are smaller than mountains is not a !iece of rote memori<ation, but a !iece of knowled e which can be created by deduction& Therefore, most likely it is not stored in any sin le symbol in your brain, but rather it can be !roduced as a result of the activation, followed by the mutual interaction, of many symbolsfor e(am!le, those for -com!are-, -si<e-, -car-, -mountain-, and !robably, others& This means that the knowled e is stored not e(!licitly, but im!licitly, in a s!read-about manner, rather than as a local -!acket of information-& Such sim!le facts as relative si<es of ob.ects have to be assembled, rather than merely retrieved& Therefore, even in the case of a verbally accessible !iece of knowled e, there are com!le( inaccessible !rocesses which mediate its comin to the state of bein ready to be said& 3e shall continue our e(!loration of the entities called -symbols- in different 7ha!ters& In 7ha!ters TGIII and TIT, on )rtificial Intelli ence, we shall discuss some !ossible ways of im!lementin active symbols in !ro rams& )nd ne(t 7ha!ter, we shall discuss some of the insi hts that our symbol-based model of brain activity ive into the com!arison of brains&

'n!lish "rench (erman Suite

0y Aewis 7arroll"&&& &&& et 'rank A& 3arrinB&&& &&& und ;obert Scott% 1Twas brilli , and the slithy toves :id yre and imble in the wabe, )ll mimsy were the boro oves, )nd the mome raths out rabe& Il bril ue, les toves lubricilleu( Se yrent en vrillant dans le uave& EnmpmKs sont les ou ebosqueu( Et le momerade hors rave& Es brilli war& :ie schlichten Toven 3irrten and wimmelten in 3aben= Fnd aller-m`msi e 0ur oven :ie mohmen ;Nth1 aus raben& -0eware the Jabberwock, my son2 The .aws that bite, the claws that catch2 0eware the Jub.ub bird, and shun The frumious 0andersnatch2q6arde-toi du Jaseroque, mon fils2 Aa ueule qui mord= la riffe qui !rend2 6arde-toi de I1oiseau Jube, Kvite Ae frumieu( 0and-r-!rend2s s0ewahre doch vor Jammerwoch2 :ie 9Nhne knirschen, @rallen krat<en2 0ewahr1 vor Jub.ub-Go el, vor 'rumi>sen 0anderschnNt<chen2s ?e took his vor!al sword in hand, Aon time the man(ome foe he sou ht So rested he by the Tumtum tree, )nd stood awhile in thou ht&

Son laive vor!al en main, it va T-r la recherche du fauve manscant= Puis arrivK r l1arbre TK-tK, Il y reste, rKflKchissant& Er riff sein vor!als Schwertchen <u, Er suchte lan das manchsam1 :in = :ann, stehend unterm Tumtum 0aum, Er an-<u-denken-fin & )nd, as in uffish thou ht he stood, The Jabberwock, with eyes of flame& 7ame whifflin throu h the tul ey wood, )nd burbled as it came2 Pendant qu1il !ense, tout uffusK, Ae Jaseroque, a l1oeil flambant, Gient siblant !ar le bois tulle eais, Et burbule en venant& )ls stand er tief in )ndacht auf, :es Jammerwochen1s )u en-feuer :urch tur en 3ald mit 3iffek kam 'in burbelnd Fn eheuer2 Ene, two2 Ene, two2 )nd throu h and throu h The vor!al blade went snicker-snack2 ?e left it dead, and with its head ?e went alum!hin back& Fn deu(, un deu(, !ar le milieu, Ae laive vor!al fait !at-r-!an2 Aa btte dKfaite, avec sa ttte, Il rentre allom!hant& Eins, 9wei2 Eins, 9wei2 Fnd durch and durch Sein vor!als Schwert <erschnifer-schn`ck, :a blieb es todt2 Er, @o!f in ?and, 6elNumfi <o <ur`ck& -)nd hast thou slain the Jabberwock? 7ome to my arms, my beamish boy2 E frab.ous day2 7allooh2 7allay2?e chortled in his .oy&

q)s-tu tuK le Jaseroque? Giens r mon coeur, fils rayonnais2 u .our frabbe.ais2 7alleau2 7allai2s Il cortule dans sa .oie& sFnd schlu st :u .a den Jammerwoch? Fmarme mich, mein 0>hm1sches @ind2 E 'reuden-Ta 2 E ?alloo-Schla 2q Er schortelt froh- esinnt& 1Twas brilli , and the slithy toves :id yre and imble in the wabe, )ll mimsy were the boro oves, )nd the mome raths out rabe& Il bril ue, les toves lubricilleu( Se yrent en vrillant dans le uave& EnmpmKs sont les ou ebosqueu( Et le momerade hors rave& Es brilli war& :ie schlichten Toven 3irrten and wimmelten in 3aben= Fnd aller-m`msi e 0ur oven :ie mohmen ;Nth1 aus raben&

C/A,TE$ <II

Minds and Thoughts

Can Minds Be Ma!!ed onto Each 0ther7
8E3 T?)T 3E have hy!othesi<ed the e(istence of very hi h-level active subsystems of the brain 4symbols5, we may return to the matter of a !ossible isomor!hism, or !artial isomor!hism, between two brains& Instead of askin about an isomor!hism on the neural level 4which surely does not e(ist5, or on the macrosco!ic subor an level 4which surely does e(ist but does not tell us very much5, we ask about the !ossibility of an isomor!hism between brains on the symbol level, a corres!ondence which not only ma!s symbols in one brain onto symbols in another brain, but also ma!s tri erin !atterns onto tri erin !atterns& This means that corres!ondin symbols in the two brains are linked in corres!ondin ways& This would be a true functional isomor!hismthe same ty!e of isomor!hism as we s!oke of when tryin to characteri<e what it is that is invariant about all butterflies, It is clear from the outset that such an isomor!hism does not e(ist between any !air of human bein s& If it did, they would be com!letely indistin uishable in their thou hts= but in order for that to be true, they would have to have com!letely indistin uishable memories, which would mean they would have to have led one and the same life& Even identical twins do not a!!roach, in the remotest de ree, this ideal& ?ow about a sin le individual? 3hen you look back over thin s which you yourself wrote a few years a o, you think -?ow awful2- and smile with amusement at the !erson you once were& 3hat is worse is when you do the same thin with somethin you wrote or said five minutes a o& 3hen this ha!!ens, it shows that you do not fully understand the !erson you were moments a o& The isomor!hism from your brain now to your brain then is im!erfect& 3hat, then, of the isomor!hisms to other !eo!le, other s!ecies&&& The o!!osite side of the coin is shown by the !ower of the communication that arises between the unlikeliest !artners& Think of the barriers s!anned when you read lines of !oetry !enned in .ail by 'ranvois Gillon, the 'rench !oet of the "C++1s& )nother human bein , in another era, ca!tive in .ail, s!eakin another lan ua e&&& ?ow can you ever ho!e to have a sense of the connotations behind the facade of his words, translated into En lish? Het a wealth of meanin comes throu h& Thus, on the one hand, we can dro! all ho!es of findin e(actly isomor!hic software in humans, but on the other, it is clear that some !eo!le think more alike than others do& It would seem an obvious conclusion that there is some sort of !artial software isomor!hism connectin the brains of !eo!le whose style of thinkin is similarin !articular, a corres!ondence of 4"5 the re!ertoire of symbols, and 4B5 the tri erin !atterns of symbols&

73G5=E CK! * tin) portion of the author's "semantic network"!

Com!aring

ifferent Semantic ;et'or.s

0ut what is a partial isomor!hism? This is a most difficult question to answer& It is made even more difficult by the fact that no one has found an adequate way to re!resent the network of symbols and their tri erin !atterns& Sometimes a !icture of a small !art of such a network of symbols is drawn, where each symbol is re!resented as a node into which, and out of which, lead some arcs& The lines re!resent tri erin relationshi!sin some sense& Such fi ures attem!t to ca!ture somethin of the intuitively sensible notion of -conce!tual nearness-& ?owever, there are many different kinds of nearness, and different ones are relevant in different conte(ts& ) tiny !ortion of my own -semantic network- is shown in 'i ure $+& The !roblem is that re!resentin a com!le( interde!endency of many symbols cannot be carried out very easily with .ust a few lines .oinin vertices& )nother !roblem with such a dia ram is that it is not accurate to think of a symbol as sim!ly -on- or -off-& 3hile this is true of neurons, it does not carry u!wards, to collections of them& In this res!ect, symbols are quite a bit more com!licated than neuronsas you mi ht e(!ect, since they are made u! of many neurons& The messa es that are e(chan ed between symbols are more com!le( than the mere fact, -I am now activated-& That is more like the neuron-level messa es& Each symbol can be activated in many different ways, and the ty!e of activation will be influential in determinin which other symbols it tries to activate& ?ow these intertwinin tri erin relationshi!s can be re!resented in a !ictorial mannerindeed, whether they can be at allis not clear& 0ut for the moment, su!!ose that issue had been solved& Su!!ose we now a ree that there are certain drawin s of nodes, connected by links 4let us say they come in various colors, so that various ty!es of conce!tual nearness can be distin uished from each other5, which ca!ture !recisely the way in which symbols tri er other symbols& Then under what conditions would we feel that two such drawin s were isomor!hic, or nearly isomor!hic? Since we are dealin with a visual re!resentation of the network of symbols, let us consider an analo ous visual !roblem& ?ow would you try to determine whether two s!iderwebs had been s!un by s!iders belon in to the same s!ecies? 3ould you try to identify individual vertices which corres!ond e(actly, thereby settin u! an e(act ma! of one web onto the other, verte( by verte(, fiber by fiber, !erha!s even an le by an le? This would be a futile effort& Two webs are never e(actly the same, yet there is still some sort of -style-, -form-, what-have-you, that infallibly brands a iven s!ecies1 web& In any network-like structure, such as a s!iderweb, one can look at local !ro!erties and lobal !ro!erties& Aocal !ro!erties require only a very nearsi hted observerfor e(am!le an observer who can only see one verte( at a time= and lobal !ro!erties require only a swee!in vision, without attention to detail& Thus, the overall sha!e of a s!iderweb is a lobal !ro!erty, whereas the avera e number of lines meetin at a verte( is a local !ro!erty& Su!!ose we a ree that the most reasonable criterion for callin two s!iderwebs -isomor!hic- is that they should have been s!un by s!iders of the same s!ecies& Then it is interestin to ask which kind of observationlocal or lobal tends to be a more reliable uide in determinin whether two s!iderwebs are isomor!hic&

3ithout answerin the question for s!iderwebs, let us now return to the question of the closenessor isomor!hicness, if you willof two symbol networks&

Translations of (3abber'oc.y(
Ima ine native s!eakers of En lish, 'rench, and 6erman, all of whom have e(cellent command of their res!ective native lan ua es, and all of whom en.oy word!lay in their own lan ua e& 3ould their symbol networks be similar on a local level, or on a lobal level? Er is it meanin ful to ask such a question? The question becomes concrete when you look at the !recedin translations of Aewis 7arroll1s famous -Jabberwocky-& I chose this e(am!le because it demonstrates, !erha!s better than an e(am!le in ordinary !rose, the !roblem of tryin to find -the same node- in two different networks which are, on some level of analysis, e(tremely nonisomor!hic& In ordinary lan ua e, the task of translation is more strai htforward, since to each word or !hrase in the ori inal lan ua e, there can usually be found a corres!ondin word or !hrase in the new lan ua e& 0y contrast, in a !oem of this ty!e, many -words- do not carry ordinary meanin , but act !urely as e(citers of nearby symbols& ?owever, what is nearby in one lan ua e may be remote in another& Thus, in the brain of a native s!eaker of En lish, -slithy- !robably activates such symbols as -slimy-, -slither-, -sli!!ery-, -lithe-, and -sly-, to varyin e(tents& :oes -lubricilleu(- do the corres!ondin thin in the brain of a 'renchman? 3hat indeed would be -the corres!ondin thin -? 3ould it be to activate symbols which are the ordinary translations of those words? 3hat if there is no word, real or fabricated, which will accom!lish that? Er what if a word does e(ist, but is very intellectual-soundin and Aatinate 4-lubricilleu(-5, rather than earthy and )n lo-Sa(on 4-slithy-5? Perha!s -huilasse- would be better than -lubricilleu(-? Er does the Aatin ori in of the word -lubricilleu(- not make itself felt to a s!eaker of 'rench in the way that it would if it were an En lish word 4-lubricilious-, !erha!s5? )n interestin feature of the translation into 'rench is the trans!osition into the !resent tense& To kee! it in the !ast would make some unnatural turns of !hrase necessary, and the !resent tense has a much fresher flavor in 'rench than the !ast& The translator sensed that this would be -more a!!ro!riate-in some ill-defined yet com!ellin senseand made the switch& 3ho can say whether remainin faithful to the En lish tense would have been better? In the 6erman version, the droll !hrase -er an-<u-denken-fin - occurs= it does not corres!ond to any En lish ori inal& It is a !layful reversal of words, whose flavor va uely resembles that of the En lish !hrase -he out-to-!onder set-, if I may ha<ard a reverse translation& /ost likely this funny turnabout of words was ins!ired by the similar !layful reversal in the En lish of one line earlier, -So rested he by the Tumtum tree-& It corres!onds, yet doesn1t corres!ond& Incidentally, why did the Tumtum tree et chan ed into an -arbre TK-tK- in 'rench? 'i ure it out for yourself& The word -man(ome- in the ori inal, whose -(- imbues it with many rich overtones, is weakly rendered in 6erman by -manchsam-, which hack-translates into

En lish as -maniful-& The 'rench -manscant- also lacks the manifold overtones of -man(ome-& There is no end to the interest of this kind of translation task& 3hen confronted with such an e(am!le, one reali<es that it is utterly im!ossible to make an e(act translation& Het even in this !atholo ically difficult case of translation, there seems to be some rou h equivalence obtainable& 3hy is this so, if there really is no isomor!hism between the brains of !eo!le who will read the different versions? The answer is that there is a kind of rou h isomor!hism, !artly lobal, !artly local, between the brains of all the readers of these three !oems&

AS&%s
)n amusin eo ra!hical fantasy will ive some intuition for this kind of quasiisomor!hism& 4Incidentally, this fantasy is somewhat similar to a eo ra!hical analo y devised by /& /insky in his article on -frames-, which can be found in P& ?& 3inston1s book The &s)cholog) of 8omputer <ision&5 Ima ine that you are iven a stran e atlas of the FS), with all natural eolo ical features !remarkedsuch as rivers, mountains, lakes, and so onbut with nary a !rinted word& ;ivers are shown as blue lines, mountains by color, and so on& 8ow you are told to convert it into a road atlas for a tri! which you will soon make& Hou must neatly fill in the names of all states, their boundaries, time <ones, then all counties, cities, towns, all freeways and hi hways and toll routes, all county roads, all state and national !arks, cam! rounds, scenic areas, dams, air!orts, and so on&&& )ll of this must be carried out down to the level that would a!!ear in a detailed road atlas& )nd it must be manufactured out of your own head& Hou are not allowed access to any information which would hel! you for the duration of your task& Hou are told that it will !ay off, in ways that will become clear at a later date, to make your ma! as true as you can& Ef course, you will be in by fillin in lar e cities and ma.or roads, etc&, which you know& )nd when you have e(hausted your factual knowled e of an area, it will be to your advanta e to use your ima ination to hel! you re!roduce at least the flavor of that area, if not its true eo ra!hy, by makin u! fake town names, fake !o!ulations, fake roads, fake !arks, and so on& This arduous task will take months& To make thin s a little easier, you have a carto ra!her on hand to !rint everythin in neatly& The end !roduct will be your !ersonal ma! of the -)lternative Structure of the Fnion-your own !ersonal -)SF-& Hour !ersonal )SF will be very much like the FS) in the area where you rew u!& 'urthermore, wherever your travels have chanced to lead you, or wherever you have !erused ma!s with interest, your )SF will have s!ots of strikin a reement with the FS), a few small towns in 8orth :akota or /ontana, !erha!s, or the whole of metro!olitan 8ew Hork, mi ht be quite faithfully re!roduced in your )SF&

A Sur!rise $e"ersal
3hen your )SF is done, a sur!rise takes !lace& /a ically, the country you have desi ned comes into bein , and you are trans!orted there& ) friendly committee !resents

you with your favorite kind of automobile, and e(!lains that, -)s a reward for your desi nin efforts, you may now en.oy an all-e(!ense-!aid tri!, at a leisurely !ace, around the ood old )& S& of F& Hou may o wherever you want, do whatever you wish to do, takin as lon as you wishcom!liments of the 6eo ra!hical Society of the )SF& )nd to uide you aroundhere is a road atlas&- To your sur!rise, you are iven not the atlas which you desi ned, but a re ular road atlas of the FS)& 3hen you embark on your tri!, all sorts of curious incidents will take !lace& ) road atlas is bein used to uide you throu h a country which it only !artially fits& )s lon as you stick to ma.or freeways, you will !robably be able to cross the country without ross confusions& 0ut the moment you wander off into the byways of 8ew /e(ico or rural )rkansas, there will be adventure in store for you& The locals will not reco ni<e any of the towns you1re lookin for, nor will they know the roads you1re askin about& They will only know the lar e cities you name, and even then the routes to those cities will not be the same as are indicated on your ma!& It will ha!!en occasionally that some of the cities which are considered hu e by the locals are none(istent on your ma! of the FS)= or !erha!s they e(ist, but their !o!ulation accordin to the atlas is wron by an order of ma nitude&

Centrality and &ni"ersality

3hat makes an )SF and the FS), which are so different in some ways, nevertheless so similar? It is that their most im!ortant cities and routes of communication can be ma!!ed onto each other& The differences between them are found in the less frequently traveled routes, the cities of smaller si<e, and so on& 8otice that this cannot be characteri<ed either as a local or a lobal isomor!hism& Some corres!ondences do e(tend down to the very local levelfor instance, in both 8ew Horks, the main street may be 'ifth )venue, and there may be a Times Square in both as wellyet there may not be a sin le town that is found in both /ontanas& So the local- lobal distinction is not relevant here& 3hat is relevant is the centrality of the city, in terms of economics, communication, trans!ortation, etc& The more vital the city is, in one of these ways, the more certain it will be to occur in both the )SF and the FS)& In this eo ra!hic analo y, one as!ect is very crucial, that there are certain definite, absolute !oints of reference which will occur in nearly all )SF1s, 8ew Hork, San 'rancisco, 7hica o, and so on& 'rom these it is then !ossible to orient oneself& In other words, if we be in com!arin my )SF with yours, I can use the known a reement on bi cities to establish !oints of reference with which I can communicate the location of smaller cities in my )SF& )nd if I hy!othesi<e a voya e from @ankakee to 'ruto and you don1t know where those towns are, I can refer to somethin we have in common, and thereby uide you& )nd if I talk about a voya e from )tlanta to /ilwaukee, it may o alon different freeways or smaller roads, but the voya e itself can still be carried out in both countries& )nd if you start describin a tri! from ?orsemilk to Jan<o, I can !lot out what seems to me to be an analo ous tri! in my )SF, des!ite not havin towns by those names, as lon as you constantly kee! me oriented by describin your !osition with res!ect to nearby lar er towns which are found in my )SF as well as in yours&

/y roads will not be e(actly the same as yours, but, with our se!arate ma!s, we can each et from a !articular !art of the country to another& 3e can do this, thanks to the e(ternal, !redetermined eolo ical factsmountain chains, streams, etc&facts which were available to us both as we worked on our ma!s& 3ithout those e(ternal features, we would have no !ossibility of reference !oints in common& 'or instance, if you had been iven only a ma! of 'rance, and I had been iven a ma! of 6ermany, and then we had both filled them in in reat detail, there would he no way to try to find -the same !lacein our fictitious lands& It is necessary to be in with identical e(ternal conditions otherwise nothin will match& 8ow that we have carried our eo ra!hical analo y quite far, we return to the question of isomor!hisms between brains& Hou mi ht well wonder why this whole question of brain isomor!hisms has been stressed so much& 3hat does it matter if two brains are isomor!hic, or quasi-isomor!hic, or not isomor!hic at all? The answer is that we have an intuitive sense that, althou h other !eo!le differ from us in im!ortant ways, they are still -the same- as we are in some dee! and im!ortant ways& It would be instructive to be able to !in!oint what this invariant core of human intelli ence is, and then to be able to describe the kinds of -embellishments- which can be added to it, makin each one of us a unique embodiment of this abstract and mysterious quality called -intelli ence-& In our eo ra!hic analo y, cities and towns were the analo ues of symbols, while roads and hi hways were analo ous to !otential tri erin !aths& The fact that all )SF1s have some thin s in common, such as the East 7oast, the 3est 7oast, the /ississi!!i ;iver, the 6reat Aakes, the ;ockies, and many ma.or cities and roads is analo ous to the fact that we are all forced, by e(ternal realities, to construct certain class symbols and tri erin !aths in the same way& These core symbols are like the lar e cities, to which everyone can make reference without ambi uity& 4Incidentally, the fact that cities are locali<ed entities should in no way be taken as indicative that symbols in a brain are small, almost !oint-like entities& They are merely symboli<ed in that manner in a network&5 The fact is that a lar e !ro!ortion of every human1s network of symbols is universal& 3e sim!ly take what is common to all of us so much for ranted that it is hard to see how much we have in common with other !eo!le& It takes the conscious effort of ima inin how muchor how littlewe have in common with other ty!es of entities, such as stones, cars, restaurants, ants, and so forth, to make evident the lar e amount of overla! that we have with randomly chosen !eo!le& 3hat we notice about another !erson immediately is not the standard overla!, because that is taken for ranted as soon as we reco ni<e the humanity of the other !erson= rather, we look beyond the standard overla! and enerally find some ma.or differences, as well as some une(!ected, additional overla!& Eccasionally, you find that another !erson is missin some of what you thou ht was the standard, minimal coreas if 7hica o were missin from their )SF, which is almost unima inable& 'or instance, someone mi ht not know what an ele!hant is, or who is President, or that the earth is round& In such cases, their symbolic network is likely to

be so fundamentally different from your own that si nificant communication will be difficult& En the other hand, !erha!s this same !erson will share some s!eciali<ed kind of knowled e with yousuch as e(!ertise in the ame of dominoesso that you can communicate well in a limited domain& This would be like meetin someone who comes from the very same rural area of 8orth :akota as you do, so that your two )SF1s coincide in reat detail over a very small re ion, which allows you to describe how to et from one !lace to another very fluently&

/o' Much

o Language and Culture Channel Thought7

If we now o back to com!arin our own symbol network with those of a 'renchman and a 6erman, we can say that we e(!ect them to have the standard core of class symbols, des!ite the fact of different native lan ua es& 3e do not e(!ect to share hi hly s!eciali<ed networks with them, but we do not e(!ect such sharin with a randomly chosen !erson who shares our native lan ua e, either& The tri erin !atterns of !eo!le with other lan ua es will be somewhat different from our own, but still the ma.or class symbols, and the ma.or routes between them, will be universally available, so that more minor routes can be described with reference to them& 8ow each of our three !eo!le may in addition have some command of the lan ua es of the other two& 3hat is it that marks the difference between true fluency, and a mere ability to communicate? 'irst of all, someone fluent in En lish uses most words at rou hly their re ular frequencies& ) non-native s!eaker will have !icked u! some words from dictionaries, novels, or classeswords which at some time may have been !revalent or !referable, but which are now far down in frequencyfor e(am!le, -fetchinstead of - et-, -quite- instead of -very-, etc& Thou h the meanin usually comes throu h, there is an alien quality transmitted by the unusual choice of words& 0ut su!!ose that a forei ner learns to use all words at rou hly the normal frequencies& 3ill that make his s!eech truly fluent? Probably not& ?i her than the word level, there is an association level, which is attached to the culture as a wholeits history, eo ra!hy, reli ion, children1s stories, literature, technolo ical level, and so on& 'or instance, to be able to s!eak modern ?ebrew absolutely fluently, you need to know the 0ible quite well in ?ebrew, because the lan ua e draws on a stock of biblical !hrases and their connotations& Such an association level !ermeates each lan ua e very dee!ly& Het there is room for all sorts of variety inside fluencyotherwise the only truly fluent s!eakers would be !eo!le whose thou hts were the most stereoty!ed !ossible2 )lthou h we should reco ni<e the de!th to which culture affects thou ht, we should not overstress the role of language in moldin thou hts& 'or instance, what we mi ht call two -chairs- mi ht be !erceived by a s!eaker of 'rench as ob.ects belon in to two distinct ty!es, -chaise- and -fauteuil- 4-chair- and -armchair-5& Peo!le whose native lan ua e is 'rench are more aware of that difference than we arebut then !eo!le who row u! in a rural area are more aware of, say, the difference between a !icku! and a truck, than a city dweller is& ) city dweller may call them both -trucks-& It is not the difference in native lan ua e, but the difference in culture 4or subculture5, that ives rise to this !erce!tual difference&

The relationshi!s between the symbols of !eo!le with different native lan ua es have every reason to be quite similar, as far as the core is concerned, because everyone lives in the same world& 3hen you come down to more detailed as!ects of the tri erin !atterns, you will find that there is less in common& It would he like com!arin rural areas in 3isconsin in )SF1s which had been made u! by !eo!le who had never lived in 3isconsin& This will be quite irrelevant, however, as lon as there is sufficient a reement on the ma.or cities and ma.or routes, so that there are common !oints of reference all over the ma!&

Tri!s and Itineraries in AS&%s

3ithout makin it e(!licit, I have been usin an ima e of what a -thou ht- is in the )SFanalo ynamely, I have been im!lyin that a thought corres!onds to a trip& The towns which are !assed throu h re!resent the symbols which are e(cited& This is not a !erfect analo y, but it is quite stron & Ene !roblem with it is that when a thou ht recurs in someone1s mind sufficiently often, it can et chunked into a sin le conce!t& This would corres!ond to quite a stran e event in an )SF, a commonly taken tri! would become, in some stran e fashion, a new town or city2 If one is to continue to use the )SF-meta!hor, then, it is im!ortant to remember that the cities re!resent not only the elementar) symbols, such as those for - rass-, -house-, and -car-, but also symbols which et created as a result of the chunkin ability of a brain-symbols for such so!histicated conce!ts as -crab canon-, -!alindrome-, or -)SF-& 8ow if it is ranted that the notion of takin a tri! is a fair counter!art to the notion of havin a thou ht& then the followin difficult issue comes u!, virtually any route leadin from one city to a second, then to a third, and so on, can be ima ined, as lon as one remembers that some intervenin cities are also !assed throu h& This would corres!ond to the activation of an ar(itrar) se"uence of s)m(ols, one after another, makin allowance for some e(tra symbolsthose which lie en route& 8ow if virtually any sequence of symbols can be activated in any desired order, it may seem that a brain is an indiscriminate system, which can absorb or !roduce any thou ht whatsoever& 0ut we all know that that is not so& In fact, there are certain kinds of thou hts which we call knowledge, or (eliefs, which !lay quite a different role from random fancies, or humorously entertained absurdities& ?ow can we characteri<e the difference between dreams, !assin thou hts, beliefs, and !ieces of knowled e?

,ossible) ,otential) and ,re!osterous ,ath'ays

There are some !athwaysyou can think of them as !athways either in an )SF or in a brainwhich are taken routinely in oin from one !lace to another& There are other !athways which can only be followed if one is led throu h them by the hand& These !athways are -!otential !athways-, which would be followed only if s!ecial e(ternal circumstances arose& The !athways which one relies on over and over a ain are !athways which incor!orate knowled eand here I mean not only knowled e of facts 4declarative

knowled e5, but also knowled e of how%to's 4!rocedural knowled e5& These stable, reliable !athways are what constitute knowled e& Pieces of knowled e mer e radually with beliefs, which are also re!resented by reliable !athways, but !erha!s ones which are more susce!tible to re!lacement if, so to s!eak, a brid e oes out, or there is heavy fo & This leaves us with fancies, lies, falsities, absurdities, and other variants& These would corres!ond to !eculiar routes such as, 8ew Hork 7ity to 8ewark via 0an or, /aine and Aubbock, Te(as& They are indeed !ossible !athways, but ones which are not likely to be stock routes, used in everyday voya es& ) curious, and amusin , im!lication of this model is that all of the -aberrantkinds of thou hts listed above are com!osed, at rock bottom, com!letely out of beliefs or !ieces of knowled e& That is, any weird and snaky indirect route breaks u! into a number of non-weird, non-snaky direct stretches, and these short, strai htforward symbolconnectin routes re!resent sim!le thou hts that one can rely onbeliefs and !ieces of knowled e& En reflection, this is hardly sur!risin , however, since it is quite reasonable that we should only be able to ima ine fictitious thin s that are somehow rounded in the realities we have e(!erienced, no matter how wildly they deviate from them& :reams are !erha!s .ust such random meanderin s about the )SF1s of our minds& Aocally, they make sense-but lobally&&&

ifferent Styles of Translating ;o"els

) !oem like -Jabberwocky- is like an unreal .ourney around an )SF, ho!!in from one state to another very quickly, followin very curious routes& The translations convey this as!ect of the !oem, rather than the !recise sequence of symbols which are tri ered, althou h they do their best in that res!ect& In ordinary !rose, such lea!s and bounds are not so common& ?owever, similar !roblems of translation do occur& Su!!ose you are translatin a novel from ;ussian to En lish, and come across a sentence whose literal translation is, -She had a bowl of borscht&- 8ow !erha!s many of your readers will have no idea what borscht is& Hou could attem!t to re!lace it by the -corres!ondin - item in their culturethus, your translation mi ht run, -She had a bowl of 7am!bell1s sou!&8ow if you think this is a silly e(a eration, take a look at the first sentence of :ostoevsky1s novel 8rime and &unishment in ;ussian and then in a few different En lish translations& I ha!!ened to look at three different En lish !a!erback translations, and found the followin curious situation& The first sentence em!loys the street name -S& Pereulok- 4as transliterated5& 3hat is the meanin of this? ) careful reader of :ostoevsky1s work who knows Aenin rad 4which used to be called -St& Petersbur -or should I say -Petro rad-?5 can discover by doin some careful checkin of the rest of the eo ra!hy in the book 4which incidentally is also iven only by its initials5 that the street must be -Stoliarny Pereulok-& :ostoevsky !robably wished to tell his story in a realistic way, yet not so realistically that !eo!le would take literally the addresses at which crimes and other events were su!!osed to have occurred& In any case, we have a translation !roblem= or to be more !recise, we have several translation !roblems, on several different levels&

'irst of all, should we kee! the initial so as to re!roduce the aura of semi-mystery which a!!ears already in this first sentence of the book? 3e would et -S& Aane- 4-lanebein the standard translation of -!ereulok-5& 8one of the three translators took this tack& ?owever, one chose to write -S& Place-& The translation of 8rime and &unishment which I read in hi h school took a similar o!tion& I will never for et the disoriented feelin I e(!erienced when I be an readin the novel and encountered those streets with only letters for names& I had some sort of intan ible malaise about the be innin of the book= I was sure that I was missin somethin essential, and yet I didn1t know what it was&&& I decided that all ;ussian novels were very weird& 8ow we could be frank with the reader 4who, it may be assumed, !robably won1t have the sli htest idea whether the street is real or fictitious anyway25 and ive him the advanta e of our modern scholarshi!, writin -Stoliarny Aane- 4or -Place-5& This was the choice of translator number B, who ave the translation as -Stoliarny Place-& 3hat about number %? This is the most interestin of all& This translation says -7ar!enter1s Aane-& )nd why not, indeed? )fter all, -stoliar- means -car!enter- and -nyis an ad.ectival endin & So now we mi ht ima ine ourselves in Aondon, not Petro rad, and in the midst of a situation invented by :ickens, not :ostoevsky& Is that what we want? Perha!s we should .ust read a novel by :ickens instead, with the .ustification that it is -the corres!ondin work in En lish-& 3hen viewed on a sufficiently hi h level, it is a -translation- of the :ostoevsky novelin fact, the best !ossible one2 3ho needs :ostoevsky? 3e have come all the way from attem!ts at reat literal fidelity to the author1s style, to hi h-level translations of flavor& 8ow if this ha!!ens already in the first sentence, can you ima ine how it must o on in the rest of the book? 3hat about the !oint where a 6erman landlady be ins shoutin in her 6erman-style ;ussian? ?ow do you translate broken ;ussian s!oken with a 6erman accent, into En lish? Then one may also consider the !roblems of how to translate slan and colloquial modes of e(!ression& Should one search for an -analo ous- !hrase, or should one settle for a word-by-word translation? If you search for an analo ous !hrase, then you run the risk of committin a -7am!bell1s sou!- ty!e of blunder= but if you translate every idiomatic !hrase word by word, then the En lish will sound alien& Perha!s this is desirable, since the ;ussian culture is an alien one to s!eakers of En lish& 0ut a s!eaker of En lish who reads such a translation will constantly be e(!eriencin , thanks to the unusual turns of !hrase, a sensean artificial senseof stran eness, which was not intended by the author, and which is not e(!erienced by readers of the ;ussian ori inal& Problems such as these ive one !ause in considerin such statements as this one, made by 3arren 3eaver, one of the first advocates of translation by com!uter, in the late "#C+1s, -3hen I look at an article in ;ussian, I say, 1This is really written in En lish, but it has been coded in some stran e symbols& I will now !roceed to decode&1 - 3eaver1s remark sim!ly cannot be taken literally= it must rather be considered a !rovocative way of sayin that there is an ob.ectively describable meanin hidden in the symbols, or at least somethin !retty close to ob.ective= therefore, there would be no reason to su!!ose a com!uter could not ferret it out, if sufficiently well !ro rammed&

/igh-Le"el Com!arisons bet'een ,rograms

3eaver1s statement is about translations between different natural lan ua es& Aet1s consider now the !roblem of translatin between two com!uter lan ua es& 'or instance, su!!ose two !eo!le have written !ro rams which run on different com!uters, and we want to know if the two !ro rams carry out the same task& ?ow can we find out? 3e must com!are the !ro rams& 0ut on what level should this be done? Perha!s one !ro rammer wrote in a machine lan ua e, the other in a com!iler lan ua e& )re two such !ro rams com!arable? 7ertainly& 0ut how to com!are them? Ene way mi ht be to com!ile the com!iler lan ua e !ro ram, !roducin a !ro ram in the machine lan ua e of its home com!uter& 8ow we have two machine lan ua e !ro rams& 0ut there is another !roblem, there are two com!uters, hence two different machine lan ua es-and they may be e(tremely different& Ene machine may have si(teen-bit words= the other thirty-si(-bit words& Ene machine may have built-in stack-handlin instructions 4!ushin and !o!!in 5, while the other lacks them& The differences between the hardware of the two machines may make the two machine lan ua e !ro rams seem incom!arableand yet we sus!ect they are !erformin the same task, and we would like to see that at a lance& 3e are obviously lookin at the !ro rams from much too close a distance& 3hat we need to do is to ste! back, away from machine lan ua e, towards a hi her, more chunked view& 'rom this vanta e !oint, we ho!e we will be able to !erceive chunks of !ro ram which make each !ro ram seem rationally !lanned out on a lobal, rather than a local, scale-that is, chunks which fit to ether in a way that allows one to !erceive the oals of the !ro rammer& Aet us assume that both !ro rams were ori inally written in hi h-level lan ua es& Then some chunkin has already been done for us& 0ut we will run into other troubles& There is a !roliferation of such lan ua es, 'ortran, )l ol, AISP, )PA, and many others& ?ow can you com!are a !ro ram written in )PA with one written in )l ol, 7ertainly not by matchin them u! line by line& Hou will a ain chunk these !ro rams in your mind, lookin for conce!tual, functional units which corres!ond& Thus, you are not com!arin hardware, you are not com!arin softwareyou are com!arin -etherware-the !ure conce!ts which lie back of the software& There is some sort of abstract -conce!tual skeleton- which must be lifted out of low levels before you can carry out a meanin ful com!arison of two !ro rams in different com!uter lan ua es, of two animals, or of two sentences in different natural lan ua es& 8ow this brin s us back to an earlier question which we asked about com!uters and brains, ?ow can we make sense of a low-level descri!tion of a com!uter or a brain? Is there, in any reasonable sense, an ob.ective way to !ull a hi h-level descri!tion out of a low-level one, in such com!licated systems? In the case of a com!uter, a full dis!lay of the contents of memorya so-called memor) dumpis easily available& :um!s were commonly !rinted out in the early days of com!utin , when somethin went wron with a !ro ram& Then the !ro rammer would have to o home and !ore over the memory dum! for hours, tryin to understand what each minuscule !iece of memory re!resented& In essence, the !ro rammer would be doin the o!!osite of what a com!iler does, he would be translatin from machine lan ua e into a hi her-level lan ua e, a conce!tual

lan ua e& In the end, the !ro rammer would understand the oals of the !ro ram and could describe it in hi h-level termsfor e(am!le, -This !ro ram translates novels from ;ussian to En lish-, or -This !ro ram com!oses an ei ht-voice fu ue based on any theme which is fed in-&

/igh-Le"el Com!arisons bet'een Brains

8ow our question must be investi ated in the case of brains& In this case, we are askin , -)re !eo!le1s brains also ca!able of bein 1read1, on a hi h level? Is there some ob.ective descri!tion of the content of a brain?- In the )nt 'u ue, the )nteater claimed to be able to tell what )unt ?illary was thinkin about, by lookin at the scurryin s of her com!onent ants& 7ould some su!erbein a 8euroneater, !erha!sconceivably look down on our neurons, chunk what it sees, and come u! with an analysis of our thou hts? 7ertainly the answer must be yes, since we are all quite able to describe, in chunked 4i&e&, non-neural5 terms, the activity of our minds at any iven time& This means that we have a mechanism which allows us to chunk our own brain state to some rou h de ree, and to ive a functional descri!tion of it& To be more !recise, we do not chunk all of the brain statewe only chunk those !ortions of it which are active& ?owever, if someone asks us about a sub.ect which is coded in a currently inactive area of our brain, we can almost instantly ain access to the a!!ro!riate dormant area and come u! with a chunked descri!tion of itthat is, some belief on that sub.ect& 8ote that we come back with absolutely <ero information on the neural level of that !art of the brain, our descri!tion is so chunked that we don1t even have any idea what !art of our brain it is a descri!tion of& This can be contrasted with the !ro rammer whose chunked descri!tion comes from conscious analysis of every !art of the memory dum!& 8ow if a !erson can !rovide a chunked descri!tion of any !art of his own brain, why shouldn1t an outsider too, iven some nondestructive means of access to the same brain, not only be able to chunk limited !ortions of the brain, but actually to ive a com!lete chunked descri!tion of itin other words, a com!lete documentation of the beliefs of the !erson whose brain is accessible? It is obvious that such a descri!tion would have an astronomical si<e, but that is not of concern here& 3e are interested in the question of whether, in !rinci!le, there e(ists a well-defined, hi h-level descri!tion of a brain, or whether, conversely, the neuron-level descri!tionor somethin equally !hysiolo ical and intuitively unenli htenin is the best descri!tion that in !rinci!le e(ists& Surely, to answer this question would be of the hi hest im!ortance if we seek to know whether we can ever understand ourselves&

,otential Beliefs) ,otential Symbols

It is my contention that a chunked descri!tion is !ossible, but when we et it, all will not suddenly be clear and li ht& The !roblem is that in order to !ull a chunked descri!tion out of the brain state, we need a lan ua e to describe our findin s& 8ow the most a!!ro!riate way to describe a brain, it would seem, would be to enumerate the kinds of thou hts it could entertain, and the kinds of thou hts it could not entertainor, !erha!s, to

enumerate its beliefs and the thin s which it does not believe& If that is the kind of oal we will be strivin for in a chunked descri!tion, then it is easy to see what kinds of troubles we will run u! a ainst& Su!!ose you wanted to enumerate all !ossible voya es that could be taken in an )SF= there are infinitely many& ?ow do you determine which ones are !lausible, thou h? 3ell, what does -!lausible- mean? 3e will have !recisely this kind of difficulty in tryin to establish what a -!ossible !athway- from symbol to symbol in a brain is& 3e can ima ine an u!side-down do flyin throu h the air with a ci ar in its mouthor a collision between two iant fried e s on a freewayor any number of other ridiculous ima es& The number of far-fetched !athways which can be followed in our brains is without bound, .ust as is the number of insane itineraries that could be !lanned on an )SF& 0ut .ust what constitutes a -sane- itinerary, iven an )SF? )nd .ust what constitutes a -reasonable- thou ht, iven a brain state? The brain state itself does not forbid any !athway, because for any !athway there are always circumstances which could force the followin of that !athway& The !hysical status of a brain, if read correctly, ives information tellin not which !athways could be followed, but rather how much resistance would be offered alon the way& 8ow in an )SF, there are many tri!s which could be taken alon two or more reasonable alternative routes& 'or e(am!le, the tri! from San 'rancisco to 8ew Hork could o alon either a northern route or a southern route& Each of them is quite reasonable, but !eo!le tend to take them under different circumstances& Aookin at a ma! at a iven moment in time does not tell you anythin about which route will be !referable at some remote time in the futurethat de!ends on the e(ternal circumstances under which the tri! is to be taken& Aikewise, the -readin - of a brain state will reveal that several reasonable alternative !athways are often available, connectin a iven set of symbols& ?owever, the tri! amon these symbols need not be imminent= it may be sim!ly one of billions of -!otential- tri!s, all of which fi ure in the readout of the brain state& 'rom this follows an im!ortant conclusion, there is no information in the brain state itself which tells which route will be chosen& The e(ternal circumstances will !lay a lar e determinin role in choosin the route& 3hat does this im!ly? It im!lies that thou hts which clash totally may be !roduced by a sin le brain, de!endin on the circumstances& )nd any hi h-level readout of the brain state which is worth its salt must contain all such conflictin versions& )ctually this is quite obviousthat we all are bundles of contradictions, and we mana e to han to ether by brin in out only one side of ourselves at a iven time& The selection cannot be !redicted in advance, because the conditions which will force the selection are not known in advance& 3hat the brain state can !rovide, if !ro!erly read, is a conditional descri!tion of the selection of routes& 7onsider, for instance, the 7rab1s !li ht, described in the &relude& ?e can react in various ways to the !layin of a !iece of music& Sometimes he will be nearly immune to it, because he knows it so well& Ether times, he will be quite e(cited by it, but this reaction requires the ri ht kind of tri erin from the outsidefor instance, the !resence of an enthusiastic listener, to whom the work is new& Presumably, a hi h-level readin of the 7rab1s brain state would reveal the !otential thrill 4and conditions which would

induce it5, as well as the !otential numbness 4and conditions which would induce it5& The brain state itself would not tell which one would occur on the ne(t hearin of the !iece, however, it could only say, -If such-w-such conditions obtain, then a thrill will result= otherwise&&&Thus a chunked descri!tion of a brain state would ive a catalo ue of beliefs which could be evoked conditionally, de!endent on circumstances& Since not all !ossible circumstances can be enumerated, one would have to settle for those which one thinks are -reasonable-& 'urthermore, one would have to settle for a chunked descri!tion of the circumstances themselves, since they obviously cannotand should notbe s!ecified down to the atomic level2 Therefore, one will not be able to make an e(act, deterministic !rediction sayin which beliefs will be !ulled out of the brain state by a iven chunked circumstance& In summary, then, a chunked descri!tion of a brain state will consist of a !robabilistic catalo ue, in which are listed those beliefs which are most likely to be induced 4and those symbols which are most likely to be activated5 by various sets of -reasonably likely- circumstances, themselves described on a chunked level& Tryin to chunk someone1s beliefs without referrin to conte(t is !recisely as silly as tryin to describe the ran e of a sin le !erson1s -!otential !ro eny- without referrin to the mate& The same sorts of !roblems arise in enumeratin all the symbols in a iven !erson1s brain& There are !otentially not only an infinite number of pathwa)s in a brain, but also an infinite number of s)m(ols& )s was !ointed out, new conce!ts can always be formed from old ones, and one could ar ue that the symbols which re!resent such new conce!ts are merely dormant symbols in each individual, waitin to be awakened& They may never et awakened in the !erson1s lifetime, but it could be claimed that those symbols are nonetheless always there, .ust waitin for the ri ht circumstances to tri er their synthesis& ?owever, if the !robability is very low, it would seem that -dormantwould be a very unrealistic term to a!!ly in the situation& To make this clear, try to ima ine all the -dormant dreams- which are sittin there inside your skull while you1re awake& Is it conceivable that there e(ists a decision !rocedure which could tell -!otentially dreamable themes- from -undreamable themes-, iven your brain state?

>here Is the Sense of Self7

Aookin back on what we have discussed, you mi ht think to yourself, -These s!eculations about brain and mind are all well and ood, but what about the feelin s involved in consciousness, These symbols may tri er each other all they want, but unless someone percei#es the whole thin , there1s no consciousness&This makes sense to our intuition on some level, but it does not make much sense lo ically& 'or we would then be com!elled to look for an e(!lanation of the mechanism which does the !erceivin of all the active symbols, if it is not covered by what we have described so far& Ef course, a -soulist- would not have to look any furtherhe would merely assert that the !erceiver of all this neural action is the soul, which cannot be described in !hysical terms, and that is that& ?owever, we shall try to ive a -nonsouliste(!lanation of where consciousness arises&

Eur alternative to the soulist e(!lanationand a disconcertin one it is, toois to sto! at the symbol level and say, -This is itthis is what consciousness is& 7onsciousness is that !ro!erty of a system that arises whenever there e(ist symbols in the system which obey tri erin !atterns somewhat like the ones described in the !ast several sections&Put so starkly, this may seem inadequate& ?ow does it account for the sense of -I-, the sense of self?

Subsystems
There is no reason to e(!ect that -I-, or -the self-1, should not be re!resented by a symbol& In fact, the symbol for the self is !robably the most com!le( of all the symbols in the brain& 'or this reason, I choose to !ut it on a new level of the hierarchy and call it a su(s)stem, rather than a symbol& To be !recise, by -subsystem-, I mean a constellation of symbols, each of which can be se!arately activated under the control of the subsystem itself& The ima e I wish to convey of a subsystem is that it functions almost as an inde!endent -subbrain-, equi!!ed with its own re!ertoire of symbols which can tri er each other internally& Ef course, there is also much communication between the subsystem and the -outside- worldthat is, the rest of the brain& -Subsystem- is .ust another name for an over rown symbol, one which has otten so com!licated that it has many subsymbols which interact amon themselves& Thus, there is no strict level distinction between symbols and subsystems& 0ecause of the e(tensive links between a subsystem and the rest of the brain 4some of which will be described shortly5, it would be very difficult to draw a shar! boundary between the subsystem and the outside= but even if the border is fu<<y, the subsystem is quite a real thin & The interestin thin about a subsystem is that, once activated and left to its own devices, it can work on its own& Thus, two or more subsystems of the brain of an individual may o!erate simultaneously& I have noticed this ha!!enin on occasion in my own brain, sometimes I become aware that two different melodies are runnin throu h my mind, com!etin for -my- attention& Somehow, each melody is bein manufactured, or -!layed-, in a se!arate com!artment of my brain& Each of the systems res!onsible for drawin a melody out of my brain is !resumably activatin a number of symbols, one after another, com!letely oblivious to the other system doin the same thin & Then they both attem!t to communicate with a third subsystem of my brainmy self-symboland it is at that !oint that the -I- inside my brain ets wind of what1s oin on, in other words, it starts !ickin u! a chunked descri!tion of the activities of those two subsystems&

Subsystems and Shared Code

Ty!ical subsystems mi ht be those that re!resent the !eo!le we know intimately& They are re!resented in such a com!le( way in our brains that their symbols enlar e to the rank of subsystem, becomin able to act autonomously, makin use of some resources in our brains for su!!ort& 0y this, I mean that a subsystem symboli<in a friend can activate many of the symbols in my brain .ust as I can& 'or instance, I can fire u! my subsystem

for a ood friend and virtually feel myself in his shoes, runnin throu h thou hts which he mi ht have, activatin symbols in sequences which reflect his thinkin !atterns more accurately than my own& It could be said that my model of this friend, as embodied in a subsystem of my brain, constitutes my own chunked descri!tion of his brain& :oes this subsystem include, then, a symbol for every symbol which I think is in his brain? That would be redundant& Probably the subsystem makes e(tensive use of symbols already !resent in my brain& 'or instance, the symbol for -mountain- in my brain can be borrowed by the subsystem, when it is activated& The way in which that symbol is then used by the subsystem will not necessarily be identical to the way it is used by my full brain& In !articular, if I am talkin with my friend about the Tien Shan mountain ran e in 7entral )sia 4neither of us havin been there5, and I know that a number of years a o he had a wonderful hikin e(!erience in the )l!s, then my inter!retation of his remarks will be colored in !art by my im!orted ima es of his earlier )l!ine e(!erience, since I will be tryin to ima ine how he visuali<es the area& In the vocabulary we have been buildin u! in this 7ha!ter, we could say that the activation of the -mountain- symbol in me is under control of my subsystem re!resentin him& The effect of this is to o!en u! a different window onto to my memories from the one which I normally usenamely, my -default o!tion- switches from the full ran e of my memories to the set of my memories of his memories& 8eedless to say, my re!resentations of his memories are only a!!ro(imations to his actual memories, which are com!le( modes of activation of the symbols in his brain, inaccessible to me& /y re!resentations of his memories are also com!le( modes of activation of my own symbolsthose for -!rimordial- conce!ts, such as rass, trees, snow, sky, clouds, and so on& These are conce!ts which I must assume are re!resented in him -identicallyto the way they are in me& I must also assume a similar re!resentation in him of even more !rimordial notions, the e(!eriences of ravity, breathin , fati ue, color, and so forth& Aess !rimordial but !erha!s a nearly universal human quality is the en.oyment of reachin a summit and seein a view& Therefore, the intricate !rocesses in my brain which are res!onsible for this en.oyment can be taken over directly by the friendsubsystem without much loss of fidelity& 3e could o on to attem!t to describe how I understand an entire tale told by my friend, a tale filled with many com!le(ities of human relationshi!s and mental e(!eriences& 0ut our terminolo y would quickly become inadequate& There would be tricky recursions connected with re!resentations in him of re!resentations in me of re!resentations in him of one thin and another& If mutual friends fi ured in the tale bein told, I would unconsciously look for com!romises between my ima e of his re!resentations of them, and my own ima es of them& Pure recursion would sim!ly be an ina!!ro!riate formalism for dealin with symbol amal ams of this ty!e& )nd I have barely scratched the surface2 3e !lainly lack the vocabulary today for describin the com!le( interactions that are !ossible between symbols& So let us sto! before we et bo ed down& 3e should note, however, that com!uter systems are be innin to run into some of the some kinds of com!le(ity, and therefore some of these notions have been iven names& 'or instance, my -mountain- symbol is analo ous to what in com!uter .ar on is

called shared 4or reentrant5 codecode which can be used by two or more se!arate timesharin !ro rams runnin on a sin le com!uter& The fact that activation of one symbol can have different results when it is !art of different subsystems can be e(!lained by sayin that its code is bein !rocessed by different inter!reters& Thus, the tri erin !atterns in the -mountain- symbol are not absolute= they are relative to the system within which the symbol is activated& The reality of such -subbrains- may seem doubtful to some& Perha!s the followin quote from /& 7& Escher, as he discusses how he creates his !eriodic !lanefillin drawin s, will hel! to make clear what kind of !henomenon I am referrin to,
3hile drawin I sometimes feel as if I were a s!iritualist medium, controlled by the creatures which I am con.urin u!& It is as if they themselves decide on the sha!e in which they choose to a!!ear& They take little account of my critical o!inion durin their birth and I cannot e(ert much influence on the measure of their develo!ment& They are usually very difficult and obstinate creatures&B

?ere is a !erfect e(am!le of the near-autonomy of certain subsystems of the brain, once they are activated& Escher1s subsystems seemed to him almost to be able to override his esthetic .ud ment& Ef course, this o!inion must be taken with a rain of salt, since those !owerful subsystems came into bein as a result of his many years of trainin and submission to !recisely the forces that molded his esthetic sensitivities& In short, it is wron to divorce the subsystems in Escher1s brain from Escher himself or from his esthetic .ud ment& They constitute a vital !art of his esthetic sense, where -he- is the com!lete bein of the artist&

The Self-Symbol and Consciousness

) very im!ortant side effect of the self-subsystem is that it can !lay the role of -soul-, in the followin sense, in communicatin constantly with the rest of the subsystems and symbols in the brain, it kee!s track of what symbols are active, and in what way& This means that it has to have symbols for mental activityin other words, symbols for symbols, and symbols for the actions of symbols& Ef course, this does not elevate consciousness or awareness to any -ma ical-, non!hysical level& )wareness here is a direct effect of the com!le( hardware and software we have described& Still, des!ite its earthly ori in, this way of describin awarenessas the monitorin of brain activity by a subsystem of the brain itselfseems to resemble the nearly indescribable sensation which we all know and call -consciousness-& 7ertainly one can see that the com!le(ity here is enou h that many une(!ected effects could be created& 'or instance, it is quite !lausible that a com!uter !ro ram with this kind of structure would make statements about itself which would have a reat deal of resemblance to statements which !eo!le commonly make about themselves& This includes insistin that it has free will, that it is not e(!licable as a -sum of its !arts-, and so on& 4En this sub.ect, see the article -/atter, /ind, and /odels- by /& /insky in his book -emantic 3nformation &rocessing&5

3hat kind of uarantee is there that a subsystem, such as I have here !ostulated, which re!resents the self, actually e(ists in our brains? 7ould a whole com!le( network of symbols such as has been described above evolve without a self-symbol evolvin ? ?ow could these symbols and their activities !lay out -isomor!hic- mental events to real events in the surroundin universe, if there were no symbol for the host or anism, )ll the stimuli comin into the system are centered on one small mass in s!ace& It would be quite a larin hole in a brain1s symbolic structure not to have a symbol for the !hysical ob.ect in which it is housed, and which !lays a lar er role in the events it mirrors than any other ob.ect& In fact, u!on reflection, it seems that the only way one could make sense of the world surroundin a locali<ed animate ob.ect is to understand the role of that ob.ect in relation to the other ob.ects around it& This necessitates the e(istence of a self-symbol= and the ste! from symbol to subsystem is merely a reflection of the im!ortance of the self-symbol, and is not a qualitative chan e&

0ur +irst Encounter 'ith Lucas

The E(ford !hiloso!her J& ;& Aucas 4not connected with the Aucas numbers described earlier5 wrote a remarkable article in "#D", entitled -/inds, /achines, and 6>del-& ?is views are quite o!!osite to mine, and yet he mana es to mi( many of the same in redients to ether in comin u! with his o!inions& The followin e(cer!t is quite relevant to what we have .ust been discussin ,
)t one1s first and sim!lest attem!ts to !hiloso!hi<e, one becomes entan led in questions of whether when one knows somethin one knows that one knows it, and what, when one is thinkin of oneself, is bein thou ht about, and what is doin the thinkin & )fter one has been !u<<led and bruised by this !roblem for a lon time, one learns not to !ress these questions, the conce!t of a conscious bein is, im!licitly, reali<ed to be different from that of an unconscious ob.ect& In sayin that a conscious bein knows somethin , we are sayin not only that he knows it, but that he knows that he knows it, and that he knows that he knows that he knows it, and so on, as lon as we care to !ose the question, there is, we reco ni<e, an infinity here, but it is not an infinite re ress in the bad sense, for it is the questions that !eter out, as bein !ointless, rather than the answers& The questions are felt to be !ointless because the conce!t contains within itself the idea of bein able to o on answerin such questions indefinitely& )lthou h conscious bein s have the !ower of oin on, we do not wish to e(hibit this sim!ly as a succession of tasks they are able to !erform, nor do we see the mind as an infinite sequence of selves and su!er-selves and su!er-su!er-selves& ;ather, we insist that a conscious bein is a unity, and thou h we talk about !arts of the mind, we do so only as a meta!hor, and will not allow it to be taken literally& The !arado(es of consciousness arise because a conscious bein can be aware of itself as well as of other thin s, and yet cannot really be construed as bein divisible into !arts& It means that a conscious bein can deal with 6>delian questions in a way in which a machine cannot, because a conscious bein can both consider itself and its !erformance and yet not be other than that which did the !erformance& ) machine can be made in a manner of s!eakin to -consider- its !erformance, but it cannot take this -into account- without thereby becomin a different machine, namely the old machine with a -new !art- added& 0ut it is inherent in our idea of a conscious mind that it can reflect u!on itself and critici<e its own !erformances, and no e(tra !art is required to do this, it is already com!lete, and has no )chilles1 heel&

The thesis thus be ins to become more of a matter of conce!tual analysis than mathematical discovery& This is borne out by considerin another ar ument !ut forward by Turin & So far, we have constructed only fairly sim!le and !redictable artifacts& 3hen we increase the com!le(ity of our machines, there may, !erha!s, be sur!rises in store for us& ?e draws a !arallel with a fission !ile& 0elow a certain -critical- si<e, nothin much ha!!ens, but above the critical si<e, the s!arks be in to fly& So too, !erha!s, with brains and machines& /ost brains and all machines are, at !resent, -sub-critical-they react to incomin stimuli in a stod y and uninterestin way, have no ideas of their own, can !roduce only stock res!onses but a few brains at !resent, and !ossibly some machines in the future, are su!er-critical, and scintillate on their own account& Turin is su estin that it is only a matter of com!le(ity, and that above a certain level of com!le(ity a qualitative difference a!!ears, so that -su!er-criticalmachines will be quite unlike the sim!le ones hitherto envisa ed& This may be so& 7om!le(ity often does introduce qualitative differences& )lthou h it sounds im!lausible, it mi ht turn out that above a certain level of com!le(ity, a machine ceased to be !redictable, even in !rinci!le, and started doin thin s on its own account, or, to use a very revealin !hrase, it mi ht be in to have a mind of its own& It mi ht be in to have a mind of its own& It would be in to have a mind of its own when it was no lon er entirely !redictable and entirely docile, but was ca!able of doin thin s which we reco ni<ed as intelli ent, and not .ust mistakes or random shots, but which we had not !ro rammed into it& 0ut then it would cease to be a machine, within the meanin of the act& 3hat is at stake in the mechanist debate is not how minds are, or mi ht be, brou ht into bein , but how they o!erate& It is essential for the mechanist thesis that the mechanical model of the mind shall o!erate accordin to -mechanical !rinci!les,- that is, that we can understand the o!eration of the whole in terms of the o!erations of its !arts, and the o!eration of each !art either shall be determined by its initial state and the construction of the machine, or shall be a random choice between a determinate number of determinate o!erations& If the mechanist !roduces a machine which is so com!licated that this ceases to hold ood of it, then it is no lon er a machine for the !ur!oses of our discussion, no matter how it was constructed& 3e should say, rather, that he had created a mind, in the same sort of sense as we !rocreate !eo!le at !resent& There would then be two ways of brin in new minds into the world, the traditional way, by be ettin children born of women, and a new way by constructin very, very com!licated systems of, say, valves and relays& 3hen talkin of the second way& we should take care to stress that althou h what was created looked like a machine, it was not one really, because it was not .ust the total of its !arts& Ene could not tell what it was oin to do merely by knowin the way in which it was built u! and the initial state of its !arts, one could not even tell the limits of what it could do, for even when !resented with a 6>del-ty!e question, it ot the answer ri ht& In fact we should say briefly that any system which was not floored by the 6>del question was eo ipso not a Turin machine, i&e& not a machine within the meanin of the act&%

In readin this !assa e, my mind constantly bo les at the ra!id succession of to!ics, allusions, connotations, confusions, and conclusions& 3e .um! from a 7arrollian !arado( to 6>del to Turin to )rtificial Intelli ence to holism and reductionism, all in the s!an of two brief !a es& )bout Aucas one can say that he is nothin if not stimulatin & In the followin 7ha!ters, we shall come back to many of the to!ics touched on so tantali<in ly and fleetin ly in this odd !assa e&

Aria with )iverse *ariations

*chilles has (een una(le to sleep these past few nights! ,is friend the Tortoise has come o#er tonight, to keep him compan) during these anno)ing hours! Tortoise, I am so sorry to hear of the troubles that have been !la uin you, my dear )chilles& I ho!e my com!any will !rovide a welcome relief from all the unbearable stimulation which has ke!t you awake& Perha!s I will bore you sufficiently that you will at lon last o to slee!& In that way, I will be of some service& *chilles, Eh, no, I am afraid that I have already had some of the world1s finest bores try their hand at borin me to slee!and all, sad to say, to no avail& So you will be no match for them& 8o, /r& T, I invited you over ho!in that !erha!s you could entertain me with a little this or that, taken from number theory, so that I could at least while away these lon hours in an a reeable fashion& Hou see, I have found that a little number theory does wonders for my troubled !syche& Tortoise, ?ow quaint an idea2 Hou know, it reminds me, .ust a wee bit, of the story of !oor 7ount @aiserlin & *chilles, 3ho was he? Tortoise, Eh, he was a 7ount in Sa(ony in the ei hteenth centurya 7ount of no account, to tell the truthbut because of himwell, shall I tell you the story? It is quite entertainin & *chilles, In that case, by all means, do2 Tortoise, There was a time when the ood 7ount was sufferin from slee!lessness, and it .ust so ha!!ened that a com!etent musician lived in the same town, and so 7ount @aiserlin commissioned this musician to com!ose a set of variations to be !layed by the 7ount1s court har!sichordist for him durin his slee!less ni hts, to make the hours !ass by more !leasantly& *chilles, 3as the local com!oser u! to the challen e? Tortoise, I su!!ose so, for after they were done, the 7ount rewarded him most lucratively he !resented him with a old oblet containin one hundred Aouis d1or& *chilles, Hou don1t say2 I wonder where he came u!on such a oblet and all those Aouis d1or, in the first !lace& Tortoise& Perha!s he saw it in a museum, and took a fancy to it& *chilles, )re you su estin he absconded with it? Tortoise, 8ow, now, I wouldn1t !ut it e(actly that way, but&&& Those days, 7ounts could et away with most anythin & )nyway, it is clear that the 7ount was most !leased with the music, for he was constantly entreatin his har!sichordista mere lad of a fellow, name of 6oldber to !lay one or another of these thirty variations& 7onsequently 4and somewhat ironically5 the variations became attached to the name of youn 6oldber , rather than to the distin uished 7ount1s name& *chilles, Hou mean, the com!oser was 0ach, and these were the so-called -6oldber Gariations-?

Tortoise, :o I ever2 )ctually, the work was entitled *ria with .i#erse <ariations, of which there are thirty& :o you know how 0ach structured these thirty ma nificent variations? *chilles, :o tell& Tortoise, )ll the !iecese(ce!t the final oneare based on a sin le theme, which he called an -aria-& )ctually, what binds them all to ether is not a common melody, but a common harmonic round& The melodies may vary, but underneath, there is a constant theme& Enly in the last variation did 0ach take liberties& It is a sort of -!ostendin endin -& It contains e(traneous musical ideas havin little to do with the ori inal Themein fact, two 6erman folk tunes& That variation is called a -quodlibet-& *chilles, 3hat else is unusual about the 6oldber Gariations? Tortoise, 3ell, every third variation is a canon& 'irst a canon in two canoni<in voices enter on the S)/E note& Second, a canon in which one of the canoni<in voices enters E8E 8ETE ?I6?E; than the first& Third, one voice enters T3E notes hi her than the other& )nd so on, until the final canon has entries .ust e(actly one ninth a!art& Ten canons, all told& )nd *chilles, 3ait a minute& :on1t I recall readin somewhere or other about fourteen recently discovered 6oldber canons&&& Tortoise, :idn1t that a!!ear in the same .ournal where they recently re!orted the discovery of fourteen !reviously unknown days in 8ovember? *chilles, 8o, it1s true& ) fellow named 3olffa musicolo istheard about a s!ecial co!y of the 6oldber Gariations in Strasbour & ?e went there to e(amine it, and to his sur!rise, on the back !a e, as a sort of -!ost-endin endin -, he found these fourteen new canons, all based on the first ei ht notes of the theme of the 6oldber Gariations& So now it is known that there are in reality forty-four 6oldber Gariations, not thirty& Tortoise, That is, there are forty-four of them, unless some other musicolo ist discovers yet another batch of them in some unlikely s!ot& )nd althou h it seems im!robable, it is still !ossible, even if unlikely, that still another batch will be discovered, and then another one, and on and on and on&&& 3hy, it mi ht never sto!2 3e may never know if or when we have the full com!lement of 6oldber Gariations& *chilles, That is a !eculiar idea& Presumably, everybody thinks that this latest discovery was .ust a fluke, and that we now really do have all the 6oldber Gariations& 0ut .ust su!!osin that you are ri ht, and some turn u! sometime, we shall start to e(!ect this kind of thin & )t that !oint, the name -6oldber Gariations- will start to shift sli htly in meanin , to include not only the known ones, but also any others which mi ht eventually turn u!& Their numbercall it 1g1is certain to be finite, wouldn1t you a ree?but merely knowin that g is finite isn1t the same as knowin how bi g is& 7onsequently, this information won1t tell us when the last 6oldber Gariation has been located& Tortoise, That is certainly true& *chilles, Tell mewhen was it that 0ach wrote these celebrated variations? Tortoise, It all ha!!ened in the year "$CB, when he was 7antor in Aei!<i &

*chilles, "$CB? ?mm&&& That number rin s a bell& Tortoise, It ou ht to, for it ha!!ens to be a rather interestin number, bein a sum of two odd !rimes, "$B# and "%& *chilles, 0y thunder2 3hat a curious fact2 I wonder how often one runs across an even number with that !ro!erty& Aet1s see&&& D X %^% * X %^L "+ X %^$ X L^L "B X L^$ "C X %^"" X $^$ "D X % ^ "% X L ^ "" "* X L ^ "% X $ ^ "" B+ X % ^ "$ X $^ "% BB X % ^ "# X L ^ "$ X "" ^ "" BC X L ^ "# X $ ^ "$ X "" ^ "% BD X % ^ B% X $ ^ "# X "% ^ "% B* X L ^ B% X "" ^ "$ %+ X $ ^ B% X "" ^ "# X "% ^ "$ 8ow what do you knowaccordin to my little table here, it seems to be quite a common occurrence& Het I don1t discern any sim!le re ularity in the table so far& Tortoise, Perha!s there is no re ularity to be discerned& *chilles, 0ut of course there is2 I am .ust not clever enou h to s!ot it ri ht off the bat& Tortoise, Hou seem quite convinced of it& *chilles, There1s no doubt in my mind& I wonder&&& 7ould it be that )AA even numbers 4e(ce!t C5 can be written as a sum of two odd !rimes? Tortoise, ?mm&&& That question rin s a bell&&& )h, I know why2 Hou1re not the first !erson to ask that question& 3hy, as a matter of fact, in the year "$CB, a mathematical amateur !ut forth this very question in a *chilles, :id you say "$CB? E(cuse me for interru!tin , but I .ust noticed that "$CB ha!!ens to be a rather interestin number, bein a difference of two odd !rimes, "$C$ and L& Tortoise, 0y thunder2 3hat a curious fact2 I wonder how often one runs across an even number with that !ro!erty& *chilles, 0ut !lease don1t let me distract you from your story& Tortoise, Eh, yesas I was sayin , in "$CB& a certain mathematical amateur, whose name esca!es me momentarily, sent a letter to Euler, who at the time was at the court of @in 'rederick the 6reat in Potsdam, andwell, shall I tell you the story? It is not without charm& *chilles, In that case, by all means, do2 Tortoise, Gery well& In his letter, this dabbler in number theory !ro!ounded an un!roved con.ecture to the reat Euler, -Every even number can he re!resented as a sum of two odd !rimes&- 8ow what was that fellow1s name?

*chilles, I va uely recollect the story, from some number theory book or other& 3asn1t the fellow named -@u!fer >del-? Tortoise, ?mm&&& 8o, that sounds too lon & *chilles, 7ould it have been -Silberescher-? Tortoise, 8o, that1s not it, either& There1s a name on the ti! of my ton ueahahoh yes2 It was -6oldbach-2 6oldbach was the fellow& *chilles, I knew it was somethin like that& Tortoise, Hesyour uesses hel!ed .o my memory& It1s quite odd, how one occasionally has to hunt around in one1s memory as if for a book in a library without call numbers&&& 0ut let us et back to "$CB& *chilles, Indeed, let1s& I wanted to ask you, did Euler ever !rove that this uess by 6oldbach was ri ht? Tortoise, 7uriously enou h, he never even considered it worthwhile workin on& ?owever, his disdain was not shared by all mathematicians& In fact, it cau ht the fancy of many, and became known as the -6oldbach 7on.ecture-& *chilles, ?as it ever been !roven correct? Tortoise, 8o, it hasn1t& 0ut there have been some remarkable near misses& 'or instance, in "#%" the ;ussian number theorist Schnirelmann !roved that any numbereven or oddcan be re!resented as the sum of not more than %++,+++ !rimes& *chilles, 3hat a stran e result& Ef what ood is it? Tortoise, It has brou ht the !roblem into the domain of the finite& Previous to Schnirelmann1s !roof, it was conceivable that as you took lar er and lar er even numbers, they would require more and more !rimes to re!resent them& Some even number mi ht take a trillion !rimes to re!resent it2 8ow it is known that that is not soa sum of %++,+++ !rimes 4or fewer5 will always suffice& *chilles, I see& Tortoise, Then in "#%$, a sly fellow named Gino radova ;ussian toomana ed to establish somethin far closer to the desired result, namely, every sufficiently lar e E:: number can be re!resented as a sum of no more than T?;EE odd !rimes& 'or e(am!le, "#%$ X DC" ^ DC% ^ DL%& 3e could say that an odd number which is re!resentable as a sum of three odd !rimes has -the Gino radov !ro!erty& Thus, all sufficiently lar e odd numbers have the Gino radov !ro!erty& *chilles, Gery wellbut what does -sufficiently lar e- mean? Tortoise, It means that some finite number of odd numbers may fail to have the Gino radov !ro!erty, but there is a numbercall it 1#1beyond which all odd numbers have the Gino radov !ro!erty& 0ut Gino radov was unable to say how bi # is& So in a way, # is like g, the finite but unknown number of 6oldber Gariations& /erely knowin that # is finite isn1t the same as knowin how bi # is& 7onsequently, this information won1t tell us when the last odd number which needs more than three !rimes to re!resent it has been located& *chilles, I see& )nd so any sufficiently lar e even number BN can be re!resented as a sum of 'EF; !rimes, by first re!resentin BN - % as a sum of three !rimes, and then addin back the !rime number %&

Tortoise, Precisely& )nother close a!!roach is contained in the Theorem which says, -)ll even numbers can be re!resented as a sum of one !rime and one number which is a !roduct of at most two !rimes&*chilles, This question about sums of two !rimes certainly leads you into stran e territory& I wonder where you would be led if you looked at :I''E;E87ES of two odd !rimes& I1ll bet I could lean some insi ht into this teaser by makin a little table of even numbers, and their re!resentations as differences of two odd !rimes, .ust as I did for sums& Aet1s see &&& B X L - %, C X $ - %, D X "" - L, * X "" - %, "+ X"% - %, $ - L, "" - $, "% - $, "% - L, "$ - $, "% - "", "$ - "%, "$ - "", "# - "", B% - "%, "# - "$, etc& B% - "#, etc& "# - "%, etc& %" - B%, etc& B# - "#, etc&

/y racious2 There seems to be no end to the number of different re!resentations I can find for these even numbers& Het I don1t discern any sim!le re ularity in the table so far& Tortoise, Perha!s there is no re ularity to be discerned& *chilles, Eh, you and your constant rumblin s about chaos2 I1ll hear none of that, thank you& Tortoise, :o you su!!ose that EGE;H even number can be re!resented somehow as the difference of two odd !rimes? *chilles, The answer certainly would a!!ear to be yes, from my table& 0ut then a ain, I su!!ose it could also be no& That doesn1t really et us very far, does it? Tortoise, 3ith all due res!ect, I would say there are dee!er insi hts to be had on the matter& *chilles, 7urious how similar this !roblem is to 6oldbach1s ori inal one& Perha!s it should be called a -6oldbach Gariation-& Tortoise, Indeed& 0ut you know, there is a rather strikin difference between the 6oldbach 7on.ecture, and this 6oldbach Gariation, which I would like to tell you about& Aet us say that an even number B N has the -6oldbach !ro!erty- if it is the SF/ of two odd !rimes, and it has the -Tortoise !ro!erty- if it is the :I''E;E87E of two odd !rimes& *chilles, I think you should call it the -)chilles !ro!erty-& )fter all, I su ested the !roblem& Tortoise, I was .ust about to !ro!ose that we should say that a number which A)7@S the Tortoise !ro!erty has the -)chilles !ro!erty-& *chilles, 3ell, all ri ht&&& Tortoise, 8ow consider, for instance, whether " trillion has the 6oldbach !ro!erty or the Tortoise !ro!erty& Ef course, it may have both& *chilles, I can consider it, but I doubt whether I can ive you an answer to either question&

Tortoise, :on1t ive u! so soon& Su!!ose I asked you to answer one or the other question& 3hich one would you !ick to work on? *chilles, I su!!ose I would fli! a coin& I don1t see much difference between them& Tortoise, )ha2 0ut there1s a world of difference2 If you !ick the 6oldbach !ro!erty, involvin SF/S of !rimes, then you are limited to usin !rimes which are bounded between B and " trillion, ri ht? *chilles, Ef course& Tortoise, So your search for a re!resentation for " trillion as a sum of two !rimes is 6F);)8TEE: TE TE;/I8)TE& *chilles, )hhh2 I see your !oint& 3hereas if I chose to work on re!resentin " trillion as the :I''E;E87E of two !rimes, I would not have any bound on the si<e of the !rimes involved& They mi ht be so bi that it would take me a trillion years to find them& Tortoise, Er then a ain, they mi ht not even ETIST2 )fter all, that1s what the question was askin do such !rimes e(ist, It wasn1t of much concern how bi they mi ht turn out to be& *chilles, Hou1re ri ht& If they didn1t e(ist, then a search !rocess would lead on forever, never answerin yes, and never answerin no& )nd nevertheless, the answer would be no& Tortoise, So if you have some number, and you wish to test whether it has the 6oldbach !ro!erty or the Tortoise !ro!erty, the difference between the two tests will be this, in the former, the search involved is 6F);)8TEE: TE TE;/I8)TE= in the latter, it is PETE8TI)AAH E8:AESSthere are no uarantees of any ty!e& It mi ht .ust o merrily on forever, without yieldin an answer& )nd yet, on the other hand, in some cases, it mi ht sto! on the first ste!& *chilles, I see there is a rather vast difference between the 6oldbach and Tortoise !ro!erties& Tortoise, Hes, the two similar !roblems concern these vastly different !ro!erties& The 6oldbach 7on.ecture is to the effect that all even numbers have the 6oldbach !ro!erty= the 6oldbach Gariation su ests that all even numbers have the Tortoise !ro!erty& 0oth !roblems are unsolved, but what is interestin is that althou h they sound very much alike, they involve !ro!erties of whole numbers which are quite different& *chilles, I see what you mean& The 6oldbach !ro!erty is a detectable, or reco ni<able !ro!erty of any even number, since I know how to test for its !resence.ust embark on a search& It will automatically come to an end with a yes or no answer& The Tortoise !ro!erty, however, is more elusive, since a brute force search .ust may never ive an answer& Tortoise, 3ell, there may be cleverer ways of searchin in the case of the Tortoise !ro!erty, and maybe followin one of them would always come to an end, and yield an answer& *chilles, 7ouldn1t the search only end if the answer were yes? Tortoise, 8ot necessarily& There mi ht be some way of !rovin that whenever the search lasts lon er than a certain len th of time, then the answer must be no& There mi ht even be some ET?E; way of searchin for the !rimes, not such a brute force way,

which is uaranteed to find them if they e(ist, and to tell if they don1t& In either case, a finite search would be able to yield the answer no& 0ut I don1t know if such a thin can be !roven or not& Searchin throu h infinite s!aces is always a tricky matter, you know& *chilles, So as thin s stand now, you know of no test for the Tortoise !ro!erty which is uaranteed to terminateand yet there /I6?T e(ist such a search& Tortoise, ;i ht& I su!!ose one could embark on a search for such a search, but I can ive no uarantee that that -meta-search- would terminate, either& *chilles, Hou know, it strikes me as quite !eculiar that if some even numberfor e(am!le, a trillionfailed to have the Tortoise !ro!erty, it would be caused by an infinite number of se!arate !ieces of information& It1s funny to think of wra!!in all that information u! into one bundle, and callin it, as you so allantly su ested, -the )chilles !ro!erty- of " trillion& It is really a !ro!erty of the number system as a 3?EAE, not .ust of the number " trillion& Tortoise, That is an interestin observation, )chilles, but I maintain that it makes a ood deal of sense to attach this fact to the number " trillion nevertheless& 'or !ur!oses of illustration, let me su est that you consider the sim!ler statement -B# is !rime-& 8ow in fact, this statement really means that B times B is not B#, and L times D is not B#, and so forth, doesn1t it? *chilles, It must, I su!!ose& Tortoise, 0ut you are !erfectly ha!!y to collect all such facts to ether, and attach them in a bundle to the number B#, sayin merely, -B# is !rime-? *chilles, Hes&&& Tortoise, )nd the number of facts involved is actually infinite, isn1t it,, )fter all, such facts as -CCCC times %%%% is not B#- are all !art of it, aren1t they? *chilles, Strictly s!eakin , I su!!ose so& 0ut you and I both know that you can1t !roduce B# by multi!lyin two numbers which are both bi er than B#& So in reality, sayin -B# is !rime- is only summari<in a 'I8ITE number of facts about multi!lication& Tortoise, Hou can !ut it that way if you want, but think of this, the fact that two numbers which are bi er than B# can1t have a !roduct equal to B# involves the entire structure of the number system& In that sense, that fact in itself is a summary of an infinite number of facts& Hou can1t et away from the fact, )chilles, that when you say -B# is !rime-, you are actually statin an infinite number of thin s& *chilles, /aybe so, but it feels like .ust one fact to me& Tortoise, That1s because an infinitude of facts are contained in your !rior knowled e they are embedded im!licitly in the way you visuali<e thin s& Hou don1t see an e(!licit infinity because it is ca!tured im!licitly inside the ima es you mani!ulate& *chilles, I uess that you1re ri ht& It still seems odd to lum! a !ro!erty of the entire number system into a unit, and label the unit -!rimeness of B#-& Tortoise, Perha!s it seems odd, but it is also quite a convenient way to look at thin s& 8ow let us come back to your hy!othetical idea& If, as you su ested, the number " trillion has the )chilles !ro!erty, then no matter what !rime you add to it, you do not et another !rime& Such a state of affairs would be caused by an infinite number of se!arate mathematical -events-& 8ow do all these -events- necessarily s!rin from

the same source? :o they have to have a common cause? 0ecause if they don1t, then some sort of -infinite coincidence- has created the fact, rather than an underlyin re ularity& *chilles, )n -infinite coincidence-? )mon the natural numbers, 8ET?I86 is coincidental nothin ha!!ens without there bein some underlyin !attern& Take $, instead of a trillion& I can deal with it more easily, because it is smaller& $ has the )chilles !ro!erty& Tortoise, Hou1re sure? *chilles, Hes& ?ere1s why& If you add B to it, you et #, which isn1t !rime& )nd if you add any other !rime to $, you are addin two odd numbers, resultin in an even number thus you a ain fail to et a !rime& So here the -)chilleanity- of $, to coin a term, is a consequence of .ust T3E reasons, a far cry from any -infinite coincidence-& 3hich .ust oes to su!!ort my assertion, that it never takes an infinite number of reasons to account for some arithmetical truth& If there 3E;E some arithmetical fact which were caused by an infinite collection of unrelated coincidences, then you could never ive a finite !roof for that truth& )nd that is ridiculous& Tortoise, That is a reasonable o!inion, and you are in ood com!any in makin it& ?owever *chilles, )re there actually those who disa ree with this view? Such !eo!le would have to believe that there are -infinite coincidences-, that there is chaos in the midst of the most !erfect, harmonious, and beautiful of all creations, the system of natural numbers& Tortoise, Perha!s they do= but have you ever considered that such chaos mi ht be an inte ral !art of the beauty and harmony? *chilles, 7haos, !art of !erfection? Erder and chaos make a !leasin unity? ?eresy2 Tortoise, Hour favorite artist, /& 7& Escher, has been known to su est such a heretical !oint of view in one of his !ictures&&& )nd while we1re on the sub.ect of chaos, I believe that you mi ht be interested in hearin about two different cate ories of search, both of which are uaranteed to terminate& *chilles, 7ertainly& Tortoise, The first ty!e of searchthe non-chaotic ty!eis e(em!lified by the test involved in checkin for the 6oldbach !ro!erty& Hou .ust look at !rimes less than BN, and if some !air adds u! to B N, then BN has the 6oldbach !ro!erty= otherwise, it doesn1t& This kind of test is not only sure to terminate, but you can !redict 0H 3?E8 it will terminate, as well& *chilles, So it is a P;E:I7T)0AH TE;/I8)TI86 test& )re you oin to tell me that checkin for some number-theoretical !ro!erties involves tests which are uaranteed to terminate, but about which there is no way to know in advance how lon they will take? Tortoise, ?ow !ro!hetic of you, )chilles& )nd the e(istence of such tests shows that there is intrinsic chaos, in a certain sense, in the natural number system& *chilles, 3ell, in that case, I would have to say that !eo!le .ust don1t know enou h about the test& If they did a little more research, they could fi ure out how lon it will take, at most, before it terminates& )fter all, there must always be some rhyme or reason to

73G5=E CB! Erder and 7haos, () '! 8! Escher lithograph, BIGK/! the !atterns amon inte ers& There can1t .ust be chaotic !atterns which defy !rediction2 Tortoise, I can understand your intuitive faith, )chilles& ?owever, it1s not always .ustified& Ef course, in many cases you are e(actly ri ht.ust because somebody doesn1t know somethin , one can1t conclude that it is unknowable2 0ut there are certain !ro!erties of inte ers for which terminatin tests can be !roven to e(ist, and yet about which it can also be P;EGE8 that there is no way to !redict in advance how lon they will take& *chilles, I can hardly believe that& It sounds as if the devil himself mana ed to sneak in and throw a monkey wrench into 6od1s beautiful realm of natural numbers2 Tortoise, Perha!s it will comfort you to know that it is by no means easy, or natural, to define a !ro!erty for which there is a terminatin but not P;E:I7T)0AH terminatin test& /ost -natural- !ro!erties of inte ers do admit of !redictably terminatin tests& 'or e(am!le, !rimeness& squareness, bein a !ower of ten, and so on& *chilles, Hes, I can see that those !ro!erties are com!letely strai htforward to test for& 3ill you tell me a !ro!erty for which the only !ossible test is a terminatin but non!redictable one?

Tortoise, That1s too com!licated for me in my slee!y state& Aet me instead show you a !ro!erty which is very easy to define, and yet for which no terminatin test is known& I1m not sayin there won1t ever be one discovered, mind you .ust that none is known& Hou be in with a numberwould you care to !ick one? *chilles, ?ow about "L? Tortoise, )n e(cellent choice& 3e be in with your number, and if it is E::, we tri!le it, and add "& If it is EGE8, we take half of it& Then we re!eat the !rocess& 7all a number which eventually reaches " this way a 3E8:;EFS number, and a number which doesn1t, an F83E8:;EFS number *chilles, Is "L wondrous, or unwondrous? Aet1s see, "L CD B% $+ %L "+D L% "D+ *+ C+ B+ "+ L "D * C B is E::, so I make %n ^ ", is EGE8, so I take half, is E::, so I make %n ^ ", is EGE8, so I take half, is E::, so I make %n ^ ", is EGE8, so I take half, is E::, so I make %n ^ ", is EGE8, so I take half, is EGE8, so I take half, is EGE8, so I take half, is EGE8, so I take half, is EGE8, so I take half, is E::, so I make %n ^ ", is EGE8, so I take half, is EGE8, so I take half, is EGE8, so I take half, is EGE8, so I take half, CD B% $+ %L "+D L% "D+ *+ C+ B+ "+ L "D * C B "

3ow2 That1s quite a roundabout .ourney, from "L to "& 0ut I finally reached it& That shows that "L has the !ro!erty of bein wondrous& I wonder what numbers are F8wondrous&&& Tortoise, :id you notice how the numbers swun u! and down, in this sim!ly defined !rocess? *chilles, Hes& I was !articularly sur!rised, after thirteen turns, to find myself at "D, only one reater than "L, the number I started with& In one sense, I was almost back where I startedyet in another sense, I was nowhere near where I had started& )lso, I found it quite curious that I had to o as hi h as "D+ to resolve the question& I wonder how come& Tortoise, Hes, there is an infinite -sky- into which you can sail, and it is very hard to know in advance how hi h into the sky you will wind u! sailin & Indeed, it is quite !lausible that you mi ht .ust sail u! and u! and u!, and never come down&

*chilles, ;eally? I uess that is conceivablebut what a weird coincidence it would require2 Hou1d .ust have to hit odd number after odd number, with only a few evens mi(ed in& I doubt if that would ever ha!!enbut I .ust don1t know for sure& Tortoise, 3hy don1t you try startin with B$? /ind you, I don1t !romise anythin & 0ut sometime, .ust try it, for your amusement& )nd I1d advise you to brin alon a rather lar e sheet of !a!er& *chilles, ?mm&&& Sounds interestin & Hou know, it still makes me feel funny to associate the wondrousness 4or unwondrousness5 with the startin number, when it is so obviously a !ro!erty of the entire number system& Tortoise, I understand what you mean, but it1s not that different from sayin -B# is !rimeor - old is valuable-both statements attribute to a sin le entity a !ro!erty which it has only by virtue of bein embedded in a !articular conte(t& *chilles, I su!!ose you1re ri ht& This -wondrousness- !roblem is wondrous tricky, because of the way in which the numbers oscillatenow increasin , now decreasin & The !attern EF6?T to be re ular, yet on the surface it a!!ears to be quite chaotic& Therefore, I can well ima ine why, as of yet, no one knows of a test for the !ro!erty of wondrousness which is uaranteed to terminate& Tortoise, S!eakin of terminatin and nonterminatin !rocesses, and those which hover in between, I am reminded of a friend of mine, an author, who is at work on a book& *chilles, Eh, how e(citin 2 3hat is it called? Tortoise, 8opper, -il#er, Gold: an 3ndestructi(le 'etallic *llo)& :oesn1t that sound interestin ? *chilles, 'rankly, I1m a little confused by the title& )fter all, what do 7o!!er, Silver, and 6old have to do with each other? Tortoise, It seems clear to me& *chilles, 8ow if the title were, say, Giraffes, -il#er, Gold, or 8opper, Elephants, Gold, why, I could see it&&& Tortoise, Perha!s you would !refer 8opper, -il#er, Ba(oons? *chilles, Eh, absolutely2 0ut that ori inal title is a loser& 8o one would understand it& Tortoise, I1ll tell my friend& ?e1ll be deli hted to have a catchier title 4as will his !ublisher5& *chilles, I1m lad& 0ut how were you reminded of his book by our discussion? Tortoise, )h, yes& Hou see, in his book there will be a :ialo ue in which he wants to throw readers off by makin them SE);7? for the endin & *chilles, ) funny thin to want to do& ?ow is it done? Tortoise, Hou1ve undoubtedly noticed how some authors o to so much trouble to build u! reat tension a few !a es before the end of their storiesbut a reader who is holdin the book !hysically in his hands can 'EEA that the story is about to end& ?ence, he has some e(tra information which acts as an advance warnin , in a way& The tension is a bit s!oiled by the !hysicality of the book& It would be so much better if, for instance, there were a lot of !addin at the end of novels& *chilles, Paddin ?

Tortoise, Hes= what I mean is, a lot of e(tra !rinted !a es which are not !art of the story !ro!er, but which serve to conceal the e(act location of the end from a cursory lance, or from the feel of the book& *chilles, I see& So a story1s true endin mi ht occur, say, fifty or a hundred !a es before the !hysical end of the book? Tortoise, Hes& This would !rovide an element of sur!rise, because the reader wouldn1t know in advance how many !a es are !addin , and how many are story& *chilles, If this were standard !ractice, it mi ht be quite effective& 0ut there is a !roblem& Su!!ose your !addin were very obvioussuch as a lot of blanks, or !a es covered with T1s or random letters& Then, it would be as ood as absent& Tortoise, 6ranted& Hou1d have to make it resemble normal !rinted !a es& *chilles, 0ut even a cursory lance at a normal !a e from one story will often suffice to distin uish it from another story& So you will have to make the !addin resemble the enuine story rather closely& Tortoise, That1s quite true& The way I1ve always envisioned it is this, you brin the story to an end= then without any break, you follow it with somethin which looks like a continuation but which is in reality .ust !addin , and which is utterly unrelated to the true theme& The !addin is, in a way, a -!ost-endin endin -& It may contain e(traneous literary ideas, havin little to do with the ori inal theme& *chilles, Sneaky2 0ut then the !roblem is that you won1t be able to tell when the real endin comes& It1ll .ust blend ri ht into the !addin & Tortoise, That1s the conclusion my author friend and I have reached as well& It1s a shame, for I found the idea rather a!!ealin & *chilles, Say, I have a su estion& The transition between enuine story and !addin material could be made in such a way that, by sufficiently assiduous ins!ection of the te(t, an intelli ent reader will be able to detect where one leaves off and the other be ins& Perha!s it will take him quite a while& Perha!s there will be no way to !redict how lon it will take&&& 0ut the !ublisher could ive a uarantee that a sufficiently assiduous search for the true endin will always terminate, even if he can1t say how lon it will be before the test terminates& Tortoise, Gery wellbut what does -sufficiently assiduous- mean? *chilles, It means that the reader must be on the lookout for some small but telltale feature in the te(t which occurs at some !oint& That would si nal the end& )nd he must be in enious enou h to think u!, and hunt for, many such features until he finds the ri ht one& Tortoise, Such as a sudden shift of letter frequencies or word len ths? Er a rash of rammatical mistakes? *chilles, Possibly& Er a hidden messa e of some sort mi ht reveal the true end to a sufficiently assiduous reader& 3ho knows? Ene could even throw in some e(traneous characters or events which are inconsistent with the s!irit of the fore oin story& ) naPve reader would swallow the whole thin , whereas a so!histicated reader would be able to s!ot the dividin line e(actly& Tortoise, That1s a most ori inal idea, )chilles& I1ll relay it to my friend, and !erha!s he can incor!orate it in his :ialo ue&

*chilles, I would be hi hly honored& Tortoise, 3ell, I am afraid that I myself am rowin a little ro y, )chilles& It would be well for me to take my leave, while I am still ca!able of navi atin my way home& *chilles, I am most flattered that you have stayed u! for so lon , and at such an odd hour of the ni ht, .ust for my benefit& I assure you that your number-theoretical entertainment has been a !erfect antidote to my usual tossin and turnin & )nd who knows!erha!s I may even be able to o to slee! toni ht& )s a token of my ratitude, /r& T, I would like to !resent you with a s!ecial ift& Tortoise, Eh, don1t be silly, )chilles& *chilles, It is my !leasure, /r& T& 6o over to that dresser= on it, you will see an )sian bo(& The Tortoise mose)s o#er to *chilles' dresser!/ Tortoise& Hou don1t mean this very old )sian bo(, do you? *chilles, That1s the one& Please acce!t it, /r& T, with my warmest com!liments& Tortoise, Thank you very much indeed, )chilles& ?mm&&& 3hy are all these mathematicians1 names en raved on the to!? 3hat a curious list, De/or an )bel 0oole 0r o u w e r Sier inski 3eierstrass *chilles, I believe it is su!!osed to be a 7om!lete Aist of )ll 6reat /athematicians& 3hat I haven1t been able to fi ure out is why the letters runnin down the dia onal are so much bolder& Tortoise, )t the bottom it says, -Subtract " from the dia onal, to find 0ach in Aei!<i -& *chilles, I saw that, but I couldn1t make head or tail of it& Say, how about a shot of e(cellent whiskey? I ha!!en to have some in that decanter on my shelf& Tortoise, 8o, thanks& I1m too tired& I1m .ust oin to head home& 8asuall), he opens the (o9!/ Say, wait a moment, )chillesthere are one hundred Aouis d1or in here2 *chilles, I would be most !leased if you would acce!t them, /r& T& Tortoise, 0utbut *chilles, 8o ob.ections, now& The bo(, the oldthey1re yours& )nd thank you for an evenin without !arallel& Tortoise, 8ow whatever has come over you, )chilles? 3ell, thank you for your outstandin enerosity and I ho!e you have sweet dreams about the stran e 6oldbach 7on.ecture, and its Gariation& 6ood ni ht& *nd he picks up the #er) gold *sian (o9 filled with the one hundred +ouis d'or, and walks towards the door! *s he is a(out to lea#e, there is a loud knock!/

3ho could be knockin at this un odly hour, )chilles? *chilles, I haven1t the fo iest idea& It seems sus!icious to me& 3hy don1t you o hide behind the dresser, in case there1s any funny business& Tortoise, 6ood idea& -cram(les in (ehind the dresser!/ *chilles, 3ho1s there? <oice, E!en u!it1s the co!s& *chilles, 7ome in, it1s o!en& Two (url) policemen walk in, wearing shin) (adges!/ 8op, I1m Silva& This is 6ould& 4&oints at his (adge!5 Is there an )chilles at this address? *chilles, That1s me2 8op, 3ell, )chilles, we have reason to believe that there is a very old )sian bo( here, filled with one hundred Aouis d1or& Someone absconded with it from the museum this afternoon& *chilles, ?eavens to 0etsy2 8op, If it is here, )chilles, since you would be the only !ossible sus!ect, I re ret to say that I should have to take you into custody& 8ow I have here a search warrant *chilles, Eh, sirs, am I ever lad you arrived2 )ll evenin lon , I have been bein terrori<ed by /r& Tortoise and his very )sian old bo(& 8ow at last you have come to liberate me2 Please, sirs, .ust take a look behind that dresser, and there you will find the cul!rit2 The cops look (ehind the dresser and sp) the Tortoise huddled (ehind it, holding his #er) gold *sian (o9, and trem(ling!/ 8op, So there it is2 )nd so /r& Tortoise is the varmint, eh? I never would have sus!ected ?I/& 0ut he1s cau ht, red-handed& *chilles, ?aul the villain away, kind sirs2 Thank oodness, that1s the last I1ll have to hear of him, and the Gery )sian 6old 0o(2

C/A,TE$ <III

Bloo, and +loo, and Gloo,

Self-A'areness and Chaos
0AEEP, 'AEEP, )8: 6AEEP are not trolls, talkin ducks, or the sounds made by a sinkin shi!they are three com!uter lan ua es, each one with its own s!ecial !ur!ose& These lan ua es were invented s!ecially for this cha!ter& They will be of use in e(!lainin some new senses of the word -recursive-in !articular, the notions of primiti#e recursi#it) and general recursi#it)& They will !rove very hel!ful in clarifyin the machinery of selfreference in T8T& 3e seem to be makin a rather abru!t transition from brains and minds to technicalities of mathematics and com!uter science& Thou h the transition is abru!t in some ways, it makes some sense& 3e .ust saw how a certain kind of self-awareness seems to be at the cru( of consciousness& 8ow we are oin to scrutini<e -self-awareness- in more formal settin s, such as T8T& The ulf between T8T and a mind is wide, but some of the ideas will be most illuminatin , and !erha!s meta!horically trans!ortable back to our thou hts about consciousness& Ene of the ama<in thin s about T8T1s self-awareness is that it is intimately connected to questions about order versus chaos amon the natural numbers& In !articular, we shall see that an orderly system of sufficient com!le(ity that it can mirror itself cannot be totall) orderlyit must contain some stran e, chaotic features& 'or readers who have some )chilles in them, this will be hard to take& ?owever, there is a -ma ical- com!ensation, there is a kind of order to the disorder, which is now its own field of study, called -recursive function theory-& Fnfortunately, we will not be able to do much more than hint at the fascination of this sub.ect&

$e!resentability and $efrigerators

Phrases such as -sufficiently com!le(-, -sufficiently !owerful- and the like have cro!!ed u! quite often earlier& Just what do they mean? Aet us o back to the battle of the 7rab and Tortoise, and ask, -3hat qualifies somethin as a record !layer?- The 7rab mi ht claim that his refri erator is a -Perfect- record !layer& Then to !rove it, he could set any record whatsoever ato! it, and say, -Hou seeit1s !layin it2- The Tortoise, if he wanted to counter this 9en-like act, would have to re!ly, -8oyour refri erator is too lowfidelity to be counted as a !hono ra!h, it cannot re!roduce sounds at all 4let alone its self-breakin sound5&- The Tortoise can only make a record called -I 7annot 0e Played on ;ecord Player T- !rovided that ;ecord Player T is really a record !layer2 The Tortoise1s method is quite insidious, as it !lays on the stren th, rather than on the weakness, of the system& )nd therefore he requires -sufficiently hi-fi- record !layers&

:itto for formal versions of number theory& The reason that T8T is a formali<ation of 8 is that its symbols act the ri ht way, that is, its theorems are not silent like a refri eratorthey s!eak actual truths of 8& Ef course, so do the theorems of the !q-system& :oes it, too, count as -a formali<ation of number theory-, or is it more like a refri erator? 3ell, it is a little better than a refri erator, but it is still !retty weak& The !qsystem does not include enou h of the core truths of 8 to count as -a number theory-& 3hat, then, are these -core truths- of 8? They are the primiti#e recursi#e truths= that means they involve only predicta(l) terminating calculations& These core truths serve for 8 as Euclid1s first four !ostulates served for eometry, they allow you to throw out certain candidates before the ame be ins, on the rounds of -insufficient !ower-& 'rom here on out, the representa(ilit) of all primiti#e recursi#e truths will be the criterion for callin a system -sufficiently !owerful-&

Gant%s A- in Metamathematics
The si nificance of the notion is shown by the followin key fact, If you have a sufficiently !owerful formali<ation of number theory, then 6>del1s method is a!!licable, and consequently your system is incomplete& If, on the other hand, your system is not sufficiently !owerful 4i&e&, not all !rimitive recursive truths are theorems5, then your system is, !recisely by virtue of that lack, incomplete& ?ere we have a reformulation of -6antR1s )(- in metamathematics, whatever the system does, 6>del1s )( will cho! its head off2 8otice also how this com!letely !arallels the hi h-fidelity-versus-low-fidelity battle in the 8ontracrostipunctus& )ctually, it turns out that much weaker systems are still vulnerable to the 6>del method= the criterion that all !rimitive recursive truths need be re!resented as theorems is far too strin ent& It is a little like a thief who will only rob -sufficiently rich- !eo!le, and whose criterion is that the !otential victim should be carryin at least a million dollars in cash& In the case of T8T, luckily, we will be able to act in our ca!acity as thieves, for the million in cash is therewhich is to say, T8T does indeed contain all !rimitive recursive truths as theorems& 8ow before we !lun e into a detailed discussion of !rimitive recursive functions and !redicates, I would like to tie the themes of this 7ha!ter to themes from earlier 7ha!ters, so as to !rovide a bit better motivation&

+inding 0rder by Choosing the $ight +ilter

3e saw at a very early sta e that formal systems can be difficult and unruly beasts because they have len thenin and shortenin rules, which can !ossibly lead to neverendin searches amon strin s& The discovery of 6>del-numberin showed that any search for a strin havin a s!ecial ty!o ra!hical !ro!erty has an arithmetical cousin, an isomor!hic search for an inte er with a corres!ondin s!ecial arithmetical !ro!erty& 7onsequently, the quest for decision !rocedures for formal systems involves solvin the mystery of un!redictably lon searcheschaosOamon the inte ers& 8ow in the *ria with .i#erse <ariations, I ave !erha!s too much wei ht to a!!arent manifestations of

chaos in !roblems about inte ers& )s a matter of fact, !eo!le have tamed wilder e(am!les of a!!arent chaos than the -wondrousness- !roblem, findin them to be quite entle beasts after all& )chilles1 !owerful faith in the re ularity and !redictability of numbers should therefore be accorded quite a bit of res!ectes!ecially as it reflects the beliefs of nearly all mathematicians u! till the "#%+1s& To show why order versus chaos is such a subtle and si nificant issue, and to tie it in with questions about the location and revelation of meanin , I would like to quote a beautiful and memorable !assa e from *re 4uanta =eal$Oa 6alilean :ialo ue by the late J& /& Jauch,
S)AGI)TI Su!!ose I ive you two sequences of numbers, such as $ * L % # * " D % % # $ C C * % + # D " L D D + * C&&& and ", -"I%, ^"IL, -"I$, ^"I#, -"I"", ^"I"%, -"I"L,&&& If I asked you, Sim!licio, what the ne(t number of the first sequence is, what would you say? SI/PAI7IE I could not tell you& I think it is a random sequence and that there is no law in it& S)AGI)TI )nd for the second sequence? SI/PAI7IE That would be easy& It must be ^"I"$& S)AGI)TI ;i ht& 0ut what would you say if I told you that the first sequence is also constructed by a law and this law is in fact identical with the one you have .ust discovered for the second sequence? SI/PAI7IE This does not seem !robable to me& S)AGI)TI 0ut it is indeed so, since the first sequence is sim!ly the be innin of the decimal fraction Ue(!ansionV of the sum of the second& Its value is mIC& SI/PAI7IE Hou are full of such mathematical tricks, but I do not see what this has to do with abstraction and reality& S)AGI)TI The relationshi! with abstraction is easy to see& The first sequence looks random unless one has develo!ed throu h a !rocess of abstraction a kind of filter which sees a sim!le structure behind the a!!arent randomness& It is e(actly in this manner that laws of nature are discovered& 8ature !resents us with a host of !henomena which a!!ear mostly as chaotic randomness until we select some si nificant events, and abstract from their !articular, irrelevant circumstances so that they become ideali<ed& Enly then can they e(hibit their true structure in full s!lendor& S)6;E:E This is a marvelous idea2 It su ests that when we try to understand nature, we should look at the !henomena as if they were messages to be understood& E(ce!t that each messa e a!!ears to be random until we establish a code to read it& This code takes the form of an abstraction, that is, we choose to i nore certain thin s as irrelevant and we thus !artially select the content of the messa e by a free choice& These irrelevant si nals form the -back round noise,- which will limit the accuracy of our messa e& 0ut since the code is not absolute there may be several messa es in the same raw material of the data, so chan in the code will result in a messa e of equally dee! si nificance in somethin that was merely noise before, and conversely, In a new code a former messa e may be devoid of meanin & Thus a code !resu!!oses a free choice amon different, com!lementary as!ects, each of which has equal claim to realit), if I may use this dubious word& Some of these as!ects may be com!letely unknown to us now but they may reveal themselves to an observer with a different system of abstractions&

0ut tell me, Salviati, how can we then still claim that we disco#er somethin out there in the ob.ective real world? :oes this not mean that we are merely creatin thin s accordin to our own ima es and that reality is only within ourselves? S)AGI)TI I don1t think that this is necessarily so, but it is a question which requires dee!er reflection&"

Jauch is here dealin with messa es that come not from a -sentient bein - but from nature itself& The questions that we raised in 7ha!ter GI on the relation of meanin to messa es can be raised equally well with messa es from nature& Is nature chaotic, or is nature !atterned? )nd what is the role of intelli ence in determinin the answer to this question? To back off from the !hiloso!hy, however, we can consider the !oint about the dee! re ularity of an a!!arently random sequence& /i ht the function M4n5 from 7ha!ter G have a sim!le, nonrecursive e(!lanation, too? 7an every !roblem, like an orchard, be seen from such an an le that its secret is revealed? Er are there some !roblems in number theory which, no matter what an le they are seen from, remain mysteries? 3ith this !rolo ue, I feel it is time to move ahead to define the !recise meanin of the term -!redictably lon search-& This will be accom!lished in terms of the lan ua e 0looP&

,rimordial Ste!s of the Language Bloo,

Eur to!ic will be searches for natural numbers which have various !ro!erties& In order to talk about the length of any search, we shall have to define some !rimordial steps, out of which all searches are built, so that len th can be measured in terms of number of ste!s& Some ste!s which we mi ht consider !rimordial are, addin any two natural numbers= multi!lyin any two natural numbers= determinin if two numbers are equal= determinin the lar er 4smaller5 of two numbers&

Loo!s and &!!er Bounds

If we try to formulate a test for, say, !rimality in terms of such ste!s, we shall soon see that we have to include a control structureOthat is, descri!tions of the order to do thin s in, when to branch back and try somethin a ain, when to ski! over a set of ste!s, when to sto!, and similar matters& It is ty!ical of any algorithmOthat is, a s!ecific delineation of how to carry out a taskOthat it includes a mi(ture of 4"5 s!ecific o!erations to be !erformed, and 4B5 control statements& Therefore, as we develo! our lan ua e for e(!ressin !redictably lon calculations, we shall have to incor!orate !rimordial control structures also& In fact, the hallmark of 0looP is its limited set of control structures& It does not allow you to branch to arbitrary ste!s, or to re!eat rou!s of ste!s without limit= in 0looP, essentially the only

control structure is the bounded loo!, a set of instructions which can be e(ecuted over and over a ain, u! to a !redefined ma(imum number of times, called the upper (ound, or ceiling, of the loo!& If the ceilin were %++, then the loo! mi ht be e(ecuted +, $, or %++ timesObut not %+"& 8ow the e(act values of all the u!!er bounds in a !ro ram need not be !ut in numerically by the !ro rammerOindeed, they may not be known in advance& Instead, any u!!er bound may be determined by calculations carried out (efore its loo! is entered& 'or instance, if you wanted to calculate the value of B % , there would be two loo!s& 'irst, you evaluate %n, which involves n multi!lications& Then, you !ut B to that !ower, which involves %n multi!lications& Thus, the u!!er bound for the second loo! is the result of the calculation of the first loo!& ?ere is how you would e(!ress this in a 0looP !ro ram,
n

E+I;E ,$0CE &$E (T>0-T0-T/E-T/$EE-T0-T/E( A;B5 BL0CK ?5 BEGI; CELLC?@ DE 6F L00, ; TIMES5 BL0CK 65 BEGI; CELLC?@ DE G - CELLC?@F BL0CK 65 E; F CELLC6@ DE 6F L00, CELLC?@ TIMES5 BL0CK H5 BEGI; CELLC6@ DE H < CELLCl @F BL0CK H5 E; F 0&T,&T DE CELLC6@F BL0CK ?5 E; 1

Con"entions of Bloo,
8ow it is an acquired skill to be able to look at an al orithm written in a com!uter lan ua e, and fi ure out what it is doin & ?owever, I ho!e that this al orithm is sim!le enou h that it makes sense without too much scrutiny& ) procedure is defined, havin one in!ut !arameter, 8= its out!ut is the desired value& This !rocedure definition has what is called (lock structure, which means that certain !ortions of it are to be considered as units, or (locks& )ll the statements in a block et e(ecuted as a unit& Each block has a number 4the outermost bein BL0CK ?5, and is delimited by a BEGI; and an E; & In our e(am!le, BL0CK 6 and BL0CK H contain .ust one statement each but shortly you will see lon er blocks& ) L00, statement always means to e(ecute the block immediately under it re!eatedly& )s can be seen above, blocks can be nested&

The strate y of the above al orithm is as described earlier& Hou be in by takin an au(iliary variable, called CELLC?@= you set it initially to ", and then, in a loo!, you multi!ly it re!eatedly by % until you1ve done so e(actly ; times& 8e(t, you do the analo ous thin for CELLC6@+set it to ", multi!ly by B e(actly CELLC?@ times, then quit& 'inally, you set 0&T,&T to the value of CELLC6@& This is the value returned to the outside worldOthe only e(ternally visible behavior of the !rocedure& ) number of !oints about the notation should be made here& 'irst, the meanin of the left-arrow 1gX1 is this, Evaluate the e(!ression to its ri ht, then take the result and set the CELL 4or 0&T,&T5 on its left to that value& So the meanin of a command such as CELLC6@ gX G ! CELLC6@ is to tri!le the value stored in CELLC6@& Hou may think of each CELL as bein a se!arate word in the memory of some com!uter& The only difference between a CELL and a true word is that the latter can only hold inte ers u! to some finite limit, whereas we allow a CELL to hold any natural number, no matter how bi & Every !rocedure in 0looP, when called, yields a value Onamely the value of the variable called 0&T,&T& )t the be innin of e(ecution of any !rocedure, it is assumed as a default o!tion that 0&T,&T has the value +& That way, even if the !rocedure never resets 0&T,&T at all, 0&T,&T has a well-defined value at all times&

I+-Statements and Branching

8ow let us look at another !rocedure which will show us some other features of 0looP which ive it more enerality& ?ow do you find out, knowin only how to add, what the value of M - ; is? The trick is to add various numbers onto ; until you find the one which yields M& ?owever, what ha!!ens if M is smaller than ;? 3hat if we are tryin to take L from B? In the domain of natural numbers, there is no answer& 0ut we would like our 0looP !rocedure to ive an answer anywayOlet1s say +& ?ere, then, is a 0looP !rocedure which does subtraction, E+I;E ,$0CE &$E (MI;&S( AM);B5 BL0CK ?5 BEGI; I+ M D ;) T/E;5 8&IT BL0CK ?F L00, AT M0ST M I 6 TIMES5 BL0CK 65 BEGI; I+ 0&T,&T I ; E M) T/E;5 AB0$T L00, 6F 0&T,&T) DE 0&T,&T I 6F

BL0CK 65 E; F BL0CK ?5 E; 1 ?ere we are makin use of the im!licit feature that 0&T,&T be ins at +& If M is less than ;, then the subtraction is im!ossible, and we sim!ly .um! to the bottom of BL0CK ? ri ht away, and the answer is +& That is what is meant by the line 8&IT BL0CK ?& 0ut if M is not less than ;, then we ski! over that 8&IT-statement, and carry out the ne(t command in sequence 4here, a L00,-statement5& That is how I'-statements always work in 0looP& So we enter L00, 6, so called because the block which it tells us to re!eat is BL0CK 6& 3e try addin + to ;, then ", B, etc&, until we find a number that ives M& )t that !oint, we AB0$T the loo! we are in, meanin we .um! to the statement immediately followin the E; which marks the bottom of the loo!1s block& In this case, that .um! brin s us .ust below BL0CK 65 E; , which is to say, to the last statement of the al orithm, and we are done& 0&T,&T now contains the correct answer& 8otice that there are two distinct instructions for .um!in downwards, 8&IT, and AB0$T& The former !ertains to blocks, the latter to loo!s& 8&IT BL0CK n means to .um! to the last line of BL0CK n, whereas AB0$T L00, n means to .um! .ust (elow the last line of BL0CK n& This distinction only matters when you are inside a loo! and want to continue loo!in but to quit the block this time around& Then you can say 8&IT and the !ro!er thin will ha!!en& )lso notice that the words AT M0ST now !recede the u!!er bound of the loo!, which is a warnin that the loo! may be aborted before the u!!er bound is reached&

Automatic Chun.ing
8ow there are two last features of 0looP to e(!lain, both of them very im!ortant& The first is that, once a !rocedure has been defined, it may be called inside later !rocedure definitions& The effect of this is that once an operation has (een defined in a procedure, it is considered as simple as a primordial step& Thus, 0looP features automatic chunkin & Hou mi ht com!are it to the way a ood ice skater acquires new motions, not by definin them as lon sequences of !rimordial muscle-actions, but in terms of !reviously learned motions, which were themselves learned as com!ounds of earlier learned motions, etc& O and the nestedness, or chunkedness, can o back many layers until you hit !rimordial muscle-actions )nd thus, the re!ertoire of 0looP !ro rams, like the re!ertoire of a skater1s tricks, rows, quite literally, by loo!s and bounds&

Bloo, Tests
The other feature of 0looP is that certain !rocedures can have =ES or ;0 as their out!ut, instead of an inte er value& Such !rocedures are tests, rather than functions& To indicate

the difference, the name of a test must terminate in a question mark& )lso, in a test, the default o!tion for 0&T,&T is not +, of course, but ;0& Aet us see an e(am!le of these last two features of 0looP in an al orithm which tests its ar ument for !rimality, E+I;E ,$0CE &$E (,$IME7( A;B5 BL0CK ?5 BEGI; I+ ; E ?) T/E;5 8&IT BL0CK ?F CELLC?@ DE HF L00, AT M0ST MI;&S A;)HB TIMES5 BL0CK 65 BEGI; I+ $EMAI; E$ A;)CELLC?@B E ?) T/E;5 8&IT BL0CK ?F CELLC?@ DE CELLC?@ I 6F BL0CK 65 E; F 0&T,&T DE =ESF BL0CK ?5 E; 1 8otice that I have called two !rocedures inside this al orithm, MI;&S and $EMAI; E$& 4The latter is !resumed to have been !reviously defined, and you may work out its definition yourself&5 8ow this test for !rimality works by tryin out !otential factors of ; one by one, startin at B and increasin to a ma(imum of ; - "& In case any of them divides ; e(actly 4i&e&, ives remainder +5, then we .um! down to the bottom, and since 0&T,&T still has its default value at this sta e, the answer is ;0& Enly if ; has no e(act divisors will it survive the entirety of L00, "= then we will emer e smoothly at the statement 0&T,&T DE =ES, which will et e(ecuted, and then the !rocedure is over&

Bloo, ,rograms Contain Chains of ,rocedures

3e have seen how to define !rocedures in 0looP= however, a !rocedure definition is only a !art of a !ro ram& ) program consists of a chain of procedure definitions 4each only callin !reviously defined !rocedures5, o!tionally followed by one or more calls on the !rocedures defined& Thus, an e(am!le of a full 0looP !ro ram would be the definition of the !rocedure T>0-T0-T/E-T/$EE-T0-T/E, followed by the call T>0-T0-T/E-T/$EE-T0-T/E AHB which would yield an answer of L"B& If you have only a chain of !rocedure definitions, then nothin ever ets e(ecuted= they are all .ust waitin for some call, with s!ecific numerical values, to set them in

motion& It is like a meat rinder waitin for some meat to rind Oor rather, a chain of meat rinders all linked to ether, each of which is fed from earlier ones&&& In the case of meat rinders, the ima e is !erha!s not so savory= however, in the case of 0looP !ro rams, such a construct is quite im!ortant, and we will call it a -call-less !ro ram-& This notion is illustrated in 'i ure $B& 8ow 0looP is our lan ua e for definin !redictably terminatin calculations& The standard name for functions which are 0looP-com!utable is primiti#e recursi#e functionsR and the standard name for properties which can be detected by 0looP-tests is primiti#e recursi#e predicates& Thus, the function B% is a !rimitive recursive function= and the statement -n is a !rime number- is a !rimitive recursive !redicate& It is clear intuitively that the 6oldbach !ro!erty is !rimitive recursive, and to make that quite e(!licit, here is a !rocedure definition in 0looP, showin how to test for its !resence or absence,
n

E+I;E ,$0CE &$E (G0L BAC/7( A;B5 BL0CK ?5 BEGI; CELLC?@ DE HF L00, AT M0ST ; TIMES5 BL0CK 65 BEGI; I+ J,$IME7 ACELLC?@B A; ,$IME7 AMI;&S A;)CELLC?@BBK) T/E;5 BL0CK H5 BEGI; 0&T,&T DE =ESF 8&IT BL0CK ?-) BL0CK H5 E; CELLC?@ DE CELLC?@ I BL0CK 65 E; F BL0CK ?5 E; 1 )s usual, we assume ;0 until !roven =ES, and we do a brute force search amon !airs of numbers which sum u! to ;& If both are !rime, we quit the outermost block= otherwise we .ust o back and try a ain, until all !ossibilities are e(hausted& 43arnin , The fact that the 6oldbach !ro!erty is !rimitive recursive does not make the question -:o all numbers have the 6oldbach !ro!erty?- a sim!le questionfar from it25

73G5=E CF! The structure of a call%less Bloo& program! 7or this program to (e self%contained, each procedure definition ma) onl) call procedures defined a(o#e it!

Suggested E-ercises
7an you write a similar 0looP !rocedure which tests for the !resence or absence of the Tortoise !ro!erty 4or the )chilles !ro!erty5? If so, do it& If not, is it merely because you are i norant about u!!er bounds, or could it be that there is a fundamental obstacle !reventin the formulation of such an al orithm in 0looP? )nd what about the same questions, with res!ect to the !ro!erty of wondrousness, defined in the :ialo ue? 0elow, I list some functions and !ro!erties, and you ou ht to take the time to determine whether you believe they are !rimitive recursive 40looP-!ro rammable5 or not& This means that you must carefully consider what kinds of o!erations will be involved in the calculations which they require, and whether ceilin s can be iven for all the loo!s involved& +ACT0$IAL A;B E ;* 4the factorial of ;5 4e& &, +ACT0$IAL ALB E HL5 $EMAI; E$ AM);B = the remainder u!on dividin M by ; 4e& &, $EMAI; E$ AHL)MB E G5 ,I- IGIT A;B = the 8th di it of !i, after the decimal !oint 4e& & ,I- IGIT A6B E 6,

,I- IGIT AHB = L, ,I- IGIT A6??????B = 6 +IB0 A;B = the ;th 'ibonacci number (e& ., +IB0 ANB = GL5 ,$IME-BE=0; A;B = the lowest !rime beyond ; 4e& &, ,$IME-BE=0; AGGB = GM5 ,E$+ECT A;B = the ;th -!erfect- number 4a number such as B* whose divisors sum u! to itself, B* X " ^ B ^ C ^ $ ^ "C" 4e& &, ,E$+ECT AHB = #O5 ,$IME7 A;B = =ES if ; is !rime, otherwise ;0& ,E$+ECT7 A;B = =ES if ; is !erfect, otherwise ;0. T$I4IAL7 AA)B)C);B = =ES if A;IB; = C; is correct= otherwise ;0& 4e& &, T$I4IAL7 AG)L)P)HB = =ES, T$I4IAL7 AG)L)P)GB = ;05 ,IE$$E7 AA)B)CB = =ES if A;IB; = C; is satisfiable for some value of ; reater than ", otherwise ;0& 4e& &, ,IE$$E7 AG)L)PB = =ES, ,IE$$E7 A6)H)GB = ;05 +E$MAT7 A;B = =ES if A;IB; = C; is satisfied by some !ositive values of A, B, C= otherwise ;0& 4e& &, +E$MAT7 AHB = =ES5 T0$T0ISE-,AI$7 AM);B = =ES if both M and M + ; are !rime, otherwise ;0& 4e& &, T0$T0ISE-,AI$ AP)6MLHB = =ES, T0$T0ISE-,AI$ AP)6??B = ;05 T0$T0ISE7 A;B = =ES if ; is the difference of two !rimes, otherwise ;0& 4e& &, T0$T0ISE A6MLHB = =ES, T0$T0ISE AMB = ;05 MI&->ELL-+0$ME 7 A;B = =ES if ;, when seen as a strin of the /IF-System, is wellformed= otherwise ;0& 4e& &, MI&->ELL-+0$ME 7 AG6?B = =ES, MI&->ELL-+0$ME 7 AL6PB = ;05 MI&-,$00+-,AI$7 AM);B = =ES If M, as seen as a sequence of strin s of the /IF-system, is a derivation of 8, as seen as a strin of the /IF-system= otherwise ;0& 4e& &, MI&-,$00+-,AI$7 AG6G66G6666G?6)G?6B = =ES, MI&-,$00+-,AI$7 AG666G?)G?B = ;05 MI&-T/E0$EM7 A;B = =ES if ;, seen as a /IF-system strin , is a theorem= otherwise ;0. 4e& &, MI&-T/E0$EM7 AG66B = =ES, MI&-T/E0$EM7 AG?B = ;0,

MI&-T/E0$EM7 AM?6B = ;05 T;T-T/E0$EM7 A;B = =ES if ;, seen as a T;T-strin , is a theorem& 4e& &, T;T-T/E0$EM7 AQQQ666QQQB = =ES, T;T-T/E0$EM7 A6HGQQQ666QQQB = ;0, T;T-T/E0$EM7 AM?6LB = ;05 +ALSE7 A;B X =ES if ;, seen as a T8T-strin , is a false statement of number theory= otherwise ;0& 4e& &, +ALSE7 AQQQ6 6 6QQQB = ;0, +ALSE7 AHHGQQQ6 6 6QQQB = =ES, +ALSE7 AM?6LB = ;05 The last seven e(am!les are !articularly relevant to our future metamathematical e(!lorations, so they hi hly merit your scrutiny&

E-!ressibility and $e!resentability

8ow before we o on to some interestin questions about 0looP and are led to its relative, 'looP, let us return to the reason for introducin 0looP in the first !lace, and connect it to T8T& Earlier, I stated that the critical mass for 6>del1s method to be a!!licable to a formal system is attained when all !rimitive recursive notions are re!resentable in that system& E(actly what does this mean? 'irst of all, we must distin uish between the notions of re!resentability and e(!ressibility& E9pressing a !redicate is a mere matter of translation from En lish into a strict formalism& It has nothin to do with theoremhood& 'or a !redicate to be represented, on the other hand, is a much stron er notion& It means that 4"5 )ll true instances of the !redicate are theorems= 4B5 )ll false instances are nontheorems& 0y -instance-, I mean the strin !roduced when you re!lace all free variables by numerals& 'or e(am!le, the !redicate mPnQk is re!resented in the !q-system, because each true instance of the !redicate is a theorem, each false instance is a nontheorem& Thus any s!ecific addition, whether true or false, translates into a decida(le strin of the !qsystem& ?owever, the !q-system is unable to e(!resslet alone re!resentany other !ro!erties of natural numbers& Therefore it would be a weak candidate indeed in a com!etition of systems which can do number theory& 8ow T8T has the virtue of bein able to e9press virtually any number-theoretical !redicate= for e(am!le, it is easy to write a T8T-strin which e(!resses the !redicate - b has the Tortoise !ro!erty-& Thus, in terms of e(!ressive !ower, T8T is all we want& ?owever, the question -3hich !ro!erties are represented in T8T?- is !recisely the question -?ow !owerful an a(iomatic system is T8T?- )re all !ossible !redicates re!resented in T8T? If so, then T8T can answer any question of number theory= it is com!lete&

,rimiti"e $ecursi"e ,redicates Are $e!resented in T;T

8ow althou h com!leteness will turn out to be a chimera& T8T is at least com!lete with res!ect to primiti#e recursi#e !redicates& In other words, any statement of number theory whose truth or falsity can be decided by a com!uter within a !redictable len th of time is also decidable inside T8T& Er, one final restatement of the same thin , If a 0looP test can be written for some !ro!erty of natural numbers, then that !ro!erty is re!resented in T8T&

Are There +unctions >hich Are ;ot ,rimiti"e $ecursi"e7

8ow the kinds of !ro!erties which can be detected by 0looP tests are widely varied, includin whether a number is !rime or !erfect, has the 6oldbach !ro!erty, is a !ower of B, and so on and so forth& It would not be cra<y to wonder whether e#er) !ro!erty of numbers can be detected by some suitable 0looP !ro ram& The fact that, as of the !resent moment, we have no way of testin whether a number is wondrous or not need not disturb us too much, for it mi ht merely mean that we are i norant about wondrousness, and that with more di in around, we could discover a universal formula for the u!!er bound to the loo! involved& Then a 0looP test for wondrousness could be written on the s!ot& Similar remarks could be made about the Tortoise !ro!erty& So the question really is, -7an u!!er bounds always be iven for the len th of calculationsor, is there an inherent kind of .umbliness to the natural number system, which sometimes !revents calculation len ths from bein !redictable in advance?- The strikin thin is that the latter is the case, and we are about to see why& It is the sort of thin that would have driven Pytha oras, who first !roved that the square root of B is irrational, out of his mind& In our demonstration, we will use the celebrated diagonal method, discovered by 6eor 7antor, the founder of set theory&

,ool B) Inde- ;umbers) and Blue ,rograms

3e shall be in by ima inin a curious notion, the !ool of all !ossible 0looP !ro rams& 8eedless to say, this !ool-Pool 0-is an infinite one& 3e want to consider a sub!ool of Pool 0, obtained by three successive filterin o!erations& The first filter will retain for us only call%less !ro rams& 'rom this sub!ool we then eliminate all tests, leavin only functions& 40y the way, in call-less !ro rams, the last !rocedure in the chain determines whether the !ro ram as a whole is considered a test, or a function&5 The third filter will retain onl) functions which ha#e e9actl) one input parameter& 4) ain referrin to the final !rocedure in the chain&5 3hat is left? ) com!lete !ool of all call-less 0looP !ro rams which calculate functions of e(actly one in!ut !arameter&

Aet us call these s!ecial 0looP !ro rams Blue &rograms& 3hat we would like to do now is to assi n an unambi uous inde9 num(er to each 0lue Pro ram& ?ow can this be done? The easiest waywe shall use itis to list them in order of len th, the shortest !ossible& 0lue Pro ram bein O ", the second shortest bein O B, etc& Ef course, there will be many !ro rams tied for each len th& To break such ties, we use al!habetical order& ?ere, -al!habetical order- is taken in an e(tended sense, where the al!habet includes all the s!ecial characters of 0looP, in some arbitrary order, such as the followin , AB 0, ? 6 C @ C E + 8 $ S T H G L P A B J K G & Q / 4 M % I > O 7 3 < N 5 K = F L 2 E ) M ; I D R 1

and at the end comes the lowly blank2 )lto ether, fifty-si( characters& 'or convenience1s sake, we can !ut all 0lue Pro rams of len th " in Golume ", !ro rams of B characters in Golume B, etc& 8eedless to say, the first few volumes will be totally em!ty, while later volumes will have many, many entries 4thou h each volume will only have a finite number5& The very first 0lue Pro ram would be this one, E+I;E ,$0CE &$E (A( ABB5 BL0CK ?5 BEGI; BL0CK ?5 E; 1 This rather silly meat rinder returns a value of + no matter what its in!ut is& It occurs in Golume LD, since it has LD characters 4countin necessary blanks, includin blanks se!aratin successive lines5& Soon after Golume LD, the volumes will et e(tremely fat, because there are .ust so many millions of ways of combinin symbols to make 0lue 0looP !ro rams& 0ut no matterwe are not oin to try to !rint out this infinite catalo ue& )ll that we care about is that, in the abstract, it is well-defined, and that each 0lue 0looP !ro ram therefore has a unique and definite inde( number& This is the crucial idea& Aet us desi nate the function calculated by the kth 0lue Pro ram this way, 0lue!ro ramkO kb U;V ?ere, k is the inde( number of the !ro ram, and ; is the sin le in!ut !arameter& 'or instance, 0lue Pro ram O "B mi ht return a value twice the si<e of its in!ut, 0lue!ro ramkO"Bb U;V X B ( ; The meanin of the equation above is that the program named on the left-hand side returns the same value as a human would calculate from the ordinary al ebraic

e(!ression on the ri ht-hand side& )s another e(am!le, !erha!s the L+++th 0lue Pro ram calculates the cube of its in!ut !arameter, 0lue!ro ramkO L+++b U;V X ;
G

The

iagonal Method

Gery well-now we a!!ly the -twist-, 7antor1s dia onal method& 3e shall take this catalo ue of 0lue Pro rams and use it to define a new function of one variable Bluediag U8Vwhich will turn out not to be anywhere in the list 4which is why its name is in italics5& Het Bluediag will clearly be a well-defined, calculable function of one variable, and so we will have to conclude that functions e(ist which sim!ly are not !ro rammable in 0looP& ?ere is the definition of Bluediag U;V, Equation 4"5&&& Bluediag U;V X " ^ 0lue!ro ramkO ;b U;V The strate y is, feed each meat rinder with its own inde( number, then add " to the out!ut& To illustrate, let us find Bluediag U"BV& 3e saw that 0lue!ro ramkO"Bb is the function B;= therefore, Bluediag U"BV must have the value " ^ B W "B, or BL& Aikewise, Bluediag UL+++V would have the value "BL,+++,+++,++", since that is " more than the cube of L+++& Similarly, you can find Bluediag of any !articular ar ument you wish& The !eculiar thin about Bluediag U;V is that it is not re!resented in the catalo ue of 0lue Pro rams& It cannot be& The reason is this& To be a 0lue Pro ram, it would have to have an inde( number-say it were 0lue Pro ram O S& This assum!tion is e(!ressed by writin Equation 4B5&&& Bluediag U;V X 0lue!ro ramkO SU U;V 0ut there is an inconsistency between the equations 4"5 and 4B5& It becomes a!!arent at the moment we try to calculate the value of 0luedia U SV, for we can do so by lettin ; take the value of S in either of the two equations& If we substitute into equation 4"5, we et, Bluediag U SV X " ^ 0lue!ro ramkO Sb U SV 0ut if we substitute into equation 4B5 instead, we et, Bluediag U SV X 0lue!ro ramkO Sb U SV 8ow Bluediag U SV cannot be equal to a number and also to the successor of that number& 0ut that is what the two equations say& So we will have to o back and erase some assum!tion on which the inconsistency is based& The only !ossible candidate for erasure is the assum!tion e(!ressed by Equation 4B5, that the function Bluediag U;V is able to be coded u! as a 0lue 0looP !ro ram& )nd that is the !roof that Bluediag lies outside the

realm of primiti#e recursi#e functions & Thus, we have achieved our aim of destroyin )chilles1 cherished but naPve notion that every number-theoretical function must be calculable within a !redictable number of ste!s& There are some subtle thin s oin on here& Hou mi ht !onder this, for instance, the number of ste!s involved in the calculation of Bluediag U;V,for each specific value of ;, is !redictablebut the different methods of !rediction cannot all be united into a general reci!e for !redictin the len th of calculation of Bluediag U;V& This is an -infinite cons!iracy-, related to the Tortoise1s notion of -infinite coincidences-, and also to \-incom!leteness& 0ut we shall not trace out the relations in detail&

Cantor%s 0riginal

iagonal Argument

3hy is this called a diagonal ar ument? The terminolo y comes from 7antor1s ori inal dia onal ar ument, u!on which many other ar uments 4such as ours5 have subsequently been based& To e(!lain 7antor1s ori inal ar ument will take us a little off course, but it is worthwhile to do so& 7antor, too, was concerned with showin that some item is not in a certain list& S!ecifically, what 7antor wanted to show was that if a -directory- of real numbers were made, it would inevitably leave some real numbers outso that actually, the notion of a complete directory of real numbers is a contradiction in terms& It must be understood that this !ertains not .ust to directories of finite si<e, but also to directories of infinite si<e& It is a much dee!er result than the statement -the number of reals is infinite, so of course they cannot be listed in a finite directory-& The essence of 7antor1s result is that there are 4at least5 two distinct t)pes of infinity, one kind of infinity describes how many entries there can be in an infinite directory or table, and another describes how many real numbers there are 4i&e&, how many !oints there are on a line, or line se ment5and this latter is -bi er-, in the sense that the real numbers cannot be squee<ed into a table whose len th is described by the former kind of infinity& So let us see how 7antor1s ar ument involves the notion of dia onal, in a literal sense& Aet us consider .ust real numbers between + and "& )ssume, for the sake of ar ument, that an infinite list could be iven, in which each !ositive inte er N is matched u! with a real number r4N5 between + and ", and in which each real number between + and " occurs somewhere down the line& Since real numbers are iven by infinite decimals, we can ima ine that the be innin of the table mi ht look as follows, r4"5, r4B5, r4%5, r4C5, r4L5, &$ &% &$ &C &L C % " " + " % & C + L % B # + # % * " ' B % " % + D % * L + L % B D + % % * B +

The di its that run down the dia onal are in boldface, ", %, *, B, +&&&& 8ow those dia onal di its are oin to be used in makin a s!ecial real number d, which is between + and " but which, we will see, is not in the list& To make d, you take the dia onal di its in order, and chan e each one of them to some other di it& 3hen you !refi( this sequence of di its

by a decimal !oint you have d& There are of course many ways of chan in a di it to some other di it, and corres!ondin ly many different d1s& Su!!ose, for e(am!le, that we su(tract B from the diagonal digits 4with the convention that " taken from + is #5& Then our number d will be, &+ B $ " #&&& 8ow, because of the way we constructed it, d's "st di it is not the same as the "st di it of r4"5= d's Bnd di it is not the same as the Bnd di it of r4B5= d's %rd di it is not the same as the %rd di it of r4%5= &&& and so on& ?ence, d is different from r4"5= d is different from r4B5= d is different from r4%5= &&& and soon& In other words, d is not in the list2

>hat

oes a

iagonal Argument ,ro"e7

8ow comes the crucial difference between 7antor1s !roof and our !roofit is in the matter of what assum!tion to o back and undo& In 7antor1s ar ument, the shaky assum!tion was that such a table could be drawn u!& Therefore, the conclusion warranted by the construction of d is that no e(haustive table of reals can be drawn u! after all which amounts to sayin that the set of inte ers is .ust not bi enou h to inde( the set of reals& En the other hand, in our !roof, we know that the directory of 0lue 0looP !ro rams can be drawn u!the set of inte ers is bi enou h to inde( the set of 0lue 0looP !ro rams& So, we have to o back and retract some shakier idea which we used& )nd that idea is that Bluediag U;V is calculable by some !ro ram in 0looP& This is a subtle difference in the a!!lication of the dia onal method& It may become clearer if we a!!ly it to the alle ed -Aist of )ll 6reat /athematicians- in the :ialo uea more concrete e(am!le& The dia onal itself is -:bou!s-& If we !erform the desired dia onal-subtraction, we will et -7antor-& 8ow two conclusions are !ossible& If you have an unshakable belief that the list is complete, then you must conclude that 7antor is not a 6reat /athematician, for his name differs from all those on the list& En the other hand, if you have an unshakable belief that 7antor is a 6reat /athematician, then you must conclude that the Aist of )ll 6reat /athematicians is incom!lete, for 7antor1s name is not on the list2 43oe to those who have unshakable

beliefs on both sides25 The former case corres!onds to our !roof that Bluediag U;V is not !rimitive recursive= the latter case corres!onds to 7antor1s !roof that the list of reals is incom!lete&

73G5=E CH! Georg 8antor

7antor1s !roof uses a dia onal in the literal sense of the word& Ether -dia onal- !roofs are based on a more eneral notion, which is abstracted from the eometric sense of the word& The essence of the dia onal method is the fact of usin one inte er in two different waysor, one could say, using one integer on two different le#elsO thanks to which one can construct an item which is outside of some !redetermined list& Ene time, the inte er serves as a #ertical inde(, the other time as a hori@ontal inde(& In 7antor1s construction this is very clear& )s for the function Bluediag U;V, it involves usin one inte er on two different levelsfirst, as a 0lue Pro ram inde( number= and second, as an in!ut !arameter&

The Insidious $e!eatability of the

iagonal Argument

)t first, the 7antor ar ument may seem less than fully convincin & Isn1t there some way to et around it? Perha!s by throwin in the dia onally constructed number d, one mi ht obtain an e(haustive list& If you consider this idea, you will see it hel!s not a bit to throw in the number d, for as soon as you assi n it a s!ecific !lace in the table, the dia onal method becomes a!!licable to the new table, and a new missin number d' can be constructed, which is not in the new table& 8o matter how many times you re!eat the o!eration of constructin a number by the dia onal method and then throwin it in to make a -more com!lete- table, you still are cau ht on the ineradicable hook of 7antor1s method& Hou mi ht even try to build a table of reals which tries to outwit the 7antor

dia onal method by takin the whole trick, lock, stock, and barrel, includin its insidious re!eatability, into account somehow& It is an interestin e(ercise& 0ut if you tackle it, you will see that no matter how you twist and turn tryin to avoid the 7antor -hook-, you are still cau ht on it& Ene mi ht say that any self-!roclaimed -table of all reals- is hoist by its own !etard& The re!eatability of 7antor1s dia onal method is similar to the re!eatability of the Tortoise1s diabolic method for breakin the 7rab1s !hono ra!hs, one by one, as they ot more and more -hi-fi- andat least so the 7rab ho!edmore -Perfect-& This method involves constructin , for each !hono ra!h, a !articular son which that !hono ra!h cannot re!roduce& It is not a coincidence that 7antor1s trick and the Tortoise1s trick share this curious re!eatability= indeed, the 8ontracrostipunctus mi ht well have been named -7antorcrosti!unctus- instead& /oreover, as the Tortoise subtly hinted to the innocent )chilles, the events in the 8ontracrostipunctus are a !ara!hrase of the construction which 6>del used in !rovin his Incom!leteness Theorem= it follows that the 6>del construction is also very much like a dia onal construction& This will become quite a!!arent in the ne(t two 7ha!ters&

+rom Bloo, to +loo,

3e have now defined the class of !rimitive recursive functions and !rimitive recursive !ro!erties of natural numbers by means of !ro rams written in the lan ua e 0looP& 3e have also shown that 0looP doesn1t ca!ture all the functions of natural numbers which we can define in words& 3e even constructed an -un0looPable- function, 0luedia U8V, by 7antor1s dia onal method& 3hat is it about 0looP that makes 0luedia unre!resentable in it? ?ow could 0looP be im!roved so that 0luedia became re!resentable? 0looP1s definin feature was the boundedness of its loo!s& 3hat if we dro! that requirement on loo!s, and invent a second lan ua e, called -'looP- 41'1 for -free-5? 'looP will be identical to 0looP e(ce!t in one res!ect, we may have loo!s without ceilin s, as well as loo!s with ceilin s 4althou h the only reason one would include a ceilin when writin a loo!-statement in 'looP would be for the sake of ele ance5& These new loo!s will be called M&-L00,S& This follows the convention of mathematical lo ic, in which -free- searches 4searches without bounds5 are usually indicated by a symbol called a -]-o!erator- 4mu-o!erator5& Thus, loo! statements in 'looP may look like this, M&-L00,5 BL0CK n5 BEGI;

1 1
BL0CK n5 E;

This feature will allow us to write tests in 'looP for such !ro!erties as wondrousness and the Tortoise !ro!erty-tests which we did not know how to !ro ram in 0looP because of the !otential o!en-endedness of the searches involved& I shall leave it to interested readers to write a 'looP test for wondrousness which does the followin thin s, 4"5 If its in!ut, ;, is wondrous, the !ro ram halts and ives the answer =ES& 4B5 If 8 is unwondrous, but causes a closed cycle other than "-C-B-"-C-B-"-&&& , the !ro ram halts and ives the answer ;0& 4%5 If 8 is unwondrous, and causes an -endlessly risin !ro ression-, the !ro ram never halts& This is 'looP1s way of answerin by not answerin & 'looP1s nonanswer bears a stran e resemblance to JRshl1s nonanswer -/F-& The irony of case % is that 0&T,&T always has the value ;0, but it is always inaccessible, since the !ro ram is still rindin away& That troublesome third alternative is the !rice that we must !ay for the ri ht to write free loo!s& In all 'looP !ro rams incor!oratin the M&-L00, o!tion, nontermination will always be one theoretical alternative& Ef course there will be many 'looP !ro rams which actually terminate for all !ossible in!ut values& 'or instance, as I mentioned earlier, it is sus!ected by most !eo!le who have studied wondrousness that a 'looP !ro ram such as su ested above will always terminate, and moreover with the answer =ES each time&

Terminating and ;onterminating +loo, ,rograms

It would seem e(tremely desirable to be able to se!arate 'looP !rocedures into two classes, terminators and nonterminators& ) terminator will eventually halt no matter what its in!ut, des!ite the -/F-ness- of its loo!s& ) nonterminator will o on and on forever, for at least one choice of in!ut& If we could always tell, by some kind of com!licated ins!ection of a 'looP !ro ram, to which class it belon ed, there would be some remarkable re!ercussions 4as we shall shortly see5& 8eedless to say, the o!eration of class-checkin would itself have to be a terminatin o!erationotherwise one would ain nothin 2

Turing%s Tric.ery
The idea s!rin s to mind that we mi ht let a 0looP !rocedure do the ins!ection& 0ut 0looP !rocedures only acce!t numerical in!ut, not !ro rams2 ?owever, we can et around that&&& by codin !ro rams into numbers2 This sly trick is .ust 6>del-numberin in another of its many manifestations& Aet the fifty-si( characters of the 'looP al!habet et the -codons- #+", #+B&&& #LD, res!ectively& So each 'looP !ro ram now ets a very lon 6>del number& 'or instance, the shortest 0looP function 4which is also a terminatin 'looP !ro ram5

E+I;E ,$0CE &$E (A( ABB5 BL0CK ?5 BEGI; BL0CK ?5 E; 1 would et the 6>del number !artially shown below, #+C, #+L, #+D, #+#, #"C, #+L E + I ; E #+L, #"C&#+C, #LL, E ; 1

8ow our scheme would be to write a 0looP test called TE$MI;AT0$? which says =ES if its in!ut number codes for a terminatin 'looP !ro ram, ;0 if not& This way we could hand the task over to a machine and, with luck, distin uish terminators from nonterminators& ?owever, an in enious ar ument iven by )lan Turin shows that no 0looP !ro ram can make this distinction infallibly& The trick is actually much the same as 6>del1s trick, and therefore closely related to the 7antor dia onal trick& 3e shall not ive it heresuffice it to say that the idea is to feed the termination tester its own 6>del number& This is not so sim!le, however, for it is like tryin to quote an entire sentence inside itself& Hou have to quote the quote, and so forth= it seems to lead to an infinite re ress& ?owever, Turin fi ured out a trick for feedin a !ro ram its own 6>del number& ) solution to the same !roblem in a different conte(t will be !resented ne(t 7ha!ter& In the !resent 7ha!ter, we shall take a different route to the same oal, which is namely to !rove that a termination tester is im!ossible& 'or readers who wish to see an ele ant and sim!le !resentation of the Turin a!!roach, I recommend the article by ?oare and )llison, mentioned in the 0iblio ra!hy&

A Termination Tester >ould Be Magical

0efore we destroy the notion, let us delineate .ust why havin a termination tester would be a remarkable thin & In a sense, it would be like havin a ma ical dowsin rod which could solve all !roblems of number theory in one swell 'looP& Su!!ose, for instance, that we wished to know if the 6oldbach Gariation is a true con.ecture or not& That is, do all numbers have the Tortoise !ro!erty? 3e would be in by writin a 'looP test called T0$T0ISE? which checks whether its in!ut has the Tortoise !ro!erty& 8ow the defect of this !rocedurenamely that it doesn1t terminate if the Tortoise !ro!erty is absenthere turns into a virtue2 'or now we run the termination tester on the !rocedure T0$T0ISE?& If it says =ES, that means that T0$T0ISE? terminates for all values of its in!utin other words, all numbers have the Tortoise !ro!erty& If it says ;0, then we know there e(ists a number which has the )chilles !ro!erty& The irony is that we never actually use the !ro ram T0$T0ISE at allwe .ust ins!ect it& This idea of solvin any !roblem in number theory by codin it into a !ro ram and then wavin a termination tester over the !ro ram is not unlike the idea of testin a kRan for enuineness by codin it into a folded strin and then runnin a test for 0uddha-

nature on the strin instead& )s )chilles su ested, !erha!s the desired information lies -closer to the surface- in one re!resentation than in another&

,ool +) Inde- ;umbers) and Green ,rograms

3ell, enou h daydreamin & ?ow can we !rove that the termination tester is im!ossible? Eur ar ument for its im!ossibility will hin e on tryin to a!!ly the dia onal ar ument to 'looP, .ust as we did to 0looP& 3e shall see that there are some subtle and crucial differences between the two cases& )s we did for 0looP, ima ine the !ool of all 'looP !ro rams& 3e shall call it -Pool '-& Then !erform the same three filterin o!erations on Pool ', so that you et, in the end, ) com!lete !ool of all call-less 'looP !ro rams which calculate functions of e(actly one in!ut !arameter& Aet us call these s!ecial 'looP-!ro rams Green &rograms 4since they may o forever5& 8ow .ust as we assi ned inde( numbers to all 0lue Pro rams, we can assi n inde( numbers to 6reen Pro rams, by orderin them in a catalo ue, each volume of which contains all 6reen Pro rams of a fi(ed len th, arran ed in al!habetical order& So far, the carry-over from 0looP to 'looP has been strai htforward& 8ow let us see if we can also carry over the last !art, the dia onal trick& 3hat if we try to define a dia onal function? Greendiag U;V X " ^ 6reen!ro ramkO ;b U;V Suddenly, there is a sna , this function Greendiag U;V may not have a well-defined out!ut value for all in!ut values ;. This is sim!ly because we have not filtered out the nonterminator !ro rams from Pool ', and therefore we have no uarantee that we can calculate Greendiag U;V for all values of ;& Sometimes we may enter calculations which never terminate& )nd the dia onal ar ument cannot be carried throu h in such a case, for it de!ends on the dia onal function havin a value for all !ossible in!uts&

The Termination Tester Gi"es &s $ed ,rograms

To remedy this, we would have to make use of a termination tester, if one e(isted& So let us deliberately introduce the shaky assum!tion that one e(ists, and let us use it as our fourth filter& 3e run down the list of 6reen Pro rams, eliminatin one by one all nonterminators, so that in the end we are left with, ) com!lete !ool of all call-less 'looP !ro rams which calculate functions of e(actly one in!ut !arameter, and which terminate for all values of their in!ut&

Aet us call these s!ecial 'looP !ro rams =ed &rograms 4since they all must sto!5& 8ow, the dia onal ar ument will o throu h& 3e define =eddiag U;V X " ^ ;ed!ro ramkO ;b U;V and in an e(act !arallel to Bluediag, we are forced to conclude that =eddiag U;V is a welldefined, calculable function of one variable which is not in the catalo ue of ;ed Pro rams, and is hence not even calculable in the !owerful lan ua e 'looP& Perha!s it is time to move on to 6looP?

Gloo,111
Hes, but what is 6looP? If 'looP is 0looP unchained, then 6looP must be 'looP unchained& 0ut how can you take the chains off twice? ?ow do you make a lan ua e whose !ower transcends that of 'looP? In =eddiag, we have found a function whose values we humans know how to calculatethe method of doin so has been e(!licitly described in En lishbut which seemin ly cannot be !ro rammed in the lan ua e 'looP& This is a serious dilemma because no one has ever found any more !owerful com!uter lan ua e than 'looP& 7areful investi ation into the !ower of com!uter lan ua es has been carried out& 3e need not do it ourselves= let it .ust be re!orted that there is a vast class of com!uter lan ua es all of which can be !roven to have e9actl) the same e9pressi#e power as 'looP does, in this sense, any calculation which can be !ro rammed in any one of the lan ua es can be !ro rammed in them all& The curious thin is that almost any sensible attem!t at desi nin a com!uter lan ua e ends u! by creatin a member of this classwhich is to say, a lan ua e of !ower equal to that of 'looP& It takes some doin to invent a reasonably interestin com!uter lan ua e which is weaker than those in this class& 0looP is, of course, an e(am!le of a weaker lan ua e, but it is the e(ce!tion rather than the rule& The !oint is that there are some e(tremely natural ways to o about inventin al orithmic lan ua es= and different !eo!le, followin inde!endent routes, usually wind u! creatin equivalent lan ua es, with the only difference bein style, rather than !ower&

111 Is a Myth
In fact, it is widely believed that there cannot be any more !owerful lan ua e for describin calculations than lan ua es that are equivalent to 'looP& This hy!othesis was formulated in the "#%+1s by two !eo!le, inde!endently of each other, )lan Turin about whom we shall say more laterand )lon<o 7hurch, one of the eminent lo icians of this century& It is called the 8hurch%Turing Thesis& If we acce!t the 7T-Thesis, we have to conclude that -6looP- is a myththere are no restrictions to remove in 'looP, no ways to increase its !ower by -unshacklin - it, as we did 0looP& This !uts us in the uncomfortable !osition of assertin that people can calculate =eddiag U;V for any value of ;, but there is no way to !ro ram a computer to do so& 'or,

if it could be done at all, it could be done in 'looPand by construction, it can1t be done in 'looP& This conclusion is so !eculiar that it should cause us to investi ate very carefully the !illars on which it rests& )nd one of them, you will recall, was our shaky assum!tion that there is a decision !rocedure which can tell terminatin from nonterminatin 'looP !ro rams& The idea of such a decision !rocedure already seemed sus!ect, when we saw that its e(istence would allow all !roblems of number theory to be solved in a uniform way& 8ow we have double the reason for believin that any termination test is a myththat there is no way to !ut 'looP !ro rams in a centrifu e and se!arate out the terminators from the nonterminators& Ske!tics mi ht maintain that this is nothin like a ri orous !roof that such a termination test doesn1t e(ist& That is a valid ob.ection= however, the Turin a!!roach demonstrates more ri orously that no com!uter !ro ram can be written in a lan ua e of the 'looP class which can !erform a termination test on all 'looP !ro rams&

The Church-Turing Thesis

Aet us come back briefly to the 7hurch-Turin Thesis& 3e will talk about itand variations on itin considerable detail in 7ha!ter TGII= for now it will suffice to state it in a cou!le of versions, and !ost!one discussion of its merits and meanin s until then& ?ere, then, are three related ways to state the 7T-Thesis, 4"5 3hat is human-com!utable is machine-com!utable& 4B5 3hat is machine-com!utable is 'looP-com!utable& 4%5 3hat is human-com!utable is 'looP-com!utable 4i&e&, eneral or !artial recursive5&

Terminology5 General and ,artial $ecursi"e

3e have made a rather broad survey, in this 7ha!ter, of some notions from number theory and their relations to the theory of com!utable functions& It is a very wide and flourishin field, an intri uin blend of com!uter science and modern mathematics& 3e should not conclude this 7ha!ter without introducin the standard terminolo y for the notions we have been dealin with& )s has already been mentioned, -0looP-com!utable- is synonymous with -!rimitive recursive-& 8ow 'looP com!utable functions can be divided into two realms, 4"5 those which are com!utable by terminating 'looP !ro rams, these are said to be general recursi#e= and 4B5 those which are com!utable only by nonterminating 'looP !ro rams, these are said to be !artial recursive& 4Similarly for !redicates&5 Peo!le often .ust say -recursive- when they mean - eneral recursive-&

The ,o'er of T;T

It is interestin that T8T is so !owerful that not only are all !rimitive recursive !redicates re!resented, but moreover all eneral recursive !redicates are re!resented& 3e shall not

!rove either of these facts, because such !roofs would be su!erfluous to our aim, which is to show that T8T is incom!lete& If T8T could not re!resent some !rimitive or eneral recursive !redicates, then it would be incom!lete in an uninteresting wayso we mi ht as well assume that it can, and then show that it is incom!lete in an interestin way&

Air on (,s Strin!

The Tortoise and *chilles ha#e Must completed a tour of a porridge factor)! *chilles, Hou don1t mind if I chan e the sub.ect, do you? Tortoise, 0e my uest& *chilles, Gery well, then& It concerns an obscene !hone call I received a few days a o& Tortoise, Sounds interestin & *chilles, Hes& 3ellthe !roblem was that the caller was incoherent, at least as far as I could tell& ?e shouted somethin over the line and then hun u!or rather, now that I think of it, he shouted somethin , shouted it a ain, and then hun u!& Tortoise, :id you catch what that thin was? *chilles, 3ell, the whole call went like this, ')self, ?ello? 8aller shouting wildl)5, Hields falsehood when !receded by its quotation2 Hields falsehood when !receded by its quotation2 48lick&5 Tortoise, That is a most unusual thin to say to somebody on an obscene !hone call& *chilles, E(actly how it struck me& Tortoise, Perha!s there was some meanin to that seemin madness& *chilles, Perha!s& The) enter a spacious court)ard framed () some charming three%stor) stone houses! *t its center stands a palm tree, and to one side is a tower! Near the tower there is a staircase where a (o) sits, talking to a )oung woman in a window!/ Tortoise, 3here are you takin me, )chilles? *chilles, I would like to show you the !retty view from the to! of this tower& Tortoise, Eh, how nice& The) approach the (o), who watches them with curiosit), then sa)s something to the )oung womanOthe) (oth chuckle! *chilles and 'r! T, instead of going up the (o)'s staircase, turn left and head down a short flight of stairs which leads to a small wooden door!/ *chilles, 3e can .ust ste! inside ri ht here& 'ollow me&

73G5=E CD! )bove and 0elow, () '!8! Escher lithograph BIDC5&

*chilles opens the door! The) enter, and (egin clim(ing the steep helical staircase inside the tower!/ Tortoise 4puffing slightl)5, I1m a little out of sha!e for this sort of e(ercise, )chilles& ?ow much further do we have to o? *chilles, )nother few fli hts&&& but I have an idea& Instead of walkin on the to! side of these stairs, why don1t you walk on the underside? Tortoise, ?ow do I do T?)T? *chilles, Just hold on ti htly, and climb around underneaththere1s room enou h for you& Hou1ll find that the ste!s make .ust as much sense from below as from above&&& Tortoise 4gingerl) shifting himself a(out5, )m I doin it ri ht? *chilles, Hou1ve ot it2 Tortoise 4his #oice slightl) muffled5, Saythis little maneuver has ot me confused& Should I head u!stairs or downstairs, now? *chilles, Just continue headin in the same direction as you were before& En your side of the staircase, that means o :E38, on mine it means FP& Tortoise, 8ow you1re not oin to tell me that I can et to the to! of the tower by oin down, are you? *chilles, I don1t know, but it works&&& *nd so the) (egin spiraling in s)nchron), with * alwa)s on one side, and T matching him on the other side! -oon the) reach the end of the staircase!/ 8ow .ust undo the maneuver, /r& T& ?erelet me hel! you u!& ,e lends an arm to the Tortoise, and hoists him (ack to the other side of the stairs!/ Tortoise, Thanks& It was a little easier ettin back u!& *nd the) step out onto the roof, o#erlooking the town!/ That1s a lovely view, )chilles& I1m lad you brou ht me u! hereor rather, :E38 here& *chilles, I fi ured you1d en.oy it& Tortoise, I1ve been thinkin about that obscene !hone call& I think I understand it a little better now& *chilles, Hou do? 3ould you tell me about it? Tortoise, 6ladly& :o you !erchance feel, as I do, that that !hrase -!receded by its quotation- has a sli htly hauntin quality about it? *chilles, Sli htly, yese(tremely sli htly& Tortoise, 7an you ima ine somethin !receded by its quotation?

*chilles, I uess I can con.ure u! an ima e of 7hairman /ao walkin into a banquet room in which there already han s a lar e banner with some of his own writin on it& ?ere would be 7hairman /ao, !receded by his quotation& Tortoise, ) most ima inative e(am!le& 0ut su!!ose we restrict the word -!receded- to the idea of !recedence on a !rinted sheet, rather than elaborate entries into a banquet room& *chilles, )ll ri ht& 0ut what e(actly do you mean by -quotation- here? Tortoise, 3hen you discuss a word or a !hrase, you conventionally !ut it in quotes& 'or e(am!le, I can say, The word -!hiloso!her- has five letters& ?ere, I !ut -!hiloso!her- in quotes to show that I am s!eakin about the 3E;: -!hiloso!her- rather than about a !hiloso!her in the flesh& This is called the FSE/E8TIE8 distinction& *chilles, Eh? Tortoise, Aet me e(!lain& Su!!ose I were to say to you, Philoso!hers make lots of money& ?ere, I would be FSI86 the word to manufacture an ima e in your mind of a twinkleeyed sa e with bul in moneyba s& 0ut when I !ut this wordor any wordin quotes, I subtract out its meanin and connotations, and am left only with some marks on !a!er, or some sounds& That is called - /E8TIE8-& 8othin about the word matters, other than its ty!o ra!hical as!ectsany meanin it mi ht have is i nored& *chilles, It reminds me of usin a violin as a fly swatter& Er should I say -mentionin -? 8othin about the violin matters, other than its solidityany meanin or function it mi ht have is bein i nored& 7ome to think of it, I uess the fly is bein treated that way, too& Tortoise, Those are sensible, if sli htly unorthodo(, e(tensions of the use-mention distinction& 0ut now, I want you to think about !recedin somethin by its own quotation& *chilles, )ll ri ht& 3ould this be correct? -?F00)- ?F00) Tortoise, 6ood& Try another& *chilles, )ll ri ht& -1PAEP1 IS 8ET T?E TITAE E' )8H 0EE@& SE '); )S I @8E3-1 1PAEP1 IS 8ET T?E TITAE E' )8H 0EE@, SE '); )S I @8E3& Tortoise, 8ow this e(am!le can be modified into quite an interestin s!ecimen, sim!ly by dro!!in 1Plo!1& *chilles, ;eally? Aet me see what you mean& It becomes

-IS 8ET T?E TITAE E' )8H 0EE@, SE '); )S I @8E3IS 8ET T?E TITAE E' )8H 0EE@, SE '); )S I @8E3& Tortoise, Hou see, you have made a sentence& *chilles, So I have& It is a sentence about the !hrase -is not the title of any book, as far as I know-, and quite a silly one too& Tortoise, 3hy silly? *chilles, 0ecause it1s so !ointless& ?ere1s another one for you, -3IAA 0E 0EHS- 3IAA 0E 0EHS& 8ow what does that mean? ?onestly, what a silly ame& Tortoise, 8ot to my mind& It1s very earnest stuff, in my o!inion& In fact this o!eration of !recedin some !hrase by its quotation is so overwhelmin ly im!ortant that I think I1ll ive it a name& *chilles, Hou will? 3hat name will you di nify that silly o!eration by? Tortoise, I believe I1ll call it -to quine a !hrase-, to quine a !hrase& *chilles, -Muine-? 3hat sort of word is that? Tortoise, ) five-letter word, if I1m not in error& *chilles, 3hat " was drivin at is why you !icked those e(act five letters in that e(act order& Tortoise, Eh, now I understand what you meant when you asked me -3hat sort of word is that?- The answer is that a !hiloso!her by the name of -3illard Gan Erman Muineinvented the o!eration, so I name it in his honor& ?owever, I cannot o any further than this in my e(!lanation& 3hy these !articular five letters make u! his namenot to mention why they occur in this !articular orderis a question to which I have no ready answer& ?owever, I1d be !erfectly willin to o and *chilles, Please don1t bother2 I didn1t really want to know everythin about Muine1s name& )nyway, now I know how to quine a !hrase& It1s quite amusin & ?ere1s a quined !hrase, -IS ) SE8TE87E ';)6/E8T- IS ) SE8TE87E ';)6/E8T& It1s silly but all the same I en.oy it& Hou take a sentence fra ment, quine it, and lo and behold, you1ve made a sentence2 ) true sentence, in this case& Tortoise, ?ow about quinin the !hrase -is a kin with without no sub.ect-? *chilles, ) kin without a sub.ect would be Tortoise, an anomaly, of course& :on1t wander from the !oint& Aet1s have quines first, and kin s afterwards2 *chilles, I1m to quine that !hrase, am I? )ll ri ht -IS ) @I86 3IT? 8E SF0JE7T- IS ) @I86 3IT? 8E SF0JE7T& It seems to me that it mi ht make more sense if it said -sentence- instead of -kin -& Eh, well& 6ive me another2 Tortoise, )ll ri ht .ust one more& Try this one,

-3?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE SE86*chilles, That should be easy&&& I1d say the quinin ives this,

-3?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE SE863?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE SE86 ?mmm&&& There1s somethin .ust a little !eculiar here& Eh, I see what it is2 The sentence is talkin about itself2 :o you see that? Tortoise, 3hat do you mean? Sentences can1t talk& *chilles, 8o, but they ;E'E; to thin sand this one refers directly unambi uously unmistakablyto the very sentence which it is2 Hou .ust have to think back and remember what quinin is all about& Tortoise, I don1t see it sayin anythin about itself& 3here does it say -me-, or, -this sentence-, or the like? *chilles, Eh, you are bein deliberately thick-skulled& The beauty of it lies in .ust that, it talks about itself without havin to come ri ht out and say so2 Tortoise, 3ell, as I1m such a sim!le fellow, could you .ust s!ell it all out for me? *chilles, Eh, he is such a :oubtin Tortoise&&& )ll ri ht, let me see&&& Su!!ose I make u! a sentenceI1ll call it -Sentence P-with a blank in it& Tortoise, Such as? *chilles, Such as&&& -xxxxxxxx 3?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE SE86-& 8ow the sub.ect matter of Sentence P de!ends on how you fill in the blank& 0ut once you1ve chosen how to fill in the blank, then the sub.ect matter is determined, it is the !hrase which you et by MFI8I86 the blank& 7all that -Sentence M-, since it is !roduced by an act of quinin & Tortoise, That makes sense& If the blank !hrase were -is written on old .ars of mustard to kee! them fresh-, then Sentence M would have to be -IS 3;ITTE8 E8 EA: J);S E' /FST);: TE @EEP T?E/ ';ES?IS 3;ITTE8 E8 EA: J);S E' /FST);: TE @EEP T?E/ ';ES?& *chilles, True, and Sentence P makes the claim 4thou h whether it is valid or not, I do not know5 that Sentence M is a Tortoise1s love son & In any case, Sentence P here is not talkin about itself, but rather about Sentence M& 7an we a ree on that much? Tortoise, 0y all means, let us a reeand what a beautiful son it is, too& *chilles, 0ut now I want to make a different choice for the blank, namely -3?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE SE86-&

Tortoise, Eh, heavens, you1re ettin a little involved here& I ho!e this all isn1t oin to be too hi hbrow for my modest mind& *chilles, Eh, don1t worryyou1ll surely catch on& 3ith this choice, Sentence M becomes && & -3?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE-SE863?E8 MFI8E:, HIEA:S ) TE;TEISE1S AEGE-SE86& Tortoise, Eh, you wily old warrior you, I catch on& 8ow Sentence M is .ust the same as Sentence P& *chilles, )nd since Sentence M is always the to!ic of Sentence P, there is a loo! now, P !oints back to itself& 0ut you see, the self-reference is a sort of accident& Fsually Sentences M and P are entirely unlike each other= but with the ri ht choice for the blank in Sentence P, quinin will do this ma ic trick for you& Tortoise, Eh, how clever& I wonder why I never thou ht of that myself& 8ow tell me, is the followin sentence self-referential? -IS 7E/PESE: E' 'IGE 3E;:S- IS 7E/PESE: E' 'IGE 3E;:S& *chilles, ?mm&&& I can1t quite tell& The sentence which you .ust ave is not really about itself, but rather about the !hrase -is com!osed of five words-& Thou h, of course, that !hrase is P);T of the sentence&&& Tortoise, So the sentence refers to some !art of itselfso what? *chilles, 3ell, wouldn1t that qualify as self-reference, too? Tortoise, In my o!inion, that is still a far cry from true self-reference& 0ut don1t worry too much about these tricky matters& Hou1ll have am!le time to think about them in the future& *chilles, I will? Tortoise, Indeed you will& 0ut for now, why don1t you try quinin the !hrase -yields falsehood when !receded by its quotation-? *chilles, I see what you1re ettin atthat old obscene !hone call& Muinin it !roduces the followin , -HIEA:S ')ASE?EE: 3?E8 P;E7E:E: 0H ITS MFET)TIE8HIEA:S ')ASE?EE: 3?E8 P;E7E:E: 0H ITS MFET)TIE8& So this is what that caller was sayin 2 I .ust couldn1t make out where the quotation marks were as he s!oke& That certainly is an obscene remark2 Peo!le ou ht to be .ailed for sayin thin s like that2 Tortoise, 3hy in the world? *chilles, It .ust makes me very uneasy& Fnlike the earlier e(am!les, I can1t quite make out if it is a truth or a falsehood& )nd the more I think about it, the more I can1t unravel it& It makes my head s!in& I wonder what kind of a lunatic mind would make somethin like that u!, and torment innocent !eo!le in the ni ht with it?

Tortoise, I wonder&&& 3ell, shall we o downstairs now? *chilles, 3e needn1t o downwe1re at round level already& Aet1s o back inside you1ll see& 4The) go into the tower, and come to a small wooden door&5 3e can .ust ste! outside ri ht here& 'ollow me& Tortoise, )re you sure? I don1t want to fall three floors and break my shell& *chilles, 3ould I fool you? *nd he opens the door! 3n front of them sits, to all appearances, the same (o), talking to the same )oung woman! *chilles and 'r! T walk up what seem to (e the same stairs the) walked down to enter the tower, and find themsel#es in what looks like Must the same court)ard the) first came into!/ Thank you, /r& T, for your lucid clarification of that obscene tele!hone call& Tortoise, )nd thank you, )chilles, for the !leasant !romenade& I ho!e we meet a ain soon&

C/A,TE$ <I4

0n +ormally &ndecidable ,ro!ositions of T;T and $elated Systems

6

The T'o Ideas of the (0yster(

T?IS 7?)PTE;1S TITAE is an ada!tation of the title of 6>del1s famous "#%" !a!er-T8Thavin been substituted for -&rincipia 'athematica-& 6>del1s !a!er was a technical one, concentratin on makin his !roof waterti ht and ri orous= this 7ha!ter will be more intuitive, and in it I will stress the two key ideas which are at the core of the !roof& The first key idea is the dee! discovery that there are strin s of T8T which can be inter!reted as s!eakin about other strin s of T8T= in short, that T8T, as a lan ua e, is ca!able of -intros!ection-, or self-scrutiny& This is what comes from 6>del-numberin & The second key idea is that the !ro!erty of self scrutiny can be entirely concentrated into a sin le strin = thus that strin 1s sole focus of attention is itself& This -focusin trick- is traceable, in essence, to the 7antor dia onal method& In my o!inion, if one is interested in understandin 6>del1s !roof& in a dee! way, then one must reco ni<e that the !roof, in its essence, consists of a fusion of these two main ideas& Each of them alone is a master stroke= to !ut them to ether took an act of enius& If I were to choose, however, which of the two key ideas is dee!er, I would unhesitatin ly !ick the first onethe idea of 6>del-numberin , for that idea is related to the whole notion of what meanin and reference are, in symbol-mani!ulatin systems& This is an idea which oes far beyond the confines of mathematical lo ic, whereas the 7antor trick, rich thou h it is in mathematical consequences, has little if any relation to issues in real life&

The +irst Idea5 ,roof-,airs

3ithout further ado, then, let us !roceed to the elaboration of the !roof itself& 3e have already iven a fairly careful notion of what the 6>del isomor!hism is about, in 7ha!ter IT& 3e now shall describe a mathematical notion which allows us to translate a statement such as -The strin +X+ is a theorem of T8T- into a statement of number theory& This will involve the notion of proof%pairs& ) !roof-!air is a !air of natural numbers related in a !articular way& ?ere is the idea,

Two natural numbers, m and n res!ectively, form a T8T !roof-!air if and only if m is the 6>del number of a T8T derivation whose bottom line is the strin with 6>del number n! The analo ous notion e(ists with res!ect to the /IF-system, and it is a little easier on the intuition to consider that case first& So, for a moment, let us back off from T8T-!roof!airs, and look at /IF-!roof-!airs& Their definition is !arallel, Two natural numbers, m and n res!ectively, form a /IF-!roof !air if and only if m is the 6>del number of a /IF-system derivation whose bottom line is the strin with 6>del number n! Aet us see a cou!le of e(am!les involvin /IF-!roof-!airs& 'irst, let m X %"%""%""""%+", n X %+"& These values of m and n do indeed form a /IF-!roof-!air, because m is the 6>del number of the /IF-derivation MI MII MIIII M&I whose last line is M&I, havin 6>del number %+", which is n& 0y contrast, let m X %"%""%"""%+, and n X %+& 3hy do these two values not form a /IF-!roof-!air? To see the answer, let us write out the alle ed derivation which m codes for, MI MII MIII M& There is an invalid ste! in this alle ed derivation2 It is the ste! from the second to the third line, from MII to MIII& There is no rule of inference in the /IF-system which !ermits such a ty!o ra!hical ste!& 7orres!ondin lyand this is most crucialthere is no arithmetical rule of inference which carries you from %"" to %"""& This is !erha!s a trivial observation, in li ht of our discussion in 7ha!ter IT, yet it is at the heart of the 6>del isomor!hism& 3hat we do in any formal system has its !arallel in arithmetical mani!ulations& In any case, the values m X %"%""%"""%+, n X %+ certainly do not form a /IF!roof-!air& This in itself does not im!ly that %+ is not a /IF-number& There could be another value of m which forms a /IF !roof-!air with %+& 4)ctually, we know by earlier reasonin that M& is not a /IF-theorem, and therefore no number at all can form a /IF!roof-!air with %+&5

8ow what about T8T !roof !airs? ?ere are two !arallel e(am!les, one bein merely an alle ed T8T !roof-!air, the other bein a valid T8T !roof-!air& 7an you s!ot which is which? 4Incidentally, here is where the 1D""1 codon comes in& Its !ur!ose is to se!arate the 6>del numbers of successive lines in a T8T-derivation& In that sense, 1D""1 serves as a !unctuation mark& In the /IF-system, the initial 1%1 of all lines is sufficient no e(tra !unctuation is needed&5
4"5 m X DBD,BDB,D%D,BB%,"B%,BDB,""",DDD,D"",BB%,"B%&DDD,""",DDD n X "B%,DDD,""",DDD 4B5 m X DBD,BDB,D%D,BB%,"B%,BDB,""",DDD,D"",BB%,%%%,BDB,D%D,"B%,BDB,""",DDD n X BB%,%%%,BDB,D%D,"B%,BDB,""",DDD

It is quite sim!le to tell which one is which, sim!ly by translatin back to the old notation, and makin some routine e(aminations to see 4"5 whether the alle ed derivation coded for by m is actually a le itimate derivation= 4B5 if so, whether the last line of the derivation coincides with the strin which n codes for& Ste! B is trivial= and ste! " is also utterly strai htforward, in this sense, there are no o!enended searches involved, no hidden endless loo!s& Think of the e(am!les above involvin the /IF-system, and now .ust mentally substitute the rules of T8T for the /IF-system1s rules, and the a(ioms of T8T for the /IF-system1s one a(iom& The al orithm in both cases is the same& Aet me make that al orithm e(!licit, 6o down the lines in the derivation one by one& /ark those which are a(ioms& 'or each line which is not an a(iom, check whether it follows by any of the rules of inference from earlier lines in the alle ed derivation& If all nona(ioms follow by rules of inference from earlier lines, then you have a le itimate derivation= otherwise it is a !hony derivation& )t each sta e, there is a clear set of tasks to !erform, and the number of them is quite easily determinable in advance&

,roof-,air-ness Is ,rimiti"e $ecursi"e111

The reason I am stressin the boundedness of these loo!s is, as you may have sensed, that I am about to assert 'F8:)/E8T)A ')7T ", The !ro!erty of bein a !roof-!air is a !rimitive recursive number-theoretical !ro!erty, and can therefore be tested for by a 0looP !ro ram& There is a notable contrast to be made here with that other closely related numbertheoretical !ro!erty, that of bein a theorem%num(er& To assert that n is a theorem-number is to assert that some value of m e(ists which forms a !roof-!air with n& 4Incidentally,

these comments a!!ly equally well to T8T and to the /IF-system= it may !erha!s hel! to kee! both in mind, the /IF-system servin as a !rototy!e&5 To check whether n is a theorem-number, you must embark on a search throu h all its !otential !roof-!air -!artners- mOand here you may be ettin into an endless chase& 8o one can say how far you will have to look to find a number which forms a !roof-!air with n as its second element& That is the whole !roblem of havin len thenin and shortenin rules in the same system, they lead to a certain de ree of un!redictability& The e(am!le of the 6oldbach Gariation may !rove hel!ful at this !oint& It is trivial to test whether a !air of numbers 4 m,n5 form a Tortoise pair, that is to say, both m and n P m should be !rime& The test is easy because the !ro!erty of !rimeness is !rimitive recursive, it admits of a !redictably terminatin test& 0ut if we want to know whether n !ossesses the Tortoise !ro!erty, then we are askin , -:oes any number m form a Tortoise-!air with n as its second element?-and this, once a ain, leads us out into the wild, /F-loo!y unknown&

111 And Is Therefore $e!resented in T;T

The key conce!t at this .uncture, then, is 'undamental 'act " iven above, for from it we can conclude 'F8:)/E8T)A ')7T B, The !ro!erty of formin a !roof-!air is testable in 0looP, and consequently, it is represented in T8T by some formula havin two free variables& Ence a ain, we are bein casual about s!ecifyin which system these !roof-!airs are relative to= it really doesn1t matter, for both 'undamental 'acts hold for any formal system& That is the nature of formal systems, it is always !ossible to tell, in a !redictably terminatin way, whether a iven sequence of lines forms a !roof, or notand this carries over to the corres!ondin arithmetical notions&

The ,o'er of ,roof-,airs

Su!!ose we assume we are dealin with the /IF-system, for the sake of concreteness& Hou !robably recall the strin we called -/F/E8-, whose inter!retation on one level was the statement -M& is a theorem of the /IF-system-& 3e can show how /F/E8 would be e(!ressed in T8T, in terms of the formula which re!resents the notion of /IF!roof-!airs& Aet us abbreviate that formula, whose e(istence we are assured of by 'undamental 'act B, this way, MI&-,$00+-,AI$ Ja)a%K

Since it is a !ro!erty of two numbers, it is re!resented by a formula with two free variables& 48ote, In this 7ha!ter we shall always use austere T8Tso be careful to distin uish between the variables a, a1, a-&5 In order to assert - M& is a theorem of the /IF-system-, we would have to make the isomor!hic statement -%+ is a theorem-number of the /IF-system-, and then translate that into T8T-notation& 3ith the aid of our abbreviation, this is easy 4remember also from 7ha!ter GIII that to indicate the re!lacement of every a% by a numeral, we write that numeral followed by -Ia%- b a5MI&-,$00+-,AI$Ja)SSSSSSSSSSSSSSSSSSSSSSSSSSSSSS?Sa%K 7ount the S1s, there are %+& 8ote that this is a closed sentence of T8T, because one free variable was quantified, the other re!laced by a numeral& ) clever thin has been done here, by the way& 'undamental 'act B ave us a way to talk about proof%pairs= we have fi ured out how to talk about theorem-numbers, as well, you .ust add an e(istential quantifier in front2 ) more literal translation of the strin above would be, -There e(ists some number a that forms a /IF-!roof-!air with %+ as its second element-& Su!!ose that we wanted to do somethin !arallel with res!ect to T8Tsay, to e(!ress the statement -?E? is a theorem of T8T-& 3e may abbreviate the formula which 'undamental 'act B assures us e(ists, in an analo ous way 4with two free variables, a ain5, T;T-,$00+-,AI$Ja)a%K 4The inter!retation of this abbreviated T8T-formula is, -8atural numbers a and a% form a T8T-!roof-!air&-5 The ne(t ste! is to transform our statement into number theory, followin the /F/E8-model above& The statement becomes -There e(ists some number a which forms a T8T !roof-!air with DDD,""",DDD as its second element-& The T8T-formula which e(!resses this is, a5T;T-,$00+-,AI$Ja)SSSSS11111111SSSSS?Sa% K

man), man) S'sA in fact, JJJ,BBB,JJJ of them/

a closed sentence of T8T& 4Aet us call it -JyS?z-, for reasons to a!!ear momentarily&5 So you see that there is a way to talk not only about the !rimitive recursive notion of T8T-!roof-!airs, but also about the related but trickier notion of T8T-theorem-numbers& To check your com!rehension of these ideas, fi ure out how to translate into T8T the followin statements of meta-T8T, 4"5 ?X? is not a theorem of T8T& 4B5 -?X? is a theorem of T8T& 4%5 h?X? is not a theorem of T8T&

?ow do the solutions differ from the e(am!le done above, and from each other? ?ere are a few more translation e(ercises& 4C5 JyS?z is a theorem of T8T& 47all the T8T-strin which e(!resses this -/ET)JyS?z-&5 4L5 /ET)-JyS?z is a theorem of T8T& 47all the T8T-strin which e(!resses this -/ET)-/ET)-JyS?z-&5 4D5 /ET)-/ET)-JyS?z is a theorem of T8T& 4$5 /ET)-/ET)-/ET)-JyS?z is a theorem of T8T& 4etc&, etc&5 E(am!le L shows that statements of meta-meta-T8T can be translated into T8T-notation= e(am!le D does the same for meta-meta-meta-T8T, etc& It is im!ortant to kee! in mind the difference between e9pressing a !ro!erty, and representing it, at this !oint& The !ro!erty of bein a T8T theorem-number, for instance, is e9pressed by the formula a5T;T-,$00+-,AI$ Ja)a%K Translation, -a% is a T8T-theorem-number-& ?owever, we have no uarantee that this formula represents the notion, for we have no uarantee that this !ro!erty is !rimitive recursivein fact, we have more than a sneakin sus!icion that it isn1t& 4This sus!icion is well warranted& The !ro!erty of bein a T8T-theorem-number is not !rimitive recursive, and no T8T-formula can re!resent the !ro!erty25 0y contrast, the !ro!erty of bein a !roof-!air, by virtue of its !rimitive recursivity, is both e(!ressible and re!resentable, by the formula already introduced&

Substitution Leads to the Second Idea

The !recedin discussion ot us to the !oint where we saw how T8T can -intros!ect- on the notion of T8T-theoremhood& This is the essence of the first !art of the !roof& 3e now wish to !ress on to the second ma.or idea of the !roof, by develo!in a notion which allows the concentration of this intros!ection into a sin le formula& To do this, we need to look at what ha!!ens to the 6>del number of a formula when you modify the formula structurally in a sim!le way& In fact, we shall consider this s!ecific modification, re!lacement of all free variables by a s!ecific numeral& 0elow are shown a cou!le of e(am!les of this o!eration in the left hand column, and in the ri ht hand column are e(hibited the !arallel chan es in 6>del numbers&

7ormula aEa 3e now re!lace all free variables by the numeral for B, SS?ESS? a,a%5a(ECSSajSSa%@

Gdel num(er BDB,""",BDB

"B%,"B%,DDD,"""&"B%,"B%,DDD BB%,%%%,BDB,D%D,%%%,BDB,"D%,D%D, BDB,"D%,"D%,""",%DB,"B%,"B%,BDB, B%D,"B%,"B%,BDB,"D%,%B%

3e now re!lace all free variables by the numeral for C, ha5a%5SSSS?ECSSajSSa%@

BB%,%%%,BDB,D%D,%%%,BDB,"D%,D%D, "B%,"B%,"B%,"B%,DDD,""",%DB,"B%, "B%,BDB,B%D,"B%,"B%,BDB,"D%,%B%

)n isomor!hic arithmetical !rocess is oin on in the ri ht-hand column, in which one hu e number is turned into an even hu er number& The function which makes the new number from the old one would not be too difficult to describe arithmetically, in terms of additions, multi!lications, !owers of "+ and so onbut we need not do so& The main !oint is this, that the relation amon 4"5 the ori inal 6>del number, 4B5 the number whose numeral is inserted, and 4%5 the resultin 6>del number, is a !rimitive recursive relation& That is to say, a 0looP test could be written which, when fed as in!ut any three natural numbers, says =ES if they are related in this way, and ;0 if they aren1t& Hou may test yourself on your ability to !erform such a testand at the same time convince yourself that there are no hidden o!en-ended loo!s to the !rocessby checkin the followin two sets of three numbers,
4"5 %DB,BDB,""B,BDB,"D%,%B%,""","B%,"B%,"B%,"B%,DDD= B= %DB,"B%,"B%,DDD,""B,"B%,"B%,DDD,%B%,""","B%,"B%,"B%,"B%,DDD& BB%,%DB,"B%,DDD,B%D,"B%,DDD,%B%,""",BDB,"D%= "= BB%,%DB,BDB,B%D,BDB,%B%,""",BDB,"D%&

4B5

)s usual, one of the e(am!les checks, the other does not& 8ow this relationshi! between three numbers will be called the substitution relationshi!& 0ecause it is !rimitive recursive, it is re!resented by some formula of T8T havin three free variables& Aet us abbreviate that T8T-formula by the followin notation,

S&B Ja)a%)a%%K 0ecause this formula re!resents the substitution relationshi!, the formula shown below must be a T8T-theorem, S&BJSSSSS11111SSSSS?Sa)SS?Sa%)SSSSSS11111SSSS?Sa%%K

4This is based on the first e(am!le of the substitution relation shown in the !arallel columns earlier in this section&5 )nd a ain because the S&B formula re!resents the substitution relation, the formula shown below certainly is not a T8T-theorem, S&BJSSS?Sa)SS?Sa%)S?Sa(K

3e now have reached the crucial !oint where we can combine all of our disassembled !arts into one meanin ful whole& 3e want to use the machinery of the T;T-,$00+-,AI$ and S&B formulas in some way to construct a sin le sentence of T8T whose inter!retation is, -This very strin of T8T is not a T8T-theorem&- ?ow do we do it? Even at this !oint, with all the necessary machinery in front of us, the answer is not easy to find& ) curious and !erha!s frivolous-seemin notion is that of substitutin a formula1s own 6>del number into itself& This is quite !arallel to that other curious, and !erha!s frivolous-seemin , notion of -quinin - in the *ir on G's -tring& Het quinin turned out to have a funny kind of im!ortance, in that it showed a new way of makin a self-referential sentence& Self reference of the Muine variety sneaks u! on you from behind the first time you see itbut once you understand the !rinci!le, you a!!reciate that it is quite sim!le and lovely& The arithmetical version of quinin let1s call it arithmo"uiningOwill allow us to make a T8T-sentence which is -about itself -& Aet us see an e(am!le of arithmoquinin & 3e need a formula with at least one free variable& The followin one will do, aES? This formula1s 6>del number is BDB,""","B%,DDD, and we will stick this number into the formula itselfor rather, we will stick its numeral in& ?ere is the result, SSSSS11111SSSSS?ES?

b
FJF,BBB,FJF S1s

BFH,BFH,JJJ,BBB,BFH,BFH,JJJ S1s

Arithmoquining

FJF,BBB,BFH,JJJ S1s

This new formula a asserts a silly falsitythat BDB&"""&"B%&DDD equals ", If we had be un with the strin TaES? and then arthmoquined, we would have come u! with a true statementas you can see for yourself& 3hen you arithmoquine, you are of course !erformin a s!ecial case of the substitution o!eration we defined earlier& If we wanted to s!eak about arithmoquinin inside T8T, we would use the formula S&BJa() a() a%K where the first two variables are the same& This comes from the fact that we are usin a sin le number in two different ways 4shades of the 7antor dia onal method25& The number a- is both 4"5 the ori inal 6>del number, and 4B5 the insertion-number& Aet us invent an abbreviation for the above formula, A$IT/M08&I;EJa%%) a%K 3hat the above formula says, in En lish, is, a% is the 6>del number of the formula otten by arithmoquinin the formula with 6>del number a%%& 8ow the !recedin sentence is lon and u ly& Aet1s introduce a concise and ele ant term to summari<e it& 3e1ll say a% is the arithmo"uinification of a%% to mean the same thin & 'or instance, the arithmoquinification of BDB,""","B%,DDD is this unutterably i antic number,
"B%,"B%,"B%,&&& "B%,"B%,"B%,DDD,""","B%,DDD

4This is .ust the 6>del number of the formula we ot when we arithmoquined aES?&5 3e can s!eak quite easily about arithmoquinin inside T8T&

8ow if you look back in the *ir on G's -tring, you will see that the ultimate trick necessary for achievin self-reference in Muine1s way is to quine a sentence which itself talks about the conce!t of quinin & It1s not enou h .ust to quineyou must quine a quine mentionin sentence2 )ll ri ht, then the !arallel trick in our case must be to arithmoquine some formula which itself is talkin about the notion of arithmoquinin 2 3ithout further ado, we1ll now write that formula down, and call it G's uncle,

b
The Last Stra'

FJF,BBB,BFH,JJJ copies of 1"B%1

~a,a%5DT;T-,$00+-,AI$Ja)a%KA$IT/M08&I;EJa()a%KR Hou can see e(!licitly how arithmoquinification is thickly involved in the !lot, 8ow this -uncle- has a 6>del number, of course, which we1ll call 1 u1& The head and tail of u1s decimal e(!ansion, and even a teeny bit of its midsection, can be read off directly, u X BB%,%%%,BDB,D%D,%%%,BDB,"D%,D%D,B"B,&&& ,"D",&&& ,B"% 'or the rest, we1d have to know .ust how the formulas T;T-,$00+-,AI$ and A$IT/M08&I;E actually look when written out& That is too com!le(, and it is quite beside the !oint, in any case& 8ow all we need to do isarithmoquine this very uncle2 3hat this entails is -bootin out- all free variablesof which there is only one, namely a11and !uttin in the numeral for u everywhere& This ives us, ~a5a%5DT;T-,$00+-,AI$Ja)a%KA$IT/M08&I;EJSSS111SSS?Sa%%)a%KR

)nd this, believe it or not, is 6>del1s strin , which we can call 161& 8ow there are two questions we must answer without delay& They are 4"5 3hat is 61s 6>del number? 4B5 3hat is the inter!retation of 6? Muestion " first& ?ow did we make 6? 3ell, we be an with the uncle, and arithmoquined it& So, by the definition of arithmoquinification, 61s 6>del number is , the arithmoquinification of u& 8ow question B& 3e will translate 6 into En lish in sta es, ettin radually more com!rehensible as we o alon & 'or our first rou h try, we make a !retty literal translation, -There do not e(ist numbers a and a% such that both 4"5 they form a T8T-!roof!air& and 4B5 a% is the arithmoquinification of u&8ow certainly there is a number a% which is the arithmoquinification of uso the !roblem must lie with the other number, a& This observation allows us to re!hrase the translation of 6 as follows, -There is no number a that forms a T8T-!roof-!air with the arithmoquinification of u&-

b
u S1s

4This ste!, which can be confusin , is e(!lained below in more detail&5 :o you see what is ha!!enin ? 6 is sayin this, -The formula whose 6>del number is the arithmoquinification of u is not a theorem of T8T&0utand this should come as no sur!rise by nowthat formula is none other than 6 itself= whence we can make the ultimate translation of 6= as -6 is not a theorem of T8T&or if you !refer, -I am not a theorem of T8T&3e have radually !ulled a hi h-level inter!retationa sentence of meta-T8Tout of what was ori inally a low-level inter!retationa sentence of number theory&

T;T Says (&ncle*(

The main consequence of this ama<in construction has already been delineated in 7ha!ter IT, it is the incom!leteness of T8T& To reiterate the ar ument, Is 6 a T8T-theorem? If so, then it must assert a truth& 0ut what in fact does 6 assert? Its own nontheoremhood& Thus from its theoremhood would follow its nontheoremhood, a contradiction& 8ow what about 6 bein a nontheorem? This is acce!table, in that it doesn1t lead to a contradiction& 0ut 61s nontheoremhood is what 6 assertshence 6 asserts a truth& )nd since 6 is not a theorem, there e(ists 4at least5 one truth which is not a theorem of T8T& 8ow to e(!lain that one tricky ste! a ain& I will use another similar e(am!le& Take this strin , Ta5a%5DT0$T0ISE-,AI$Ja)a%KTE;T/-,0>E$JSS?Sa%%) a%KR where the two abbreviations are for strin s of T8T which you can write down yourself& TE;T/-,0>E$Ja%%)a%K re!resents the statement -a% is the tenth !ower of a%%-& The literal translation into En lish is then, -There do not e(ist numbers a and a% such that both 4"5 they form a Tortoise!air, and 4B5 a% is the tenth !ower of B&-

0ut clearly, there is a tenth !ower of Bnamely "+BC& Therefore, what the strin is really sayin is that -There is no number a that forms a Tortoise-!air with "+BCwhich can be further boiled down to, -"+BC does not have the Tortoise !ro!erty&The !oint is that we have achieved a way of substitutin a description of a number, rather than its numeral, into a !redicate& It de!ends on usin one e(tra quantified variable 4a%5& ?ere, it was the number "+BC that was described as the -tenth !ower of B-= above it was the number described as the arithmoquinification of a%%&

(=ields ;ontheoremhood >hen Arithmoquined(

Aet us !ause for breath for a moment, and review what has been done& The best way I know to ive some !ers!ective is to set out e(!licitly how it com!ares with the version of the E!imenides !arado( due to Muine& ?ere is a ma!, 'alsehood quotation of a !hrase !recedin a !redicate by a sub.ect !recedin a !redicate by a quoted !hrase !recedin a !redicate by itself, in quotes 4-quinin -5 yields falsehood when quined 4a !redicate without a sub.ect5 -yields falsehood when quined4the above !redicate& quoted5 -yields falsehood when quinedyields falsehood when quined 4com!lete sentence formed by quinin the above !redicate5

nontheoremhood 6>del number of a strin substitutin a numeral 4or definite term5 into an o!en formula substitutin the 6>del number of a strin into an o!en formula substitutin the 6>del number of an o!en formula into the formula itself 4-arithmoquinin -5 the -uncle- of 64an o!en formula of T8T5 the number u 4the 6>del number of the above o!en formula5 6 itself 4sentence of T8T formed by substitutin a into the uncle, i&e&, arithmoquinin the uncle5

G#del%s Second Theorem

Since 61s inter!retation is true, the inter!retation of its ne ation h6 is false& )nd we know that no false statements are derivable in T8T& Thus neither G nor its negation XG can (e a theorem of TNT& 3e have found a -hole- in our systeman undecidable !ro!osition& This has a number of ramifications& ?ere is one curious fact which follows from 61s undecidability, althou h neither 6 nor ~6 is a theorem, the formula g6~6a is a theorem, since the rules of the Pro!ositional 7alculus ensure that all well-formed formulas of the form gP~Pa are theorems& This is one sim!le e(am!le where an assertion inside the system and an assertion about the system seem at odds with each other& It makes one wonder if the system really reflects itself accurately& :oes the -reflected metamathematics- which e(ists inside T8T corres!ond well to the metamathematics which we do? This was one of the questions which intri ued 6>del when he wrote his !a!er& In !articular, he was interested in whether it was !ossible, in the -reflected metamathematics-, to !rove T8T1s consistency& ;ecall that this was a reat !hiloso!hical dilemma of the day, how to !rove a system consistent& 6>del found a sim!le way to e(!ress the statement -T8T is consistent- in a T8T formula= and then he showed that this formula 4and all others which e(!ress the same idea5 are only theorems of T8T under one condition, that T8T is inconsistent& This !erverse result was a severe blow to o!timists who e(!ected that one could find a ri orous !roof that mathematics is contradiction-free& ?ow do you e(!ress the statement -T8T is consistent- inside T8T It hin es on this sim!le fact, that inconsistency means that two formulas, ( and (, one the ne ation of the other, are both theorems& 0ut if both ( and h( are theorems, then accordin to the Pro!ositional 7alculus, all well-formed formulas are theorems& Thus, to show T8T1s consistency, it would suffice to e(hibit one sin le sentence of T8T which can be !roven to be a nontheorem& Therefore, one way to e(!ress -T8T is consistent- is to say -The formula -+X+ is not a theorem of T8T-& This was already !ro!osed as an e(ercise a few !a es back& The translation is, ~a5T;T-,$00+-,AI$Ja)SSSSS11111SSSSS?Sa%K FFH,JJJ,BBB,JJJ S1s

It can be shown, by len thy but fairly strai htforward reasonin , thatas lon as T8T is consistentthis oath-of-consistency by T8T is not a theorem of T8T& So T8T1s !owers of intros!ection are reat when it comes to e(!ressin thin s, but fairly weak when it comes to !rovin them& This is quite a !rovocative result, if one a!!lies it meta!horically to the human !roblem of self-knowled e&

T;T Is --Incom!lete
8ow what variety of incom!leteness does T8T -en.oy? 3e shall see that T8T1s incom!leteness is of the -ome a- variety-defined in 7ha!ter GIII& This means that there is some infinite !yramidal family of strin s all of which are theorems, but whose associated -summari<in strin - is a nontheorem& It is easy to e(hibit the summari<in strin which is a nontheorem, u S1s

"a5ha%5DT;T-,$00+-,AI$Ja)a%KA$IT/M08&I;EJSSS111SSS?Sa%%)a%KR

To understand why this strin is a nontheorem, notice that it is e(tremely similar to 6 itselfin fact, 6 can be made from it in one ste! 4vi<&, accordin to T8T1s ;ule of Interchan e5& Therefore, if it were a theorem, so would 6 be& 0ut since 6 isn1t a theo rem, neither can this be& 8ow we want to show that all of the strin s in the related !yramidal family are theorems& 3e can write them own easily enou h, u S1s

ha%5DT;T-,$00+-,AI$J?Sa)a%KA$IT/M08&I;EJSSS111SSS?Sa%%)a%KR ha%5DT;T-,$00+-,AI$JS?Sa)a%KA$IT/M08&I;EJSSS111SSS?Sa%%)a%KR ha%5DT;T-,$00+-,AI$JSS?Sa)a%KA$IT/M08&I;EJSSS111SSS?Sa%%)a%KR ha%5DT;T-,$00+-,AI$JSSS?Sa)a%KA$IT/M08&I;EJSSS111SSS?Sa%%)a%KR 1 1 1 3hat does each one assert? Their translations, one by one, are, -+ and the arithmoquinification of u do not form a T8T-!roof-!air&-" and the arithmoquinification of u do not form a T8T-!roof-!air&-B and the arithmoquinification of u do not form a T8T-!roof-!air&-% and the arithmoquinification of u do not form a T8T-!roof-!air&& & & 8ow each of these assertions is about whether two s!ecific inte ers form a !roof-!air or not& 40y contrast, 6 itself is about whether one s!ecific inte er is a theorem-number or not&5 8ow because 6 is a nontheorem, no inte er forms a !roof-!air with 61s 6>del number& Therefore, each of the statements of the family is true& 8ow the cru( of the matter is that the !ro!erty of bein a !roof-!air is !rimitive recursive, hence represented, so that each of the statements in the list above, bein true, must translate into a theorem of T8Twhich means that everythin in our infinite !yramidal family is a theorem& )nd that shows why T8T is \-incom!lete&

T'o

ifferent >ays to ,lug &! the /ole

Since 61s inter!retation is true, the inter!retation of its ne ation h6 is false& )nd, usin the assum!tion that T8T is consistent, we know that no false statements are derivable in T8T& Thus neither 6 nor its ne ation h6 is a theorem of T8T& 3e have found a hole in our systeman undecidable !ro!osition& 8ow this need be no source of alarm, if we are !hiloso!hically detached enou h to reco ni<e what this is a sym!tom of& It si nifies that T8T can be e(tended, .ust as absolute eometry could be& In fact, T8T can be e(tended in two distinct directions, .ust as absolute eometry could be& It can be e(tended in a standard directionwhich corres!onds to e(tendin absolute eometry in the Euclidean direction= or, it can be e(tended in a nonstandard directionwhich corres!onds, of course, to e(tendin absolute eometry in the non-Euclidean direction& 8ow the standard ty!e of e(tension would involve )ddin 6 as a new a(iom& This su estion seems rather innocuous and !erha!s even desirable, since, after all, 6 asserts somethin true about the natural number system& 0ut what about the nonstandard ty!e of e(tensions& If it is at all !arallel to the case of the !arallel !ostulate, it must involve addin the ne ation of 6 as a new a(iom& 0ut how can we even contem!late doin such a re!u nant, hideous thin ? )fter all, to !ara!hrase the memorable words of 6irolamo Saccheri, isn1t what h6 says -re!u nant to the nature of the natural numbers-?

Su!ernatural ;umbers
I ho!e the irony of this quotation strikes you& The e(act !roblem with Saccheri1s a!!roach to eometry was that he be an with a fi(ed notion of what was true and what was not true, and he set out only to !rove what he1d assessed as true to start with& :es!ite the cleverness of his a!!roachwhich involved denyin the fifth !ostulate, and then !rovin many -re!u nant- !ro!ositions of the ensuin eometrySaccheri never entertained the !ossibility of other ways of thinkin about !oints and lines& 8ow we should be wary of re!eatin this famous mistake& 3e must consider im!artially, to the e(tent that we can, what it would mean to add h6 as an a(iom to T8T& Just think what mathematics would be like today if !eo!le had never considered addin new a(ioms of the followin sorts, a5CaIa@ES? a5SaE? a5Ca(a@ESS? a5SCa%a@E?

3hile each of them is -re!u nant to the nature of !reviously known number systems-, each of them also !rovides a dee! and wonderful e(tension of the notion of whole numbers, rational numbers, ne ative numbers, irrational numbers, ima inary numbers& Such a !ossibility is what h6 is tryin to et us to o!en our eyes to& 8ow in the !ast, each new e(tension of the notion of number was reeted with hoots and catcalls& Hou can hear this !articularly loudly in the names attached to the unwelcome arrivals, such as -irrational numbers-, -ima inary numbers-& True to this tradition, we shall name the numbers which h6 is announcin to us the su!ernatural numbers, showin how we feel they violate all reasonable and commonsensical notions& If we are oin to throw h6 in as the si(th a(iom of T8T, we had better understand how in the world it could coe(ist, in one system, with the infinite !yramidal family we .ust finished discussin & To !ut it bluntly, h6 says, -There e(ists some number which forms a T8T-!roof-!air with the arithmoquinification of ubut the various members of the !yramidal family successively assert, -+ is not that number-" is not that number-B is not that number& & This is rather confusin , because it seems to be a com!lete contradiction 4which is why it is called -\-inconsistency-5& )t the root of our confusionmuch as in the case of the s!littin of eometryis our stubborn resistance to ado!t a modified inter!retation for the symbols, des!ite the fact that we are quite aware that the system is a modified system& 3e want to et away without reinter!retin any symbolsand of course that will !rove im!ossible& The reconciliation comes when we reinter!ret as -There e(ists a generali@ed natural number-, rather than as -There e(ists a natural number-& )s we do this, we shall also reinter!ret " in the corres!ondin way& This means that we are o!enin the door to some e(tra numbers besides the natural numbers& These are the supernatural num(ers& The naturals and su!ernaturals to ether make u! the totality of generali@ed naturals& The a!!arent contradiction vanishes into thin air, now, for the !yramidal family still says what it said before, -8o natural number forms a T8T-!roof-!air with the arithmoquinification of u&- The family doesn1t say anythin about su!ernatural numbers, because there are no numerals for them& 0ut now, h6 says, -There e(ists a generali@ed natural number which forms a T8T-!roof-!air with the arithmoquinification of u&- It is clear that taken to ether, the family and -6 tell us somethin , that there is a supernatural number which forms a T8T-!roof-!air with the arithmoquinification of u& That is all there is no contradiction any more& T8T^h6 is a consistent system, under an inter!retation which includes su!ernatural numbers&

Since we have now a reed to e(tend the inter!retations of the two quantifiers, this means that any theorem which involves either of them has an e(tended meanin & 'or e(am!le, the commutativity theorem
"a5"a%5CaIa%@ECa%Ia@

now tells us that addition is commutative for all enerali<ed natural numbersin other words, not only for natural numbers, but also for su!ernatural numbers& Aikewise, the T8T-theorem which says -B is not the square of a natural number ha5Caja@ESS? now tells us that B is not the square of a su!ernatural number, either& In fact, su!ernatural numbers share all the !ro!erties of natural numbers, as lon as those !ro!erties are iven to us in theorems of T8T& In other words, everythin that can be formall) pro#en about natural numbers is thereby established also for su!ernatural numbers& This means, in !articular, that su!ernatural numbers are not anythin already familiar to you, such as fractions, or ne ative numbers, or com!le( numbers, or whatever& The su!ernatural numbers are, instead, best visuali<ed as inte ers which are reater than all natural numbersas infinitel) large inte ers& ?ere is the !oint, althou h theorems of T8T can rule out ne ative numbers, fractions, irrational numbers, and com!le( numbers, still there is no way to rule out infinitely lar e inte ers& The !roblem is, there is no way even to e(!ress the statement -There are no infinite quantities-& This sounds quite stran e, at first& Just e(actly how bi is the number which makes a T8T-!roof-!air with 61s 6>del number? 4Aet1s call it '3', for no !articular reason&5 Fnfortunately, we have not ot any ood vocabulary for describin the si<es of infinitely lar e inte ers, so I am afraid I cannot convey a sense of 31s ma nitude& 0ut then .ust how bi is i 4the square root of -"5? Its si<e cannot be ima ined in terms of the si<es of familiar natural numbers& Hou can1t say, -3ell, i is about half as bi as "C, and #I"+ as bi as BC&- Hou have to say, -i squared is -"-, and more or less leave it at that& ) quote from )braham Aincoln seems a !ro!os here& 3hen he was asked, -?ow lon should a man1s le s be?- he drawled, -Aon enou h to reach the round&- That is more or less how to answer the question about the si<e of 3it should be .ust the si<e of a num(er which specifies the structure of a proof of GOno bi er, no smaller& Ef course, any theorem of T8T has many different derivations, so you mi ht com!lain that my characteri<ation of 3 is nonunique& That is so& 0ut the !arallel with i the square root of -"still holds& 8amely, recall that there is another number whose square is also minus one, -i& 8ow i and -i are not the same number& They .ust have a !ro!erty in common& The only trouble is that it is the !ro!erty which defines them2 3e have to choose one of themit doesn1t matter which oneand call it -i-& In fact there1s no way of tellin them a!art& So for all we know we could have been callin the wron one -i- for all these centuries and it would have made no difference& 8ow, like i, 3 is also nonuniquely defined& So you .ust have to think of 3 as bein some s!ecific one of the

many !ossible su!ernatural numbers which form T8T-!roof-!airs with the arithmoquinification of u!

Su!ernatural Theorems /a"e Infinitely Long

eri"ations

3e haven1t yet faced head on what it means to throw h6 in as an a(iom& 3e have said it but not stressed it& The !oint is that h6 asserts that 6 has a proof& ?ow can a system survive, when one of its a(ioms asserts that its own ne ation has a !roof? 3e must be in hot water now2 3ell, it is not so bad as you mi ht think& )s lon as we only construct finite !roofs, we will never !rove 6& Therefore, no calamitous collision between 6 and its ne ative h6 will ever take !lace& The su!ernatural number Y3 won1t cause any disaster& ?owever, we will have to et used to the idea that h6 is now the one which asserts a truth 4-6 has a !roof -5, while 6 asserts a falsity 4-6 has no !roof-5& In standard number theory it is the other way aroundbut then, in standard number theory there aren1t any su!ernatural numbers& 8otice that a su!ernatural theorem of T8Tnamely 6 may assert a falsity, but all natural theorems still assert truths&

Su!ernatural Addition and Multi!lication

There is one e(tremely curious and une(!ected fact about su!ernaturals which I would like to tell you, without !roof& 4I don1t know the !roof either&5 This fact is reminiscent of the ?eisenber uncertainty !rinci!le in quantum mechanics& It turns out that you can -inde(- the su!ernaturals in a sim!le and natural way by associatin with each su!ernatural number a trio of ordinary inte ers 4includin ne ative ones5& Thus, our ori inal su!ernatural number, 3, mi ht have the inde( set 4#,-*,%5, and its successor, 3 ^ ", mi ht have the inde( set 4#,-*,C5& 8ow there is no unique way to inde( the su!ernaturals= different methods offer different advanta es and disadvanta es& Fnder some inde(in schemes, it is very easy to calculate the inde( tri!let for the sum of two su!ernaturals, iven the indices of the two numbers to be added& Fnder other inde(in schemes, it is very easy to calculate the inde( tri!let for the product of two su!ernaturals, iven the indices of the two numbers to be multi!lied& 0ut under no inde(in scheme is it !ossible to calculate both& /ore !recisely, if the sum1s inde( can be calculated by a recursive function, then the !roduct1s inde( will not be a recursive function= and conversely, if the !roduct1s inde( is a recursive function, then the sum1s inde( will not be& Therefore, su!ernatural schoolchildren who learn their su!ernatural !lus-tables will have to be e(cused if they do not know their su!ernatural times-tablesand vice versa2 Hou cannot know both at the same time&

Su!ernaturals Are &seful111

Ene can o beyond the number theory of su!ernaturals, and consider su!ernatural fractions 4ratios of two su!ernaturals5, su!ernatural real numbers, and so on& In fact, the calculus can be !ut on a new footin , usin the notion of su!ernatural real numbers& Infinitesimals such as d9 and d), those old bu aboos of mathematicians, can be

com!letely .ustified, by considerin them to be reci!rocals of infinitely lar e real numbers2 Some theorems in advanced analysis can be !roven more intuitively with the aid of -nonstandard analysis-& 111 But Are They $eal7 8onstandard number theory is a disorientin thin when you first meet u! with it& 0ut, then, non-Euclidean eometry is also a disorientin sub.ect& In both instances, one is !owerfully driven to ask, -0ut which one of these two rival theories is correct? 3hich is the truth?- In a certain sense, there is no answer to such a question& 4)nd yet, in another senseto be discussed laterthere is an answer&5 The reason that there is no answer to the question is that the two rival theories, althou h they em!loy the same terms, do not talk about the same conce!ts& Therefore, they are only su!erficially rivals, .ust like Euclidean and non-Euclidean eometries& In eometry, the words -!oint-, -line-, and so on are undefined terms, and their meanin s are determined by the a(iomatic system within which they are used& Aikewise for number theory& 3hen we decided to formali<e T8T& we !reselected the terms we would use as inter!retation wordsfor instance, words such as -number-, -!lus-, -times-, and so on& 0y takin the ste! of formali<ation, we were committin ourselves to acce!tin whatever !assive meanin s these terms mi ht take on& 0ut.ust like Saccheriwe didn1t antici!ate any sur!rises& 3e thou ht we knew what the true, the real, the only theory of natural numbers was& 3e didn1t know that there would be some questions about numbers which T8T would leave o!en, and which could therefore be answered ad libitum by e(tensions of T8T headin off in different directions& Thus, there is no basis on which to say that number theory -really- is this way or that, .ust as one would be loath to say that the square root of -" -really- e(ists, or -really- does not&

Bifurcations in Geometry) and ,hysicists

There is one ar ument which can be, and !erha!s ou ht to be, raised a ainst the !recedin & Su!!ose e(!eriments in the real, !hysical world can be e(!lained more economically in terms of one !articular version of eometry than in terms of any other& Then it mi ht make sense to say that that eometry is -true-& 'rom the !oint of view of a !hysicist who wants to use the -correct- eometry, then it makes some sense to distin uish between the -true- eometry, and other eometries& 0ut this cannot be taken too sim!listically& Physicists are always dealin with a!!ro(imations and ideali<ations of situations& 'or instance, my own Ph&:& work, mentioned in 7ha!ter G, was based on an e(treme ideali<ation of the !roblem of a crystal in a ma netic field& The mathematics which emer ed was of a hi h de ree of beauty and symmetry& :es!iteor rather, because ofthe artificiality of the model, some fundamental features emer ed cons!icuously in the ra!h& These features then su est some uesses about the kinds of thin s that mi ht ha!!en in more realistic situations& 0ut without the sim!lifyin assum!tions which !roduced my ra!h, there could never be such insi hts& Ene can see this kind of thin over and over a ain in !hysics, where a !hysicist uses a -nonreal-

situation to learn about dee!ly hidden features of reality& Therefore, one should be e(tremely cautious in sayin that the brand of eometry which !hysicists mi ht wish to use would re!resent -the true eometry-, for in fact, !hysicists will always use a variety of different eometries, choosin in any iven situation the one that seems sim!lest and most convenient& 'urthermoreand !erha!s this is even more to the !oint!hysicists do not study .ust the %-: s!ace we live in& There are whole families of -abstract s!aces- within which !hysical calculations take !lace, s!aces which have totally different eometrical !ro!erties from the !hysical s!ace within which we live& 3ho is to say, then, that -the true eometry- is defined by the s!ace in which Franus and 8e!tune orbit around the sun? There is -?ilbert s!ace-, where quantum-mechanical wave functions undulate= there is -momentum s!ace-, where 'ourier com!onents dwell= there is -reci!rocal s!ace-, where wave-vectors cavort= there is -!hase s!ace-, where many-!article confi urations swish= and so on& There is absolutely no reason that the eometries of all these s!aces should be the same= in fact, they couldn1t !ossibly be the same2 So it is essential and vital for !hysicists that different and -rival- eometries should e(ist&

Bifurcations in ;umber Theory) and Ban.ers

So much for eometry& 3hat about number theory? Is it also essential and vital that different number theories should coe(ist with each other? If you asked a bank officer, my uess is that you would et an e(!ression of horror and disbelief& ?ow could B and B add u! to anythin but C? )nd moreover, if B and B did not make C, wouldn1t world economies colla!se immediately under the unbearable uncertainty o!ened u! by that fact? 8ot really& 'irst of all, nonstandard number theory doesn1t threaten the a e-old idea that B !lus B equals C& It differs from ordinary number theory only in the way it deals with the conce!t of the infinite& )fter all, e#er) theorem of TNT remains a theorem in an) e9tension of TNT2 So bankers need not des!air of the chaos that will arrive when nonstandard number theory takes over& )nd anyway, entertainin fears about old facts bein chan ed betrays a misunderstandin of the relationshi! between mathematics and the real world& /athematics only tells you answers to questions in the real world after you have taken the one vital ste! of choosin which kind of mathematics to a!!ly& Even if there were a rival number theory which used the symbols 1B1, 1%1, and 1^1, and in which a theorem said -B ^ B X %-, there would be little reason for bankers to choose to use that theory2 'or that theory does not fit the way money works& Hou fit your mathematics to the world, and not the other way around& 'or instance, we don1t a!!ly number theory to cloud systems, because the very conce!t of whole numbers hardly fits& There can be one cloud and another cloud, and they will come to ether and instead of there bein two clouds, there will still only be one& This doesn1t !rove that " !lus " equals "= it .ust !roves that our number theoretical conce!t of -one- is not a!!licable in its full !ower to cloud countin &

Bifurcations in ;umber Theory) and Metamathematicians

So bankers, cloud-counters, and most of the rest of us need not worry about the advent of su!ernatural numbers, they won1t affect our everyday !erce!tion of the world in the sli htest& The only !eo!le who mi ht actually be a little worried are !eo!le whose endeavors de!end in some crucial way on the nature of infinite entities& There aren1t too many such !eo!le aroundbut mathematical lo icians are members of this cate ory& ?ow can the e(istence of a bifurcation in number theory affect them? 3ell, number theory !lays two roles in lo ic, 4"5 when a(iomati<ed, it is an ob.ect of study= and 4B5 when used informally, it is an indis!ensable tool with which formal systems can be investi ated& This is the use-mention distinction once a ain, in fact, in role 4"5, number theory is mentioned, in role 4B5 it is used& 8ow mathematicians have .ud ed that number theory is a!!licable to the study of formal systems even if not to cloud-countin , .ust as bankers have .ud ed that the arithmetic of real numbers is a!!licable to their transactions& This is an e(tramathematical .ud ement, and shows that the thou ht !rocesses involved in doin mathematics, .ust like those in other areas, involve -tan led hierarchies- in which thou hts on one level can affect thou hts on any other level& Aevels are not cleanly se!arated, as the formalist version of what mathematics is would have one believe& The formalist !hiloso!hy claims that mathematicians only deal with abstract symbols, and that they couldn1t care less whether those symbols have any a!!lications to or connections with reality& 0ut that is quite a distorted !icture& 8owhere is this clearer than in metamathematics& If the theory of numbers is itself used as an aid in ainin factual knowled e about formal systems, then mathematicians are tacitly showin that they believe these ethereal thin s called -natural numbers- are actually part of realit)O not .ust fi ments of the ima ination& This is why I !arenthetically remarked earlier that, in a certain sense, there is an answer to the question of which version of number theory is -true-& ?ere is the nub of the matter, mathematical lo icians must choose which version of number theory to !ut their faith in& In !articular, they cannot remain neutral on the question of the e(istence or none(istence of su!ernatural numbers, for the two different theories may ive different answers to questions in metamathematics& 'or instance, take this question, -Is h6 finitely derivable in T8T?- 8o one actually knows the answer& 8evertheless, most mathematical lo icians would answer no without hesitation& The intuition which motivates that answer is based on the fact that if h6 were a theorem, T8T would be \%inconsistent, and this would force su!ernaturals down your throat if you wanted to inter!ret T8T meanin fullya most un!alatable thou ht for most !eo!le& )fter all, we didn1t intend or e(!ect su!ernaturals to be !art of T8T when we invented it& That is, weor most of usbelieve that it is !ossible to make a formali<ation of number theory which does not force you into believin that su!ernatural numbers are every bit as real as naturals& It is that intuition about reality which determines which -fork- of number theory mathematicians will !ut their faith in, when the chi!s are down& 0ut this faith may be wron & Perha!s every consistent formali<ation of number theory which humans invent will im!ly the e(istence of su!ernaturals, by bein co-inconsistent& This is a queer thou ht, but it is conceivable&

If this were the casewhich I doubt, but there is no dis!roof availablethen 6 would not have to be undecidable& In fact, there mi ht be no undecidable formulas of T8T at all& There could sim!ly be one unbifurcated theory of numberswhich necessarily includes su!ernaturals& This is not the kind of thin mathematical lo icians e(!ect, but it is somethin which ou ht not to be re.ected outri ht& 6enerally, mathematical lo icians believe that T8Tand systems similar to itare \-consistent, and that the 6>del strin which can be constructed in any such system is undecidable within that system& That means that they can choose to add either it or its ne ation as an a(iom&

/ilbert%s Tenth ,roblem and the Tortoise

I would like to conclude this 7ha!ter by mentionin one e(tension of 6>del1s Theorem& 4This material is more fully covered in the article -?ilbert1s Tenth Problem- by :avis and ?ersh, for which see the 0iblio ra!hy&5 'or this, I must define what a :io!hantine equation is& This is an equation in which a !olynomial with fi(ed inte ral coefficients and e(!onents is set to +& 'or instance, aX+ and L9 ^ "%)-"X+ and Lp ^ "$" - "$$ X + and a"B%,DDD,""",DDD ^ b"B%,DDD,""",DDD - c"B%,DDD,""",DDD X + are :io!hantine equations& It is in eneral a difficult matter to know whether a iven :io!hantine equation has any inte er solutions or not& In fact, in a famous lecture at the be innin of the century, ?ilbert asked mathematicians to look for a eneral al orithm by which one could determine in a finite number of ste!s if a iven :io!hantine equation has inte er solutions or not& Aittle did he sus!ect that no such al orithm e(ists2 8ow for the sim!lification of 6& It has been shown that whenever you have a sufficiently !owerful formal number theory and a 6>del-numberin for it, there is a :io!hantine equation which is equivalent to 6& The equivalence lies in the fact that this equation, when inter!reted on a metamathematical level, asserts of itself that it has no solutions& Turn it around, if you found a solution to it, you could construct from it the 6>del number of a !roof in the system that the equation has no solutions2 This is what the Tortoise did in the &relude, usin 'ermat1s equation as his :io!hantine equation& It is nice to know that when you do this, you can retrieve the sound of Eld 0ach from the molecules in the air2
L "$

.irthday Cantatatata
:ne fine 'a) da), the Tortoise and *chilles meet, wandering in the woods! The latter, all decked out handsomel), is doing a Miggish sort of thing to a tune which he himself is humming! :n his #est he is wearing a great (ig (utton with the words "Toda) is m) Birthda)A" Tortoise, ?ello there, )chilles& 3hat makes you so .oyful today? Is it your birthday, by any chance? *chilles, Hes, yes2 Hes it is, today is my birthday2 Tortoise, That is what I had sus!ected, on account of that button which you are wearin , and also because unless I am mistaken, you are sin in a tune from a 0irthday 7antata by 0ach, one written in "$B$ for the fifty-seventh birthday of )u ustus, @in of Sa(ony& *chilles, Hou1re ri ht& )nd )u ustus1 birthday coincides with mine, so T?IS 0irthday 7antata has double meanin & ?owever, I shan1t tell you my a e& Tortoise, Eh, that1s !erfectly all ri ht& ?owever, I would like to know one other thin & 'rom what you have told me so far, would it be correct to conclude that today is your birthday? *chilles, Hes, yes, it would be& Today is my birthday& Tortoise, E(cellent& That1s .ust as I sus!ected& So now, I 3IAA conclude it is your birthday, unless *chilles, Hesunless what? Tortoise, Fnless that would be a !remature or hasty conclusion to draw, you know& Tortoises don1t like to .um! to conclusions, after all& 43e don1t like to .um! at all, but es!ecially not to conclusions&5 So let me .ust ask you, knowin full well of your fondness for lo ical thou ht, whether it would be reasonable to deduce lo ically from the fore oin sentences, that today is in fact your birthday& *chilles, I do believe I detect a !attern to your questions, /r& T& 0ut rather than .um! to conclusions myself, I shall take your question at face value, and answer it strai htforwardly& The answer is, HES& Tortoise, 'ine2 'ine2 Then there is only one more thin I need to know, to be quite certain that today is *chilles, Hes, yes, yes, yes&&& I can already see the line of your questionin , /r& T& I1ll have you know that I am not so ullible as I was when we discussed Euclid1s !roof, a while back& Tortoise, 3hy, who would ever have thou ht you to be ullible? Muite to the contrary, I re ard you as an e(!ert in the forms of lo ical thou ht, an authority in the science of valid deductions& a fount of knowled e about certain correct methods of reasonin &&& To tell the truth, )chilles, you are, in my o!inion, a veritable titan in the art of rational co itation& )nd it is only for that reason that I would ask you, -:o the fore oin sentences !resent enou h evidence that I should conclude without further !u<<lement that today is your birthday?

*chilles, Hou flatten me with your wei hty !raise, /r& T'A)TTE;, I mean& 0ut I am struck by the re!etitive nature of your questionin and in my estimation, you, .ust as well as I, could have answered 1yes1 each time& Tortoise, Ef course I could have, )chilles& 0ut you see, to do so would have been to make a 3ild 6uessand Tortoises abhor 3ild 6uesses& Tortoises formulate only Educated 6uesses& )h, yesthe !ower of the Educated 6uess& Hou have no idea how many !eo!le fail to take into account all the ;elevant 'actors when they1re uessin & *chilles, It seems to me that there was only one relevant factor in this ri marole, and that was my first statement& Tortoise, Eh, to be sure, it1s at least E8E of the factors to take into account, I1d saybut would you have me ne lect Ao ic, that venerated science of the ancients? Ao ic is always a ;elevant 'actor in makin Educated 6uesses, and since I have with me a renowned e(!ert in Ao ic, I thou ht it only Ao ical to take advanta e of that fact, and confirm my hunches, by directly askin him whether my intuitions were correct& So let me finally come out and ask you !oint blank, -:o the !recedin sentences allow me to conclude, with no room for doubt, that Today is your 0irthday?*chilles, 'or one more time, HES& 0ut frankly s!eakin , I have the distinct im!ression that you could have su!!lied that answeras well as all the !revious onesyourself& Tortoise, ?ow your words stin 2 3ould I were so wise as your insinuation su ests2 0ut as merely a mortal Tortoise, !rofoundly i norant and lon in to take into account all the ;elevant 'actors, I needed to know the answers to all those questions& *chilles, 3ell then, let me clear the matter u! for once and for all& The answer to all the !revious questions, and to all the succeedin ones which you will ask alon the same line, is .ust this, HES& Tortoise, 3onderful2 In one fell swoo!, you have circumvented the whole mess, in your characteristically inventive manner& I ho!e you won1t mind if I call this in enious trick an )8S3E; S7?E/)& It rolls u! yes-answers numbers ", B, %, etc&, into one sin le ball& In fact, comin as it does at the end of the line, it deserves the title -)nswer Schema Eme a-, 1\1 bein the last letter of the 6reek al!habetas if HEF needed to be told T?)T2 *chilles, I don1t care what you call it& I am .ust very relieved that you finally a ree that it is my birthday, and we can o on to some other to!icsuch as what you are oin to ive me as a !resent& Tortoise, ?old onnot so fast& I 3IAA a ree it is your birthday, !rovided one thin & *chilles, 3hat? That I ask for no !resent? Tortoise, 8ot at all& In fact, )chilles, I am lookin forward to treatin you to a fine birthday dinner, !rovided merely that I am convinced that knowled e of all those yesanswers at once 4as su!!lied by )nswer Schema \5 allows me to !roceed directly and without any further detours to the conclusion that today is your birthday& That1s the case, isn1t it? *chilles, Hes, of course it is& Tortoise, 6ood& )nd now I have yes-answer \ ^ "& )rmed with it, I can !roceed to acce!t the hy!othesis that today is your birthday, if it is valid to do so& 3ould you be so kind as to counsel me on that matter, )chilles?

*chilles, 3hat is this? I thou ht I had seen throu h your infinite !lot& 8ow doesn1t yesanswer | ^ " satisfy you? )ll ri ht& I1ll ive you not only yes-answer \ ^ B, but also yes-answers \ ^ %, \ ^ C, and so on& Tortoise, ?ow enerous of you, )chilles& )nd here it is your birthday, when I should be ivin HEF !resents instead of the reverse& Er rather, I SFSPE7T it is your birthday& I uess I can conclude that it IS your birthday, now, armed with the new )nswer Schema, which I will call -)nswer Schema B\ -& 0ut tell me, )chilles, :oes )nswer Schema B\ ;E)AAH allow me to make that enormous lea!, or am I missin somethin ? *chilles, Hou won1t trick me any more, /r& T& I1ve seen the way to end this silly ame& I hereby shall !resent you with an )nswer Schema to end all )nswer Schemas2 That is, I !resent you simultaneously with )nswer Schemas \, B\, %\, C\, L\, etc& 3ith this /eta-)nswer-Schema, I have JF/PE: EFT of the whole system, kit and caboodle, transcended this silly ame you thou ht you had me tra!!ed inand now we are :E8E2 Tortoise, 6ood rief2 I feel honored, )chilles, to be the reci!ient of such a !owerful )nswer Schema& I feel that seldom has anythin so i antic been devised by the mind of man, and I am awestruck by its !ower& 3ould you mind if I ive a name to your ift? *chilles, 8ot at all& Tortoise, Then I shall call it -)nswer Schema \B-& )nd we can shortly !roceed to other mattersas soon as you tell me whether the !ossession of )nswer Schema \B allows me to deduce that today is your birthday& *chilles, Eh, woe is me2 7an1t I ever reach the end of this tantali<in trail? 3hat comes ne(t? Tortoise, 3ell, after )nswer Schema \B there1s answer \B ^ "& )nd then answer \B ^ B& )nd so forth& 0ut you can wra! those all to ether into a !acket, bein )nswer Schema \B ^ \& )nd then there are quite a few other answer-!ackets, such as \B ^ B\, and \B ^ %\S& Eventually you come to )nswer Schema B \B, and after a while, )nswer Schemas %\B and C\B& 0eyond them there are yet further )nswer Schemas, such as \%, \C, \L, and so on& It oes on quite a ways, you know& *chilles, I can ima ine, I su!!ose it comes to )nswer Schema || after a while& Tortoise, Ef course& *chilles, )nd then \\ , and \\ ? Tortoise, Hou1re catchin on mi hty fast, )chilles& I have a su estion for you, if you don1t mind& 3hy don1t you throw all of those to ether into a sin le )nswer Schema? *chilles, )ll ri ht, thou h I1m be innin to doubt whether it will do any ood& Tortoise, It seems to me that within our namin conventions as so far set u!, there is no obvious name for this one& So !erha!s we should .ust arbitrarily name it )nswer Schema }+& *chilles, 7onfound it all2 Every time you ive one of my answers a 8)/E, it seems to si nal the imminent shatterin of my ho!es that that answer will satisfy you& 3hy don1t we .ust leave this )nswer Schema nameless? Tortoise, 3e can hardly do that, )chilles& 3e wouldn1t have any way to refer to it without a name& )nd besides, there is somethin inevitable and rather beautiful about this
\ \\

!articular )nswer Schema& It would be quite un raceful to leave it nameless2 )nd you wouldn1t want to do somethin lackin in race on your birthday, would you? Er is it your birthday? Say, s!eakin of birthdays, today is /H birthday2 *chilles, It is? Tortoise, Hes, it is& 3ell, actually, it1s my uncle1s birthday, but that1s almost the same& ?ow would you like to treat me to a delicious birthday dinner this evenin ? *chilles, 8ow .ust a cotton-!ickin minute, /r& T& Today is /H birthday& Hou should do the treatin 2 Tortoise, )h, but you never did succeed in convincin me of the veracity of that remark& Hou ke!t on beatin around the bush with answers, )nswer Schemas, and whatnot& )ll I wanted to know was if it was your birthday or not, but you mana ed to befuddle me entirely& Eh, well, too bad& In any case, I1ll be ha!!y to let you treat me to a birthday dinner this evenin & *chilles, Gery well& I know .ust the !lace& They have a variety of delicious sou!s& )nd I know e(actly what kind we should have&&&

C/A,TE$ <4

3um!ing out of the System

A More ,o'erful +ormal System
E8E E' T?E thin s which a thou htful critic of 6>del1s !roof mi ht do would be to e(amine its enerality& Such a critic mi ht, for e(am!le, sus!ect that 6>del has .ust cleverly taken advanta e of a hidden defect in one !articular formal system, T8T& If this were the case, then !erha!s a formal system su!erior to T8T could be develo!ed which would not be sub.ect to the 6>delian trick, and 6>del1s Theorem would lose much of its stin & In this 7ha!ter we will carefully scrutini<e the !ro!erties of T8T which made it vulnerable to the ar uments of last 7ha!ter& ) natural thou ht is this, If the basic trouble with T8T is that it contains a -hole-in other words, a sentence which is undecidable, namely 6then why not sim!ly !lu u! the hole? 3hy not .ust tack 6 onto T8T as a si(th a(iom? Ef course, by com!arison to the other a(ioms, 6 is a ridiculously hu e iant, and the resultin system T8T^6would have a rather comical as!ect due to the dis!ro!ortionateness of its a(ioms& 0e that as it may, addin 6 is a reasonable su estion& Aet us consider it done& 8ow, it is to be ho!ed, the new system, T8T^6, is a su!erior formal systemone which is not only su!ernatural-free, but also complete& It is certain that T8T^6 is su!erior to T8T in at least one res!ect, the strin 6 is no lon er undecidable in this new system, since it is a theorem& 3hat was the vulnerability of T8T due to? The essence of its vulnerability was that it was ca!able of e(!ressin statements about itselfin !articular, the statement -I 7annot 0e Proven in 'ormal System T8Tor, e(!anded a bit, -There does not e(ist a natural number which forms a T8T-!roof-!air with the 6>del number of this strin &Is there any reason to e(!ect or ho!e that T8T^6 would be invulnerable to 6>del1s !roof? 8ot really& Eur new system is .ust as e(!ressive as T8T& Since 6>del1s !roof relies !rimarily on the e(!ressive !ower of a formal system, we should not be sur!rised to see our new system succumb, too& The trick will be to find a strin which e(!resses the statement -I 7annot 0e Proven in 'ormal System T8T^6&)ctually, it is not much of a trick, once you have seen it done for T8T& )ll the same !rinci!les are em!loyed, only, the conte(t shifts sli htly& 4'i uratively s!eakin , we take a tune we know and sim!ly sin it a ain, only in a hi her key&5 )s before, the strin

which we are lookin forlet us call it -61 -is constructed by the intermediary of an -uncle-, 0ut instead of bein based on the formula which re!resents T8T-!roof-!airs, it is based on the similar but sli htly more com!licated notion of T8T^6-!roof-!airs& This notion of T8T^6-!roof-!airs is only a sli ht e(tension of the ori inal notion of T8T!roof-!airs& ) similar e(tension could be envisa ed for the /IF-system& 3e have seen the unadulterated form of /IF-!roof-!airs& 3ere we now to add M& as a second a(iom, we would be dealin with a new system-the /IF^ M& system& ) derivation in this e(tended system is !resented, M& M&& a(iom rule B

There is a /IF^M&-!roof-!air which corres!ondsnamely, m X %+%++, n X %++& Ef course, this !air of numbers does not form a /IF-!roof-!air-only a /IF^ M&-!roof-!air& The addition of an e(tra a(iom does not substantially com!licate the arithmetical !ro!erties of !roof-!airs& The si nificant fact about themthat bein a !roof-!air is !rimitive recursiveis !reserved&

The G#del Method $ea!!lied

8ow, returnin to T8T^6, we will find a similar situation& T8T^6 !roof-!airs, like their !redecessors, are !rimitive recursive, so they are re!resented inside T8T^6 by a formula which we abbreviate in an obvious manner& CT;TIG@-,$00+-,AI$Ja)a%K 8ow we .ust do everythin all over a ain& 3e make the counter!art of 6 by be innin with an -uncle-, .ust as before, a5 a%5DCT;TIG@-,$00+-,AI$Ja)a%KA$IT/M08&I;EJa%%)a%KR Aet us say its 6>del-number is u1& 8ow we arithmoquine this very uncle& That will ive us 61, a5a%5DCT;TIG@-,$00+-,AI$Ja)a%KA$IT/M08&I;EJSSS1111SSS?Sa%%)a%KR

b
u1 S1s

Its inter!retation is -There is no number a that forms a T8T ^ 6-!roof-!air with the arithmoquinification of u1&-

/ore concisely, -I 7annot 0e Proven in 'ormal System T8T^6&-

Multifurcation
3ell 4yawn5, the details are quite borin from here on out& 61 is !recisely to T8T^6 as 6 was to T8T Itself& Ene finds that either 61 or -61 can be added to T8T^6, to yield a further s!littin of number theory& )nd, lest you think this only ha!!ens to the - ood uys-, this very same dastardly trick can be !layed u!on T8T^-61that is, u!on the nonstandard e(tension of T8T otten by addin 61s ne ation& So now we see 4'i & $L5 that there are all sorts of bifurcations in number theory,

73G5=E CG! "'ultifurcation" of TNT! Each e9tension of TNT has its #er) own Gdel sentenceR that sentence, or its negation, can (e added on, so that from each e9tension there sprouts a pair of further e9tensions, a process which goes on ad infinitum!

Ef course, this is .ust the be innin & Aet us ima ine movin down the leftmost branch of this downwards-!ointin tree, where we always toss in the 6>del sentences 4rather than their ne ations5& This is the best we can do by way of eliminatin su!ernaturals& )fter addin 6, we add 61& Then we add 6-, and 6-1, and so on& Each time we make a new e(tension of T8T, its vulnerability to the Tortoise1s method!ardon me, I mean 6>del1s methodallows a new strin to be devised, havin the inter!retation& -I cannot be !roven in formal system T&-

8aturally, after a while, the whole !rocess be ins to seem utterly !redictable and routine& 3hy, all the -holes- are made by one sin le technique2 This means that, viewed as ty!o ra!hical ob.ects, they are all cast from one sin le mold, which in turn means that one sin le a(iom schema suffices to re!resent all of them2 So if this is so, why not !lu u! all the holes at once and be done with this nasty business of incom!leteness once and for all? This would be accom!lished by addin an a(iom schema to T8T, instead of .ust one a(iom at a time& S!ecifically, this a(iom schema would be the mold in which all of 6, 61, 6-, 6-1, etc&, are cast& 0y addin this a(iom schema 4let1s call it -6 \-5, we would be outsmartin the -6>deli<ation- method& Indeed, it seems quite clear that addin 6 \ to T8T would be the last ste! necessary for the com!lete a(iomati<ation of all of numbertheoretical truth& It was at about this !oint in the 8ontracrostipunctus that the Tortoise related the 7rab1s invention of -;ecord Player Eme a-& ?owever, readers were left dan lin as to the fate of that device, since before com!letin his tale, the tuckered-out Tortoise decided that he had best o home to slee! 4but not before tossin off a sly reference to 6>del1s Incom!leteness Theorem5& 8ow, at last, we can et around to clearin u! that dan lin detail&&& Perha!s you already have an inklin , after readin the Birthda) 8antatatata&

Essential Incom!leteness
)s you !robably sus!ected, even this fantastic advance over T8T suffers the same fate& )nd what makes it quite weird is that it is still for, in essence, the same reason& The a(iom schema is not !owerful enou h, and the 6>del construction can a ain be effected& Aet me s!ell this out a little& 4Ene can do it much more ri orously than I shall here&5 If there is a way of ca!turin the various strin s 6, 61, 6-, 6-1&&& in a sin le t)pographical mold, then there is a way of describin their 6>del numbers in a sin le arithmetical mold& )nd this arithmetical !ortrayal of an infinite class of numbers can then be re!resented inside T8T^6& by some formula 0MEGA-A<I0MJaK whose inter!retation is, -a is the 6>del number of one of the a(ioms comin from 6\-& 3hen a is re!laced by any s!ecific numeral, the formula which results will be a theorem of T8T^6\& if and only if the numeral stands for the 6>del number of an a(iom comin from the schema& 3ith the aid of this new formula, it becomes !ossible to re!resent even such a com!licated notion as T8T^6\-!roof-!airs inside T8T^6\, CT;TIG\@-,$00+-,AI$Ja) a%K Fsin this formula, we can construct a new uncle, which we !roceed to )rithmoquine in the by now thorou hly familiar way, makin yet another undecidable strin , which will be called -T8T^6\^"-& )t this !oint, you mi ht well wonder, -3hy isn1t 6 \^" amon the a(ioms created by the a(iom schema 6\?- The answer is that 6\ was not clever enou h to foresee its own embeddability inside number theory& In the 8ontracrostipunctus, one of the essential ste!s in the Tortoise1s makin an -un!layable record- was to et a hold of a manufacturer1s blue!rint of the record !layer which he was out to destroy& This was necessary so that he could fi ure out to what kinds

of vibrations it was vulnerable, and then incor!orate into his record such rooves as would code for sounds which would induce those vibrations& It is a close analo ue to the 6>del trick, in which the system1s own !ro!erties are reflected inside the notion of !roof!airs, and then used a ainst it& )ny system, no matter how com!le( or tricky it is, can be 6>del-numbered, and then the notion of its !roof-!airs can be defined-and this is the !etard by which it is hoist& Ence a system is well-defined, or -bo(ed-, it becomes vulnerable& This !rinci!le is e(cellently illustrated by the 7antor dia onal trick, which finds an omitted real number for each well-defined list of reals between + and "& It is the act of ivin an e(!licit lista -bo(- of reals which causes the downfall& Aet us see how the 7antor trick can be re!eated over and over a ain& 7onsider what ha!!ens if, startin with some list A, you do the followin , 4la5 Take list A, and construct its dia onal number d& 4lb5 Throw d somewhere into list A, makin a new list A^d& 4Ba5 Take list A^d, and construct its dia onal number d'& 4Bb5 Throw d' somewhere into list A^d, makin a new list A^d^d'& & & 8ow this ste!-by-ste! !rocess may seem a doltish way to !atch u! A, for we could have made the entire list d, d', d", d-1,&&& at once, iven A ori inally& 0ut if you think that makin such a list will enable you to com!lete your list of reals, you are very wron & The !roblem comes at the moment you ask, -3here to incor!orate the list of dia onal numbers inside A?- 8o matter how diabolically clever a scheme you devise for ensconcin the d-numbers inside A, once you have done it, then the new list is still vulnerable& )s was said above, it is the act of ivin an e(!licit lista -bo(- of reals that causes the downfall& 8ow in the case of formal systems, it is the act of ivin an e(!licit reci!e for what su!!osedly characteri<es number-theoretical truth that causes the incom!leteness& This is the cru( of the !roblem with T8T^6\& Ence you insert all the 61s in a welldefined way into T8T, there is seen to be some other 6some unforeseen 6which you didn1t ca!ture in your a(iom schema& )nd in the case of the T7-battle inside the 8ontracrostipunctus, the instant a record !layer1s -architecture- is determined, the record !layer becomes ca!able of bein shaken to !ieces& So what is to be done? There is no end in si ht& It a!!ears that T8T, even when e(tended ad infinitum, cannot be made com!lete& T8T is therefore said to suffer from essential incompleteness because the incom!leteness here is !art and !arcel of T8T= it is an essential !art of the nature of T8T and cannot be eradicated in any way, whether sim!leminded or in enious& 3hat1s more, this !roblem will haunt any formal version of number theory, whether it is an e(tension of T8T, a modification of T8T, or an alternative to T8T& The fact of the matter is this, the !ossibility of constructin , in a iven system, an undecidable strin via 6>del1s self-reference method, de!ends on three basic conditions,

4"5 That the system should be rich enou h so that all desired statements about numbers, whether true or false, can be e(!ressed in it& 4'ailure on this count means that the system is from the very start too weak to be counted as a rival to T8T, because it can1t even e(!ress number-theoretical notions that T8T can& In the meta!hor of the 8ontracrostipunctus, it is as if one did not have a !hono ra!h but a refri erator or some other kind of ob.ect&5 4B5 That all eneral recursive relations should be represented by formulas in the system& 4'ailure on this count means the system fails to ca!ture in a theorem some eneral recursive truth, which can only be considered a !athetic bellyflo! if it is attem!tin to !roduce all of number theory1s truths& In the 8ontracrostipunctus meta!hor, this is like havin a record !layer, but one of low fidelity&5 4%5 That the a(ioms and ty!o ra!hical !atterns defined by its rules be reco ni<able by some terminatin decision !rocedure& 4'ailure on this count means that there is no method to distin uish valid derivations in the system from invalid ones thus that the -formal system- is not formal after all, and in fact is not even welldefined& In the 8ontracrostipunctus meta!hor, it is a !hono ra!h which is still on the drawin board, only !artially desi ned&5 Satisfaction of these three conditions uarantees that any consistent system will be incom!lete, because 6>del1s construction is a!!licable& The fascinatin thin is that any such system di s its own hole= the system1s own richness brin s about its own downfall& The downfall occurs essentially because the system is !owerful enou h to have self-referential sentences& In !hysics, the notion e(ists of a -critical mass- of a fissionable substance, such as uranium& ) solid lum! of the substance will .ust sit there, if its mass is less than critical& 0ut beyond the critical mass, such a lum! will under o a chain reaction, and blow u!& It seems that with formal systems there is an analo ous critical !oint& 0elow that !oint, a system is -harmless- and does not even a!!roach definin arithmetical truth formally= but beyond the critical !oint, the system suddenly attains the ca!acity for self-reference, and thereby dooms itself to incom!leteness& The threshold seems to be rou hly when a system attains the three !ro!erties listed above& Ence this ability for self-reference is attained, the system has a hole which is tailor-made for itself= the hole takes the features of the system into account and uses them a ainst the system&

The ,assion According to Lucas

The bafflin re!eatability of the 6>del ar ument has been used by various !eo!le notably J& ;& Aucasas ammunition in the battle to show that there is some elusive and ineffable quality to human intelli ence, which makes it unattainable by -mechanical

automata-that is, com!uters& Aucas be ins his article -/inds, /achines, and 6>delwith these words,
6>del1s theorem seems to me to !rove that /echanism is false, that is, that minds cannot be e(!lained as machines&"

Then he !roceeds to ive an ar ument which, !ara!hrased, runs like this& 'or a com!uter to be considered as intelli ent as a !erson is, it must be able to do every intellectual task which a !erson can do& 8ow Aucas claims that no com!uter can do -6>deli<ation- 4one of his amusin ly irreverent terms5 in the manner that !eo!le can& 3hy not? 3ell, think of any !articular formal system, such as T8T, or T8T^6, or even T8T^6\& Ene can write a com!uter !ro ram rather easily which will systematically enerate theorems of that system, and in such a manner that eventually, any !reselected theorem will be !rinted out& That is, the theorem- eneratin !ro ram won1t ski! any !ortion of the -s!ace- of all theorems& Such a !ro ram would be com!osed of two ma.or !arts, 4"5 a subroutine which stam!s out a(ioms, iven the -molds- of the a(iom schemas 4if there are any5, and 4B5 a subroutine which takes known theorems 4includin a(ioms, of course5 and a!!lies rules of inference to !roduce new theorems& The !ro ram would alternate between runnin first one of these subroutines, and then the other& 3e can anthro!omor!hically say that this !ro ram -knows- some facts of number theorynamely, it knows those facts which it !rints out& If it fails to !rint out some true fact of number theory, then of course it doesn1t -know- that fact& Therefore, a com!uter !ro ram will be inferior to human bein s if it can be shown that humans know somethin which the !ro ram cannot know& 8ow here is where Aucas starts rollin & ?e says that we humans can always do the 6>del trick on any formal system as !owerful as T8Tand hence no matter what the formal system, we know more than it does& 8ow this may only sound like an ar ument about formal systems, but it can also be sli htly modified so that it becomes, seemin ly, an invincible ar ument a ainst the !ossibility of )rtificial Intelli ence ever re!roducin the human level of intelli ence& ?ere is the ist of it, ;i id internal codes entirely rule com!uters and robots= er o&&& 7om!uters are isomor!hic to formal systems& 8ow&&& )ny com!uter which wants to be as smart as we are has ot to be able to do number theory as well as we can, so&&& )mon other thin s, it has to be able to do !rimitive recursive arithmetic& 0ut for this very reason&&& It is vulnerable to the 6>delian -hook-, which im!lies that&&& 3e, with our human intelli ence, can concoct a certain statement of number theory which is true, but the computer is blind to that statement1s truth 4i&e&, will never !rint it out5, !recisely because of 6>del1s boomeran in ar ument& This im!lies that there is one thin which com!uters .ust cannot be !ro rammed to do, but which we can do& So we are smarter& Aet us en.oy, with Aucas, a transient moment of anthro!ocentric lory,

?owever com!licated a machine we construct, it will, if it is a machine, corres!ond to a formal system, which in turn will be liable to the 6>del !rocedure for findin a formula un!rovable-inthat-system& This formula the machine will be unable to !roduce as bein true, althou h a mind can see it is true& )nd so the machine will still not be an adequate model of the mind& 3e are tryin to !roduce a model of the mind which is mechanical-which is essentially -dead-but the mind, bein in fact -alive,- can always o one better than any formal, ossified, dead system can& Thanks to 6>del1s theorem& the mind always has the last word&B

En first si ht, and !erha!s even on careful analysis, Aucas1 ar ument a!!ears com!ellin & It usually evokes rather !olari<ed reactions& Some sei<e onto it as a nearly reli ious !roof of the e(istence of souls, while others lau h it off as bein unworthy of comment& I feel it is wron , but fascinatin ly soand therefore quite worthwhile takin the time to rebut& In fact, it was one of the ma.or early forces drivin me to think over the matters in this book& I shall try to rebut it in one way in this 7ha!ter, and in other ways in 7ha!ter TGII& 3e must try to understand more dee!ly why Aucas says the com!uter cannot be !ro rammed to -know- as much as we do& 0asically the idea is that we are always outside the system, and from out there we can always !erform the -6>deli<in o!eration, which yields somethin which the !ro ram, from within, can1t see is true& 0ut why can1t the -6>deli<in o!erator-, as Aucas calls it, be !ro rammed and added to the !ro ram as a third ma.or com!onent, Aucas e(!lains,
The !rocedure whereby the 6>delian formula is constructed is a standard !rocedureonly so could we be sure that a 6>delian formula can be constructed for every formal system& 0ut if it is a standard !rocedure, then a machine should be able to be !ro rammed to carry it out too&&& This would corres!ond to havin a system with an additional rule of inference which allowed one to add, as a theorem, the 6>delian formula of the rest of the formal system, and then the 6>delian formula of this new, stren thened, formal system, and so on& It would be tantamount to addin to the ori inal formal system an infinite sequence of a(ioms, each the 6>delian formula of the system hitherto obtained&&& 3e mi ht e(!ect a mind, faced with a machine that !ossessed a 6>deli<in o!erator, to take this into account, and out-6>del the new machine, 6>deli<in o!erator and all& This has, in fact, !roved to be the case& Even if we ad.oin to a formal system the infinite set of a(ioms consistin of the successive 6>delian formulae, the resultin system is still incom!lete, and contains a formula which cannot be !roved-in-thesystem, althou h a rational bein can, standin outside the system, see that it is true& 3e had e(!ected this, for even if an infinite set of a(ioms were added, they would have to be s!ecified by some finite rule or s!ecification, and this further rule or s!ecification could then be taken into account by a mind considerin the enlar ed formal system& In a sense, .ust because the mind has the last word, it can always !ick a hole in any formal system !resented to it as a model of its own workin s& The mechanical model must be, in some sense, finite and definite, and then the mind can always o one better&%

3um!ing &! a

imension

) visual ima e !rovided by /& 7& Escher is e(tremely useful in aidin the intuition here, his drawin .ragon 4'i & $D5& Its most salient feature is, of course, its sub.ect mattera dra on bitin its tail, with all the 6>delian connotations which that carries& 0ut there is a

dee!er theme to this !icture& Escher himself wrote the followin most interestin comments& The first comment is about a set of his drawin s all of which are concerned with -the conflict between the flat and the s!atial-= the second comment is about :ra on in !articular&
I& Eur three-dimensional s!ace is the only true reality we know& The two-dimensional is every bit as fictitious as the four-dimensional, for nothin is flat, not even the most finely !olished mirror& )nd yet we stick to the convention that a wall or a !iece of !a!er is flat, and curiously enou h, we still o on, as we have done since time immemorial, !roducin illusions of s!ace on .ust such !lane surfaces as these& Surely it is a bit absurd to draw a few lines and then claim, -This is a house-& This odd situation is the theme of the ne(t five !ictures 4includin :ra on5& C II& ?owever much this dra on tries to be s!atial, he remains com!letely flat& Two incisions are made in the !a!er on which he is !rinted& Then it is folded in such a way as to leave two square o!enin s& 0ut this dra on is an obstinate beast, and in s!ite of his two dimensions he !ersists in assumin that he has three= so he sticks his head throu h one of the holes and his tail throu h the others&L

This second remark es!ecially is a very tellin remark& The messa e is that no matter how cleverly you try to simulate three dimensions in two, you are always missin some -essence of three-dimensionality-& The dra on tries very hard to fi ht his twodimensionality& ?e defies the two-dimensionality of the !a!er on which he thinks he is drawn, by stickin his head throu h it= and yet all the while, we outside the drawin can see the !athetic futility of it all, for the dra on and the holes and the folds are all merely two-dimensional simulations of those conce!ts, and not a one of them is real& 0ut the dra on cannot ste! out of his two-dimensional s!ace, and cannot know it as we do& 3e could, in fact, carry the Escher !icture any number of ste!s further& 'or instance, we could tear it out of the book, fold it, cut holes in it, !ass it throu h itself, and !hoto ra!h the whole mess, so that it a ain becomes two-dimensional& )nd to that !hoto ra!h, we could once a ain do the same trick& Each time, at the instant that it becomes twodimensionalno matter how cleverly we seem to have simulated three dimensions inside twoit becomes vulnerable to bein cut and folded a ain& 8ow with this wonderful Escherian meta!hor, let us return to the !ro ram versus the human& 3e were talkin about tryin to enca!sulate the -6>deli<in o!erator- inside the !ro ram itself& 3ell, even if we had written a !ro ram which carried the o!eration out, that !ro ram would not ca!ture the essence of 6>del1s method& 'or once a ain, we, outside the system, could still -<a!- it in a way which it couldn1t do& 0ut then are we ar uin with, or a ainst, Aucas?

The Limits of Intelligent Systems

) ainst& 'or the very fact that we cannot write a !ro ram to do -6>deli<in - must make us somewhat sus!icious that we ourselves could do it in every case& It is one thin to make the ar ument in the abstract that 6>deli<in -can be done-= it is another thin to know how to do it in every !articular case& In fact, as the formal systems 4or !ro rams5 escalate in com!le(ity, our own ability to -6>deli<e- will eventually be in to waver& It must, since, as we have said above, we do not have any al orithmic way of describin

73G5=E CJ! :ra on, () '! 8! Escher wood%engra#ing, "#LB5&

how to !erform it& If we can1t tell e(!licitly what is involved in a!!lyin the 6>del method in all cases, then for each of us there will eventually come some case so com!licated that we sim!ly can1t fi ure out how to a!!ly it& Ef course, this borderline of one1s abilities will be somewhat ill-defined, .ust as is the borderline of wei hts which one can !ick u! off the round& 3hile on some days you may not be able to !ick u! a BL+-!ound ob.ect, on other days maybe you can& 8evertheless, there are no days whatsoever on which you can !ick u! a BL+-ton ob.ect& )nd in this sense, thou h everyone1s 6>deli<ation threshold is va ue, for each !erson, there are systems which lie far beyond his ability to 6>deli<e& This notion is illustrated in the 0irthday 7antatatata& )t first, it seems obvious that the Tortoise can !roceed as far as he wishes in !esterin )chilles& 0ut then )chilles tries

to sum u! all the answers in a sin le swoo!& This is a move of a different character than any that has one before, and is iven the new name 1 \1& The newness of the name is quite im!ortant& It is the first e(am!le where the old namin schemewhich only included names for all the natural numbershad to be transcended& Then come some more e(tensions, some of whose names seem quite obvious, others of which are rather tricky& 0ut eventually, we run out of names once a ainat the !oint where the answerschemas \, \\ , \\
\!!!

are all subsumed into one outra eously com!le( answer schema& The alto ether new name 1}+1 is su!!lied for this one& )nd the reason a new name is needed is that some fundamentally new kind of ste! has been takena sort of irre ularity has been encountered& Thus a new name must be a!!lied ad hoc&

There Is ;o $ecursi"e $ule for ;aming 0rdinals1

8ow offhand you mi ht think that these irre ularities in the !ro ression from ordinal to ordinal 4as these names of infinity are called5 could be handled by a com!uter !ro ram& That is, there would be a !ro ram to !roduce new names in a re ular way, and when it ran out of as, it would invoke the -irre ularity handler-, which would su!!ly a new name, and !ass control back to the sim!le one& 0ut this will not work& It turns out that irre ularities themselves ha!!en in irre ular ways, and one would need also a secondorder !ro ramthat is, a !ro ram which makes new !ro rams which make new names& )nd even this is not enou h& Eventually, a third-order !ro ram becomes necessary& )nd so on, and so on& )ll of this !erha!s ridiculous-seemin com!le(ity stems from a dee! theorem, due to )lon<o 7hurch and Ste!hen 7& @leene, about the structure of these -infinite ordinals-, which says, There is no recursively related notation-system which ives a name to every constructive ordinal& 3hat -recursively related notation-systems- are, and what -constructive ordinals- are, we must leave to the more technical sources, such as ?artley ;o ers1 book, to e(!lain& 0ut the intuitive idea has been !resented& )s the ordinals et bi er and bi er, there are irre ularities, and irre ularities in the irre ularities, and irre ularities in the irre ularities in the irre ularities, etc& 8o sin le scheme, no matter how com!le(, can name all the ordinals& )nd from this, it follows that no al orithmic method can tell how to a!!ly the method of 6>del to all !ossible kinds of formal systems& )nd unless one is rather mystically inclined, therefore one must conclude that any human bein sim!ly will reach the limits of his own ability to 6>deli<e at some !oint& 'rom there on out, formal systems of that com!le(ity, thou h admittedly incom!lete for the 6>del reason, will have as much !ower as that human bein &

0ther $efutations of Lucas

8ow this is only one way to ar ue a ainst Aucas1 !osition& There are others, !ossibly more !owerful, which we shall !resent later& 0ut this counterar ument has s!ecial interest because it brin s u! the fascinatin conce!t tryin to create a com!uter !ro ram which can et outside of itself, see itself com!letely from the outside, and a!!ly the 6>del <a!!in -trick to itself& Ef course this is .ust as im!ossible as for a record !layer to be able to !lay records which would cause it to break& 0utone should not consider T8T defective for that reason& If there a defect anywhere, it is not in T8T, but in our e(!ectations of what it should he able to do& 'urthermore, it is hel!ful to reali<e that we are equally vulnerable to the word trick which 6>del trans!lanted into mathematical formalisms, the E!imenides !arado(& This was quite cleverly !ointed out by 7& ?& 3hitely, when he !ro!osed the sentence -Aucas cannot consistently assert this sentence&- If you think about it, you will see that 4"5 it is true, and yet 4B5 Aucas cannot consistently assert it& So Aucas is also -incom!lete- with res!ect to truths about the world& The way in which he mirrors the world in his brain structures !revents him from simultaneously bein -consistent- and assertin that true sentence& 0ut Aucas is no more vulnerable than any of us& ?e is .ust on a !ar with a so!histicated formal system& )n amusin way to see the incorrectness of Aucas1 ar ument is to translate it into a battle between men and women&&& In his wanderin s, Aoocus the Thinker one day comes across an unknown ob.ecta woman& Such a thin he has never seen before, and at first he is wondrous thrilled at her likeness to himself, but then, sli htly scared of her as well, he cries to all the men about him, -0ehold2 I can look u!on her face, which is somethin she cannot dotherefore women can never be like me2- )nd thus he !roves man1s su!eriority over women, much to his relief, and that of his male com!anions& Incidentally, the same ar ument !roves that Aoocus is su!erior to all other males, as well but he doesn1t !oint that out to them& The woman ar ues back, -Hes, you can see my face, which is somethin I can1t dobut I can see your face, which is somethin )ou can1t do2 3e1re even&- ?owever, Aoocus comes u! with an une(!ected counter, -I1m sorry, you1re deluded if you think you can see my face& 3hat you women do is not the same as what we men doit is, as I have already !ointed out, of an inferior caliber, and does not deserve to be called by the same name& Hou may call it 1womanseein 1& 8ow the fact that you can 1womansee1 my face is of no im!ort, because the situation is not symmetric& Hou see?- -I womansee,- womanre!lies the woman, and womanwalks away&&& 3ell, this is the kind of -heads-in-the-sand- ar ument which you have to be willin to stomach if you are bent on seein men and women runnin ahead of com!uters in these intellectual battles&

Self-Transcendence - A Modern Myth

It is still of reat interest to !onder whether we humans ever can .um! out of ourselves or whether com!uter !ro rams can .um! out of themselves& 7ertainly it is !ossible for a !ro ram to modify itselfbut such modifiability has to be inherent in the !ro ram to

start with, so that cannot be counted as an e(am!le of -.um!in out of the system-& 8o matter how a !ro ram twists and turns to et out of itself, it is still followin the rules inherent in itself& It is no more !ossible for it to esca!e than it is for a human bein to decide voluntarily not to obey the laws of !hysics& Physics is an overridin system, from which there can be no esca!e& ?owever, there is a lesser ambition which it is !ossible to achieve, that is, one can certainly .um! from a subsystem of one1s brain into a wider subsystem& Ene can ste! out of ruts on occasion& This is still due to the interaction of various subsystems of one1s brain, but it can feel very much like ste!!in entirely out of oneself& Similarly, it is entirely conceivable that a !artial ability to -ste! outside of itselfcould be embodied in a com!uter !ro ram& ?owever, it is im!ortant to see the distinction between !erceivin oneself, and transcending oneself& Hou can ain visions of yourself in all sorts of waysin a mirror, in !hotos or movies, on ta!e, throu h the descri!tions if others, by ettin !sychoanaly<ed, and so on& 0ut you cannot quite break out of your own skin and be on the outside of yourself 4modern occult movements, !o! !sycholo y fads, etc& notwithstandin 5& T8T can talk about itself, but it cannot .um! out of itself& ) com!uter !ro ram can modify itself but it cannot violate its own instructionsit can at best chan e some !arts of itself by o(e)ing its own instructions& This is reminiscent of the numerous !arado(ical question, -7an 6od make a stone so heavy that he can1t lift it?-

Ad"ertisement and +raming

e"ices

This drive to .um! out of the system is a !ervasive one, and lies behind all !ro ress in art, music, and other human endeavors& It also lies behind such trivial undertakin s as the makin of radio and television commercials& This insidious trend has been beautifully !erceived and described by Irvin 6offman in his book 7rame *nal)sis,
'or e(am!le, an obviously !rofessional actor com!letes a commercial !itch and, with the camera still on him, turns in obvious relief from his task, now to take real !leasure in consumin the !roduct he had been advertisin & This is, of course, but one e(am!le of the way in which TG and radio commercials are comin to e(!loit framin devices to ive an a!!earance of naturalness that 4it is ho!ed5 will override the reserve auditors have develo!ed& Thus, use is currently bein made of children1s voices, !resumably because these seem unschooled= street noises, and other effects to ive the im!ression of interviews with un!aid res!ondents= false starts, filled !auses, by!lays, and overla!!in s!eech to simulate actual conversation= and, followin 3elles, the interce!tion of a firm1s .in le commercials to ive news of its new !roduct, alternatin occasionally with interce!tion by a !ublic interest s!ot, this !resumably kee!in the faith of the auditor alive& The more that auditors withdraw to minor e(!ressive details as a test of enuineness, the more that advertisers chase after them& 3hat results is a sort of interaction !ollution, a disorder that is also s!read by the !ublic relations consultants of !olitical fi ures, and, more modestly, by micro-sociolo y&D

?ere we have yet another e(am!le of an escalatin -T7-battle-the anta onists this time bein Truth and 7ommercials&

Sim!licio) Sal"iati) Sagredo5 >hy Three7

There is a fascinatin connection between the !roblem of .um!in out of the system and the quest for com!lete ob.ectivity& 3hen I read Jauch1s four dialo ues in *re 4uanta =eal$ based on 6alileo1s four .ialogues 8oncerning Two New -ciences, I found myself wonderin why there were three characters !artici!atin & Sim!licio, Salviati and Sa redo& 3hy wouldn1t two have sufficed, Sim!licio, the educated sim!leton, and Salviati, the knowled eable thinker? 3hat function does Sa redo have? 3ell, he is su!!osed to be a sort of neutral third !arty, dis!assionately wei hin the two sides and comin out with a -fair- and -im!artial- .ud ment& It sounds very balanced, and yet there is a !roblem, Sa redo is always a reein with Salviati, not with Sim!licio& ?ow come Eb.ectivity Personified is !layin favorites? Ene answer, of course, is that Salviati is enunciatin correct views, so Sa redo has no choice& 0ut what, then, of fairness or -equal time-? 0y addin Sa redo, 6alileo 4and Jauch5 stacked the deck more a ainst Sim!licio, rather than less& Perha!s there should be added a yet hi her level Sa redosomeone who will be ob.ective about this whole situation&&& Hou can see where it is oin & 3e are ettin into a never-endin series of -escalations in ob.ectivity-, which have the curious !ro!erty of never ettin any more ob.ective than at the first level, where Salviati is sim!ly ri ht, and Sim!licio wron & So the !u<<le remains, why add Sa redo at all? )nd the answer is, it ives the illusion of ste!!in out of the system, in some intuitively a!!ealin sense&

2en and (Ste!!ing 0ut(

In 9en, too, we can see this !reoccu!ation with the conce!t of transcendin the system& 'or instance, the kRan in which TR<an tells his monks that -the hi her 0uddhism is not 0uddha-& Perha!s, self-transcendence is even the central theme of 9en& ) 9en !erson is always tryin to understand more dee!ly what he is, by ste!!in more and more out of what he sees himself to be, by breakin every rule and convention which he !erceives himself to be chained byneedless to say, includin those of 9en itself& Somewhere alon this elusive !ath may come enli htenment& In any case 4as I see it5, the ho!e is that by radually dee!enin one1s self-awareness, by radually widenin the sco!e of -the system-, one will in the end come to a feelin of bein at one with the entire universe&

'difyin! Thou!hts of a Tobacco Smo/er

*chilles has (een in#ited to the 8ra('s home! *chilles, I see you have made a few additions since I was last here, /r& 7rab& Hour new !aintin s are es!ecially strikin & 8ra(, Thank you& I am quite fond of certain !ainterses!ecially ;ene /a ritte& /ost of the !aintin s I have are by him& ?e1s my favorite artist& *chilles, They are very intri uin ima es, I must say& In some ways, these !aintin s by /a ritte remind me of works by /H favorite artist, /& 7& Escher& 8ra(, I can see that& 0oth /a ritte and Escher use reat realism in e(!lorin the worlds of !arado( and illusion= both have a sure sense for the evocative !ower of certain visual symbols, andsomethin which even their admirers often fail to !oint out both of them have a sense of the raceful line& *chilles, 8evertheless, there is somethin quite different about them& I wonder how one could characteri<e that difference& 8ra(, It would be fascinatin to com!are the two in detail& *chilles, I must say, /a ritte1s command of realism is astonishin & 'or instance, I was quite taken in by that !aintin over there of a tree with a iant !i!e behind it&
73G5=E CC! The Shadows, () =ene 'agritte BIJJ/!

8ra(, Hou mean a normal !i!e with a tiny tree in front of it2 *chilles, Eh, is that what it is? 3ell, in any case, when I first s!otted it, I was convinced I was smellin !i!e smoke2 7an you ima ine how silly I felt?

8ra(, I quite understand& /y uests are often taken in by that one& -o sa)ing, he reaches up, remo#es the pipe from (ehind the tree in the painting, turns o#er and taps it against the ta(le, and the room (egins to reek of pipe to(acco! ,e (egins packing in a new wad of to(acco!/ This is a fine old !i!e, )chilles& 0elieve it or not, the bowl has a co!!er linin , which makes it a e wonderfully& *chilles, ) co!!er linin 2 Hou don1t say2 8ra( pulls out a (o9 of matches, and lights his pipe/, 3ould you care for a smoke, )chilles? *chilles, 8o, thank you& I only smoke ci ars now and then& 8ra(, 8o !roblem2 I have one ri ht here2 =eaches out towards another 'agritte painting, featuring a (ic)cle mounted upon a lit cigar!/ *chilles, Fhhno thank you, not now& 8ra(, )s you will& I myself am an incurable tobacco smoker& 3hich reminds meyou undoubtedly know of Eld 0ach1s !redilection for !i!e smokin ? *chilles, I don1t recall e(actly& 8ra(, Eld 0ach was fond of versifyin , !hiloso!hi<in , !i!e smokin , and music makin
73G5=E CE! State of 6race, () =ene 'agritte BIGI/!

4not necessarily in that order5& ?e combined all four into a droll !oem which he set to music& It can be found in the famous musical notebook he ke!t for his wife, )nna /a dalena, and it is called Edif)ing Thoughts of a To(acco -moker" 3hene1er I take my !i!e and stuff it )nd smoke to !ass the time away, /y thou hts, as I sit there and !uff it, :well on a !icture sad and ray, It teaches me that very like )m I myself unto my !i!e& Aike me, this !i!e so fra rant burnin Is made of nau ht but earth and clay= To earth I too shall be returnin & It falls and, ere I1d think to say, It breaks in two before my eyes= In store for me a like fate lies& 8o stain the !i!e1s hue yet doth darken= It remains white& Thus do I know That when to death1s call I must harken /y body, too, all !ale will row& To black beneath the sod 1twill turn, Aikewise the !i!e, if oft it burn& Er when the !i!e is fairly lowin , 0ehold then, instantaneously, The smoke off into thin air oin , Till nau ht but ash is left to see& /an1s fame likewise away will burn )nd unto dust his body turn& ?ow oft it ha!!ens when one1s smokin , The sto!!er1s missin from its shelf, )nd one oes with one1s fin er !okin Into the bowl and burns oneself& If in the !i!e such !ain doth dwell, ?ow hot must be the !ains of hell& Thus o1er my !i!e, in contem!lation Ef such thin s, I can constantly Indul e in fruitful meditation, )nd so, !uffin contentedly,

En land, on sea, at home, abroad I smoke my !i!e and worshi! 6od& ) charmin !hiloso!hy, is it not? *chilles, Indeed& Eld 0ach was a turner of !hrases quite !leasin1& 8ra(, Hou took the very words from my mouth& Hou know, in my time I have tried to write clever verses& 0ut I fear mine don1t measure u! to much& I don1t have such a way with words& *chilles, Eh, come now, /r& 7rab& Hou havehow to !ut it?quite a !enchant for trick1ry and teasin1& I1d be honored if you1d sin me one of your son s, /r& 7& 8ra(, I1m most flattered& ?ow about if I !lay you a record of myself sin in one of my efforts? I don1t remember when it dates from& Its title is -) Son 3ithout Time or Season-& *chilles, ?ow !oetic2 The 8ra( pulls a record from his shel#es, and walks o#er to a huge, comple9 piece of apparatus! ,e opens it up, and inserts the record into an ominous%looking mechanical mouth! -uddenl) a (right flash of greenish light sweeps o#er the surface of the record, and after a moment, the record is silentl) whisked into some hidden (ell) of the fantastic machine! * moment passes, and then the strains of the 8ra('s #oice ring out!/ ) turner of !hrases quite !leasin1, ?ad a !enchant for trick1ry and teasin1& In his son s, the last line /i ht seem sans desi n= 3hat I mean is, without why or wherefore& *chilles, Aovely2 Enly, I1m !u<<led by one thin & It seems to me your son , the last line is 8ra(, Sans desi n? *chilles, 8o&&& 3hat I mean is, without rhyme or reason& 7rab, Hou could be ri ht& *chilles, Ether than that, it1s a very nice son , but I must say I am even more intri ued by this monstrously com!le( contra!tion& Is it merely an oversi<ed record !layer? 8ra(, Eh, no, it1s much more than that& This is my Tortoise-chom!in record !layer& *chilles, 6ood rief2 8ra(, 3ell, I don1t mean that it chom!s u! Tortoises& 0ut it chom!s u! records !roduced by /r& Tortoise& *chilles, 3hew2 That1s a little milder& Is this !art of that weird musical battle that evolved between you and /r& T some time a o? 8ra(, In a way& Aet me e(!lain a little more fully& Hou see, /r& Tortoise1s so!histication had reached the !oint where he seemed to be able to destroy almost any record !layer I would obtain&

*chilles, 0ut when I heard about your rivalry, it seemed to me you had at last come into !ossession of an invincible !hono ra!hone with a built-in TG camera, minicom!uter and so on, which could take itself a!art and rebuild itself in such a way that it would not be destroyed& 8ra(, )lack and alas2 /y !lan was foiled& 'or /r& Tortoise took advanta e of one small detail which I had overlooked, the subunit which directed the disassembly and reassembly !rocesses was itself stable durin the entire !rocess& That is, for obvious reasons, it could not take itself a!art and rebuild itself, so it stayed intact& *chilles, Hes, but what consequences did that have& 8ra(, Eh, the direst ones2 'or you see, /r& T focused his method down onto that subunit entirely& *chilles, ?ow is that? 8ra(, ?e sim!ly made a record which would induce fatal vibrations in the one structure he knew would never chan ethe disassembly-reassembly subunit& *chilles, Eh, I see&&& Gery sneaky& 8ra(, Hes, so I thou ht, too& )nd his strate y worked& 8ot the first time, mind you& I thou ht I had outwitted him when my !hono ra!h survived his first onslau ht& I lau hed leefully& 0ut the ne(t time, he returned with a steely lint in his eye, and I knew he meant business& I !laced his new record on my turntable& Then, both of us ea erly watched the com!uter-directed subunit carefully scan the rooves& then dismount the record, disassemble the record !layer, reassemble it in an astonishin ly different way, remount the recordand then slowly lower the needle into the outermost roove& *chilles, 6olly2 8ra(, 8o sooner had the first strains of sound issued forth than a loud S/)S?2 filled the room& The whole thin fell a!art, but !articularly badly destroyed was the assemblerdisassembler& In that !ainful instant I finally reali<ed, to my cha rin, that the Tortoise would )A3)HS be able to focus down u!onif you1ll !ardon the !hrasethe )chilles1 heel of the system& *chilles, F!on my soul2 Hou must have felt devastated& 8ra(, Hes, I felt rather forlorn for a while& 0ut, ha!!ily, that was not the end of the story& There is a sequel to the tale, which tau ht me a valuable lesson, which I may !ass on to you& En the Tortoise1s recommendation, I was browsin throu h a curious book filled with stran e :ialo ues about many sub.ects, includin molecular biolo y, fu ues, 9en 0uddhism, and heaven knows what else& *chilles, Probably some crack!ot wrote it& 3hat is the book called? 8ra(, If I recall correctly, it was called 8opper, -il#er, Gold: an 3ndestructi(le 'etallic *llo)& *chilles, Eh, /r& Tortoise told me about it, too& It1s by a friend of his, who, it a!!ears, is quite taken with metal-lo ic& 8ra(, I wonder which friend it is&&& )nyway, in one of the :ialo ues, I encountered some Edifyin Thou hts on the Tobacco /osaic Girus, ribosomes, and other stran e thin s I had never heard of&

73G5=E CI! To(acco 'osaic <irus! U7rom *! +ehninger, 0iochemistry New York: >orth &u(lishers, BICJ/!V

~]]]
+&"m

*chilles, 3hat is the Tobacco /osaic Girus? 3hat are ribosomes? 8ra(, I can1t quite say, for I1m a total dunce when it comes to biolo y& )ll I know is what I athered from that :ialo ue& There, it said that Tobacco /osaic Giruses are tiny ci arette-like ob.ects that cause a disease in tobacco !lants& *chilles, 7ancer? 8ra(, 8o, not e(actly, but *chilles, 3hat ne(t? ) tobacco !lant smokin and ettin cancer2 Serves it ri ht2 8ra(, I believe you1ve .um!ed to a hasty conclusion, )chilles& Tobacco !lants don1t S/E@E these -ci arettes-& The nasty little -ci arettes- .ust come and attack them, uninvited& *chilles, I see& 3ell, now that I know all about Tobacco /osaic Giruses, tell me what a ribosome is& 8ra(, ;ibosomes are a!!arently some sort of subcellular entities which take a messa e in one form and convert it into a messa e in another form& *chilles, Somethin like a teeny ta!e recorder or !hono ra!h? 8ra(, /eta!horically, I su!!ose so& 8ow the thin which cau ht my eye was a line where this one e(ceedin ly droll character mentions the fact that ribosomesas well as Tobacco /osaic Giruses and certain other bi<arre biolo ical structures!ossess -the bafflin ability to s!ontaneously self-assemble&- Those were his e(act words& *chilles, That was one of his droller lines, I take it&

8ra(, That1s .ust what the other character in the :ialo ue thou ht& 0ut that1s a !re!osterous inter!retation of the statement& 4 The 8ra( draws deepl) from his pipe, and puffs se#eral (illows of smoke into the air!/ *chilles, 3ell, what does -s!ontaneous self-assembly- mean, then? 8ra(, The idea is that when some biolo ical units inside a cell are taken a!art, they can s!ontaneously reassemble themselveswithout bein directed by any other unit& The !ieces .ust come to ether, and !resto2they stick& *chilles, That sounds like ma ic& 3ouldn1t it be wonderful if a full-si<ed record !layer could have that !ro!erty? I mean, if a miniature -record !layer- such as a ribosome can do it, why not a bi one? That would allow you to create an indestructible !hono ra!h, ri ht? )ny time it was broken, it would .ust !ut itself to ether a ain& 8ra(, E(actly my thou ht& I breathlessly rushed a letter off to my manufacturer e(!lainin the conce!t of self-assembly, and asked him if he could build me a record !layer which could take itself a!art and s!ontaneously self-assemble in another form& *chilles, ) hefty bill to fill& 8ra(, True= but after several months, he wrote to me that he had succeeded, at lon lastand indeed he sent me quite a hefty bill& Ene fine day, ho2 /y 6rand Self-assemblin ;ecord Player arrived in the mail, and it was with reat confidence that I tele!honed /r& Tortoise, and invited him over for the !ur!ose of testin my ultimate record !layer& *chilles, So this ma nificent ob.ect before us must be the very machine of which you s!eak& 8ra(, I1m afraid not, )chilles& *chilles, :on1t tell me that once a ain&&& 8ra(, 3hat you sus!ect, my dear friend, is unfortunately the case& I don1t !retend to understand the reasons why& The whole thin is too !ainful to recount& To see all those s!rin s and wires chaotically strewn about on the floor, and !uffs of smoke here and thereoh, me&&& *chilles, There, there, /r& 7rab, don1t take it too badly& 8ra(, I1m quite all ri ht= I .ust have these s!ells every so often& 3ell, to o on, after /r& Tortoise1s initial loatin , he at last reali<ed how sorrowful I was feelin , and took !ity& ?e tried to comfort me by e(!lainin that it couldn1t be hel!edit all had to do with somebody-or-other1s -Theorem-, but I couldn1t follow a word of it& It sounded like -Turtle1s Theorem-& *chilles, I wonder if it was that -6>del1s Theorem- which he s!oke of once before to me&&& It has a rather sinister rin to it& 8ra(, It could be& I don1t recall& *chilles, I can assure you, /r& 7rab, that I have followed this tale with the utmost em!athy for your !osition& It is truly sad& 0ut, you mentioned that there was a silver linin & Pray tell, what was that? 8ra(, Eh, yesthe silver linin & 3ell eventually, I abandoned my quest after -Perfection- in !hono ra!hs, and decided that I mi ht do better to ti hten u! my defenses a ainst the Tortoise1s records& I concluded that a more modest aim than a record !layer which can !lay anythin is sim!ly a record !layer that can SF;GIGE, one

that will avoid ettin destroyedeven if that means that it can only !lay a few !articular records& *chilles, So you decided you would develo! so!histicated anti-Tortoise mechanisms at the sacrifice of bein able to re!roduce every !ossible sound, eh? 8ra(, 3ell&&& I wouldn1t e(actly say I -decided- it& /ore accurate would be to say that I was 'E;7E: into that !osition& *chilles, Hes, I can see what you mean& 8ra(, /y new idea was to !revent all -alien- records from bein !layed on my !hono ra!h& I knew my own records are harmless, and so if I !revented anyone else from infiltratin T?EI; records, that would !rotect my record !layer, and still allow me to en.oy my recorded music& *chilles, )n e(cellent strate y for your new oal& 8ow does this iant thin before us re!resent your accom!lishments to date alon those lines? 8ra(, That it does& /r& Tortoise, of course, has reali<ed that he must chan e ?IS strate y, as well& ?is main oal is now to devise a record which can sli! !ast my censorsa new ty!e of challen e& *chilles, 'or your !art, how are you !lannin to kee! his and other -alien- records out? 8ra(, Hou !romise you won1t reveal my strate y to /r& T, now? *chilles, Tortoise1s honor& 8ra(, 3hat2? *chilles, Ehit1s .ust a !hrase I1ve !icked u! from /r& T& :on1t worryI swear your secret will remain secret with me& 8ra(, )ll ri ht, then& /y basic !lan is to use a A)0EAI86 technique& To each and every one of my records will be attached a secret label& 8ow the !hono ra!h before you contains, as did its !redecessors, a television camera for scannin the records, and a com!uter for !rocessin the data obtained in the scan and controllin subsequent o!erations& /y idea is sim!ly to chom! all records which do not bear the !ro!er label2 *chilles, )h, sweet reven e2 0ut it seems to me that your !lan will be easy to foil& )ll /r& T needs to do is to et a hold of one of your records, and co!y its label2 8ra(, 8ot so sim!le, )chilles& 3hat makes you think he will be able to tell the label from the rest of the record? It may be better inte rated than you sus!ect& *chilles, :o you mean that it could be mi(ed u! somehow with the actual music? 8ra(, Precisely& 0ut there is a way to disentan le the two& It requires suckin the data off the record visually and then *chilles, Is that what that bri ht reen ?ash was for? 8ra(, That1s ri ht& That was the TG camera scannin the rooves& The roove-!atterns were sent to the minicom!uter, which analy<ed the musical style of the !iece I had !ut onall in silence& 8othin had been !layed yet& *chilles, Then is there a screenin !rocess, which eliminates !ieces which aren1t in the !ro!er styles? 8ra(, Hou1ve ot it, )chilles& The only records which can !ass this second test are records of !ieces in my own styleand it will be ho!elessly difficult for /r& T to imitate that& So you see, I am convinced I will win this new musical battle& ?owever, I should

mention that /r& T is equally convinced that somehow, he will mana e to sli! a record !ast my censors& *chilles, )nd smash your marvelous machine to smithereens? 8ra(, Eh, nohe has !roved his !oint on that& 8ow he .ust wants to !rove to me that he can sli! a recordan innocuous oneby me, no matter what measures I take to !revent it& ?e kee!s on mutterin thin s about son s with stran e titles, such as -I 7an 0e Played on ;ecord Player T-& 0ut he can1t scare /E2 The only thin that worries me a little is that, as before, he seems to have some murky ar uments which&&& which&&& ,e trails off into silence! Then, looking "uite pensi#e, he takes a few puffs on his pipe!/ *chilles, ?mm&&& I1d say /r& Tortoise has an im!ossible task on his hands& ?e1s met his match, at lon last2 8ra(, 7urious that you should think so&&& I don1t su!!ose that you know ?enkin1s Theorem forwards and backwards, do you? *chilles, @now 3?ESE Theorem forwards and backwards? I1ve never heard of anythin that sounds like that& I1m sure it1s fascinatin , but I1d rather hear more about -music to infiltrate !hono ra!hs by-& It1s an amusin little story& )ctually, I uess I can fill in the end& Ebviously, /r& T will find out that there is no !oint in oin on, and so he will shee!ishly admit defeat, and that will be that& Isn1t that e(actly it? 8ra(, That1s what I1m ho!in , at least& 3ould you like to see a little bit of the inner workin s of my defensive !hono ra!h? *chilles, 6ladly& I1ve always wanted to see a workin television camera& 8ra(, 8o sooner said than done, my friend& =eaches into the gaping "mouth" of the large phonograph, undoes a couple of snaps, and pulls out a neatl) packaged instrument!/ Hou see, the whole thin is built of inde!endent modules, which can be detached and used inde!endently& This TG camera, for instance, works very well by itself& 3atch the screen over there, beneath the !aintin with the flamin tuba& ,e points the camera at *chilles, whose face instantl) appears on the large screen!/ *chilles, Terrific2 /ay I try it out? 8ra(, 7ertainly& *chilles, pointing the camera at the 8ra(!/ There HEF are, /r& 7rab, on the screen& 8ra(, So I am& *chilles, Su!!ose I !oint the camera at the !aintin with the burnin tuba& 8ow it is on the screen, too2 8ra(, The camera can <oom in and out, )chilles& Hou ou ht to try it& *chilles, 'abulous2 Aet me .ust focus down onto the ti! of those flames, where they meet the !icture frame&&& It1s such a funny feelin to be able to instantaneously -co!yanythin in the roomanythin I wantonto that screen& I merely need to !oint the camera at it, and it !o!s like ma ic onto the screen& 8ra(, )8HT?I86 in the room, )chilles? *chilles, )nythin in si ht, yes& That1s obvious& 8ra(, 3hat ha!!ens, then, if you !oint the camera at the flames on the TG screen? *chilles, ?ey, that1s funny2 That very act makes the flames :IS)PPE); from the screen2 3here did they o?

73G5=E EK! The 'air 7a!tive, () =ene 'agritte BIDC/!

*chilles shifts the camera so that it points directl) at that part of the tele#ision screen on which the flames areOor wereOdispla)ed!/ 8ra(, Hou can1t kee! an ima e still on the screen and move the camera at the same time& *chilles, So I see&&& 0ut I don1t understand what1s on the screen nownot at all2 It seems to be a stran e lon corridor& Het I1m certainly not !ointin the camera down any corridor& I1m merely !ointin it at an ordinary TG screen& 8ra(, Aook more carefully, )chilles& :o you really see a corridor? *chilles, )hhh, now I see& It1s a set of nested co!ies of the TG screen itself, ettin smaller and smaller and smaller&&& Ef course2 The ima e of the flames ?): to o away, because it came from my !ointin the camera at the P)I8TI86& 3hen I !oint the camera at the S7;EE8, then the screen itself a!!ears, with whatever is on the screen at the time which is the screen itself, with whatever is on the screen at the time which is the screen itself, with 8ra(, I believe I can fill in the rest, )chilles& 3hy don1t you try rotatin the camera? *chilles, Eh2 I et a beautiful s!iralin corridor2 Each screen is rotated inside its framin screen, so that the littler they et, the more rotated they are, with res!ect& to the outermost screen& This idea of havin a TG screen -en ulf itself- is weird& 8ra(, 3hat do you mean by -self-en ulfin -, )chilles? *chilles, I mean, when I !oint the camera at the screenor at !art of the screen& T ?)T1S self-en ulfin &

a/ The simplest case!

d/ * "failed self%engulfing"

(/ *chilles' "corridor"!

e/ >hat happens when )ou @oom in!

c/ >hat happens when )ou rotate the camera!

f/ 8om(ined effect of rotation and @ooming!

73G5=E EB! Twel#e self%engulfing T< screens! 3 would ha#e included one more, had BH not (een prime!

g/ -tarting to get weird!

M/ The late stages of a gala9)! 8ount the num(er of spokesA

h/ * "gala9)" is (orn!

k/ The gala9) has (urned out, and (ecomeO a (lack holeA

i/ The gala9) e#ol#es

l/ * "pulsating petal pattern", caught in the middle of one of its pulsations!

8ra(, :o you mind if I !ursue that a little further? I1m intri ued by this new notion& *chilles, So am I& 8ra(, Gery well, then& If you !oint the camera at a 7E;8E; of the screen, is that still what you mean by -self-en ulfin -? *chilles, Aet me try it& ?mmthe -corridor- of screens seems to o off the ed e, so there isn1t an infinite nestin any more& It1s !retty, but it doesn1t seem to me to have the s!irit of self-en ulfin & It1s a -failed self-en ulfin -& 8ra(, If you were to swin the TG camera back towards the center of the screen, maybe you could fi( it u! a ain&&& *chilles slowl) and cautiousl) turning the camera/, Hes2 The corridor is ettin lon er and lon er&&& There it is2 8ow it1s all back& I can look down it so far that it vanishes in the distance& The corridor became infinite a ain !recisely at the moment when the camera took in the 3?EAE screen& ?mmthat reminds me of somethin /r& Tortoise was sayin a while back, about self-reference only occurrin when a sentence talks about )AA of itself&&& 8ra(, Pardon me? *chilles, Eh, nothin .ust mutterin to myself& *s *chilles pla)s with the lens and other controls on the camera, a profusion of new kinds of self%engulfing images appear: swirling spirals that resem(le gala9ies, kaleidoscopic flower%like shapes, and other assorted patterns!!!/ 8ra(, Hou seem to be havin a rand time& *chilles turns awa) from the camera/, I1ll say2 3hat a wealth of ima es this sim!le idea can !roduce2 ,e glances (ack at the screen, and a look of astonishment crosses his face!/ 6ood rief, /r& 7rab2 There1s a !ulsatin !etal-!attern on the screen2 3here do the !ulsations come from? The TG is still, and so is the camera& 8ra(, Hou can occasionally set u! !atterns which chan e in time& This is because there is a sli ht delay in the circuitry between the moment the camera -sees- somethin , and the moment it a!!ears on the screen around a hundredth of a second& So if you have a nestin of de!th fifty or so, rou hly a half-second delay will result& If somehow a movin ima e ets onto the screenfor e(am!le, by you !uttin your fin er in front of the camerathen it takes a while for the more dee!ly nested screens to -find outabout it& This delay then reverberates throu h the whole system, like a visual echo& )nd if thin s are set u! so the echo doesn1t die away, then you can et !ulsatin !atterns& *chilles, )ma<in 2 Saywhat if we tried to make a TET)A self-en ulfin ? 8ra(, 3hat !recisely do you mean by that? *chilles, 3ell, it seems to me that this stuff with screens within screens is interestin , but I1d like to et a !icture of the TG camera )8: the screen, E8 the screen& Enly then would I really have made the system en ulf itself& 'or the screen is only P);T of the total system& 8ra(, I see what you mean& Perha!s with this mirror, you can achieve the effect you want&

The 8ra( hands him a mirror, and *chilles maneu#ers the mirror and camera in such a wa) that the camera and the screen are (oth pictured on the screen!/ *chilles, There2 I1ve created a TET)A self-en ulfin 2 8ra(, It seems to me you only have the front of the mirrorwhat about its back? If it weren1t for the back of the mirror, it wouldn1t be reflectiveand you wouldn1t have the camera in the !icture& *chilles, Hou1re ri ht& 0ut to show both the front and back of this mirror, I need a second mirror& 8ra(, 0ut then you1ll need to show the back of that mirror, too& )nd what about includin the back of the television, as well as its front? )nd then there1s the electric cord, and the inside of the television, and *chilles, 3hoa, whoa2 /y head1s be innin to s!in2 I can see that this -total selfen ulfin !ro.ect- is oin to !ose a wee bit of a !roblem& I1m feelin a little di<<y& 8ra(, I know e(actly how you feel& 3hy don1t you sit down here and take your mind off all this self-en ulfin ? ;ela(2 Aook at my !aintin s, and you1ll calm down& *chilles lies down, and sighs!/ Eh!erha!s my !i!e smoke is botherin you? ?ere, I1ll !ut my !i!e away& 4 Takes the pipe from his mouth, and carefull) places it a(o#e some written words in another 'agritte painting!5 There2 'eelin any better? *chilles, I1m still a little woo<y& &oints at the 'agritte!/ That1s an interestin !aintin & I like the way it1s framed, es!ecially the shiny inlay inside the wooden frame&

73G5=E EF! The )ir and the Son , () =en2 'agritte BIJD/!

8ra(, Thank you& I had it s!ecially doneit1s a old linin &

*chilles, ) old linin ? 3hat ne(t? 3hat are those words below the !i!e? They aren1t in En lish, are they? 8ra(, 8o, they are in 'rench& They say, -7eci n1est !as une !i!e&- That means, -This is not a !i!e-& 3hich is !erfectly true& *chilles, 0ut it is a !i!e2 Hou were .ust smokin it2 8ra(, Eh, you misunderstand the !hrase, I believe& The word -ceci- refers to the !aintin , not to the !i!e& Ef course the !i!e is a !i!e& 0ut a !aintin is not a !i!e& *chilles, I wonder if that -ceci- inside the !aintin refers to the 3?EAE !aintin , or .ust to the !i!e inside the !aintin & Eh, my racious2 That would be )8ET?E; self-en ulfin 2 I1m not feelin at all well, /r& 7rab& I think I1m oin to be sick&&&

C/A,TE$ <4I

Self-$ef and Self-$e!

I8 T?IS 7?)PTE;, we will look at some of the mechanisms which create self-reference in various conte(ts, and com!are them to the mechanisms which allow some kinds of systems to re!roduce themselves& Some remarkable and beautiful !arallels between these mechanisms will come to li ht&

Im!licitly and E-!licitly Self-$eferential Sentences

To be in with, let us look at sentences which, at first lance, may seem to !rovide the sim!lest e(am!les of self-reference& Some such sentences are these, 4"5 This sentence contains five words& 4B5 This sentence is meanin less because it is self-referential& 4%5 This sentence no verb& 4C5 This sentence is false& 4E!imenides !arado(5 4L5 The sentence I am now writin is the sentence you are now readin & )ll but the last one 4which is an anomaly5 involve the sim!le-seemin mechanism contained in the !hrase -this sentence-& 0ut that mechanism is in reality far from sim!le& )ll of these sentences are -floatin - in the conte(t of the En lish lan ua e& They can be com!ared to iceber s, whose ti!s only are visible& The word sequences are the ti!s of the iceber s, and the !rocessin which must be done to understand them is the hidden !art& In this sense their meanin is im!licit, not e(!licit& Ef course, no sentence1s meanin is com!letely e(!licit, but the more e(!licit the self-reference is, the more e(!osed will be the mechanisms underlyin it& In this case, for the self-reference of the sentences above to be reco ni<ed, not only has one to be comfortable with a lan ua e such as En lish which can deal with lin uistic sub.ect matter, but also one has to be able to fi ure out the referent of the !hrase -this sentence-& It seems sim!le, but it de!ends on our very com!le( yet totally assimilated ability to handle En lish& 3hat is es!ecially im!ortant here is the ability to fi ure out the referent of a noun !hrase with a demonstrative ad.ective in it& This ability is built u! slowly, and should by no means be considered trivial& The difficulty is !erha!s underlined when a sentence such as number C is !resented to someone naPve about !arado(es and lin uistic tricks, such as a child& They may say, -3hat sentence is false?- and it may take a bit of !ersistence to et across the idea that the sentence is talkin about itself& The whole idea is a little mind bo lin at first& ) cou!le of !ictures may hel! 4'i s& *%, *C5& 'i ure *% is a !icture which can be inter!reted on two levels& En one level, it is a sentence !ointin at itself= on the other level, it is a !icture of E!imenides e(ecutin his own death sentence&

73G5=E EH!

'i ure *C, showin visible and invisible !ortions of the iceber , su ests the relative !ro!ortion of sentence to !rocessin required for the reco nition of self-reference,

waterline

En lish lan ua e sea

73G5=E ED!

It is amusin to try to create a self-referrin sentence without usin the trick of sayin -this sentence-& Ene could try to quote a sentence inside itself& ?ere is an attem!t, The sentence -The sentence contains five words- contains five words& 0ut such an attem!t must fail, for any sentence that could be quoted entirely inside itself would have to be shorter than itself& This is actually !ossible, but only if you are willin to entertain infinitely lon sentences, such as,

The sentence
-The sentence
-The sentence
-The sentence
)
)
etc&, etc&

is infinitely lon -

is infinitely lon -

is infinitely lon -

is infinitely lon & 0ut this cannot work for finite sentences& 'or the same reason, 6>del1s strin 6 could not contain the e(!licit numeral for its 6>del number, it would not fit& 8o strin of T8T can contain the T8T-numeral for its own 6>del number, for that numeral always contains more symbols than the strin itself does& 0ut you can et around this by havin 6 contain a description of its own 6>del number, by means of the notions of -sub- and -arithmoquinification-& Ene way of achievin self-reference in an En lish sentence by means of descri!tion instead of by self-quotin or usin the !hrase -this sentence- is the Muine method, illustrated in the dialo ue *ir on G's -tring& The understandin of the Muine sentence requires less subtle mental !rocessin than the four e(am!les cited earlier& )lthou h it may a!!ear at first to be trickier, it is in some ways more e(!licit& The Muine construction is quite like the 6>del construction, in the way that it creates self-reference by describin another ty!o ra!hical entity which, as it turns out, is isomor!hic to the Muine sentence itself& The descri!tion of the new ty!o ra!hical entity is carried out by two !arts of the Muine sentence& Ene !art is a set of instructions tellin how to build a certain !hrase, while the other !art contains the construction materials to be used= that is, the other !art is a template& This resembles a floatin cake of soa! more than it resembles an iceber 4See 'i & *L5&
73G5=E EG!

The self-reference of this sentence is achieved in a more direct way than in the E!imenides !arado(= less hidden !rocessin is needed& 0y the way, it is interestin to !oint out that the !hrase -this sentence- a!!ears in the !revious sentence= yet it is not there to cause self-reference, you !robably understood that its referent was the Muine sentence, rather than the sentence in which it occurs& This .ust oes to show how !ointer

!hrases such as -this sentence- are inter!reted accordin to conte(t, and hel!s to show that the !rocessin of such !hrases is indeed quite involved&

A Self-$e!roducing ,rogram
The notion of quinin , and its usa e in creatin self-reference, have already been e(!lained inside the :ialo ue itself, so we need not dwell on such matters here& Aet us instead show how a com!uter !ro ram can use !recisely the same technique to re!roduce itself& The followin self-re!roducin !ro ram is written in a 0looP-like lan ua e and is based on followin a !hrase by its own quotation 4the o!!osite order from quinin , so I reverse the name -quine- to make -eniuq-5,
E+I;E ,$0CE &$E (E;I&8( ATEM,LATEB5 ,$I;T ATEM,LATE) LE+T -B$ACKET) 8&0TEMA$K) TEM,LATE) 8&0TE-MA$K) $IG/T-B$ACKET) ,E$I0 B1 E;I&8 A% E+I;E ,$0CE &$E (E;I&8( ATEM,LATEB5 ,$I;T ATEM,LATE) LE+T -B$ACKET) 8&0TEMA$K) TEM,LATE) 8&0TE-MA$K) $IG/T-B$ACKET) ,E$I0 B1 E;I&8%B1

E;I&8 is a !rocedure defined in the first two lines, and its in!ut is called - TEM,LATE-& It is understood that when the !rocedure is called, TEM,LATE1s value will be some strin of ty!o ra!hical characters& The effect of E;I&8 is to carry out a !rintin o!eration, in which TEM,LATE ets !rinted twice, the first time .ust !lain= the second time wra!!ed in 4sin le5 quotes and brackets, and arnished with a final !eriod& Thus, if TEM,LATE1s value were the strin 0&BLE-B&BBLE, then !erformin E;I&8 on it would yield, 0&BLE-B&BBLE A% 0&BLE-B&BBLE%B1 8ow in the last four lines of the !ro ram above, the !rocedure E;I&8 is called with a s!ecific value of TEM,LATEnamely the lon strin inside the sin le quotes, E+I;E&&& E;I&8& That value has been carefully chosen= it consists of the definition of E;I&8, followed by the word E;I&8& This makes the !ro ram itselfor, if you !refer, a !erfect co!y of it et !rinted out& It is very similar to Muine1s version of the E!imenides sentence, -yields falsehood when !receded by its quotationyields falsehood when !receded by its quotation& It is very im!ortant to reali<e that the character strin which a!!ears n quotes in the last three lines of the !ro ram abovethat is, the value of TEM,LATEis never inter!reted as a sequence of instructions& That it ha!!ens to be one is, in a sense& .ust an accident& )s was !ointed out above, it could .ust as well have been 0&BLE-B&BBLE or any other strin of characters& The beauty of the scheme is that when the same strin

a!!ears in the to! two lines of this !ro ram, it is treated as a !ro ram 4because it is not in quotes5& Thus in this !ro ram, one strin functions in two ways, first as !ro ram, and second as data& This is the secret of self-re!roducin !ro rams, and, as we shall see, of self-re!roducin molecules& It is useful, incidentally, to call any kind of self-re!roducin ob.ect or entity a self%rep= and likewise to call any self-referrin ob.ect or entity a self%ref& I will use those terms occasionally from here on& The !recedin !ro ram is an ele ant e(am!le of a self-re!roducin !ro ram written in a lan ua e which was not desi ned to make the writin of self-re!s !articularly easy& Thus, the task had to be carried out usin those notions and o!erations which were assumed to be !art of the lan ua esuch as the word 8&0TE-MA$K, and the command ,$I;T& 0ut su!!ose a lan ua e were desi ned e(!ressly for makin self-re!s easy to write& Then one could write much shorter self-re!s& 'or e(am!le, su!!ose that the o!eration of eniuq-in were a built-in feature of the lan ua e, needin no e(!licit definition 4as we assumed P;I8T was5& Then a teeny self-re! would be this, E;I&8 A%E;I&8%B1 It is very similar to the Tortoise1s version of Muine1s version of the E!imenides self-ref, where the verb -to quine- is assumed to be known, -yields falsehood when quined- yields falsehood when quined 0ut self-re!s can be even shorter& 'or instance, in some com!uter lan ua e it mi ht be a convention that any !ro ram whose first symbol is an asterisk is to be co!ied before bein e(ecuted normally& Then the !ro ram consistin of merely one asterisk is a self-re!2 Hou may com!lain that this is silly and de!ends on a totally arbitrary convention& In doin so, you are echoin my earlier !oint that it is almost cheatin to use the !hrase -this sentence- to achieve self-referenceit relies too much on the !rocessor, and not enou h on e(!licit directions for self-reference& Fsin an asterisk as an e(am!le of a self-re! is like usin the word -I- as an e(am!le of a self-ref, both conceal all the interestin as!ects of their res!ective !roblems& This is reminiscent of another curious ty!e of self-re!roduction, via !hotoco!y machine& It mi ht be claimed that any written document is a self-re! because it can cause a co!y of itself to be !rinted when it is !laced in a !hotoco!y machine and the a!!ro!riate button is !ushed& 0ut somehow this violates our notion of self-re!roduction= the !iece of !a!er is not consulted at all, and is therefore not directin its own re!roduction& ) ain, everythin is in the !rocessor& 0efore we call somethin a self-re!, we want to have the feelin that, to the ma(imum e(tent !ossible, it e(!licitly contains the directions for co!yin itself& To be sure, e(!licitness is a matter of de ree= nonetheless there is an intuitive borderline on one side of which we !erceive true selfdirected self-re!roduction, and on the other side of which we merely see co!yin bein carried out by an infle(ible and autonomous co!yin machine&

>hat Is a Co!y7
8ow in any discussion of self-refs and self-re!s, one must sooner or later come to ri!s with the essential issue, what is a co!y? 3e already dealt with that question quite seriously in 7ha!ters G and GI= and now we come back to it& To ive the flavor of the issue, let us describe some hi hly fanciful, yet !lausible, e(am!les of self-re!s& A Self-$e!roducing Song Ima ine that there is a nickelodeon in the local bar which, if you !ress buttons ""-F, will !lay a son whose lyrics o this way, Put another nickel in, in the nickelodeon, )ll I want is ""-F, and music, music, music& 3e could make a little dia ram of what ha!!ens one evenin 4'i & *D5&

73G5=E EJ! * self%reproducing song!

)lthou h the effect is that the son re!roduces itself, it would feel stran e to call the son a self-re!, because of the fact that when it !asses throu h the ""-F sta e, not all of the information is there& The information only ets !ut back by virtue of the fact that it is fully stored in the nickelodeon that is, in one of the arrows in the dia ram, not in one of the ovals& It is questionable whether this son contains a com!lete descri!tion of how to et itself !layed a ain, because the symbol !air -""-F- is only a tri er, not a co!y&

A (Crab( ,rogram
7onsider ne(t a com!uter !ro ram which !rints itself out backwards& 4Some readers mi ht en.oy thinkin about how to write such a !ro ram in the 0looP-like lan ua e above, usin the iven self-re! as a model&5 3ould this funny !ro ram count as a selfre!? Hes, in a way, because a trivial transformation !erformed on its out!ut will restore the ori inal !ro ram& It seems fair to say that the out!ut contains the same information as the !ro ram itself, .ust recast in a sim!le way& Het it is clear that someone mi ht look at

the out!ut and not reco ni<e it as a !ro ram !rinted backwards& To recall terminolo y from 7ha!ter GI, we could say that the -inner messa es- of the out!ut and the !ro ram itself are the same, but they have different -outer messa es-that is, they must be read by usin different decodin mechanisms& 8ow if one counts the outer messa e as !art of the informationwhich seems quite reasonablethen the total information is not the same after all, so the !ro ram can1t be counted as a self-re!& ?owever, this is a disquietin conclusion, because we are accustomed to considerin somethin and its mirror ima e as containin the same information& 0ut recall that in 7ha!ter GI, we made the conce!t of -intrinsic meanin - de!endent on a hy!othesi<ed universal notion of intelli ence& The idea was that, in determinin the intrinsic meanin of an ob.ect, we could disre ard some ty!es of outer messa ethose which would be universally understood& That is, if the decodin mechanism seems fundamental enou h, in some still ill-defined sense, then the inner messa e which it lets be revealed is the only meanin that counts& In this e(am!le, it seems reasonably safe to uess that a -standard intelli ence- would consider two mirror ima es to contain the same information as each other= that is, it would consider the isomor!hism between the two to be so trivial as to be i norable& )nd thus our intuition that the !ro ram is in some sense a fair self-re!, is allowed to stand&

E!imenides Straddles the Channel

8ow another far-fetched e(am!le of a self-re! would be a !ro ram which !rints itself out, but translated into a different com!uter lan ua e& Ene mi ht liken this to the followin curious version of the Muine version of the E!imenides self-ref, -est une e(!ression qui, quand elle est !rKcKdKe de sa traduction, mise entre uillemets, dans la lan ue !rovenant de l1autre cotK de la /anche& crKe une faussetK- is an e(!ression which, when it is !receded by its translation, !laced in quotation marks, into the lan ua e ori inatin on the other side of the 7hannel, yields a falsehood& Hou mi ht try to write down the sentence which is described by this weird concoction& 4?int, It is not itselfor at least it is not if -itself- is taken in a naPve sense&5 If the notion of -self-re! by retro rade motion- 4i&e&, a !ro ram which writes itself out backwards5 is reminiscent of a crab canon, the notion of -self-re! by translation- is no less reminiscent of -a canon which involves a trans!osition of the theme into another key&-

A ,rogram That ,rints 0ut Its 0'n G#del ;umber

The idea of !rintin out a translation instead of an e(act co!y of the ori inal !ro ram may seem !ointless& ?owever, if you wanted to write a self-re! !ro ram in 0looP or 'looP, you would have to resort to some such device, for in those lan ua es, 0&T,&T is always a number, rather than a ty!o ra!hical strin & Therefore, you would have to make the !ro ram !rint out its own 6>del number, a very hu e inte er whose decimal

e(!ansion codes for the !ro ram, character by character, by usin three di it codons& The !ro ram is comin as close as it can to !rintin itself, within the means available to it, it !rints out a co!y of itself in another -s!ace-, and it is easy to switch back and forth between the s!ace of inte ers and the s!ace of strin s& Thus, the value of 0&T,&T is not a mere tri er, like -""-"B-& Instead, all the information of the ori inal !ro ram lies -close to the surface- of the out!ut&

G#delian Self-$eference
This comes very close to describin the mechanism of 6>del1s self-ref 6& )fter all, that strin of T8T contains a descri!tion not of itself, but of an inte er 4the arithmoquinification of u5& It .ust so ha!!ens that that inte er is an e(act -ima e- of the strin 6, in the s!ace of natural numbers& Thus, 6 refers to a translation of itself into another s!ace& 3e still feel comfortable in callin 6 a self-referential strin , because the isomor!hism between the two s!aces is so ti ht that we can consider them to be identical& This isomor!hism that mirrors T8T inside the abstract realm of natural numbers can be likened to the quasi-isomor!hism that mirrors the real world inside our brains, by means of symbols& The symbols !lay quasi-isomor!hic roles to the ob.ects, and it is thanks to them that we can think& Aikewise, the 6>del numbers !lay isomor!hic roles to strin s, and it is thanks to them that we can find metamathematical meanin s in statements about natural numbers& The ama<in , nearly ma ical, thin about 6 is that it mana es to achieve self-reference des!ite the fact that the lan ua e in which it is written, T8T, seems to offer no ho!e of referrin to its own structures, unlike En lish, in which it is the easiest thin in the world to discuss the En lish lan ua e& So 6 is an outstandin e(am!le of a self-ref via translationhardly the most strai htforward case& Ene mi ht also think back to some of the :ialo ues, for some of them, too, are self-refs via translation& 'or instance, take the -onata for 5naccompanied *chilles& In that :ialo ue, several references are made to the 0ach Sonatas for unaccom!anied violin, and the Tortoise1s su estion of ima inin har!sichord accom!animents is !articularly interestin & )fter all, if one a!!lies this idea to the :ialo ue itself, one invents lines which the Tortoise is sayin = but if one assumes that )chilles1 !art stands alone 4as does the violin5, then it is quite wron to attribute any lines at all to the Tortoise& In any case, here a ain is a self-ref by means of a ma!!in which ma!s :ialo ues onto !ieces by 0ach& )nd this ma!!in is left, of course, for the reader to notice& Het even if the reader does not notice it, the ma!!in is still there, and the :ialo ue is still a self-ref&

A Self-$e! by Augmentation
3e have been likenin self-re!s to canons& 3hat, then, would be a fair analo ue to a canon by au mentation? ?ere is a !ossibility, consider a !ro ram which contains a dummy loo! whose only !ur!ose is to slow u! the !ro ram& ) !arameter mi ht tell how often to re!eat the loo!& ) self-re! could be made which !rints out a co!y of itself, but with the !arameter chan ed, so that when that co!y is run, it will run at half the s!eed of

its !arent !ro ram= and its -dau hter- will in turn run at half a ain the s!eed, and so on&&& 8one of these !ro rams !rints itself out !recisely= yet all clearly belon to a sin le -family-& This is reminiscent of the self-re!roduction of livin or anisms& 7learly, an individual is never identical to either of its !arents= why, then, is the act of makin youn called -self-re!roduction1? The answer is that there is a coarse- rained isomor!hism between !arent and child= it is an isomor!hism which !reserves the information about s!ecies& Thus, what is re!roduced is the class, rather than the instance& This is also the case in the recursive !icture 6!lot, in 7ha!ter G, that is, the ma!!in between -ma netic butterflies- of various si<es and sha!es is coarse- rained= no two are identical, but they all belon to a sin le -s!ecies-, and the ma!!in !reserves !recisely that fact& In terms of self-re!licatin !ro rams, this would corres!ond to a famil) of !ro rams, all written in -dialects- of a sin le com!uter lan ua e= each one can write itself out, but sli htly modified, so that it comes out in a dialect of its ori inal lan ua e&

A Kimian Self-$e!
Perha!s the sneakiest e(am!le of a self-re! is the followin , instead of writin a le al e(!ression in the com!iler lan ua e, you ty!e one of the com!iler1s own error messa es& 3hen the com!iler looks at your -!ro ram-, the first thin it does is et confused, because your -!ro ram- is un rammatical= hence the com!iler !rints out an error messa e& )ll you need to do is arran e that the one it !rints out will be the one you ty!ed in& This kind of self-re!, su ested to me by Scott @im, e(!loits a different level of the system from the one you would normally a!!roach& )lthou h it may seem frivolous, it may have counter!arts in com!le( systems where self-re!s vie a ainst each other for survival, as we shall soon discuss&

>hat Is the 0riginal7

0esides the question -3hat constitutes a co!y?-, there is another fundamental !hiloso!hical question concernin self-re!s& That is the obverse side of the coin, -3hat is the ori inal?- This can best be e(!lained by referrin to some e(am!les, 4"5 a !ro ram which, when inter!reted by some inter!reter runnin on some com!uter, !rints itself out= 4B5 a !ro ram which, when inter!reted by some inter!reter runnin on some com!uter& !rints itself out alon with a com!lete co!y of the inter!reter 4which, after all, is also a !ro ram5= 4%5 a !ro ram which, when inter!reted by some inter!reter runnin on some com!uter, not only !rints itself out alon with a com!lete co!y of the inter!reter, but also directs a mechanical assembly !rocess in which a second

com!uter, identical to the one on which the inter!reter and !ro ram are runnin , is !ut to ether& It is clear that in 4"5, the !ro ram is the self-re!& 0ut in 4%5, is it the !ro ram which is the self-re!, or the com!ound system of !ro ram !lus inter!reter, or the union of !ro ram, inter!reter, and !rocessor? 7learly, a self-re! can involve more than .ust !rintin itself out& In fact, most of the rest of this 7ha!ter is a discussion of self-re!s in which data, !ro ram, inter!reter, and !rocessor are all e(tremely intertwined, and in which self-re!lication involves re!licatin all of them at once&

Ty!ogenetics
3e are now about to broach one of the most fascinatin and !rofound to!ics of the twentieth century, the study of -the molecular lo ic of the livin state-, to borrow )lbert Aehnin er1s richly evocative !hrase& )nd lo ic it is, toobut of a sort more com!le( and beautiful than any a human mind ever ima ined& 3e will come at it from a sli htly novel an le, via an artificial solitaire ame which I call T)pogeneticsshort for -Ty!o ra!hical 6enetics-& In Ty!o enetics I have tried to ca!ture some ideas of molecular enetics in a ty!o ra!hical system which, on first si ht, resembles very much the formal systems e(em!lified by the /IF-system& Ef course, Ty!o enetics involves many sim!lifications, and therefore is useful !rimarily for didactic !ur!oses& I should e(!lain immediately that the field of molecular biolo y is a field in which !henomena on several levels interact, and that Ty!o enetics is only tryin to illustrate !henomena from one or two levels& In !articular, !urely chemical as!ects have been com!letely avoidedthey belon to a level lower than is here dealt with= similarly, all as!ects of classical enetics 4vi<&, nonmolecular enetics5 have also been avoided they belon to a level hi her than is here dealt with& I have intended in Ty!o enetics only to ive an intuition for those !rocesses centered on the celebrated 8entral .ogma of 'olecular Biolog), enunciated by 'rancis 7rick 4one of the co-discoverers of the doubleheli( structure of :8)5, :8) ;8) !roteins& It is my ho!e that with this very skeletal model I have constructed the reader will !erceive some sim!le unifyin !rinci!les of the field !rinci!les which mi ht otherwise be obscured by the enormously intricate inter!lay of !henomena at many different levels& 3hat is sacrificed is, of course, strict accuracy= what is ained is, I ho!e, a little insi ht&

Strands) Bases) En:ymes

The ame of Ty!o enetics involves ty!o ra!hical mani!ulation on sequences of letters& There are four letters involved,

A C G T1 )rbitrary sequences of them are called strands& Thus, some strands are, GGGG ATTACCA CATCATCATCAT Incidentally, -ST;)8:- s!elled backwards be ins with -:8)-& This is a!!ro!riate since strands, in Ty!o enetics, !lay the role of !ieces of :8) 4which, in real enetics, are often called -strands-5& 8ot only this, but -ST;)8:- fully s!elled out backwards is -:8) ;TS-, which may be taken as an acronym for -:8) ;a!id Transit Service-& This, too, is a!!ro!riate, for the function of -messen er ;8)-which in Ty!o enetics is re!resented by strands as wellis quite well characteri<ed by the !hrase -;a!id Transit Service- for :8), as we shall see later& I will sometimes refer to the letters A, C, G, T as bases, and to the !ositions which they occu!y as units& Thus, in the middle strand, there are seven units, in the fourth of which is found the base A& If you have a strand, you can o!erate on it and chan e it in various ways& Hou can also !roduce additional strands, either by co!yin , or by cuttin a strand in two& Some o!erations len then strands, some shorten them, and some leave their len th alone& E!erations come in !acketsthat is, several to be !erformed to ether, in order& Such a !acket of o!erations is a little like a !ro rammed machine which moves u! and down the strand doin thin s to it& These mobile machines are called -ty!o ra!hical en<ymes-en<ymes for short& En<ymes o!erate on strands one unit at a time, and are said to be -bound- to the unit they are o!eratin on at any iven moment& I will show how some sam!le en<ymes act on !articular strin s& The first thin to know is that each en<yme likes to start out bound to a !articular letter& Thus, there are four kinds of en<ymethose which !refer A, those which !refer C, etc& 6iven the sequence of o!erations which an en<yme !erforms, you can fi ure out which letter it !refers, but for now I1ll .ust ive them without e(!lanation& ?ere1s a sam!le en<yme, consistin of three o!erations,

4"5 :elete the unit to which the en<yme is bound 4and then bind to the ne(t unit to the ri ht5& 4B5 /ove one unit to the ri ht& 4%5 Insert a T 4to the immediate ri ht of this unit5&

This en<yme ha!!ens to like to bind to ) initially& )nd here1s a sam!le strand, ACA

3hat ha!!ens if our en<yme binds to the left A and be ins actin ? Ste! " deletes the A, so we are left with CAand the en<yme is now bound to the C& Ste! B slides the en<yme ri htwards, to the A, and Ste! % a!!ends a T onto the end to form the strand CAT& )nd the en<yme has done its com!lete duty, it has transformed ACA into CAT& 3hat if it had bound itself to the ri ht A of ACA? It would have deleted that A and moved off the end of the strand& 3henever this ha!!ens, the en<yme quits 4this is a eneral !rinci!le5& So the entire effect would .ust be to lo! off one symbol& Aet1s see some more e(am!les& ?ere is another en<yme,

4"5 Search for the nearest !yrimidine to the ri ht of this unit& 4B5 6o into 7o!y mode& 4%5 Search for the nearest !urine to the ri ht of this unit& 4C5 7ut the strand here 4vi<&, to the ri ht of the !resent unit5&

8ow this contains the terms -!yrimidine- and -!urine-& They are easy terms& A and G are called purines, and C and T are called p)rimidines& So searchin for a !yrimidine merely means searchin for the nearest C or T.

Co!y Mode and

ouble Strands

The other new term is 8op) mode& )ny strand can be -co!ied- onto another strand, but in a funny way& Instead of co!yin A onto A, you co!y it onto T, and vice versa& )nd instead of co!yin C onto C, you co!y it onto G, and vice versa& 8ote that a !urine co!ies onto a !yrimidine, and vice versa& This is called complementar) (ase pairing& The com!lements are shown below&

Purines

7om!lement A T G C

Pyrimidines

Hou can !erha!s remember this molecular !airin scheme by recallin that Achilles is !aired with the Tortoise, and the Crab with his Genes& 3hen -co!yin - a strand, therefore, you don1t actually co!y it, but you manufacture its complementar) strand& )nd this one will be written u!side down above the ori inal strand& Aet1s see this in concrete terms& Aet the !revious en<yme act on the followin strand 4and that en<yme also ha!!ens to like to start at A5, CAAAGAGAATCCTCTTTGAT There are many !laces it could start& Aet1s take the second ), for e(am!le& The en<yme binds to it, then e(ecutes ste! ", Search for the nearest !yrimidine to the ri ht& 3ell, this

means a C or a T. The first one is a T somewhere near the middle of the strand, so that1s where we o& 8ow ste! B, 7o!y mode& 3ell, we .ust !ut an u!side-down A above our T& 0ut that1s not all, for 7o!y mode remains in effect until it is shut offor until the en<yme is done, whichever comes first& This means that every base which is !assed throu h by the en<yme while 7o!y mode is on will et a com!lementary base !ut above it& Ste! % says to look for a !urine to the ri ht of our T& That is the G two symbols in from the ri hthand end& 8ow as we move u! to that G, we must -co!y-that is, create a com!lementary strand& ?ere1s what that ives, CAAAGAGAATCCTCTTTGAT The last ste! is to cut the strand& This will yield two !ieces, CAAAGAGAATCCTCTTTG and AT& )nd the instruction !acket is done& 3e are left with a double strand, however& 3henever this ha!!ens, we se!arate the two com!lementary strands from each other 4 eneral !rinci!le5= so in fact our end !roduct is a set of three strands, AT, CAAAGAGGA, and CAAAGAGAATCCTCTTTG CAAAGAGGA CAAAGAGGA

8otice that the u!side-down strand has been turned ri ht side u!, and thereby ri ht and left have been reversed& 8ow you have seen most of the ty!o ra!hical o!erations which can be carried out on strands& There are two other instructions which should be mentioned& Ene shuts off 7o!y mode= the other switches the en<yme from a strand to the u!side-down strand above it& 3hen this ha!!ens, if you kee! the !a!er ri ht side u!, then you must switch -left- and -ri ht- in all the instructions& Er better, you can kee! the wordin and .ust turn the !a!er around so the to! strand becomes le ible& If the -switch- command is iven, but there is no com!lementary base where the en<yme is bound at that instant, then the en<yme .ust detaches itself from the strand, and its .ob is done& It should be mentioned that when a -cut- instruction is encountered, this !ertains to both strands 4if there are two5, however, -delete- !ertains only to the strand on which the en<yme is workin & If 7o!y mode is on, then the -insert- command !ertains to both strandsthe base itself into the strand the en<yme is workin on, and its com!lement into the other strand& If 7o!y mode is off, then the -insert- command !ertains only to the one strand, so a blank s!ace must he inserted into the com!lementary strand& )nd, whenever 7o!y mode is on, -move- and -search- commands require that one manufacture com!lementary bases to all bases which the slidin en<yme touches&

Incidentally, 7o!y mode is always off when an en<yme starts to work& If 7o!y mode is off, and the command -Shut off co!y mode- is encountered, nothin ha!!ens& Aikewise, If 7o!y mode is already on, and the command -Turn co!y mode on- is encountered, then nothin ha!!ens&

Amino Acids
There are fifteen ty!es of command, listed below, cm del swi mvr mvl co! off ina inc in int r!y r!u l!y l!u cut strand4s5 delete a base from strand switch en<yme to other strand move one unit to the ri ht move one unit to the left turn on 7o!y mode turn off 7o!y mode insert A to the ri ht of this unit insert C to the ri ht of this unit insert G to the ri ht of this unit insert T to the ri ht of this unit search for the nearest !yrimidine to the ri ht search for the nearest !urine to the ri ht search for the nearest !yrimidine to the left search for the nearest !urine to the left

Each one has a three-letter abbreviation& 3e shall refer to the three-letter abbreviations of commands as amino acids& Thus, e#er) en@)me is made up of a se"uence of amino acids & Aet us write down an arbitrary en<yme, r!u - inc - co! - mvr - mvl - swi - l!u - int and an arbitrary strand, TAGATCCAGTCCATCGA and see how the en<yme acts on the strand& It so ha!!ens that the en<yme binds to G only& Aet us bind to the middle G and be in& Search ri htwards for a !urine 4vi<&, A or G5& 3e 4the en<yme5 ski! over TCC and land on A& Insert a C& 8ow we have TAGATCCAGTCCACTCGA

where the arrow !oints to the unit to which the en<yme is bound& Set 7o!y mode& This !uts an u!side-down G above the C& /ove ri ht, move left, then switch to the other strand& ?ere1s what we have so far, AG

TAGATCCAGTCCACTCGA Aet1s turn it u!side down, so that the en<yme is attached to the lower strand, TAGATCCAGTCCACTCGA AG 8ow we search leftwards for a !urine and find A& 7o!y mode is on, but the com!lementary bases are already there, so nothin is added& 'inally, we insert a T 4in 7o!y mode5, and quit, TAGATCCAGTCCACATCGA ATG Eur final !roduct is thus two strands, ATG, and TAGATCCAGTCCACATCGA The old one is of course one&

Translation and the Ty!ogenetic Code

8ow you mi ht be wonderin where the en<ymes and strands come from, and how to tell the initial bindin -!reference of a iven en<yme& Ene way mi ht be .ust to throw some random strands and some random en<ymes to ether, and see what ha!!ens when those en<ymes act on those strands and their !ro eny& This has a similar flavor to the /F!u<<le, where there were some iven rules of inference and an a(iom, and you .ust be an& The only difference is that here, every time a strand is acted on, its ori inal form is one forever& In the /F-!u<<le, actin on MI to make MI& didn1t destroy MI1 0ut in Ty!o enetics, as in real enetics, the scheme is quite a bit trickier& 3e do be in with some arbitrary strand, somewhat like an a(iom in a formal system& 0ut we have, initially, no -rules of inference-that is, no en<ymes& ?owever, we can translate each strand into one or more en<ymes2 Thus, the strands themselves will dictate the o!erations which will be !erformed u!on them, and those o!erations will in turn !roduce new strands which will dictate further en<ymes, etc& etc&2 This is mi(in levels with a ven eance2 Think, for the sake of com!arison, how different the /F-!u<<le would have been if each new theorem !roduced could have been turned into a new rule of inference by means of some code&

?ow is this -translation- done? It involves a T)pogenetic 8ode by which ad.acent !airs of basescalled -du!lets-in a sin le strand re!resent different amino acids& There are si(teen !ossible du!lets, AA, AC, AG, AT, CA, CC, etc& )nd there are fifteen amino acids& The Ty!o enetic 7ode is shown in 'i ure *$&

73G5=E EC! The T)pogenetic 8ode, () which each duplet in a strand codes for one of fifteen "amino acids" or a punctuation mark!/

)ccordin to the table, the translation of the du!let GC is -inc- 4-insert a C-5= that of AT is -swi- 4-switch strands-5= and so on& Therefore it becomes clear that a strand can dictate an en<yme very strai htforwardly& 'or e(am!le, the strand TAGATCCAGTCCACATCGA breaks u! into du!lets as follows, TA GA TC CA GT CC AC AT CG A with the A left over at the end& Its translation into an en<yme is, r!y - ina - r!u - mvr - int - mvl - cut - swi - co!& 48ote that the leftover A contributes nothin &5

Tertiary Structure of En:ymes

3hat about the little letters 1s1, 1l1, and 1r1 in the lower ri hthand corner of each bo(? They are crucial in determinin the en<yme1s bindin -!reference, and in a !eculiar way& In order to fi ure out what letter an en<yme likes to bind to, you have to fi ure out the en<yme1s -tertiary structure-, which is itself determined by the en<yme1s -!rimary structure-& 0y its primar) structure is meant its amino acid sequence& 0y its tertiary structure is meant the way it likes to -fold u!-& The !oint is that en<ymes don1t like bein in strai ht lines, as we have so far e(hibited them& )t each internal amino acid 4all but the two ends5, there is a !ossibility of a -kink-, which is dictated by the letters in the corners&

In !articular, 1"1 and 1r1 stand for -left- and -ri ht-, and 1s1 stands for -strai ht-& So let us take our most recent sam!le en<yme, and let it fold itself u! to show its tertiary structure& 3e will start with the en<yme1s !rimary structure, and move alon it from left to ri ht& )t each amino acid whose corner-letter is 1"1 we1ll !ut a left turn, for those with 1r1, we1ll !ut a ri ht turn, and at 1s1 we1ll !ut no turn& In 'i ure ** is shown the two-dimensional conformation for our en<yme& co! swi cut mvl int mvr r!y ina r!u
73G5=E EE! The tertiar) structure of a t)poen@)me!

8ote the left-kink at -r!u-, the ri ht-kink at -swi-, and so on& 8otice also that the first se ment 4-r!y ina-5 and the last se ment 4-swi co!-5 are !er!endicular& This is the key to the bindin -!reference& In fact, the relati#e orientation of the first and last segments of an en<yme1s tertiary structure determines the bindin -!reference of the en<yme& 3e can always orient the en<yme so that its first se ment !oints to the ri ht& If we do so, then the last se ment determines the bindin -!reference, as shown in 'i ure *#&
73G5=E EI! Ta(le of (inding%preferences for t)poen@)mes!

7irst -egment

+ast -egment

Binding%letter
A C G T

So in our case, we have an en<yme which likes the letter C& If, in foldin u!, an en<yme ha!!ens to cross itself, that1s okay.ust think of it as oin under or over itself& 8otice that all its amino acids !lay a role in the determination of an en<yme1s tertiary structure&

,unctuation) Genes) and $ibosomes

8ow one thin remains to he e(!lained& 3hy is there a blank in bo( AA of the ty!o enetic 7ode? The answer is that the du!let AA acts as a !unctuation mark inside a strand, and it si nals the end of the code for an en<yme& That is to say, one strand may code for two or more en<ymes if it has one or more du!lets AA in it& 'or e(am!le, the strand CG GA TA CT AA AC CG A 7odes for two en<ymes co! - ina - r!y - off and cut Y co! with the AA servin to divide the strand u! into two - enes-& The definition of ene is, that portion of a strand which codes for a single en@)me & 8ote that the mere !resence of AA inside a strand does not mean that the strand codes for two en<ymes& 'or instance, CAAG codes for -mvr - del-& The AA be ins on an even-numbered unit and therefore is not read as a du!let2 The mechanism which reads strands and !roduces the en<ymes which are coded inside them is called a ri(osome& 4In Ty!o enetics, the !layer of the ame does the work of the ribosomes&5 ;ibosomes are not in any way res!onsible for the tertiar) structure of en<ymes, for that is entirely determined once the primar) structure is created& Incidentally, the !rocess of translation always oes from strands to en@)mes, and never in the reverse direction&

,u::le5 A Ty!ogenetical Self-$e!

8ow that the rules of Ty!o enetics have been fully set out, you may find it interestin to e(!eriment with the ame& In !articular, it would he most interestin to devise a selfre!licatin strand& This would mean somethin alon the followin lines& ) sin le strand is written down& ) ribosome acts on it, to !roduce any or all of the en<ymes which are coded for in the strand& Then those en<ymes are brou ht into contact with the ori inal strand, and allowed to work on it& This yields a set of -dau hter strands-& The dau hter strands themselves !ass throu h the ribosomes, to yield a second eneration of en<ymes, which act on the dau hter strands= and the cycle oes on and on& This can o on for any number of sta es= the ho!e is that eventually, amon the strands which are !resent at some !oint, there

The Central

ogma of Ty!ogenetics

Ty!o enetical !rocesses can be re!resented in skeletal form in a dia ram 4'i & #+5& en<ymes translation #ia ri(osomes strands This dia ram illustrates the 8entral .ogma of T)pogenetics& It shows how strands define en<ymes 4via the Ty!o enetic 7ode5= and how in turn, en<ymes act back on the strands which ave rise to them, yieldin new strands& Therefore, the line on the left !ortrays how old information flows u!wards, in the sense that an en<yme is a translation of a strand, and contains therefore the same information as the strand, only in a different form in !articular= in an active form& The line on the ri ht, however, does not show information flowin downwards= instead, it shows how new information gets created, by the shuntin of symbols in strands& )n en<yme in Ty!o enetics, like a rule of inference in a formal system, blindly shunts symbols in strands without re ard to any -meanin - which may lurk in those symbols& So there is a curious mi(ture of levels here& En the one hand, strands are acted u!on, and therefore !lay the role of data 4as is indicated by the arrow on the ri ht5= on the other hand, they also dictate the actions which are to be !erformed on the data, and therefore they !lay the role of programs 4as is indicated by the arrow on the left5& It is the !layer of Ty!o enetics who acts as inter!reter and !rocessor, of course& The two-way street which links -u!!er- and -lower- levels of Ty!o enetics shows that, in fact, neither strands nor en<ymes can be thou ht of as bein on a hi her level than the other& 0y contrast, a !icture of the 8entral .ogma of the '35-system looks this way, rules of interference t)pographical manipulation
73G5=E IK! The ^8entral .ogma of T)pogenetics_: an e9ample of a ^Tangled ,ierarch)_

4ty!o ra!hical mani!ulation5

strin s In the /IF-system, there is a clear distinction of levels, rules of inference sim!ly belon to a hi her level than strin s& Similarly for T8T, and all formal systems&

Strange Loo!s) T;T) and $eal Genetics

?owever, we have seen that in T8T, levels are mi(ed, in another sense& In fact, the distinction between lan ua e and metalan ua e breaks down, statements about the system et mirrored inside the system& It turns out that if we make a dia ram showin the

relationshi! between T8T and its metalan ua e, we will !roduce somethin which resembles in a remarkable way the dia ram which re!resents the 7entral :o ma of /olecular 0iolo y& In fact, it is our oal to make this com!arison in detail= but to do so, we need to indicate the !laces where Ty!o enetics and true enetics coincide, and where they differ& Ef course, real enetics is far more com!le( than Ty!o eneticsbut the -conce!tual skeleton- which the reader has acquired in understandin Ty!o enetics will be very useful as a uide in the labyrinth of true enetics&

;A and ;ucleotides
3e be in by discussin the relationshi! between -strands-, and :8)& The initials -:8)stand for -deo(yribonucleic acid-& The :8) of most cells resides in the cell1s nucleus, which is a small area !rotected by a membrane& 6unther Stent has characteri<ed the nucleus as the -throne room- of the cell, with :8) actin as the ruler& :8) consists of lon chains of relatively sim!le molecules called nucleotides& Each nucleotide is made u! of three !arts, 4"5 a !hos!hate rou! stri!!ed of one s!ecial o(y en atom, whence the !refi( -deo(y-= 4B5 a su ar called -ribose-, and 4%5 a base& It is the base alone which distin uishes one nucleotide from another= thus it suffices to s!ecify its base to identify a nucleotide& The four ty!es of bases which occur in :8) nucleotides are, ), 6, 7, T, adenine uanine cytosine thymine

b b

purines

p)rimidines

4)lso see 'i & #"&5 It is easy to remember which ones are !yrimidines because the first vowel in -cytosine-, -thymine-, and -!yrimidine- is 1y1& Aater, when we talk about ;8), -uracil-also a !yrimidinewill come in and wreck the !attern, unfortunately& 48ote, Aetters re!resentin nucleotides in real enetics will not be in the 8uadrata font, as they were in Ty!o enetics&5 ) sin le strand of :8) thus consists of many nucleotides strun to ether like a chain of beads& The chemical bond which links a nucleotide to its two nei hbors is very stron = such bonds are called co#alent (onds, and the -chain of beads- is often called the co#alent (ack(one of :8)& 8ow :8) usually comes in double strandsthat is, two sin le strands which are !aired u!, nucleotide by nucleotide 4see 'i & #B5& It is the bases which are res!onsible for the !eculiar kind of !airin which takes !lace between strands& Each base in one strand faces a com!lementary base in the other strand, and binds to it& The com!lements are as in Ty!o enetics, ) !airs u! with T, and 7 with 6& )lways one !urine !airs u! with a !yrimidine&

73G5=E IB! The four constituent (ases of .N*: *denine, Guanine, Th)mine, 8)tosine! U7rom ,anawalt and ,a)nes, The 7hemical 0asis of Aife -an 7rancisco: >! ,! 7reeman, BICH/, p! BDF!V 73G5=E IF! .N* structure resem(les a ladder in which the side pieces consist of alternating units of deo9)ri(ose and phosphate! The rungs are formed () the (ases paired in a special wa), * with T and G with 8, and held together respecti#el) () two and three h)drogen (onds! U7rom ,anawalt and ,a)nes, The 7hemical 0asis of Aife, p! BDF!V

7om!ared to the stron covalent bonds alon the backbone, the interstrand bonds are quite weak& They are not covalent bonds, but h)drogen (onds& ) hydro en bond arises when two molecular com!le(es are ali ned in such a way that a hydro en atom which ori inally belon ed to one of them becomes -confused- about which one it belon s to, and it hovers between the two com!le(es, vacillatin as to which one to .oin& 0ecause the two halves of double-stranded :8) are held to ether only by hydro en bonds, they may come a!art or be !ut to ether relatively easily= and this fact is of reat im!ort for the workin s of the cell& 3hen :8) forms double strands, the two strands curl around each other like twistin vines 4'i & #%5& There are e(actly ten nucleotide !airs !er revolution= in other words, at each nucleotide, the -twist- is %D de rees& Sin le-stranded :8) does not e(hibit this kind of coilin , for it is a consequence of the base-!airin &

73G5=E IH! 'olecular model of the .N* dou(le heli9! U7rom <ernon '! 3ngram, 0iosynthesis 'enlo &ark, 8alif!: >! *! BenMamin, BICF/, p! BHV

Messenger $;A and $ibosomes

)s was mentioned above, in many cells, :8), the ruler of the cell, dwells in its !rivate -throne room-, the nucleus of the cell& 0ut most of the -livin - in a cell oes on outside

of the nucleus, namely in the c)toplasmOthe - round- to the nucleus1 -fi ure-& In !articular, en@)mes, which make !ractically every life !rocess o, are manufactured by ri(osomes in the cyto!lasm, and they do most of their work in the cyto!lasm& )nd .ust as in Ty!o enetics, the blue!rints for all en<ymes are stored inside the strandsthat is, inside the :8), which remains !rotected in its little nuclear home& So how does the information about en<yme structure et from the nucleus to the ribosomes1 ?ere is where messenger ;8)m;8)comes in& Earlier, m;8) strands were humorously said to constitute a kind of :8) ;a!id Transit Service= by this is meant not that m;8) !hysically carries :8) anywhere, but rather that it serves to carry the information, or messa e, stored in the :8) in its nuclear chambers, out to the ribosomes in the cyto!lasm& ?ow is this done? The idea is easy, a s!ecial kind of en<yme inside the nucleus faithfully co!ies lon stretches of the :8)1s base sequence onto a new stranda strand of messen er ;8)& This m;8) then de!arts from the nucleus and wanders out into the cyto!lasm, where it runs into many ribosomes which be in doin their en<ymecreatin work on it& The !rocess by which :8) ets co!ied onto m;8) inside the nucleus is called transcription= in it, the double-stranded :8) must be tem!orarily se!arated into two sin le strands, one of which serves as a tem!late for the m;8)& Incidentally, -;8)stands for -ribonucleic acid-, and it is very much like :8) e(ce!t that all of its nucleotides !ossess that s!ecial o(y en atom in the !hos!hate rou! which :8)1s nucleotides lack& Therefore the -deo(y- !refi( is dro!!ed& )lso, instead of thymine, ;8) uses the base uracil, so the information in strands of ;8) can be re!resented by arbitrary sequences of the four letters 1)1, 171, 161, 1F1& 8ow when m;8) is transcribed off of :8), the transcri!tion !rocess o!erates via the usual base-!airin 4e(ce!t with F instead of T5, so that a :8)-tem!late and its m;8)-mate mi ht look somethin like this, :8),&&&&&&&&&&&&&&&&&&&&&&&&66T)))T7))6T7) &&&&&&&&&&&&&&&&&&&&&4tem!late5 m;8),&&&&&&&&&&&&&&&&&&&&&&&667)FFF)6F7)6F&&&&&&&&&&&&&&&&&&&&4-co!y-5 ;8) does not enerally form lon double strands with itself, althou h it can& Therefore it is !revalently found not in the helical form which so characteri<es :8), but rather in lon , somewhat randomly curvin strands& Ence a strand of m;8) has esca!ed the nucleus, it encounters those stran e subcellular creatures called -ribosomes-but before we o on to e(!lain how a ribosome uses m;8), I want to make some comments about en<ymes and !roteins& En<ymes belon to the eneral cate ory of biomolecules called proteins, and the .ob of ribosomes is to make all !roteins, not .ust en<ymes& Proteins which are not en<ymes are much more !assive kinds of bein s, many of them, for instance, are structural molecules, which means that they are like irders and beams and so forth in buildin s, they hold the cell1s !arts to ether& There are other kinds of !roteins, but for our !ur!oses& the !rinci!al !roteins are en<ymes, and I will henceforth not make a shar! distinction&

Amino Acids
Proteins are com!osed of sequences of amino acids, which come in twenty !rimary varieties, each with a three-letter abbreviation, ala ar asn as! cys ln lu ly his ile leu lys met !he !ro ser thr tr! tyr val alanine ar inine as!ara ine as!artic acid cysteine lutamine lutamic acid lycine histidine isoleucine leucine lysine methionine !henylalanine !roline serine threonine try!to!han tyrosine valine

8otice the sli ht numerical discre!ancy with Ty!o enetics, where we had only fifteen -amino acids- com!osin en<ymes& )n amino acid is a small molecule of rou hly the same com!le(ity as a nucleotide= hence the buildin blocks of !roteins and of nucleic acids 4:8), ;8)5 are rou hly of the same si<e& ?owever, !roteins are com!osed of much shorter sequences of com!onents, ty!ically, about three hundred amino acids make a com!lete !rotein, whereas a strand of :8) can consist of hundreds of thousands or millions of nucleotides&

$ibosomes and Ta!e $ecorders

8ow when a strand of m;8), after its esca!e into the cyto!lasm, encounters a ribosome, a very intricate and beautiful !rocess called translation takes !lace& It could be said that this !rocess of translation is at the very heart of all of life, and there are many mysteries connected with it& 0ut in essence it is easy to describe& Aet us first ive a !icturesque ima e, and then render it more !recise& Ima ine the m;8) to be like a lon !iece of ma netic recordin ta!e, and the ribosome to be like a ta!e recorder& )s the ta!e !asses throu h the !layin head of the recorder, it is -read- and converted into music, or other sounds& Thus ma netic markin s are -translated- into notes& Similarly, when a -ta!e- of

m;8) !asses throu h the -!layin head- of a ribosome, the -notes- which are !roduced are amino acids, and the -!ieces of music- which they make u! are !roteins& This is what translation is all about, it is shown in 'i ure #D&

The Genetic Code

0ut how can a ribosome !roduce a chain of amino acids when it is readin a chain of nucleotides This mystery was solved in the early "#D+1s by the efforts of a lar e number of !eo!le, and at the core of the answer lies the 6enetic 7odea ma!!in from tri!lets of nucleotides into amino acids 4see 'i & #C5& This is in s!irit e(tremely similar to the Ty!o enetic 7ode, e(ce!t that here, three consecutive bases 4or nucleotides5 form a codon, whereas 7F) 6)F there, only two were 7 u ) ) u needed& Thus there must be CWCWC 4equals DC5 different * t)pical segment of m=N* entries in the table, instead of read first as two triplets si(teen& ) ribosome clicks a(o#e/, and second as three down a strand of ;8) three duplets (elow/: an e9ample nucleotides at a timewhich of hemiolia in (iochemistr) is to say, one codon at a time and each time it does so, 4above5, it a!!ends a sin le new amino acid toand thesecond !roteinas it three is !resently manufacturin & Thus, a !rotein comes out of the ribosome amino acid by amino acid&

Tertiary Structure
?owever, as a !rotein emer es from a ribosome, it is not only ettin lon er and lon er, but it is also continually foldin itself u! into an e(traordinary three-dimensional sha!e, very much in the way that those funny little 'ourth-of-July fireworks called -snakessimultaneously row lon er and curl u!, when they are lit& This fancy sha!e is called the !rotein1s tertiar) structure 4'i & #L5, while the amino acid sequence !er se is called the primar) structure of the !rotein& The tertiary structure is im!licit in the !rimary structure, .ust as in Ty!o enetics& ?owever, the reci!e for derivin the tertiary structure, if you know only the !rimary structure, is by far more com!le( than that iven in Ty!o enetics& In fact, it is one of the outstandin !roblems of contem!orary molecular biolo y to fi ure out some rules by which the tertiary structure of a !rotein can be !redicted if only its !rimary structure is known&

The Genetic 8ode!

*
!he !he leu leu

C
ser ser ser ser !ro !ro !ro !ro tyr tyr

+
cys cys punc! tr! ar ar ar ar F 7

punc! punc! his his in ln

*
6 F 7 ) 6

leu

leu leu Ieu

ile

thr thr thr thr

asn asn lys lys

ser ser ar ar

F 7 ) 6

ile ile met

val

ala ala ala ala

as! as! lu lu

ly ly ly ly

F 7 ) 6

val val val

73G5=E ID! The Genetic 8ode, () which each triplet in a strand of messenger =N* codes for one of twent) amino acids or a punctuation mark/!

$eductionistic E-!lanation of ,rotein +unction

)nother discre!ancy between Ty!o enetics and true eneticsand this is !robably the most serious one of allis this, whereas in Ty!o enetics, each com!onent amino acid of an en<yme is res!onsible for some s!ecific -!iece of the action-, in real en<ymes, individual amino acids cannot be assi ned such clear roles& It is the tertiary structure as a whole which determines the mode in which an en<yme will function= there is no way one can say, -This amino acid1s !resence means that such-and-such an o!eration will et !erformed-& In other words, in real enetics, an individual amino acid1s contribution to

the en<yme1s overall function is not -conte(t-free-& ?owever, this fact should not be construed in any way as ammunition for an antireductionist ar ument to the effect that -the whole Uen<ymeV cannot be e(!lained as the sum of its !arts-& That would be wholly un.ustified& 3hat is .ustified is re.ection of the sim!ler claim that -each amino acid contributes to the sum in a manner which is inde!endent of the other amino acids !resent-& In other words, the function of a !rotein cannot be considered to be built u! from conte(t-free functions of its !arts= rather, one must consider how the !arts interact& It is still !ossible in !rinci!le to write a com!uter !ro ram which takes as in!ut the !rimary structure of a !rotein, and firstly determines its tertiary structure, and secondly determines the function of the en<yme& This would be a com!letely reductionistic e(!lanation of the workin s of !roteins, but the determination of the -sum- of the !arts
73G5=E IG! The structure of m)oglo(in, deduced from high%resolution S%ra) data! The large%scale "twisted pipe" appearance is the tertiar) structureR the finer heli9 insideOthe "alpha heli9"Ois the secondar) structure! U7rom *! +ehninger, 0iochemistryV

would require a hi hly com!le( al orithm& The elucidation of the function of an en<yme, iven its !rimary, or even its tertiary, structure, is another reat !roblem of contem!orary molecular biolo y& Perha!s, in the last analysis, the function of the whole en<yme can be considered to be built u! from functions of !arts in a conte(t-free manner, but where the !arts are now considered to be individual !articles, such as electrons and !rotons, rather than -chunks-, such as amino acids& This e(em!lifies the -;eductionist1s :ilemma-, In order to e(!lain everythin in terms of conte(t free sums, one has to o down to the level of !hysics= but then the number of !articles is so hu e as to make it only a theoretical -in!rinci!le- kind of thin & So, one has to settle for a conte(t-de!endent sum, which has two disadvanta es& The first is that the !arts are much lar er units, whose behavior is describable only on a hi h level, and therefore indeterminately& The second is that the word -sum- carries the connotation that each !art can be assi ned a sim!le function and that the function of the whole is .ust a conte(t-free sum of those individual functions& This .ust cannot be done when one tries to e(!lain a whole en<yme1s function, iven its amino acids as !arts& 0ut for better or for worse, this is a eneral !henomenon which arises in the e(!lanations of com!le( systems& In order to acquire an intuitive and mana eable understandin of how !arts interactin short, in order to !roceedone often has to sacrifice the e(actness yielded by a microsco!ic, conte(t-free !icture, sim!ly because of its unmana eability& 0ut one does not sacrifice at that time the faith that such an e(!lanation e(ists in !rinci!le&

Transfer $;A and $ibosomes

;eturnin , then, to ribosomes and ;8) and !roteins, we have stated that a !rotein is manufactured by a ribosome accordin to the blue!rint carried from the :8)1s -royal chambers- by its messen er, ;8)& This seems to im!ly that the ribosome can translate from the lan ua e of codons into the lan ua e of amino acids, which amounts to sayin that the ribosome -knows- the 6enetic 7ode& ?owever, that amount of information is sim!ly not !resent in a ribosome& So how does it do it? 3here is the 6enetic 7ode stored? The curious fact is that the 6enetic 7ode is storedwhere else?in the :8) itself& This certainly calls for some e(!lanation& Aet us back off from a total e(!lanation for a moment, and ive a !artial e(!lanation& There are, floatin about in the cyto!lasm at any iven moment, lar e numbers of four-leaf-clover-sha!ed molecules= loosely fastened 4i&e&, hydro en-bonded5 to one leaf is an amino acid, and on the o!!osite leaf there is a tri!let of nucleotides called an anticodon& 'or our !ur!oses, the other two leaves are irrelevant& ?ere is how these -clovers- are used by the ribosomes in their !roduction of !roteins& 3hen a new codon of m;8) clicks into !osition in the ribosome1s -!layin head-, the ribosome reaches out into the cyto!lasm and latches onto a clover whose anticodon is com!lementary to the m;8) codon& Then it !ulls the clover into such a !osition that it can ri! off the clover1s amino acid, and stick it covalently onto the rowin !rotein& 4Incidentally, the bond between an amino acid and its nei hbor in a !rotein is a very

73G5=E IJ! * section of m=N* passing through a ri(osome! 7loating near() are t=N* molecules, carr)ing amino acids which are stripped off () the ri(osome and appended to the growing protein! The Genetic 8ode is contained in the t=N* molecules, collecti#el)! Note how the (ase%pairing *%5, 8%G/ is represented () interlocking letter%forms in the diagram! U.rawing () -cott E! ;imV

stron covalent bond, called a -!e!tide bond-& 'or this reason, !roteins are sometimes called -!oly!e!tides-&5 Ef course it is no accident that the -clovers- carry the !ro!er amino acids, for they have all been manufactured accordin to !recise instructions emanatin from the -throne room-& The real name for such a clover is transfer ;8)& ) molecule of t;8) is quite smallabout the si<e of a very small !roteinand consists of a chain of about ei hty nucleotides& Aike m;8), t;8) molecules are made by transcription off of the rand cellular tem!late, :8)& ?owever, t;8)1s are tiny by com!arison with the hu e m;8) molecules, which may contain thousands of nucleotides in lon , lon chains& )lso, t;8)1s resemble !roteins 4and are unlike strands of m;8)5 in this res!ect, they have fi(ed, well-defined tertiary structuresdetermined by their !rimary structure& ) t;8) molecule1s tertiary structure allows !recisely one amino acid to bind to its amino-acid site, to be sure, it is that one dictated accordin to the 6enetic 7ode by the anticodon on the o!!osite arm& ) vivid ima e of the function of t;8) molecules is as flashcards

floatin in a cloud around a simultaneous inter!reter, who sna!s one out of the air invariably the ri ht one2whenever he needs to translate a word& In this case, the inter!reter is the ribosome, the words are codons, and their translations are amino acids& In order for the inner messa e of :8) to et decoded by the ribosomes, the t;8) flashcards must be floatin about in the cyto!lasm& In some sense, the t;8)1s contain the essence of the outer messa e of the :8), since they are the keys to the !rocess of translation& 0ut they themselves came from the :8)& Thus, the outer messa e is tryin to be !art of the inner messa e, in a way reminiscent of the messa e-in-a-bottle which tells what lan ua e it is written in& 8aturally, no such attem!t can be totally successful, there is no way for the :8) to hoist itself by its own bootstra!s& Some amount of knowled e of the 6enetic 7ode must already be !resent in the cell beforehand, to allow the manufacture of those en<ymes which transcribe t;8)1s themselves off of the master co!y of :8)& )nd this knowled e resides in !reviously manufactured t;8) molecules& This attem!t to obviate the need for any outer messa e at all is like the Escher dra on, who tries as hard as he can, within the conte(t of the two-dimensional world to which he is constrained, to be three-dimensional& ?e seems to o a lon waybut of course he never makes it, des!ite the fine imitation he ives of three-dimensionality&

,unctuation and the $eading +rame

?ow does a ribosome know when a !rotein is done? Just as in Ty!o enetics, there is a si nal inside the m;8) which indicates the termination or initiation of a !rotein& In fact, three s!ecial codonsF)), 7)6, F6) act as punctuation marks instead of codin for amino acids& 3henever such a tri!let clicks its way into the -readin head- of a ribosome, the ribosome releases the !rotein under construction and be ins a new one& ;ecently, the entire enome of the tiniest known virus, fT"$C, has been laid bare& Ene most une(!ected discovery was made en route, some of its nine enes overla!that is, two distinct proteins are coded for () the same stretch of .N* 2 There is even one ene contained entirely inside another2 This is accom!lished by havin the readin frames of the two enes shifted relative to each other, by e(actly one unit& The density of information !ackin in such a scheme is incredible& This is, of course, the ins!iration behind the stran e -LI"$ haiku- in )chilles1 fortune cookie, in the 8anon () 3nter#allic *ugmentation,

$eca!
In brief, then, this !icture emer es, from its central throne, :8) sends off lon strands of messen er ;8) to the ribosomes in the cyto!lasm= and the ribosomes, makin use of the -flashcards- of t;8) hoverin about them, efficiently construct !roteins, amino acid by amino acid, accordin to the blue!rint contained in the m;8)& Enly the !rimary structure of the !roteins is dictated by the :8)= but this is enou h, for as they emer e from the ribosomes, the !roteins -ma ically- fold u! into com!le( conformations which then have the ability to act as !owerful chemical machines&

Le"els of Structure and Meaning in ,roteins and Music

3e have been usin this ima e of ribosome as ta!e recorder, m;8) as ta!e, and !rotein as music& It may seem arbitrary, and yet there are some beautiful !arallels& /usic is not a mere linear sequence of notes& Eur minds !erceive !ieces of music on a level far hi her than that& 3e chunk notes into !hrases, !hrases into melodies, melodies into movements, and movements into full !ieces& Similarly, !roteins only make sense when they act as chunked units& )lthou h a !rimary structure carries all the information for the tertiary structure to be created, it still -feels- like less, for its !otential is only reali<ed when the tertiary structure is actually !hysically created& Incidentally, we have been referrin only to !rimary and tertiary structures, and you may well wonder whatever ha!!ened to the secondary structure& Indeed, it e(ists, as does a quaternary structure, as well& The foldin -u! of a !rotein occurs at more than one level& S!ecifically, at some !oints alon the chain of amino acids, there may be a tendency to form a kind of heli(, called the alpha heli9 4not to be confused with the :8) double heli(5& This helical twistin of a !rotein is on a lower level than its tertiary structure& This level of structure is visible in 'i ure #L& Muaternary structure can be directly com!ared with the buildin of 1a musical !iece out of inde!endent movements, for it involves the assembly of several distinct !oly!e!tides, already in their full-blown tertiary beauty, into a lar er structure& The bindin of these inde!endent chains is usually accom!lished by hydro en bonds, rather than covalent bonds= this is of course .ust as with !ieces of music com!osed of several movements, which are far less ti htly bound to each other than they are internally, but which nevertheless form a ti ht -or anic- whole& The four levels of !rimary, secondary, tertiary, and quaternary structure can also be com!ared to the four levels of the /F-!icture 4'i & D+5 in the &relude, *nt 7ugue& The lobal structureconsistin of the letters 1/1 and 1F1is its quaternary structure= then each of those two !arts has a tertiary structure, consistin of -?EAIS/- or -;E:F7TIE8IS/-= and then the o!!osite word e(ists on the secondary level, and at bottom, the !rimary structure is once a ain the word -/F-, over and over a ain&

,olyribosomes and T'o-Tiered Canons

8ow we come to another lovely !arallel between ta!e recorders translatin ta!e into music and ribosomes translatin m;8) into !roteins& Ima ine a collection of many ta!e recorders, arran ed in a row, evenly s!aced& 3e mi ht call this array a -!olyrecorder-& 8ow ima ine a sin le ta!e !assin serially throu h the !layin heads of all the com!onent recorders& If the ta!e contains a sin le lon melody, then the out!ut will be a many-voiced canon, of course, with the delay determined by the time it takes the ta!e to et from one ta!e recorder to the ne(t& In cells, such -molecular canons- do indeed e(ist, where many ribosomes, s!aced out in lon linesformin what is called a !olyribosome all -!lay- the same strand of m;8), !roducin identical !roteins, sta ered in time 4see 'i & #$5& 8ot only this, but nature oes one better& ;ecall that m;8) is made by transcri!tion off of :8)= the en<ymes which are res!onsible for this !rocess are called

73G5=E IC! * pol)ri(osome! * single strand of m=N* passes through one ri(osome after another, like one tape passing through se#eral tape recorders in a row! The result is a set of growing proteins in #arious stages of completion: the analogue of a musical canon produced () the staggered tape recorders! U7rom *! +ehninger, 0iochemistry&V

;8) !olymerases 4--ase- is a eneral suffi( for en<ymes5& It ha!!ens often that a series of ;8) !olymerases will be at work in !arallel on a sin le strand of :8), with the result that many se!arate 4but identical5 strands of m;8) are bein !roduced, each delayed with res!ect to the other by the time required for the :8) to slide from one ;8) !olymerase to the ne(t& )t the same time, there can be several different ribosomes workin on each of the !arallel emer in m;8)1s& Thus one arrives at a double-decker, or two-tiered, -molecular canon- 4'i & #*5& The corres!ondin ima e in music is a rather fanciful but amusin scenario, several different co!yists are all at work simultaneously,
73G5=E IE! ,ere, an e#en more comple9 scheme! Not Must one (ut se#eral strands of m=N*, all emerging () transcription from a single strand of .N*, are acted upon () pol)ri(osomes! The result is a two%tiered molecular canon! U7rom ,anawalt and ,a)nes, The 7hemical 0asis of Aife, p! FCBV

each one of them co!yin the same ori inal manuscri!t from a clef which flutists cannot read into a clef which they can read& )s each co!yist finishes a !a e of the ori inal manuscri!t, he !asses it on to the ne(t co!yist, and starts transcribin a new !a e himself& /eanwhile, from each score emer in from the !ens of the co!yists, a set of flutists are readin and tootin the melody, each flutist delayed with res!ect to the others who are readin from the same sheet& This rather wild ima e ives, !erha!s, an idea of some of the com!le(ity of the !rocesses which are oin on in each and every cell of your body durin every second of every day&&&

>hich Came +irstUThe $ibosome or the ,rotein7

3e have been talkin about these wonderful beasts called ribosomes= but what are they themselves com!osed of? ?ow are they made? ;ibosomes are com!osed of two ty!es of thin s, 4"5 various kinds of !roteins, and 4B5 another kind of ;8), called ri(osomal =N* 4r;8)5& Thus, in order for a ribosome to be made, certain kinds of !roteins must be !resent, and r;8) must be !resent& Ef course, for !roteins to be !resent, ribosomes must be there to make them& So how do you et around the vicious circle? 3hich comes first the ribosome or the !rotein? 3hich makes which? Ef course there is no answer because one always traces thin s back to !revious members of the same class .ust as with the

chicken-and-the-e questionuntil everythin vanishes over the hori<on of time& In any case, ribosomes are made of two !ieces, a lar e and a small one, each of which contains some r;8) and some !roteins& ;ibosomes are about the si<e of lar e !roteins= they are much much smaller than the strands of m;8) which they take as in!ut, and alon which they move&

,rotein +unction
3e have s!oken somewhat of the structure of !roteinss!ecifically en<ymes= but we have not really mentioned the kinds of tasks which they !erform in the cell, nor how they do them& )ll en<ymes are catal)sts, which means that in a certain sense, they do no more than selecti#el) accelerate various chemical !rocesses in the cell, rather than make thin s ha!!en which without them never could ha!!en& )n en<yme reali<es certain !athways out of the myriad myriad !otentialities& Therefore, in choosin which en<ymes shall be !resent, you choose what shall ha!!en and what shall not ha!!endes!ite the fact that, theoretically s!eakin , there is a non<ero !robability for any cellular !rocess to ha!!en s!ontaneously, without the aid of catalysts& 8ow how do en<ymes act u!on the molecules of the cell? )s has been mentioned, en<ymes are folded-u! !oly!e!tide chains& In every en<yme, there is a cleft or !ocket or some other clearly-defined surface feature where the en<yme hinds to some other kind of molecule& This location is called its acti#e site, and any molecule which ets bound there is called a su(strate& En<ymes may have more than one active site, and more than one substrate& Just as in Ty!o enetics, en<ymes are indeed very choosy about what they will o!erate u!on& The active site usually is quite s!ecific, and allows .ust one kind of molecule to bind to it, althou h there are sometimes -decoys-other molecules which can fit in the active site and clo it u!, foolin the en<yme and in fact renderin it inactive& Ence an en<yme and its substrate are bound to ether, there is some disequilibrium of electric char e, and consequently char ein the form of electrons and !rotonsflows around the bound molecules and read.usts itself& 0y the time equilibrium has been reached, some rather !rofound chemical chan es may have occurred to the substrate& Some e(am!les are these, there may have been a -weldin -, in which some standard small molecule ot tacked onto a nucleotide, amino acid, or other common cellular molecule= a :8) strand may have been -nicked- at a !articular location= some !iece of a molecule may have otten lo!!ed off= and so forth& In fact, bio-en<ymes do o!erations on molecules which are quite similar to the ty!o ra!hical o!erations which Ty!o-en<ymes !erform& ?owever, most en<ymes !erform essentially only a sin le task, rather than a sequence of tasks& There is one other strikin difference between Ty!o-en<ymes and bioen<ymes, which is this, whereas Ty!o-en<ymes o!erate only on strands, bio-en<ymes can act on :8), ;8), other !roteins, ribosomes, cell membranesin short, on anythin and everythin in the cell& In other words, en<ymes are the universal mechanisms for ettin thin s done in the cell& There are en<ymes which stick thin s to ether and take them a!art and modify them and activate them and deactivate them and co!y them and re!air them and destroy them&&&

Some of the most com!le( !rocesses in the cell involve -cascades- in which a sin le molecule of some ty!e tri ers the !roduction of a certain kind of en<yme= the manufacturin !rocess be ins and the en<ymes which come off the -assembly line- o!en u! a new chemical !athway which allows a second kind of en<yme to be !roduced& This kind of thin can o on for three or four levels, each newly !roduced ty!e of en<yme tri erin the !roduction of another ty!e& In the end a -shower- of co!ies of the final ty!e of en<yme is !roduced, and all of the co!ies o off and do their s!eciali<ed thin , which may be to cho! u! some -forei n- :8), or to hel! make some amino acid for which the cell is very -thirsty-, or whatever&

;eed for a Sufficiently Strong Su!!ort System

Aet us describe nature1s solution to the !u<<le !osed for Ty!o enetics, -3hat kind of strand of :8) can direct its own re!lication?- 7ertainly not every strand of :8) is inherently a self-re!& The key !oint is this, any strand which wishes to direct its own co!yin must contain directions for assemblin !recisely those en<ymes which can carry out the task& 8ow it is futile to ho!e that a strand of :8) in isolation could be a self-re!= for in order for those !otential !roteins to be !ulled out of the :8), there must not only be ribosomes, but also ;8) !olymerase, which makes the m;8) that ets trans!orted to the ribosomes& )nd so we have to be in by assumin a kind of -minimal su!!ort system.ust sufficiently stron that it allows transcri!tion and translation to be carried out& This minimal su!!ort system will thus consist in 4"5 some !roteins, such as ;8) !olymerase, which allow m;8) to be made from :8), and 4B5 some ribosomes&

/o'

;A Self-$e!licates

It is not by any means coincidental that the !hrases -sufficiently stron su!!ort systemand -sufficiently !owerful formal system- sound alike& Ene is the !recondition for a selfre! to arise, the other for a self-ref to arise& In fact there is in essence only one !henomenon oin on in two very different uises, and we shall e(!licitly ma! this out shortly& 0ut before we do so, let us finish the descri!tion of how a strand of :8) can be a self-re!& The :8) must contain the codes for a set of !roteins which will co!y it& 8ow there is a very efficient and ele ant way to co!y a double-stranded !iece of :8), whose two strands are com!lementary& This involves two ste!s, 4"5 unravel the two strands from each other= 4B5 -mate- a new strand to each of the two new sin le strands& This !rocess will create two new double strands of :8), each identical to the ori inal one& 8ow if our solution is to be based on this idea, it must involve a set of !roteins, coded for in the :8) itself, which will carry out these two ste!s& It is believed that in cells, these two ste!s are !erformed to ether in a coordinated way, and that they require three !rinci!al en<ymes, :8) endonuclease, :8)

!olymerase, and :8) li ase& The first is an -un<i!!in en<yme-, it !eels the two ori inal strands a!art for a short distance, and then sto!s& Then the other two en<ymes come into the !icture& The :8) !olymerase is basically a co!y-and-move en<yme, it chu s down the short sin le strands of :8), co!yin them com!lementarily in a fashion reminiscent of the 7o!y mode in Ty!o enetics& In order to co!y, it draws on raw materials s!ecifically nucleotideswhich are floatin about in the cyto!lasm& 0ecause the action !roceeds in fits and starts, with some un<i!!in and some co!yin each time, some short a!s are created, and the :8) li ase is what !lu s them u!& The !rocess is re!eated over and over a ain& This !recision three-en<yme machine !roceeds in careful fashion all the way down the len th of the :8) molecule, until the whole thin has been !eeled a!art and simultaneously re!licated, so that there are now two co!ies of it&

Com!arison of

;A%s Self-$e! Method 'ith 8uining

8ote that in the en<ymatic action on the :8) strands, the fact that information is stored in the :8) is .ust !lain irrelevant= the en<ymes are merely carryin out their symbolshuntin functions, .ust like rules of inference in the /IF-system& It is of no interest to the three en<ymes that at some !oint they are actually co!yin the very enes which coded for them& The :8), to them, is .ust a tem!late without meanin or interest& It is quite interestin to com!are this with the Muine sentence1s method of describin how to construct a co!y of itself& There, too, one has a sort of -double strand-two co!ies of the same information, where one co!y acts as instructions, the other as tem!late& In :8), the !rocess is va uely !arallel, since the three en<ymes 4:8) endonuclease, :8) !olymerase, :8) li ase5 are coded for in .ust one of the two strands, which therefore acts as !ro ram, while the other strand is merely a tem!late& The !arallel is not !erfect, for when the co!yin is carried out, both strands are used as tem!late, not .ust one& 8evertheless, the analo y is hi hly su estive& There is a biochemical analo ue to the use-mention dichotomy, when :8) is treated as a mere sequence of chemicals to be co!ied, it is like mention of ty!o ra!hical symbols= when :8) is dictatin what o!erations shall he carried out, it is like use of ty!o ra!hical symbols&

Le"els of Meaning of

;A

There are several levels of meanin which can be read from a strand of :8), de!endin on how bi the chunks are which you look at, and how !owerful a decoder you use& En the lowest level, each :8) strand codes for an equivalent ;8) strandthe !rocess of decodin bein transcription& If one chunks the :8) into tri!lets, then by usin a - enetic decoder-, one can read the :8) as a sequence of amino acids& This is translation 4on to! of transcri!tion5& En the ne(t natural level of the hierarchy, :8) is readable as a code for a set of !roteins& The !hysical !ullin -out of !roteins from enes is called gene e9pression& 7urrently, this is the hi hest level at which we understand what :8) means& ?owever, there are certain to be hi her levels of :8) meanin which are harder to discern& 'or instance, there is every reason to believe that the :8) of, say, a human

bein codes for such features as nose sha!e, music talent, quickness of refle(es, and so on& 7ould one, in !rinci!le, learn to read off such !ieces of information directly from a strand of :8), without oin throu h the actual !hysical !rocess of epigenesisthe !hysical !ullin -out of !henoty!e from enoty!e? Presumably, yes, sincein theory one could have an incredibly !owerful com!uter !ro ram simulatin the entire !rocess, includin every cell, every !rotein, every tiny feature involved in the re!lication of :8), of cells, to the bitter end& The out!ut of such a pseudo%epigenesis !ro ram would be a hi h-level descri!tion of the !henoty!e& There is another 4e(tremely faint5 !ossibility, that we could learn to read the !henoty!e off of the enoty!e without doin an isomor!hic simulation of the !hysical !rocess of e!i enesis, but by findin some sim!ler sort of decodin mechanism& This could be called -shortcut !seudo-e!i enesis-& 3hether shortcut or not, !seudo-e!i enesis is, of course, totally beyond reach at the !resent timewith one notable e(ce!tion, in the s!ecies 7elis catus, dee! !robin has revealed that it is indeed !ossible to read the !henoty!e directly off of the enoty!e& The reader will !erha!s better a!!reciate this remarkable fact after directly e(aminin the followin ty!ical section of the :8) of 7elis catus, S 7)T7)T7)T7)T7)T7)T7)T7)T7)T7)T&&& 0elow is shown a summary of the levels of :8)-readability, to ether with the names of the different levels of decodin & :8) can be read as a sequence of, 4"5 bases 4nucleotides5 & & & & & & & & & & & & & & & & & & & & transcription 4B5 amino acids & & & & & & & & & & & & & & & & & & & & & & & & & translation 4%5 !roteins 4!rimary structure5 ! ! ! ! ! ! ! ! ! ! ! !gene e9pression 4C5 !roteins 4tertiary structure5 4L5 !rotein clusters & & & & & & & & & & & & & & & & & & & & & & & higher le#els of gene e9pression S& S& & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & & unknown le#els of .N* meaning S& 4N-"5 $$$ 4N5 !hysical, mental, and !sycholo ical traits & & & & & pseudo%epigenesis

The Central

ogma!

3ith this back round, now we are in a !osition to draw an elaborate com!arison between '& 7rick1s -7entral :o ma of /olecular 0iolo y- 4&:E6/) I5 u!on which all cellular !rocesses are based= and what I, with !oetic license, call the -7entral :o ma of /athematical Ao ic- 4&:E6/) II5, u!on which 6>del1s Theorem is based& The ma!!in from one onto the other is laid out in 'i ure ## and the followin chart, which to ether constitute the 8entral .ogmap&
73G5=E II! The 8entral .ogmap! *n analog) is esta(lished (etween two fundamental Tangled ,ierarchies: that of molecular (iolog) and that of mathematical logic!

&:E6/) I 4/olecular 0iolo y5 strands of :8) strands of m;8) !roteins !roteins which act on !roteins !roteins which act on !roteins which act on !roteins transcri!tion 4:8) ;8)5 Translation 4;8) !roteins5 7rick 6enetic code 4arbitrary convention5 codon 4tri!let of bases5 amino acid self-re!roduction sufficiently stron cellular su!!ort system to !ermit self-re!

&:E6/) II 4/athematical Ao ic5 strin s of T8T statements of 8 statements of meta-T8T statements about statements of meta-T8T statements about statements about statements of meta-T8T inter!retation 4T8T 85 )rithmeti<ation 4N meta-T8T5 6>del 6>del code 4arbitrary convention5 codon 4tri!let of di its5 quoted symbol of T8T used in meta-T8T self-reference sufficiently !owerful arithmetical formal system to !ermit self-ref

7entral :o ma!

8ote the base-!airin of ) and T 4)rithmeti<ation and Translation5, as well as of 6 and 7 46>del and 7rick5& /athematical lo ic ets the !urine side, and molecular biolo y ets the !yrimidine side& To com!lete the esthetic side of this ma!!in , I chose to model my 6>delnumberin scheme on the 6enetic 7ode absolutely faithfully& In fact, under the followin corres!ondence, the table of the 6enetic 7ode becomes the table of the 6>del 7ode, 4odd5 4even5 4odd5 4even5 " B % D
) 7 6 F 4!urine5 4!yrimidine5 4!urine5 4!yrimidine5

Each amino acidof which there are twentycorres!onds to e(actly one symbol of T8Tof which there are twenty& Thus, at last, my motive for concoctin -austere T8Tcomes outso that there would be e(actly twenty symbols2 The 6>del 7ode is shown in 'i ure "++& 7om!are it with the 6enetic 7ode 4'i & #C5& There is somethin almost mystical in seein the dee! sharin of such an abstract structure by these two esoteric, yet fundamental, advances in knowled e achieved in our century& This 7entral :o ma! is by no means a ri orous !roof of identity of the two theories= but it clearly shows a !rofound kinshi!, which is worth dee!er e(!loration&

Strange Loo!s in the Central

ogma!

Ene of the more interestin similarities between the two sides of the ma! is the way in which -loo!s- of arbitrary com!le(ity arise on the to! level of both, on the left, !roteins which act on !roteins which act on !roteins and so on, ad infinitum= and on the ri ht, statements about statements about statements of meta-T8T and so on, ad infinitum& These are like heterarchies, which we discussed in 7ha!ter G, where a sufficiently com!le( substratum allows hi h-level Stran e Aoo!s to occur and to cycle around, totally sealed off from lower levels& 3e will e(!lore this idea in reater detail in 7ha!ter TT& Incidentally, you may be wonderin about this question, -3hat, accordin to the 7entral :o ma!, is 6>del1s Incom!leteness Theorem itself ma!!ed onto?- This is a ood question to think about before readin ahead&

The Central

ogma! and the Contracrosti!unctus

It turns out that the central do ma! is quite similar to the ma!!in that was laid out in 7ha!ter IG between the 8ontracrostipunctus and 6>del1s Theorem& Ene can therefore draw !arallels between all three systems&

73G5=E BKK! The Gdel 8ode! 5nder this Gdel%num(ering scheme, each TNT s)m(ol gets one or more codons! The small o#als show how this ta(le su(sumes the earlier Gdel%num(ering ta(le of 8hapter 3S!

4"5 formal systems and strin s 4B5 cells and strands of :8) 4%5 record !layers and records In the followin chart, the ma!!in between systems B and % is e(!lained carefully&

8ontracrostipunctus !hono ra!h -Perfect- !hono ra!h record record !layable by a iven !hono ra!h record un!layable by that !hono ra!h !rocess of convertin record rooves into sound sounds !roduced by record !layer translation of sounds into vibration of !hono ra!h ma!!in from e(ternal sounds onto vibrations of !hono ra!h breakin of !hono ra!h Title of son s!ecially tailored for ;ecord Player T, -I 7annot 0e Played on ;ecord Player T-Im!erfect- ;ecord Player

-T>del1s Theorem-, -There always e(ists an un!layable record, iven a !articular !hono ra!h&-

/olecular 0iolo y cell -Perfect- cell strand of :8) strand of :8) re!roducible by a iven cell strand of :8) unre!roducible by that cell !rocess of transcri!tion of :8) onto m;8) strands of messen er ;8) translation of m;8) into !roteins 6enetic 7ode 4ma!!in from m;8) tri!lets onto amino acids5 destruction of the cell ?i h-level inter!retation of :8) strand s!ecially tailored for 7ell T, -I 7annot 0e ;e!licated by 7ell T7ell for which there e(ists at least one :8) strand which it cannot re!roduce Immunity Theorem, -There always e(ists an unre!roducible :8) strand, iven a !articular cell&-

The analo ue of 6>del1s Theorem is seen to be a !eculiar fact, !robably little useful to molecular biolo ists 4to whom it is likely quite obvious5, It is always !ossible to desi n a strand of :8) which, if in.ected into a cell, would, u!on bein transcribed, cause such !roteins to be manufactured as would destroy the cell 4or the :8)5, and thus result in the non-re!roduction of that :8)& This con.ures ti! a somewhat droll scenario, at least if taken in li ht of evolution, an invadin s!ecies of virus enters a cell by some surre!titious means, and then carefully ensures the manufacture of !roteins which will have the effect of destroyin the virus itself2 It is a sort of suicideor E!imenides sentence, if you willon the molecular level& Ebviously it would not !rove advanta eous from the !oint of view of survival of the s!ecies& ?owever, it demonstrates the s!irit, if not the letter, of the mechanisms of !rotection and subversion which cells and their invaders have develo!ed&

73G5=E BKB! The TD (acterial #irus is an assem(l) of protein components a/! The "head" is a protein mem(rane, shaped like a kind of prolate icosahedron with thirt) facets and filled with .N*! 3t is attached () a neck to a tail consisting of a hollow core surrounded () a contractile sheath and (ased on a spiked end plate to which si9 fi(ers are attached! The spikes and fi(ers affi9 the #irus to a (acterial cell wall (/! The sheath contracts, dri#ing the core through the wall, and #iral .N* enters the cell! U7rom ,anawalt and ,a)nes, The 7hemical 0asis of Aife, p! FHK!B

E1 Coli #s! TL
Aet us consider the biolo ists1 favorite cell, that of the bacterium Escherichia coli 4no relation to /& 7& Escher5, and one of their favorite invaders of that cell, the sinister and eerie TC !ha e, !ictures of which you can see in 'i ure "+"& 4Incidentally, the words -!ha e- and -virus- are synonymous and mean -attacker of bacterial cells-&5 The weird tidbit looks like a little like a cross between a AE/ 4Aunar E(cursion /odule5 and a mosquitoand it is much more sinister than the latter& It has a -head- wherein is stored all its -knowled e-namely its :8)= and it has si( -le s- wherewith to fasten itself to the cell it has chosen to invade= and it has a -stin in tube- 4more !ro!erly called its -tail-5 like a mosquito& The ma.or difference is that unlike a mosquito, which uses its stin er for suckin blood, the TC !ha e uses its stin er for in.ectin its hereditary substance into the cell a ainst the will of its victim& Thus the !ha e commits -ra!e- on a tiny scale&

73G5=E BKF! <iral infection (egins when #iral .N* enters a (acterium! Bacterial .N* is disrupted and #iral .N* replicated! -)nthesis of #iral structural proteins and their assem(l) into #irus continues until the cell (ursts, releasing particles! U7rom ,anawalt and ,a)nes, The 7hemical 0asis of Aife, p! FHK!V

A Molecular TroVan /orse

3hat actually ha!!ens when the viral :8) enters a cell? The virus -ho!es-, to s!eak anthro!omor!hically, that its :8) will et e(actly the same treatment as the :8) of the host cell& This would mean ettin transcribed and translated, thus allowin it to direct the synthesis of its own s!ecial !roteins, alien to the host cell, which will then be in to do their thin & This amounts to secretly trans!ortin alien !roteins -in code- 4vi<&, the 6enetic 7ode5 into the cell, and then -decodin - 4vi<&, !roducin 5 them& In a way this resembles the story of the Tro.an horse, accordin to which hundreds of soldiers were sneaked into Troy inside a harmless seemin iant wooden horse= but once inside the city, they broke loose and ca!tured it& The alien !roteins, once they have been -decoded4synthesi<ed5 from their carrier :8), now .um! into action& The sequence of actions directed by the TC !ha e has been carefully studied, and is more or less as follows 4see also 'i s& "+B and "+%5,
Time elapsed *ction taking place

+ min& " min&

L min& * min& "% min& BL min&

In.ection of viral :8)& 0reakdown of host :8)& 7essation of !roduction of native !roteins and initiation of !roduction of alien 4TC5 !roteins& )mon the earliest !roduced !roteins are those which direct the re!lication of the alien 4TC5 :8)& ;e!lication of viral :8) be ins& Initiation of !roduction of structural !roteins which will form the -bodiesof new !ha es& 'irst com!lete re!lica of TC invader is !roduced& Ayso<yme 4a !rotein5 attacks host cell wall& breakin o!en the bacterium, and the -bicentu!lets- emer e&

73G5=E BKH! The morphogenic path% wa) of the TD #irus has three prin% cipal (ranches leading independentl) to the formation of heads, tails and tail fi(ers, which then com(ine to form complete #irus particles! U7rom ,anawalt and ,a)nes, The 8hemical Basis of +ife, p! FHCV

Thus, when a TC !ha e invades an E! coli cell, after the brief s!an of about twenty-four or twenty-five minutes, the cell has been com!letely subverted, and breaks o!en& Eut !o! about two hundred e(act co!ies of the ori inal virus-bicentu!lets-ready to o attack more bacterial cells, the ori inal cell havin been lar ely consumed in the !rocess& )lthou h from a bacterium1s !oint of view this kind of thin is a deadly serious menace, from our lar e-scale vanta e !oint it can be looked u!on as an amusin ame between two !layers, the invader, or -T- !layer 4named after the T-even class of !ha es, includin the TB, TC, and others5, and the -7- !layer 4standin for -7ell-5& The ob.ective of the T !layer is to invade and take over the cell of the 7 !layer from within, for the !ur!ose of re!roducin itself& The ob.ective of the 7 !layer is to !rotect itself and destroy the invader& 3hen described this way, the molecular T7- ame can be seen to be quite !arallel to the macrosco!ic T7- ame described in the !recedin :ialo ue& 4The reader can doubtless fi ure out which !layerT or 7corres!onds to the Tortoise, and which to the 7rab&5

$ecognition)

isguises) Labeling

This - ame- em!hasi<es the fact that reco nition is one of the central themes of cellular and subcellular biolo y& ?ow do molecules 4or hi her-level structures5 reco ni<e each other? It is essential for the functionin of en<ymes that they should be able to latch onto s!ecial -bindin sites- on their substrates= it is essential that a bacterium should be able to distin uish its own :8) from that of !ha es= it is essential that two cells should be able to reco ni<e each other and interact in a controlled way& Such reco nition !roblems may remind you of the ori inal, key !roblem about formal systems, ?ow can you tell if a strin has, or does not have, some !ro!erty such as theoremhood? Is there a decision !rocedure? This kind of question is not restricted to mathematical lo ic, it !ermeates com!uter science and, as we are seein , molecular biolo y& The labelin technique described in the :ialo ue is in fact one of E! coli1s tricks for outwittin its !ha e invaders& The idea is that strands of :8) can be chemically labeled by tackin on a small moleculemethylto various nucleotides& 8ow this labelin o!eration does not chan e the usual biolo ical !ro!erties of the :8)= in other words, methylated 4labeled5 :8) can be transcribed .ust as well as unmethylated 4unlabeled5 :8), and so it can direct the synthesis of !roteins& 0ut if the host cell has some s!ecial mechanisms for e(aminin whether :8) is labeled or not, then the label may make all the difference in the world& In !articular, the host cell may have an en<yme system which looks for unlabeled :8), and destroys any that it finds by unmercifully cho!!in it to !ieces& In that case, woe to all unlabeled invaders2 The methyl labels on the nucleotides have been com!ared to serifs on letters& Thus, usin this meta!hor, we could say that the E! coli cell is lookin for :8) written in its -home scri!t-, with its own !articular ty!efaceand will cho! u! any strand of :8) written in an -alien- ty!eface& Ene counterstrate y, of course, is for !ha es to learn to label themselves, and thereby become able to fool the cells which they are invadin into re!roducin them&

This T7-battle can continue to arbitrary levels of com!le(ity, but we shall not !ursue it further& The essential fact is that it is a battle between a host which is tryin to re.ect all invadin :8), and a !ha e which is tryin to infiltrate its :8) into some host which will transcribe it into m;8) 4after which its re!roduction is uaranteed5& )ny !ha e :8) which succeeds in ettin itself re!roduced this way can be thou ht of as havin this hi h-level inter!retation, -I 8an 0e ;e!roduced in 7ells of Ty!e T-& This is to be distin uished from the evolutionarily !ointless kind of !ha e mentioned earlier, which codes for !roteins that destroy it, and whose hi h-level inter!retation is the selfdefeatin sentence, -I 8annot 0e ;e!roduced in 7ells of Ty!e T-&

/en.in Sentences and 4iruses

8ow both of these contrastin ty!es of self-reference in molecular biolo y have their counter!arts in mathematical lo ic& 3e have already discussed the analo ue of the selfdefeatin !ha esnamely, strin s of the 6>del ty!e, which assert their own un!roducibility within s!ecific formal systems& 0ut one can also make a counter!art sentence to a real !ha e, the !ha e asserts its own !roducibility in a s!ecific cell, and the sentence asserts its own !roducibility in a s!ecific formal system& Sentences of this ty!e are called ?enkin sentences, after the mathematical lo ician Aeon ?enkin& They can be constructed e(actly alon the lines of 6>del sentences, the only difference bein the omission of a ne ation& Ene be ins with an -uncle-, of course, a5a%5DT;T-,$00+-,AI$Ja)a%KA$IT/M08&I;EJa%%)a%KR and then !roceeds by the standard trick& Say the 6>del number of the above -uncle- is h& 8ow by arithmoquinin this very uncle, you et a ?enkin sentence, a5a%5DT;T-,$00+-,AI$Ja)a%KA$IT/M08&I;EJSSS1111SSS?Sa%%)a%KR

40y the way, can you s!ot how this sentence differs from h6?5 The reason I show it e(!licitly is to !oint out that a ?enkin sentence does not ive a full reci!e for its own derivation= it .ust asserts that there e(ists one& Hou mi ht well wonder whether its claim is .ustified& :o ?enkin sentences indeed !ossess derivations? )re they, as they claim, theorems? It is useful to recall that one need not believe a !olitician who says& -I am honest-he may be honest, and yet he may not be& )re ?enkin sentences any more trustworthy than !oliticians? Er do ?enkin sentences, like !oliticians, lie in cast-iron sinks? It turns out that these ?enkin sentences are invariably truth tellers& 3hy this is so is not obvious= but we will acce!t this curious fact without !roof&

b
h1 S1s

Im!licit #s! E-!licit /en.in Sentences

I mentioned that a ?enkin sentence tells nothin about its own derivation= it .ust asserts that one e(ists& 8ow it is !ossible to invent a variation on the theme of ?enkin sentences namely sentences which e9plicitl) descri(e their own derivations& Such a sentence1s hi h-level inter!retation would not be -Some Sequence of Strin s E(ists 3hich is a :erivation of /e-, but rather, -The ?erein-described Sequence of Strin s Is a :erivation of /e-& Aet us call the first ty!e of sentence an implicit ?enkin sentence& The new sentences will be called e9plicit ?enkin sentences, since they e(!licitly describe their own derivations& 8ote that, unlike their im!licit brethren, e(!licit ?enkin sentences need not (e theorems& In fact, it is quite easy to write a strin which asserts that its own derivation consists of the sin le strin ?E?a false statement, since ?E? is not a derivation of anythin & ?owever, it is also !ossible to write an e(!licit ?enkin sentence which is a theoremthat is, a sentence which in fact ives a reci!e for its own derivation&

/en.in Sentences and Self-Assembly

The reason I brin u! this distinction between e(!licit and im!licit ?enkin sentences is that it corres!onds very nicely to a si nificant distinction between ty!es of virus& There are certain viruses, such as the so-called -tobacco mosaic virus-, which are called self% assem(ling viruses= and then there are others, such as our favorite T-evens, which are non%self%assem(ling& 8ow what is this distinction? It is a direct analo ue to the distinction between im!licit and e(!licit ?enkin sentences& The :8) of a self-assemblin virus codes only for the !arts of a new virus, but not for any en<ymes& Ence the !arts are !roduced, the sneaky virus relies u!on them to link u! to each other without hel! from any en<ymes& Such a !rocess de!ends on chemical affinities which the !arts have for each other, when swimmin in the rich chemical brew of a cell& 8ot only viruses, but also some or anellessuch as ribosomes assemble themselves& Sometimes en<ymes may be neededbut in such cases, they are recruited from the host cell, and enslaved& This is what is meant by self-assembly& 0y contrast, the :8) of more com!le( viruses, such as the T-evens, codes not only for the !arts, but in addition for various en<ymes which !lay s!ecial roles in the assembly of the !arts into wholes& Since the assembly !rocess is not s!ontaneous but requires -machines-, such viruses are not considered to be self-assemblin & The essence of the distinction, then, between self-assemblin units and non-self-assemblin units is that the former et away with self-re!roduction without tellin the cell anythin about their construction, while the latter need to ive instructions as to how to assemble themselves& 8ow the !arallel to ?enkin sentences, im!licit and e(!licit, ou ht to be quite clear& Im!licit ?enkin sentences are self-!rovin but do not tell anythin at all about their !roofsthey are analo ous to self-assemblin viruses= e(!licit ?enkin sentences direct the construction of their own !roofsthey are analo ous to more com!le( viruses which direct their host cells in !uttin co!ies of themselves to ether&

The conce!t of self-assemblin biolo ical structures as com!le( as viruses raises the !ossibility of com!le( self-assemblin machines as well& Ima ine a set of !arts which, when !laced in the !ro!er su!!ortin environment, s!ontaneously rou! themselves in such a way as to form a com!le( machine& It seems unlikely, yet this is quite an accurate way to describe the !rocess of the tobacco mosaic virus1 method of selfre!roduction via self-assembly& The information for the total conformation of the or anism 4or machine5 is s!read about in its !arts= it is not concentrated in some sin le !lace& 8ow this conce!t can lead in some stran e directions, as was shown in the Edif)ing Thoughts of a To(acco -moker& There, we saw how the 7rab used the idea that information for self-assembly can be distributed around, instead of bein concentrated in a sin le !lace& ?is ho!e was that this would !revent his new !hono ra!hs from succumbin to the Tortoise1s !hono ra!h-crashin method& Fnfortunately, .ust as with the most so!histicated a(iom schemata, once the system is all built and !acka ed into a bo(, its well-definedness renders it vulnerable to a sufficiently clever -6>deli<er-= and that was the sad tale related by the 7rab& :es!ite its a!!arent absurdity, the fantastic scenario of that :ialo ue is not so far from reality, in the stran e, surreal world of the cell&

T'o 0utstanding ,roblems5

ifferentiation and Mor!hogenesis

8ow self-assembly may be the trick whereby certain subunits of cells are constructed, and certain virusesbut what of the most com!le( macrosco!ic structures, such as the body of an ele!hant or a s!ider, or the sha!e of a Genus1s-flytra!? ?ow are homin instincts built into the brain of a bird, or huntin instincts into the brain of a do ? In short, how is it that merely by dictatin which !roteins are to be !roduced in cells, :8) e(ercises such s!ectacularly !recise control over the e(act structure and function of macrosco!ic livin ob.ects? There are two ma.or distinct !roblems here& Ene is that of cellular differentiation, how do different cells, sharin e(actly the same :8), !erform different rolessuch as a kidney cell, a bone marrow cell, and a brain cell? The other is that of morphogenesis 4-birth of form-5, how does intercellular communication on a local level ive rise to lar e-scale, lobal structures and or ani<ationssuch as the various or ans of the body, the sha!e of the face, the subor ans of the brain, and so on? )lthou h both cellular differentiation and mor!ho enesis are !oorly understood at !resent& the trick a!!ears to reside in e(quisitely fine-tuned feedback and -feedforward- mechanisms within cells and between cells, which tell a cell when to -turn on- and when to -turn off!roduction of various !roteins&

+eedbac. and +eedfor'ard

'eedback takes !lace when there is too much or too little of some desired substance in the cell, then the cell must somehow re ulate the !roduction line which is assemblin that substance& 'eedforward also involves the re ulation of an assembly line, but not accordin to the amount of end !roduct !resent, rather, accordin to the amount of some

precursor of the end !roduct of that assembly line& There are two ma.or devices for achievin ne ative feedforward or feedback& Ene way is to !revent the relevant en<ymes from bein able to !erformthat is, to -clo u!- their active sites& This is called inhi(ition& The other way is to !revent the relevant en<ymes from ever bein manufactured2 This is called repression& 7once!tually, inhibition is sim!le, you .ust block u! the active site of the first en<yme in the assembly line, and the whole !rocess of synthesis ets sto!!ed dead&

$e!ressors and Inducers

;e!ression is trickier& ?ow does a cell sto! a ene from bein e(!ressed? The answer is, it !revents it from ever ettin transcribed& This means that it has to !revent ;8) !olymerase from doin its .ob& This can be accom!lished by !lacin a hu e obstacle in its !ath, alon the :8), !recisely in front of that ene which the cell wants not to et transcribed& Such obstacles do e(ist, and are called repressors& They are themselves !roteins, and they bind to s!ecial obstacle-holdin sites on the :8), called 4I am not sure why5 o!erators& )n o!erator therefore is a site of control for the ene 4or enes5 which immediately follow it, those enes are called its o!eron& 0ecause a series of en<ymes often act in concert in carryin out a lon chemical transformation, they are often coded for in sequence= and this is why o!erons often contain several enes, rather than .ust one& The effect of the successful re!ression of an o!eron is that a whole series of enes is !revented from bein transcribed, which means that a whole set of related en<ymes remains unsynthesi<ed& 3hat about positi#e feedback and feedforward? ?ere a ain, there are two o!tions, 4"5 unclo the clo ed en<ymes, or 4B5 sto! the re!ression of the relevant o!eron& 48otice how nature seems to love double-ne ations2 Probably there is some very dee! reason for this&5 The mechanism by which re!ression is re!ressed involves a class of molecules called inducers& The role of an inducer is sim!le, it combines with a re!ressor !rotein before the latter has had a chance to bind to an o!erator on a :8) molecule= the resultin -re!ressor-inducer com!le(- is inca!able of bindin to an o!erator, and this leaves the door o!en for the associated o!eron to be transcribed into m;8) and subsequently translated into !rotein& Eften the end !roduct or some !recursor of the end !roduct can act as an inducer&

+eedbac. and Strange Loo!s Com!ared

Incidentally, this is a ood time to distin uish between sim!le kinds of feedback, as in the !rocesses of inhibition and re!ression, and the loo!in -hack between different informational levels, shown in the 7entral :o ma!& 0oth are -feedback- in some sense= but the latter is much dee!er than the former& 3hen an amino acid, such as try!to!han or isoleucine, acts as feedback 4in the form of an inducer5 by bindin to its re!ressor so that more of it ets made, it is not tellin how to construct itself= it is .ust tellin en<ymes to make more of it& This could be com!ared to a radio1s volume, which, when fed throu h a listener1s ears, may cause itself to be turned down or u!& This is another thin entirely

from the case in which the broadcast itself tells you e(!licitly to turn your radio on or off, or to tune to another wavelen thor even how to build another radio2 The latter is much more like the loo!in -back between informational levels, for here, information inside the radio si nal ets -decoded- and translated into mental structures& The radio si nal is com!osed of symbolic constituents whose symbolic meanin mattersa case of use, rather than mention& En the other hand, when the sound is .ust too loud, the symbols are not conveyin meanin , they are merely bein !erceived as loud sounds, and mi ht as well be devoid of meanin a case of mention, rather than use& This case more resembles the feedback loo!s by which !roteins re ulate their own rates of synthesis& It has been theori<ed that the difference between two nei hborin cells which share the e(act same enoty!e and yet have different functions is that different se ments of their enome have been re!ressed, and therefore they have different working sets of !roteins& ) hy!othesis like this could account for the !henomenal differences between cells in different or ans of the body of a human bein &

T'o Sim!le E-am!les of

ifferentiation

The !rocess by which one initial cell re!licates over and over, ivin rise to a myriad of differentiated cells with s!eciali<ed functions, can be likened to the s!read of a chain letter from !erson to !erson, in which each new !artici!ant is asked to !ro!a ate the messa e faithfully, but also to add some e(tra !ersonal touch& Eventually, there will be letters which are tremendously different from each other& )nother illustration of the ideas of differentiation is !rovided by this e(tremely sim!le com!uter analo ue of a differentiatin self-re!& 7onsider a very short !ro ram which is controlled by an u!-down switch, and which has an internal !arameter 8a natural number& This !ro ram can run in two modesthe u!-mode, and the down-mode& 3hen it runs in the u!mode, it self-re!licates into an ad.acent !art of the com!uter1s memorye(ce!t it makes the internal !arameter 8 of its -dau hter- one reater than in itself& 3hen it runs in the down-mode, it does not self-re!, but instead calculates the number
4-"5NI4BN ^ "5

and adds it to a runnin total& 3ell, su!!ose that at the be innin , there is one co!y of the !ro ram in memory, 8 X +, and the mode is u!& Then the !ro ram will co!y itself ne(t door in memory, with 8 X "& ;e!eatin the !rocess, the new !ro ram will self-re! ne(t door to itself, with a co!y havin 8 X B& )nd over and over a ain&&& 3hat ha!!ens is that a very lar e !ro ram is rowin inside memory& 3hen memory is full, the !rocess quits& 8ow all of memory can be looked u!on as bein filled with one bi !ro ram, com!osed of many similar, but differentiated, modulesor -cells-& 8ow su!!ose we switch the mode to down, and run this bi !ro ram& 3hat ha!!ens? The first -cell- runs, and calculates "I"& The second -cell- runs, calculatin -"I%, and addin it to the !revious result& The third -cell- runs,

calculatin ^ "IL and addin it on&&& The end result is that the whole -or anism-the bi !ro ramcalculates the sum l - "I% ^ "IL - "I$ ^ "I# - "I"" ^ "I"% - "I"L ^ &&& to a lar e number of terms 4as many terms as -cells- can fit inside memory5& )nd since this series conver es 4albeit slowly5 to mIC, we have a -!henoty!e- whose function is to calculate the value of a famous mathematical constant&

Le"el Mi-ing in the Cell

I ho!e that the descri!tions of !rocesses such as labelin , self-assembly, differentiation, mor!ho enesis, as well as transcri!tion and translation, have hel!ed to convey some notion of the immensely com!le( system which is a cellan information-!rocessin system with some strikin ly novel features& 3e have seen, in the 7entral :o ma!, that althou h we can try to draw a clear line between !ro ram and data, the distinction is somewhat arbitrary& 7arryin this line of thou ht further, we find that not only are program and data intricately woven to ether, but also the interpreter of !ro rams, the !hysical processor, and even the language are included in this intimate fusion& Therefore, althou h it is !ossible 4to some e(tent5 to draw boundaries and se!arate out the levels, it is .ust as im!ortantand .ust as fascinatin to reco ni<e the level-crossin s and mi(in s& Illustrative of this is the ama<in fact that in biolo ical systems, all the various features necessary for self-re! 4vi<&, lan ua e, !ro ram, data, inter!reter, and !rocessor5 coo!erate to such a de ree that all of them are re!licated simultaneouslywhich shows how much dee!er is biolo ical self-re!1in than anythin yet devised alon those lines by humans& 'or instance, the self-re! !ro ram e(hibited at the be innin of this 7ha!ter takes for ranted the !re-e(istence of three e(ternal as!ects, a lan ua e, an inter!reter, and a !rocessor, and does not re!licate those& Aet us try to summari<e various ways in which the subunits of a cell can be classified in com!uter science terms& 'irst, let us take :8)& Since :8) contains all the information for construction of !roteins, which are the active a ents of the cell, :8) can be viewed as a program written in a hi her-level lan ua e, which is subsequently translated 4or inter!reted5 into the -machine lan ua e- of the cell 4!roteins5& En the other hand, :8) is itself a !assive molecule which under oes mani!ulation at the hands of various kinds of en<ymes= in this sense, a :8) molecule is e(actly like a lon !iece of data, as well& Thirdly, :8) contains the tem!lates off of which the t;8) -flashcardsare rubbed, which means that :8) also contains the definition of its own hi her-level lan ua e& Aet us move on to !roteins& Proteins are active molecules, and carry out all the functions of the cell= therefore it is quite a!!ro!riate to think of them as programs in the -machine lan ua e- of the cell 4the cell itself bein the !rocessor5& En the other hand, since !roteins are hardware and most !ro rams are software, !erha!s it is better to think of the !roteins as processors& Thirdly, !roteins are often acted u!on by other !roteins, which means that !roteins are often data& 'inally, one can view !roteins as interpreters=

this involves viewin :8) as a collection of hi h-level lan ua e !ro rams, in which case en<ymes are merely carryin out the !ro rams written in the :8) code, which is to say, the !roteins are actin as inter!reters& Then there are ribosomes and t;8) molecules& They mediate the translation from :8) to !roteins, which can be com!ared to the translation of a !ro ram from a hi hlevel lan ua e to a machine lan ua e= in other words, the ribosomes are functionin as interpreters and the t;8) molecules !rovide the definition of the hi her-level language& 0ut an alternative view of translation has it that the ribosomes are processors, while the t;8)1s are interpreters& 3e have barely scratched the surface in this analysis of interrelations between all these biomolecules& 3hat we have seen is that nature feels quite comfortable in mi(in levels which we tend to see as quite distinct& )ctually, in com!uter science there is already a visible tendency to ni( all these seemin ly distinct as!ects of an information!rocessin system& This is !articularly so in )rtificial Intelli ence research, which is usually at the forefront of com!uter lan ua e desi n&

The 0rigin of Life

) natural and fundamental question to ask, on learnin of these incredibly intricately interlockin !ieces of software and hardware is, -?ow did they ever et started in the first !lace?- It is truly a bafflin thin & Ene has to ima ine some sort of a bootstra! !rocess occurrin , somewhat like that which is used in the develo!ment of new com!uter lan ua esbut a bootstra! from sim!le molecules to entire cells is almost beyond one1s !ower to ima ine& There are various theories on the ori in of life& They all run a round on this most central of all central questions, -?ow did the 6enetic 7ode, alon with the mechanisms for its translation 4ribosomes and t;8) molecules5, ori inate- 'or the moment, we will have to content ourselves with a sense of wonder and awe, rather than with an answer& )nd !erha!s e(!eriencin that sense of wonder and awe is more satisfyin than havin an answerat least for a while&

The $a!nificrab# Indeed

3t is spring, and the Tortoise and *chilles are taking a -unda) promenade in the woods together! The) ha#e decided to clim( a hill at the top of which, it is said, there is a wonderful teahouse, with all sorts of delicious pastries! *chilles, /an oh man2 If a crab Tortoise, If a crab?? *chilles, I was about to say, if a crab ever were intelli ent, then surely it would be our mutual friend the 7rab& 3hy, he must be at least two times as smart as any crab alive& Er maybe even three times as smart as any crab alive& Er !erha!s Tortoise, /y soul2 ?ow you ma nify the 7rab2 *chilles, 3ell, I .ust ha!!en to be an admirer of his&&& Tortoise, 8o need to a!olo i<e& I admire him, too& S!eakin of 7rab admirers, did I tell you about the curious fan letter which the 7rab received not too lon a o? *chilles, I don1t believe so& 3ho sent it? Tortoise, It bore a !ostmark from India, and was from someone neither of us had ever heard of beforea /r& 8a.unamar, I believe& *chilles, I wonder why someone who never knew /r& 7rab would send him a letter or for that matter, how they would et his address& Tortoise, )!!arently whoever it was was under the illusion that the 7rab is a mathematician& It contained numerous results, all of which were0ut, ho2 S!eak of the devil2 ?ere comes /r& 7rab now, down the hill& 8ra(, 6ood-bye2 It was nice to talk with you a ain& 3ell, I uess I had best be off& 0ut I1m utterly stuffedcouldn1t eat one more bite if I had to2 I1ve .ust been u! there myselfrecommend it hi hly& ?ave you ever been to the teahouse at the crest of the hill? ?ow are you, )chilles? Eh, there1s )chilles& ?ello, hello& 3ell, well, if it isn1t /r& T2 Tortoise, ?ello, /r& 7& )re you headed u! to the hillto! teahouse? 8ra(, 3hy, yes indeed, I am= how did you uess it? I1m quite lookin forward to some of their s!ecial na!oleonsscrum!tious little morsels& I1m so hun ry I could eat a fro & Eh, there1s )chilles& ?ow are you, )chilles? *chilles, 7ould be worse, I su!!ose& 8ra(, 3onderful2 3ell, don1t let me interru!t your discussion& I1ll .ust ta alon & Tortoise, 7uriously enou h, I was .ust about to describe your mysterious letter from that Indian fellow a few weeks backbut now that you1re here& I1ll let )chilles et the story from the 7rab1s mouth& 8ra(, 3ell, it was this way& This fellow 8a.unamar had a!!arently never had any formal trainin in mathematics, but had instead worked out some of his own methods for derivin new truths of mathematics& Some of his discoveries defeated me com!letely= I had never seen anythin in the least like them before& 'or instance, he e(hibited a ma! of India that he had mana ed to color usin no fewer than "$B# distinct colors&

73G5=E BKD! 7astrovalva, () '! 8! Escher lithograph, BIHK/!

*chilles, "$B#2 :id you say "$B#? 8ra(, Heswhy do you ask? *chilles, 3ell, "$B# is a very interestin number, you know& 8ra(, Indeed& I wasn1t aware of it& *chilles, In !articular, it so ha!!ens that "$B# is the number of the ta(icab which I took to /r& Tortoise1s this mornin 2 8ra(, ?ow fascinatin 2 7ould you !ossibly tell me the number of the trolley car which you1ll take to /r& Tortoise1s tomorrow mornin ? *chilles after a moment's thought/, It1s not obvious to me= however, I should think it would be very lar e& Tortoise, )chilles has a wonderful intuition for these thin s& 8ra(, Hes& 3ell, as I was sayin , 8a.unamar in his letter also !roved that every even !rime is the sum of two odd numbers, and that there are no solutions in !ositive inte ers to the equation an ^ bn X cn for n X +&

*chilles, 3hat? )ll these old classics of mathematics resolved in one fell swoo!? ?e must be a enius of the first rank2 Tortoise, 0ut )chillesaren1t you even in the sli htest ske!tical? *chilles, 3hat? Eh, yesske!tical& 3ell, of course I am& Hou don1t think I believe that /r& 7rab ot such a letter, do you? I don1t fall for .ust anythin , you know& So it must have been you, /r& T, who received the letter2 Tortoise, Eh, no, )chilles, the !art about /r& 7 receivin the letter is quite true& 3hat I meant was, aren1t you ske!tical about the content of the letterits e(trava ant claims? *chilles, 3hy should I be? ?mm&&& 3ell, of course I am& I1m a very ske!tical !erson, as both of you should well know by now& It1s very hard to convince me of anythin , no matter how true or false it is& Tortoise, Gery well !ut, )chilles& Hou certainly have a first-class awareness of your own mental workin s& *chilles, :id it ever occur to you, my friends, that these claims of 8a.unamar mi ht be incorrect? 8ra(, 'rankly, )chilles, bein rather conservative and orthodo( myself, I was a bit concerned about that very !oint on first receivin the letter& In fact, I sus!ected at first that here was an out-and-out fraud& 0ut on second thou ht, it occurred to me that not many ty!es of !eo!le could manufacture such stran e-soundin and com!le( results !urely from their ima ination& In fact, what it boiled down to was this question, -3hich is the more likely, a charlatan of such e(traordinary in enuity, or a mathematician of reat enius?- )nd before lon , I reali<ed that the !robabilities clearly favored the former& *chilles, :idn1t you directly check out any of his ama<in claims, however? 8ra(, 3hy should I? The !robability ar ument was the most convincin thin I had ever thou ht of= no mathematical !roof would have equaled it& 0ut /r& T here insisted on ri or& I finally ave in to his insistence, and checked all of 8a.unamar1s results& To my

reat sur!rise, each one of them was ri ht& ?ow he discovered them, I1ll never know, however& ?e must have some ama<in and inscrutable Eriental ty!e of insi ht which we here in the Eccident can have no inklin of& )t !resent, that1s the only theory which makes an sense to me& Tortoise, /r& 7rab has always been a little more susce!tible to mystical or fanciful e(!lanations than I am& I have full confidence that whatever 8a.unamar did in his way has a com!lete !arallel inside orthodo( mathematics& There is no way of doin mathematics which is fundamentally different from what we now know, in my o!inion& *chilles, That is an interestin o!inion& I su!!ose it has somethin to do with the 7hurchTurin Thesis and related to!ics& 8ra(, Eh, well, let us leave these technical matters aside on such a fine day, and en.oy the quiet of the forest, the chir!in of the birds, and the !lay of sunli ht on the new leaves and buds& ?o2 Tortoise, I second the motion& )fter all, all enerations of Tortoises have reveled in such deli hts of nature& 8ra(, )s have all enerations of 7rabs& *chilles, Hou don1t ha!!en to have brou ht your flute alon , by any chance, /r& 7? 8ra(, 3hy, certainly2 I take it with me everywhere& 3ould you like to hear a tune or two? *chilles, It would be deli htful, in this !astoral settin & :o you !lay from memory? 8ra(, Sad to say, that is beyond my ca!ability& I have to read my music from a sheet& 0ut that is no !roblem& I have several very !leasant !ieces here in this case& ,e opens up a thin case and draws out a few pieces of paper! The topmost one has the following s)m(ols on it:
"a5TSaE?

,e sticks the top sheet into a little holder attached to his flute, and pla)s! The tune is #er) short!/ *chilles, That was charmin & &eers o#er at the sheet on the flute, and a "ui@@ical e9pression (eclouds his face!/ 3hat is that statement of number theory doin , attached to your flute like that? The 8ra( looks at his flute, then his music, turns his head all around, and appears slightl) confused!/ 8ra(, I don1t understand& 3hat statement of number theory? *chilles, -9ero is not the successor of any natural number&- ;i ht there, in the holder on your flute2 8ra(, That1s the third Piano Postulate& There are five of them, and I1ve arran ed them all for flute& They1re obvious, but catchy&

*chilles, 3hat1s not obvious to me is how a number-theoretical statement can be !layed as music& 8ra(, 0ut I insist, it1s 8ET a number-theoretical statementit1s a Piano Postulate2 3ould you like to hear another? *chilles, I1d be enchanted& The 8ra( places another piece of paper on his flute, and this time *chilles watches more carefull)!/ 3ell, I watched your eyes, and they were lookin at that 'E;/FA) on the sheet& )re you sure that that is musical notation? I swear, it most ama<in ly resembles the notation which one mi ht use in a formali<ed version of number theory& 8ra(, ?ow odd2 0ut certainly that is music, not any kind of statement of mathematics, as far as I can tell2 Ef course, I am not a mathematician in any sense of the word& 3ould you like to hear any other tunes? *chilles, 0y all means& ?ave you some others? 8ra(, Scads& ,e takes a new sheet, and attaches it to his flute! 3t contains the following s)m(ols: Ta5b5CSSajSSb@ESSSSSSSSSSSSS? *chilles peers at it, while the 8ra( pla)s it!/ Isn1t it lovely? *chilles, Hes, it certainly is a tuneful little !iece& 0ut I have to say, it1s lookin more and more like number theory to me& 8ra(, ?eavens2 It is .ust my usual music notation, nothin more& I sim!ly don1t know how you read all these e(tramusical connotations into a strai htforward re!resentation for sounds& *chilles, 3ould you be averse to !layin a !iece of my own com!osition? 8ra(, 8ot in the least& ?ave you ot it with you? *chilles, 8ot yet, but I have a hunch I mi ht be able to com!ose some tunes all by myself& Tortoise, I must& tell you, )chilles, that /r& 7 is a harsh .ud e of music com!osed by others, so do not be disa!!ointed if, by some chance, he is not an enthusiast for your efforts& *chilles, That is very kind of you to forewarn me& Still, I1m willin to ive it a try&&& ,e writes: CCSSS?jSSS?@ICSSSS?jSSSS?@@ECSSSSS?jSSSSS?@

The 8ra( takes it, looks it o#er for a moment, then sets it in his music holder, and pipes!/ 8ra(, 3hy, that1s quite nice, )chilles& I en.oy stran e rhythms& *chilles, 3hat1s stran e about the rhythms in that !iece? 8ra(, Eh, naturally, to you as the com!oser it must seem quite bland, but to my ears, shiftin from a %I% rhythm to CIC and then to LIL is quite e(otic& If you have any other son s, I1d be lad to !lay them& *chilles, Thank you very much& I1ve never com!osed anythin before, and I must say com!osin is quite different from how I had ima ined it to be& Aet me try my hand at another one& 4?ots down a line&5 Ta5b5CSSajSSb@ESSSSSSSSSSSSSS? 8ra(, ?mmm&&& Isn1t that .ust a co!y of my earlier !iece? *chilles, Eh, no2 I1ve added one more S& 3here you had thirteen in a row, I have fourteen& 8ra(, Eh, yes& Ef course& 4,e pla)s it, and looks #er) stern&5 *chilles, I do ho!e you didn1t dislike my !iece2 8ra(, I am afraid, )chilles, that you com!letely failed to ras! the subtleties of my !iece, u!on which yours is modeled& 0ut how could I e(!ect you to understand it on first hearin ? Ene does not always understand what is at the root of beauty& It is so easy to mistake the su!erficial as!ects of a !iece for its beauty, and to imitate them, when the beauty itself is locked dee! inside the music, in a way which seems always to elude analysis& *chilles, I am afraid that you have lost me a little in your erudite commentary& I understand that my !iece does not measure u! to your hi h standards, but I do not know e(actly where I went astray& 7ould you !erha!s tell me some s!ecific way in which you find fault with my com!osition? 8ra(, Ene !ossible way to save your com!osition, )chilles, would be to insert another three S1sfive would do as wellinto that lon rou! of S1s near the end& That would create a subtle and unusual effect& *chilles, I see& 8ra(, 0ut there are other ways you mi ht choose to chan e your !iece& Personally, I would find it most a!!ealin to !ut another tilde in the front& Then there would be a nice balance between the be innin and the end& ?avin two tildes in a row never fails to ive a ay little twist to a !iece, you know& *chilles, ?ow about if I take both of your su estions, and make the followin !iece? Ta5b5CSSajSSb@ESSSSSSSSSSSSSSSSS? 8ra(: a painful grimace crossing his face/: 8ow, )chilles, it is im!ortant to learn the followin lesson, never try to !ut too much into any sin le !iece& There is always a !oint beyond which it cannot be im!roved, and further attem!ts to im!rove it will in

fact destroy it& Such is the case in this e(am!le& Hour idea of incor!oratin both of my su estions to ether does not yield the desired e(tra amount of beauty, but on the contrary creates an imbalance which quite takes away all the charm& *chilles, ?ow is it that two very similar !ieces, such as yours with thirteen S1s, and mine with fourteen S1s, seem to you to be so different in their musical worth? Ether than in that minor res!ect, the two are identical& 8ra(, 6racious2 There is a world of difference between your !iece and mine& Perha!s this is a !lace where words fail to convey what the s!irit can feel& Indeed, I would venture to say that there e(ists no set of rules which delineate what it is that makes a !iece beautiful, nor could there ever e(ist such a set of rules& The sense of 0eauty is the e(clusive domain of 7onscious /inds, minds which throu h the e(!erience of livin have ained a de!th that transcends e(!lanation by any mere set of rules& *chilles, I will always remember this vivid clarification of the nature of 0eauty& I su!!ose that somethin similar a!!lies to the conce!t of Truth, as well? 8ra(, 3ithout doubt& Truth and 0eauty are as interrelated asas *chilles, )s interrelated as, say, mathematics and music? 8ra(, Eh2 Hou took the words ri ht out of my mouth2 ?ow did you know that that is what I was thinkin ? Tortoise, )chilles is very clever, /r& 7& 8ever underestimate the !otency of his insi ht& *chilles, 3ould you say that there could conceivably be any relationshi! between the truth or falsity of a !articular statement of mathematics, and the beauty, or lack of beauty, of an associated !iece of music? Er is that .ust a far-fetched fancy of mine, with no basis in reality? 8ra(, If you are askin me, that is carryin thin s much too far& 3hen I s!oke of the interrelatedness of music and mathematics, I was s!eakin very fi uratively, you know& )s for a direct connection between s!ecific !ieces of music and s!ecific statements of mathematics, however, I harbor e(tremely rave doubts about its !ossibility& I would humbly counsel you not to ive too much time to such idle s!eculations& *chilles, Hou are no doubt ri ht& It would be most un!rofitable& Perha!s I ou ht to concentrate on shar!enin my musical sensitivity by com!osin some new !ieces& 3ould you be willin to serve as my mentor, /r& 7? 8ra(, I would be very ha!!y to aid you in your ste!s towards musical understandin & -o *chilles takes pen in hand, and, with what appears to (e a great deal of concentration, writes,
??a"'~5b+cSC=?((~d"<("Sj+(>

The 8ra( looks #er) startled!/ Hou really want me to !lay that-that-that whatever-it-is? *chilles, Eh, !lease do2

-o the 8ra( pla)s it, with e#ident difficult)!/ Tortoise, 0ravo2 0ravo2 Is John 7a e your favorite com!oser, )chilles? *chilles, )ctually, he1s my favorite anti-com!oser& )nyway, I1m lad you liked /H music& 8ra(, The two of you may find it amusin to listen to such totally meanin less caco!hony, but I assure you it is not at all !leasant for a sensitive com!oser to be sub.ected to such e(cruciatin , em!ty dissonances and meanin less rhythms& )chilles, I thou ht you had a ood feelin for music& 7ould it be that your !revious !ieces had merit merely by coincidence? *chilles, Eh, !lease for ive me, /r& 7rab& I was tryin to e(!lore the limits of your musical notation& I wanted to learn directly what kinds of sound result when I write certain ty!es of note sequences, and also how you evaluate !ieces written in various styles& 8ra(, ?arrum!h2 I am not .ust an automatic music-machine, you know& 8or am I a arba e dis!osal for musical trash& *chilles, I am very sorry& 0ut I feel that I have learned a reat deal by writin that small !iece, and I am convinced that I can now write much better music than I ever could have if I hadn1t tried that idea& )nd if you1ll .ust !lay one more !iece of mine, I have hi h ho!es that you will feel better about my musical sensitivities& 8ra(, 3ell, all ri ht& 3rite it down and I1ll ive it a chance& 4*chilles writes:
"a5"b5<Caja@ =(SS?j(bjb""aE?>

and the 8ra( pla)s!/ Hou were ri ht, )chilles& Hou seem to have com!letely re ained your musical acuity& This is a little em2 ?ow did you come to com!ose it? I have never heard anythin like it& It obeys all the rules of harmony, and yet has a certain what shall I say? irrational a!!eal to it& I can1t !ut my fin er on it, but I like it for that very reason& *chilles, I kind of thou ht you mi ht like it& Tortoise, ?ave you ot a name for it, )chilles? Perha!s you mi ht call it -The Son of Pytha oras-& Hou remember that Pytha oras and his followers were amon the first to study musical sound& *chilles, Hes, that1s true& That would be a fine title& 8ra(, 3asn1t Pytha oras also the first to discover that the ratio of two squares can never be equal to B? Tortoise, I believe you1re ri ht& It was considered a truly sinister discovery at the time, for never before had anyone reali<ed that there are numbers such as the square root of B which are not ratios of inte ers& )nd thus the discovery was dee!ly disturbin to the Pytha oreans, who felt that it revealed an unsus!ected and rotesque defect in the abstract world of numbers& 0ut I don1t know what this has to do with the !rice of tea in 7hina&

*chilles, S!eakin of tea, isn1t that the teahouse .ust u! there ahead of us? Tortoise, Hes, that1s it, all ri ht& 3e ou ht to be there in a cou!le of minutes& *chilles, ?mm&&& That1s .ust enou h time for me to whistle for you the tune which the ta(i driver this mornin had on his radio& It went like this& 8ra(, ?old on for a moment= I1ll et some !a!er from my case, and .ot down your tune& 4-crounges around inside his case, and finds a (lank sheet&5 6o ahead= I1m ready& *chilles whistles a rather long tune, and the 8ra( scram(les to keep up with him!/ 7ould you whistle the last few bars a ain? *chilles, 3hy, certainly& *fter a couple of such repeats, the session is complete, and the 8ra( proudl) displa)s his transcription, DCCSSSSS?jSSSSS?@ICSSSSS?jSSSSS?@@ECCSSSSSSS?jSSSSSSS?@ICS?jS?@@Tb5<c5 CScIb@ECCSSSSSSS?jSSSSSSS?@ICS?jS?@@ d5d%,e5e%5 <~<dEedEe%RDbECCSdjSd@ICSd%jSd%@@bECCSejSe@ICSe%jSe%@@>>>> The 8ra( then pla)s it himself!/ Tortoise, It1s !eculiar music, isn1t it? It sounds a wee bit like music from India, to me& 8ra(& Eh, I think it1s too sim!le to be from India& 0ut of course I know !recious little about such thin s& Tortoise, 3ell, here we are at the teahouse& Shall we sit outside here, on the verandah? 8ra(, If you don1t mind, I1d !refer to o inside& I1ve otten !erha!s enou h sun for the day& The) go inside the teahouse and are seated at a nice wooden ta(le, and order cakes and tea! -oon a cart of scrumptious%looking pastries is wheeled up, and each of them chooses his fa#orite!/ *chilles, Hou know, /r& 7, I would love to know what you think of another !iece which I have .ust com!osed in my head& 8ra(, 7an you show it to me? ?ere, write it down on this na!kin& *chilles writes,
"a5b5c5<~d5e5<(SSd(SSe"=b(SSd(SSe"Ec>(aIa"=(bIc">

The 8ra( and Tortoise stud) it with interest&5

Tortoise, Is it another beautiful !iece, /r& 7, in your o!inion? 8ra(, 3ell, uh&&& -hifts in his chair, and looks somewhat uncomforta(le!/ *chilles, 3hat1s the matter? Is it harder to decide whether this !iece is beautiful than it is for other !ieces? 8ra(, )hm&&& 8o, it1s not thatnot at all& It1s .ust that, well&&& I really have to ?E); a !iece before I can tell how much I like it& *chilles, So o ahead and !lay it2 I1m dyin to know whether you find it beautiful or not& 8ra(, Ef course, I1d be e(tremely lad to !lay it for you& The only thin is *chilles, 7an1t you !lay it for me? 3hat1s the matter? 3hy are you balkin ? Tortoise, :on1t you reali<e, )chilles, that for /r& 7rab to fulfill your request would be most im!olite and disturbin to the clientele and em!loyees of this fine establishment? 8ra( suddenl) looking relie#ed/, That1s ri ht& 3e have no ri ht to im!ose our music on others& *chilles deMectedl)/, Eh, P?EEEH2 )nd I so much wanted to know what he thinks of this !iece2 8ra(, 3hew2 That was a close call2 *chilles, 3hat was that remark? 8ra(, Ehnothin & It1s .ust that that waiter over there, he ot knocked into by another waiter, and almost dro!!ed a whole !ot of tea into a lady1s la!& ) narrow esca!e, I must say& 3hat do you say, /r& Tortoise? Tortoise, Gery ood teas, I1d say& 3ouldn1t you a ree, )chilles? *chilles, Eh, yes& Prime teas, in fact& 8ra(, :efinitely& 3ell, I don1t know about you two, but I should !erha!s be oin , for I1ve a lon stee! trail back to my house, on the other side of this hill& *chilles, Hou mean this is a bi bluff? 8ra(, Hou said it, )chilles& *chilles, I see& 3ell, I1ll have to remember that& 8ra(, It has been such a .olly afternoon, )chilles, and I sincerely ho!e we will e(chan e more musical com!ositions another day& *chilles, I1m lookin forward to that very much, /r& 7& 3ell, ood-bye& Tortoise, 6ood-bye, /r& 7& *nd the 8ra( heads off down his side of the hill!/ *chilles, 8ow there oes a brilliant fellow&&& In my estimation, he1s at least four times as smart as any crab alive& Er he mi ht even be five Tortoise, )s you said in the be innin , and !robably shall be sayin forevermore, words without end&

C/A,TE$ <4II

Church) Turing) Tars.i) and 0thers

+ormal and Informal Systems
3E ?)GE 7E/E to the !oint where we can develo! one of the main theses of this book, that every as!ect of thinkin can be viewed as a hi h-level descri!tion of a system which, on a low level, is overned by sim!le, even formal, rules& The -system-, of course, is a brainunless one is s!eakin of thou ht !rocesses flowin in another medium, such as a com!uter1s circuits& The ima e is that of a formal system underlyin an -informal system-a system which can, for instance, make !uns, discover number !atterns, for et names, make awful blunders in chess, and so forth& This is what one sees from the outside, its informal, overt, software level& 0y contrast, it has a formal, hidden, hardware level 4or -substrate-5 which is a formidably com!le( mechanism that makes transitions from state to state accordin to definite rules !hysically embodied in it, and accordin to the in!ut of si nals which im!in e on it& ) vision of the brain such as this has many !hiloso!hical and other consequences, needless to say& I shall try to s!ell some of them out in this 7ha!ter& )mon other thin s, this vision seems to im!ly that, at bottom, the brain is some sort of a -mathematicalob.ect& )ctually, that is at best a very awkward way to look at the brain& The reason is that, even if a brain is, in a technical and abstract sense, some sort of formal system, it remains true that mathematicians only work with sim!le and ele ant systems, systems in which everythin is e(tremely clearly defined and the brain is a far cry from that, with its ten billion or more semi-inde!endent neurons, quasi-randomly connected u! to each other& So mathematicians would never study a real brain1s networks& )nd if you define -mathematics- as what mathematicians en.oy doin , then the !ro!erties of brains are not mathematical& The only way to understand such a com!le( system as a brain is by chunkin it on hi her and hi her levels, and thereby losin some !recision at each ste!& 3hat emer es at the to! level is the -informal system- which obeys so many rules of such com!le(ity that we do not yet have the vocabulary to think about it& )nd that is what )rtificial Intelli ence research is ho!in to find& It has quite a different flavor from mathematics research& 8evertheless, there is a loose connection to mathematics, )I !eo!le often come from a stron mathematics back round, and mathematicians sometimes are intri ued by the workin s of their own brains& The followin !assa e, quoted from Stanislaw Flam1s autobio ra!hical *d#entures of a 'athematician, illustrates this !oint,
It seems to me that more could be done to elicit&&& the nature of associations, with com!uters !rovidin the means for e(!erimentation& Such a study would have to involve a radation of notions, of symbols, of classes of symbols, of classes of classes, and so on, in the same way that the com!le(ity of mathematical or !hysical structures is investi ated&

There must be a trick to the train of thou ht, a recursive formula& ) rou! of neurons starts workin automatically, sometimes without e(ternal im!ulse& It is a kind of iterative !rocess with a rowin !attern& It wanders about in the brain, and the way it ha!!ens must de!end on the memory of similar !atterns&"

Intuition and the Magnificent Crab

)rtificial Intelli ence is often referred to as -)I-& Eften, when I try to e(!lain what is meant by the term, I say that the letters -)I- could .ust as well stand for -)rtificial Intuition-, or even -)rtificial Ima ery-& The aim of )I is to et at what is ha!!enin when one1s mind silently and invisibly chooses, from a myriad alternatives, which one makes most sense in a very com!le( situation& In many real-life situations, deductive reasonin is ina!!ro!riate, not because it would ive wron answers, but because there are too many correct but irrelevant statements which can be made= there are .ust too many thin s to take into account simultaneously for reasonin alone to be sufficient& 7onsider this mini-dialo ue, -The other day I read in the !a!er that the -Ehyou were readin ? It follows that you have eyes& Er at least one eye& Er rather, that you had at least one eye then&) sense of .ud ment-3hat is im!ortant here, and what is not?-is called for& Tied u! with this is a sense of sim!licity, a sense of beauty& 3here do these intuitions come from? ?ow can they emer e from an underlyin formal system? In the 'agnificra(, some unusual !owers of the 7rab1s mind are revealed& ?is own version of his !owers is merely that he listens to music and distin uishes the (eautiful from the non%(eautiful& 4)!!arently for him there is a shar! dividin line&5 8ow )chilles finds another way to describe the 7rab1s abilities, the 7rab divides statements of number theory into the cate ories true and false& 0ut the 7rab maintains that, if he chances to do so, it is only by the !urest accident, for he is, by his own admission, incom!etent in mathematics& 3hat makes the 7rab1s !erformance all the more mystifyin to )chilles, however, is that it seems to be in direct violation of a celebrated result of metamathematics with which )chilles is familiar, 7?F;7?1S T?EE;E/, There is no infallible method for tellin theorems of T8T from nontheorems& It was !roven in "#%D by the )merican lo ician )lon<o 7hurch& 7losely related is what I call the T);S@I-7?F;7?-TF;I86 T?EE;E/, There is no infallible method for tellin true from false statements of number theory&

The Church-Turing Thesis

To understand 7hurch1s Theorem and the Tarski-7hurch-Turin Theorem better, we should first describe one of the ideas on which they are based= and that is the 7hurchTurin Thesis 4often called -7hurch1s Thesis-5& 'or the 7hurch-Turin Thesis is certainly one of the most im!ortant conce!ts in the !hiloso!hy of mathematics, brains, and thinkin & )ctually, like tea, the 7hurch-Turin Thesis can be iven in a variety of different stren ths& So I will !resent it in various versions, and we will consider what they im!ly& The first version sounds very innocentin fact almost !ointless, 7?F;7?-TF;I86 T?ESIS, T)FTEAE6I7)A GE;SIE8, /athematics !roblems can be solved only by doin mathematics& Ef course, its meanin resides in the meanin of its constituent terms& 0y -mathematics !roblem- I mean the !roblem of decidin whether some number !ossesses or does not !ossess a iven arithmetical !ro!erty& It turns out that by means of 6>del-numberin and related codin tricks, almost any !roblem in any branch of mathematics can be !ut into this form, so that -mathematics !roblem- retains its ordinary meanin & 3hat about -doin mathematics-? 3hen one tries to ascertain whether a number has a !ro!erty, there seem to be only a small number of o!erations which one uses in combination over and over a ainaddition, multi!lication, checkin for equality or inequality& That is, loo!s com!osed of such o!erations seem to be the only tool we have that allows us to !robe the world of numbers& 8ote the word -seem-& This is the critical word which the 7hurchTurin Thesis is about& 3e can ive a revision, 7?F;7?-TF;I86 T?ESIS, ST)8:);: GE;SIE8, Su!!ose there is a method which a sentient bein follows in order to sort numbers into two classes& Su!!ose further that this method always yields an answer within a finite amount of time, and that it always ives the same answer for a iven number& Then, Some terminatin 'looP !ro ram 4i&e&, some eneral recursive function5 e(ists which ives e(actly the same answers as the sentient bein 1s method does& The central hy!othesis, to make it very clear, is that any mental !rocess which divides numbers into two sorts can be described in the form of a 'looP !ro ram& The intuitive belief is that there are no other tools than those in 'looP, and that there are no ways to use those tools other than by unlimited iterations 4which 'looP allows5& The 7hurch-Turin Thesis is not a !rovable fact in the sense of a Theorem of mathematics it is a hy!othesis about the !rocesses which human brains use&

The ,ublic-,rocesses 4ersion

8ow some !eo!le mi ht feel that this version asserts too much& These !eo!le mi ht !ut their ob.ections as follows, -Someone such as the 7rab mi ht e(ist someone with an

almost mystical insi ht into mathematics, but who is .ust as much in the dark about his own !eculiar abilities as anyone elseand !erha!s that !erson1s mental mechanisms carry out o!erations which have no counter!art in 'looP&- The idea is that !erha!s we have a subconscious !otential for doin thin s which transcend the conscious !rocesses thin s which are somehow ine(!ressible in terms of the elementary 'looP o!erations& 'or these ob.ectors, we shall ive a weaker version of the Thesis, one which distin uishes between !ublic and !rivate mental !rocesses, 7?F;7?-TF;I86 T?ESIS, PF0AI7-P;E7ESSES GE;SIE8, Su!!ose there is a method which a sentient bein follows in order to sort numbers into two classes& Su!!ose further that this method always yields an answer within a finite amount of time, and that it always ives the same answer for a iven number& Proviso& Su!!ose also that this method can be communicated reliably from one sentient bein to another by means of lan ua e& Then, Some terminatin 'looP !ro ram 4i&e&, eneral recursive function5 e(ists which ives e(actly the same answers as the sentient bein s1 method does& This says that !ublic methods are sub.ect to -'looPification-, but asserts nothin about !rivate methods& It does not say that they are un-'looP-able, but it at least leaves the door o!en&

Srini"asa $amanuVan
)s evidence a ainst any stron er version of the 7hurch-Turin Thesis, let us consider the case of the famous Indian mathematician of the first quarter of the twentieth century, Srinivasa ;amanu.an 4"**$-"#B+5& ;amanu.an 4'i & "+L5 came from Tamilnadu, the southernmost !art of India, and studied mathematics a little in hi h school& Ene day, someone who reco ni<ed ;amanu.an1s talent for math !resented him with a co!y of a sli htly out-of-date te(tbook on analysis, which ;amanu.an devoured 4fi uratively s!eakin 5& ?e then be an makin his own forays into the world of analysis, and by the time he was twenty-three, he had made a number of discoveries which he considered worthwhile& ?e did not know to whom to turn, but somehow was told about a !rofessor of mathematics in faraway En land, named 6& ?& ?ardy& ;amanu.an com!iled his best results to ether in a !acket of !a!ers, and sent them all to the, unforewarned ?ardy with a coverin letter which friends hel!ed him e(!ress in En lish& 0elow are some e(cer!ts taken from ?ardy1s descri!tion of his reaction u!on receivin the bundle,
&&& It soon became obvious that ;amanu.an must !ossess much more eneral theorems and was kee!in a reat deal u! his sleeve&&& USome formulaeV defeated me com!letely= I had never seen anythin in the least like them before& ) sin le look at them is enou h to show that they could only be written down by a mathematician of the hi hest class& They must be true because, if they were not true, no one would have had the ima ination to invent them& 'inally&&& the writer must be com!letely honest, because reat mathematicians are commoner than thieves or humbu s of such incredible skill&B

73G5=E BKG! -rini#asa =amanuMan and one of his strange 3ndian melodies!

3hat resulted from this corres!ondence was that ;amanu.an came to En land in "#"%, s!onsored by ?ardy= and then followed an intense collaboration which terminated in ;amanu.an1s early demise at a e thirty-three from tuberculosis& ;amanu.an had several e(traordinary characteristics which set him a!art from the ma.ority of mathematicians& Ene was his lack of ri or& Gery often he would sim!ly state a result which, he would insist, had .ust come to him from a va ue intuitive source, far out of the realm of conscious !robin & In fact, he often said that the oddess 8ama iri ins!ired him in his dreams& This ha!!ened time and a ain, and what made it all the more mystifyin !erha!s even imbuin it with a certain mystical qualitywas the fact that many of his -intuition-theorems- were wron & 8ow there is a curious !arado(ical effect where sometimes an event which you think could not hel! but make credulous !eo!le become a little more ske!tical, actually has the reverse effect, hittin the credulous ones in some vulnerable s!ot of their minds, tantali<in them with the hint of some bafflin irrational side of human nature& Such was the case with ;amanu.an1s blunders, many educated !eo!le with a yearnin to believe in somethin of the sort considered ;amanu.an1s intuitive !owers to be evidence of a mystical insi ht into Truth, and the fact of his fallibility seemed, if anythin , to stren then, rather than weaken, such beliefs& Ef course it didn1t hurt that he was from one of the most backward !arts of India, where fakirism and other eerie Indian rites had been !racticed for millennia, and were still !racticed with a frequency !robably e(ceedin that of the teachin of hi her

mathematics& )nd his occasional wron flashes of insi ht, instead of su estin to !eo!le that he was merely human, !arado(ically ins!ired the idea that ;amanu.an1s wron ness always had some sort of -dee!er ri htness- to it an -Eriental- ri htness, !erha!s touchin u!on truths inaccessible to 3estern minds& 3hat a delicious, almost irresistible thou ht2 Even ?ardywho would have been the first to deny that ;amanu.an had any mystical !owersonce wrote about one of ;amanu.an1s failures, -)nd yet I am not sure that, in some ways, his failure was not more wonderful than any of his trium!hs&The other outstandin feature of ;amanu.an1s mathematical !ersonality was his -friendshi! with the inte ers-, as his collea ue Aittlewood !ut it& This is a characteristic that a fair number of mathematicians share to some de ree or other, but which ;amanu.an !ossessed to an e(treme& There are a cou!le of anecdotes which illustrate this s!ecial !ower& The first one is related by ?ardy,
I remember once oin to see him when he was lyin ill at Putney& I had ridden in ta(i-cab 8o& "$B#, and remarked that the number seemed to me rather a dull one, and that I ho!ed it was not an unfavorable omen& -8o,- he re!lied, -it is a very interestin number= it is the smallest number e(!ressible as a sum of two cubes in two different ways&- I asked him, naturally, whether he knew the answer to the corres!ondin !roblem for fourth !owers= and he re!lied, after a moment1s thou ht, that he could see no obvious e(am!le, and thou ht that the first such number must be very lar e&%

It turns out that the answer for fourth !owers is, D%L%"*DL$ X "%CC ^ "%%C X "L*C ^ L#C The reader may find it interestin to tackle the analo ous !roblem for squares, which is much easier& It is actually quite interestin to !onder why it is that ?ardy immediately .um!ed to fourth !owers& )fter all, there are several other reasonably natural enerali<ations of the equation u% ^ v% X (% ^ y% alon different dimensions& 'or instance, there is the question about re!resentin number in three distinct ways as a sum of two cubes, r% ^ s % X u % ^ v% X ( % ^ y% Er, one can use three different cubes, u% ^ v% ^ w% X (% ^ y% ^ <%, Er one can even make a 6rand 6enerali<ation in all dimensions at once, rC ^ sC ^ tC X uC ^ vC ^ wC X (C ^ vC ^ <C a

There is a sense, however, in which ?ardy1s enerali<ation is -the most mathematicianlike-& 7ould this sense of mathematical esthetics ever be !ro rammed? The other anecdote is taken from a bio ra!hy of ;amanu.an by his countryman S& ;& ;an anathan, where it is called -;amanu.an1s 'lash-& It is related by a Indian friend of ;amanu.an1s from his 7ambrid e days, :r& P& 7& /ahalanobis&
En another occasion, I went to his room to have lunch with him& The 'irst 3orld 3ar had started some time earlier& I had in my hand a co!y of the monthly -Strand /a a<ine- which at that time used to !ublish a number of !u<<les to be solved by readers& ;amanu.an was stirrin somethin in a !an over the fire for our lunch& I was sittin near the table, turnin over the !a es of the /a a<ine& I ot interested in a !roblem involvin a relation between two numbers& I have for otten the details= but I remember the ty!e of the !roblem& Two 0ritish officers had been billeted in Paris in two different houses in a lon street= the door numbers of these houses were related in a s!ecial way= the !roblem was to find out the two numbers& It was not at all difficult& I ot the solution in a few minutes by trial and error& /)?)A)8E0IS 4in a .okin way5, 8ow here is a !roblem for you& ;)/)8FJ)8, 3hat !roblem, tell me& ,e went on stirring the pan!/ I read out the question from the -Strand /a a<ine-& ;)/)8FJ)8, Please take down the solution& ,e dictated a continued fraction!/ The first term was the solution which I had obtained& Each successive term re!resented successive solutions for the same ty!e of relation between two numbers, as the number of houses in the street would increase indefinitely& I was ama<ed& /)?)A)8E0IS, :id you et the solution in a flash? ;)/)8FJ)8, Immediately I heard the !roblem, it was clear that the solution was obviously a continued fraction= I then thou ht, -3hich continued fraction?- and the answer came to my mind& It was .ust as sim!le as this&C

?ardy, as ;amanu.an1s closest co-worker, was often asked after ;amanu.an1s death if there had been any occult or otherwise e(otically flavored elements to ;amanu.an1s thinkin style& ?ere is one comment which he ave,
I have often been asked whether ;amanu.an had any s!ecial secret= whether his methods differed in kind from those of other mathematicians= whether there was anythin really abnormal in his mode of thou ht& I cannot answer these questions with any confidence or conviction= but I do not believe it& /y belief is that all mathematicians think, at bottom, in the same kind of way, and that ;amanu.an was no e(ce!tion &L

?ere ?ardy states in essence his own version of the 7hurch-Turin Thesis& I !ara!hrase, 7?F;7?-TF;I86 T?ESIS, ?);:H1S GE;SIE8, )t bottom, all mathematicians are isomor!hic& This does not equate the mathematical !otential of mathematicians with that of eneral recursive functions= for that, however, all you need is to show that some mathematician1s mental ca!acity is no more eneral than recursive functions& Then, if you believe ?ardy1s Gersion, you know it for all mathematicians& Then ?ardy com!ares ;amanu.an with calculatin !rodi ies,

?is memory, and his !owers of calculation, were very unusual, but they could not reasonably be called -abnormal-& If he had to multi!ly two lar e numbers, he multi!lied them in the ordinary way= he could do it with unusual ra!idity and accuracy, but not more ra!idly and accurately than any mathematician who is naturally quick and has the habit of com!utations& D

?ardy describes what he !erceived as ;amanu.an1s outstandin intellectual attributes,

3ith his memory, his !atience, and his !ower of calculation, he combined a power of generalisation, a feeling for form, and a capacit) for rapid modification of his h)potheses , that were often really startlin , and made him, in his own field, without a rival in his day&$

The !art of this !assa e which I have italici<ed seems to me to be an e(cellent characteri<ation of some of the subtlest features of intelli ence in eneral& 'inally, ?ardy concludes somewhat nostal ically,
U?is workV has not the sim!licity and inevitableness of the very reatest work= it would be reater if it were less stran e& Ene ift it has which no one can deny!rofound and invincible ori inality& ?e would !robably have been a reater mathematician if he had been cau ht and tamed a little in his youth= he would have discovered more that was new, and that, no doubt, of reater im!ortance& En the other hand he would have been less of a ;amanu.an, and more of a Euro!ean !rofessor and the loss mi ht have been reater than the ain& *

The esteem in which ?ardy held ;amanu.an is revealed by the romantic way in which he s!eaks of him&

(Idiots Sa"ants(
There is another class of !eo!le whose mathematical abilities seem to defy rational e(!lanationthe so-called -idiots savants-, who can !erform com!le( calculations at li htnin s!eeds in their heads 4or wherever they do it5& Johann /artin 9acharias :ase, who lived from "*BC to "*D" and was em!loyed by various Euro!ean overnments to !erform com!utations, is an outstandin e(am!le& ?e not only could multi!ly two numbers each of "++ di its in his head= he also had an uncanny sense of quantity& That is, he could .ust -tell-, without countin , how many shee! were in a field, or words in a sentence, and so forth, u! to about %+this in contrast to most of us, who have such a sense u! to about D, with reliability& Incidentally, :ase was not an idiot& I shall not describe the many fascinatin documented cases of -li htnin calculators-, for that is not my !ur!ose here& 0ut I do feel it is im!ortant to dis!el the idea that they do it by some mysterious, unanaly<able method& )lthou h it is often the case that such wi<ards1 calculational abilities far e(ceed their abilities to e(!lain their results, every once in a while, a !erson with other intellectual ifts comes alon who also has this s!ectacular ability with numbers& 'rom such !eo!le1s intros!ection, as well as from e(tensive research by !sycholo ists, it has been ascertained that nothin occult takes !lace durin the !erformances of li htnin calculators, but sim!ly that their minds race throu h intermediate ste!s with the kind of self-confidence that a natural athlete has in e(ecutin a com!licated motion quickly and racefully& They do not reach their answers by some sort of instantaneous flash of enli htenment 4thou h sub.ectively it may feel that

way to some of them5, butlike the rest of usby sequential calculation, which is to say, by 'looP-in 4or 0looP-in 5 alon & Incidentally, one of the most obvious clues that no -hot line to 6od- is involved is the mere fact that when the numbers involved et bi er, the answers are slower in comin & Presumably, if 6od or an -oracle- were su!!lyin the answers, he wouldn1t have to slow u! when the numbers ot bi er& Ene could !robably make a nice !lot showin how the time taken by a li htnin calculator varies with the si<es of the numbers involved, and the o!erations involved, and from it deduce some features of the al orithms em!loyed&

The Isomor!hism 4ersion of the Church-Turing Thesis

This finally brin s us to a stren thened standard version of the 7hurch-Turin Thesis, 7?F;7?-TF;I86 T?ESIS, ISE/E;P?IS/ GE;SIE8, Su!!ose there is a method which a sentient bein follows in order to sort numbers into two classes& Su!!ose further that this method always yields an answer within a finite amount of time, and that it always ives the same answer for a iven number& Then, Some terminatin 'looP !ro ram 4i&e&, eneral recursive function5 e(ists which ives e(actly the same answers as the sentient bein 1s method does& /oreover, The mental !rocess and the 'looP !ro ram are isomor!hic in the sense that on some level there is a corres!ondence between the ste!s bein carried out in both com!uter and brain& 8otice that not only has the conclusion been stren thened, but also the !roviso of communicability of the faint-hearted Public-Processes Gersion has been dro!!ed& This bold version is the one which we now shall discuss& In brief, this version asserts that when one com!utes somethin , one1s mental activity can be mirrored isomor!hically in some 'looP !ro ram& )nd let it be very clear that this does not mean that the brain is actually runnin a 'looP !ro ram, written in the 'looP lan ua e com!lete with BEGI;1s, E; 1s, AB0$T1s, and the restnot at all& It is .ust that the ste!s are taken in the same order as they could be in a 'looP !ro ram, and the lo ical structure of the calculation can be mirrored in a 'looP !ro ram& 8ow in order to make sense of this idea, we shall have to make some level distinctions in both com!uter and brain, for otherwise it could be misinter!reted as utter nonsense& Presumably the ste!s of the calculation oin on inside a !erson1s head are on the hi hest level, and are su!!orted by lower levels, and eventually by hardware& So if we s!eak of an isomor!hism, it means we1ve tacitly made the assum!tion that the hi hest level can be isolated, allowin us to discuss what oes on there inde!endently of other levels, and then to ma! that to! level into 'looP& To be more !recise, the assum!tion is that there e(ist software entities which !lay the roles of various mathematical constructs, and which are activated in ways which can be mirrored e(actly inside 'looP 4see 'i & "+D5& 3hat enables these software entities to come into e(istence is the entire infrastructure discussed in 7ha!ters TI and TI I, as well as in the Prelude, )nt 'u ue&

There is no assertion of isomor!hic activity on the lower levels of brain and com!uter 4e& &, neurons and bits5& The s!irit of the Isomor!hism Gersion, if not the letter, is otten across by sayin that what an idiot savant does in calculatin , say, the lo arithm of p, is isomor!hic to what a !ocket calculator does in calculatin itwhere the isomor!hism holds on the arithmetic-ste! level, not on the lower levels of, in the one case, neurons, and in the other, inte rated circuits& 4Ef course different routes can be followed in calculatin anythin but !resumably the !ocket calculator, if not the human, could be instructed to calculate the answer in any s!ecific manner&5
73G5=E BKJ! The (eha#ior of natural num(ers can (e mirrored in a human (rain or in the programs of a computer! These two different representations can then (e mapped onto each other on an appropriatel) a(stract le#el!

$e!resentation of Kno'ledge about the $eal >orld

8ow this seems quite !lausible when the domain referred to is number theory, for there the total universe in which thin s ha!!en is very small and clean& Its boundaries and residents and rules are well-defined, as in a hard-ed ed ma<e& Such a world is far less com!licated than the o!en-ended and ill-defined world which we inhabit& ) number theory !roblem, once stated, is com!lete in and of itself& ) real-world !roblem, on the other hand, never is sealed off from any !art of the world with absolute certainty& 'or instance, the task of re!lacin a burnt-out li ht bulb may turn out to require movin a arba e ba = this may une(!ectedly cause the s!illin of a bo( of !ills, which then forces the floor to be swe!t so that the !et do won1t eat any of the s!illed !ills, etc&, etc& The !ills and the arba e and the do and the li ht bulb are all quite distantly related !arts of the worldyet an intimate connection is created by some everyday ha!!enin s& )nd there is no tellin what else could be brou ht in by some other small variations on the e(!ected& 0y contrast, if you are iven a number theory !roblem, you never wind u! havin to consider e(traneous thin s such as !ills or do s or ba s of arba e or brooms in order to solve your !roblem& 4Ef course, your intuitive knowled e of such ob.ects may serve you in ood stead as you o about unconsciously tryin to manufacture mental ima es to hel! you in visuali<in the !roblem in eometrical termsbut that is another matter&5 0ecause of the com!le(ity of the world, it is hard to ima ine a little !ocket calculator that can answer questions !ut to it when you !ress a few buttons bearin labels such as -do -, - arba e-, -li ht bulb-, and so forth& In fact, so far it has !roven to be

e(tremely com!licated to have a full-si<e hi h-s!eed com!uter answer questions about what a!!ear to us to be rather sim!le subdomains of the real world& It seems that a lar e amount of knowled e has to be taken into account in a hi hly inte rated way for -understandin - to take !lace& 3e can liken real-world thou ht !rocesses to a tree whose visible !art stands sturdily above round but de!ends vitally on its invisible roots which e(tend way below round, ivin it stability and nourishment& In this case the roots symboli<e com!le( !rocesses which take !lace below the conscious level of the mind!rocesses whose effects !ermeate the way we think but of which we are unaware& These are the -tri erin !atterns of symbols- which were discussed in 7ha!ters TI and TII& ;eal-world thinkin is quite different from what ha!!ens when we do a multi!lication of two numbers, where everythin is -above round-, so to s!eak, o!en to ins!ection& In arithmetic, the to! level can be -skimmed off - and im!lemented equally well in many different sorts of hardware, mechanical addin machines, !ocket calculators, lar e com!uters, !eo!le1s brains, and so forth& This is what the 7hurch-Turin Thesis is all about& 0ut when it comes to real-world understandin , it seems that there is no sim!le way to skim off the to! level, and !ro ram it& alone& The tri erin !atterns of symbols are .ust too com!le(& There must be several levels throu h which thou hts may -!ercolate- and -bubble-& In !articularand this comes back to a ma.or theme of 7ha!ters TI and TIIthe re!resentation of the real world in the brain, althou h rooted in isomor!hism to some e(tent, involves some elements which have no counter!arts at all in the outer world& That is, there is much more to it than sim!le mental structures re!resentin -do -, -broom-, etc& )ll of these symbols e(ist, to be surebut their internal structures are e(tremely com!le( and to a lar e de ree are unavailable for conscious ins!ection& /oreover, one would hunt in vain to ma! each as!ect of a symbol1s internal structure onto some s!ecific feature of the real world&

,rocesses That Are ;ot So S.immable

'or this reason, the brain be ins to look like a very !eculiar formal system, for on its bottom levelthe neural levelwhere the -rules- o!erate and chan e the state, there may be no inter!retation of the !rimitive elements 4neural firin s, or !erha!s even lowerlevel events5& Het on the to! level, there emer es a meanin ful inter!retationa ma!!in from the lar e -clouds- of neural activity which we have been callin -symbols-, onto the real world& There is some resemblance to the 6>del construction, in that a hi h-level isomor!hism allows a hi h level of meanin to be read into strin s= but in the 6>del construction, the hi her-level meanin -rides- on the lower levelthat is, it is derived from the lower level, once the notion of 6>del-numberin has been introduced& 0ut in the brain, the events on the neural level are not sub.ect to real-world inter!retation= they are sim!ly not imitatin anythin & They are there !urely as the substrate to su!!ort the hi her level, much as transistors in a !ocket calculator are there !urely to su!!ort its numbermirrorin activity& )nd the im!lication is that there is no way to skim off .ust the hi hest level and make an isomor!hic co!y in a !ro ram= if one is to mirror the brain !rocesses which allow real-world understandin , then one must mirror some of the lower-level

thin s which are takin !lace, the -lan ua es of the brain-& This doesn1t necessarily mean that one must o all the way down to the level of the hardware, thou h that may turn out to be the case& In the course of develo!in a !ro ram with the aim of achievin an -intelli ent4vi<&, human-like5 internal re!resentation of what is -out there-, at some !oint one will !robably be forced into usin structures and !rocesses which do not admit of any strai htforward inter!retationsthat is, which cannot be directly ma!!ed onto elements of reality& These lower layers of the !ro ram will be able to be understood only by virtue of their catalytic relation to layers above them, rather than because of some direct connection they have to the outer world& 4) concrete ima e of this idea was su ested by the )nteater in the *nt 7ugue, the -indescribably borin ni htmare- of tryin to understand a book on the letter level&5 Personally, I would uess that such multilevel architecture of conce!t-handlin systems becomes necessary .ust when !rocesses involvin ima es and analo ies become si nificant elements of the !ro ramin contrast to !rocesses which are su!!osed to carry out strictly deductive reasonin & Processes which carry out deductive reasonin can be !ro rammed in essentially one sin le level, and are therefore skimmable, by definition& )ccordin to my hy!othesis, then, ima ery and analo ical thou ht !rocesses intrinsically require several layers of substrate and are therefore intrinsically non-skimmable& I believe furthermore that it is !recisely at this same !oint that creativity starts to emer ewhich would im!ly that creativity intrinsically de!ends u!on certain kinds of -uninter!retablelower-level events& The layers of under!innin of analo ical thinkin are, of course, of e(treme interest, and& some s!eculations on their nature will be offered in the ne(t two 7ha!ters&

Articles of $eductionistic +aith

Ene way to think about the relation between hi her and lower levels in the brain is this& Ene could assemble a neural net which, on a local 4neuron-to-neuron5 level, !erformed in a manner indistin uishable from a neural net in a brain, but which had no hi her-level meanin at all& The fact that the lower level is com!osed of interactin neurons does not necessarily force any hi her level of meanin to a!!earno more than the fact that al!habet sou! contains letters forces meanin ful sentences to be found, swimmin about in the bowl& ?i h-level meanin is an o!tional feature of a neural networkone which may emer e as a consequence of evolutionary environmental !ressures& 'i ure "+$ is a dia ram illustratin the fact that emer ence of a hi her level of meanin is o!tional& The u!wards-!ointin arrow indicates that a substrate can occur without a hi her level of meanin , but not vice versa, the hi her level must be derived from !ro!erties of a lower one&

73G5=E BKC! 7loating on neural acti#it), the s)m(ol le#el of the (rain mirrors the world! But neural acti#it) per se, which can (e simulated on a computer, does not create thoughtR that calls for higher le#els of organi@ation!

The dia ram includes an indication of a com!uter simulation of a neural network& This is in !rinci!le feasible, no matter how com!licated the network, !rovided that the behavior of individual neurons can be described in terms of com!utations which a com!uter can carry out& This is a subtle !ostulate which few !eo!le even think of questionin & 8evertheless it is a !iece of -reductionistic faith-= it could be considered a -microsco!ic version- of the 7hurch-Turin Thesis& 0elow we state it e(!licitly, 7?F;7?-TF;I86 T?ESIS, /I7;ES7EPI7 GE;SIE8, The behavior of the com!onents of a livin bein can be simulated on a com!uter& That is, the behavior of any com!onent 4ty!ically assumed to be a cell5 can be calculated by a 'looP !ro ram 4i&e&, eneral recursive function5 to any desired de ree of accuracy, iven a sufficiently !recise descri!tion of the com!onent1s internal state and local environment& This version of the 7hurch-Turin Thesis says that brain !rocesses do not !ossess any more mystiqueeven thou h they !ossess more levels of or ani<ationthan, say, stomach !rocesses& It would be unthinkable in this day and a e to su est that !eo!le di est their food, not by ordinary chemical !rocesses, but by a sort of mysterious and ma ic -assimilation-& This version of the 7T-Thesis sim!ly e(tends this kind of commonsense reasonin to brain !rocesses& In short, it amounts to faith that the brain o!erates in a way which is, in !rinci!le, understandable& It is a !iece of reductionist faith& ) corollary to the /icrosco!ic 7T-Thesis is this rather terse new macrosco!ic version, 7?F;7?-TF;I86 T?ESIS, ;E:F7TIE8IST1S GE;SIE8, )ll brain !rocesses are derived from a com!utable substrate& This statement is about the stron est theoretical under!innin one could ive in su!!ort of the eventual !ossibility of reali<in )rtificial Intelli ence&

Ef course, )rtificial Intelli ence research is not aimed at simulatin neural networks, for it is based on another kind of faith, that !robably there are si nificant features of intelli ence which can be floated on to! of entirely different sorts of substrates than those of or anic brains& 'i ure "+* shows the !resumed relations amon )rtificial Intelli ence, natural intelli ence, and the real world&

,arallel ,rogress in AI and Brain Simulation7

The idea that, if )I is to be achieved, the actual hardware of the brain mi ht one day have to be simulated or du!licated, is, for the !resent at least, quite an abhorrent thou ht to many )I workers& Still one wonders, -?ow finely will we need to co!y the brain to achieve )I?- The real answer is !robably that it all de!ends on how many of the features of human consciousness you want to simulate&

73G5=E BKE! 8rucial to the endea#or of *rtificial 3ntelligence research is the notion that the s)m(olic le#els of the mind can (e "skimmed off" of their neural su(strate and implemented in other media, such as the electronic su(strate of computers! To what depth the cop)ing of (rain must go is at present completel) unclear!

Is an ability to !lay checkers well a sufficient indicator of intelli ence? If so, then )I already e(ists, since checker-!layin !ro rams are of world class& Er is intelli ence an ability to inte rate functions symbolically, as in a freshman calculus class? If so, then )I already e(ists, since symbolic inte ration routines outdo the best !eo!le in most cases& Er is intelli ence the ability to !lay chess well? If so, then )I is well on its way, since chess!layin !ro rams can defeat most ood amateurs= and the level of artificial chess will !robably continue to im!rove slowly& ?istorically, !eo!le have been naPve about what qualities, if mechani<ed, would undeniably constitute intelli ence& Sometimes it seems as thou h each new ste! towards

)I, rather than !roducin somethin which everyone a rees is real intelli ence, merely reveals what real intelli ence is not& If intelli ence involves learnin , creativity, emotional res!onses, a sense of beauty, a sense of self, then there is a lon road ahead, and it may be that these will only be reali<ed when we have totally du!licated a livin brain&

Beauty) the Crab) and the Soul

8ow what, if anythin , does all this have to say about the 7rab1s virtuoso !erformance in front of )chilles? There are two issues clouded to ether here& They are, 4"5 7ould any brain !rocess, under any circumstances, distin uish com!letely reliably between true and false statements of T8T without bein in violation of the 7hurch-Turin Thesisor is such an act in !rinci!le im!ossible? 4B5 Is !erce!tion of beauty a brain !rocess? 'irst of all, in res!onse to 4"5, if violations of the 7hurch-Turin Thesis are allowed, then there seems to be no fundamental obstacle to the stran e events in the :ialo ue& So what we are interested in is whether a believer in the 7hurch-Turin Thesis would have to disbelieve in the 7rab1s ability& 3ell, it all de!ends on which version of the 7T-Thesis you believe& 'or e(am!le, if you only subscribe to the Public-Processes Gersion, then you could reconcile the 7rab1s behavior with it very easily by !ositin that the 7rab1s ability is not communicable& 7ontrariwise, if you believe the ;eductionist1s Gersion, you will have a very hard time believin in the 7rab1s ostensible ability 4because of 7hurch1s Theorem soon to be demonstrated5& 0elievin in intermediate versions allows you a certain amount of wishy-washiness on the issue& Ef course, switchin your stand accordin to convenience allows you to waffle even more& It seems a!!ro!riate to !resent a new version of the 7T-Thesis, one which is tacitly held by vast numbers of !eo!le, and which has been !ublicly !ut forth by several authors, in various manners& Some of the more famous ones are, !hiloso!hers ?ubert :reyfus, S& Jaki, /ortimer Taube, and J& ;& Aucas= the biolo ist and !hiloso!her /ichael Polanyi 4a holist !ar e(cellence5= the distin uished )ustralian neuro!hysiolo ist John Eccles& I am sure there are many other authors who have e(!ressed similar ideas, and countless readers who are sym!athetic& I have attem!ted below to summari<e their .oint !osition& I have !robably not done full .ustice to it, but I have tried to convey the flavor as accurately as I can, 7?F;7?-TF;I86 T?ESIS, SEFAISTS1 GE;SIE8, Some kinds of thin s which a brain can do can be va uely a!!ro(imated on a com!uter but not most, and certainly not the interestin ones& 0ut anyway, even if they all could, that would still leave the soul to e(!lain, and there is no way that com!uters have any bearin on that& This version relates to the tale of the /a nificrab in two ways& In the first !lace, its adherents would !robably consider the tale to be silly and im!lausible, butnot

forbidden in !rinci!le& In the second !lace, they would !robably claim that a!!reciation of qualities such as beauty is one of those !ro!erties associated with the elusive soul, and is therefore inherently !ossible only for humans, not for mere machines& 3e will come back to this second !oint in a moment= but first, while we are on the sub.ect of -soulists-, we ou ht to e(hibit this latest version in an even more e(treme form, since that is the form to which lar e numbers of well-educated !eo!le subscribe these days, 7?F;7?-TF;I86 T?ESIS, T?EE:E;E ;ES9)@ GE;SIE8, 7om!uters are ridiculous& So is science in eneral& This view is !revalent amon certain !eo!le who see in anythin smackin of numbers or e(actitude a threat to human values& It is too bad that they do not a!!reciate the de!th and com!le(ity and beauty involved in e(!lorin abstract structures such as the human mind, where, indeed, one comes in intimate contact with the ultimate questions of what to be human is& 6ettin back to beauty, we were about to consider whether the a!!reciation of beauty is a brain !rocess, and if so, whether it is imitable by a com!uter& Those who believe that it is not accounted for by the brain are very unlikely to believe that a com!uter could !ossess it& Those who believe it is a brain !rocess a ain divide u! accordin to which version of the 7T-Thesis they believe& ) total reductionist would believe that any brain !rocess can in !rinci!le be transformed into a com!uter !ro ram= others, however, mi ht feel that beauty is too ill-defined a notion for a com!uter !ro ram ever to assimilate& Perha!s they feel that the a!!reciation of beauty requires an element of irrationality, and therefore is incom!atible with the very fiber of com!uters&

Irrational and $ational Can Coe-ist on

ifferent Le"els

?owever, this notion that -irrationality is incom!atible with com!uters- rests on a severe confusion of levels& The mistaken notion stems from the idea that since com!uters are faultlessly functionin machines, they are therefore bound to be -lo ical- on all levels& Het it is !erfectly obvious that a com!uter can be instructed to !rint out a sequence of illo ical statementsor, for variety1s sake, a batch of statements havin random truth values& Het in followin such instructions, a com!uter would not be makin any mistakes2 En the contrary, it would only be a mistake if the com!uter !rinted out somethin other than the statements it had been instructed to !rint& This illustrates how faultless functionin on one level may underlie symbol mani!ulation on a hi her leveland the oals of the hi her level may be com!letely unrelated to the !ro!a ation of Truth& )nother way to ain !ers!ective on this is to remember that a brain, too, is a collection of faultlessly functionin elementsneurons& 3henever a neuron1s threshold is sur!assed by the sum of the incomin si nals, 0)862it fires& It never ha!!ens that a neuron for ets its arithmetical knowled ecarelessly addin its in!uts and ettin a wron answer& Even when a neuron dies, it continues to function correctly, in the sense that its com!onents continue to obey the laws of mathematics and !hysics& Het as we all

73G5=E BKI! The (rain is rationalR the mind ma) not (e! U.rawing () the author!V

know, neurons are !erfectly ca!able of su!!ortin hi h-level behavior that is wron , on its own level, in the most ama<in ways& 'i ure "+# is meant to illustrate such a clash of levels, an incorrect belief held in the software of a mind, su!!orted by the hardware of a faultlessly functionin brain& The !ointa !oint which has been made several times earlier in various conte(ts is sim!ly that meanin can e(ist on two or more different levels of a symbol-handlin system, and alon with meanin , ri htness and wron ness can e(ist on all those levels& The !resence of meanin on a iven level is determined by whether or not reality is mirrored in an isomor!hic 4or looser5 fashion on that level& So the fact that neurons always !erform correct additions 4in fact, much more com!le( calculations5 has no bearin whatsoever on the correctness of the to!-level conclusions su!!orted by their machinery& 3hether one1s to! level is en a ed in !rovin kRans of 0oolean 0uddhism or in meditatin on theorems of 9en )l ebra, one1s neurons are functionin rationally& 0y the same token, the hi h-level symbolic !rocesses which in a brain create the e(!erience of a!!reciatin beauty are !erfectly rational on the bottom level, where the faultless functionin is takin !lace= any irrationality, if there is such, is on the hi her level, and is an e!i!henomenona consequenceof the events on the lower level& To make the same !oint in a different way, let us say you are havin a hard time makin u! your mind whether to order a cheesebur er or a !inea!!lebur er& :oes this im!ly that your neurons are also balkin , havin difficulty decidin whether or not to fire? Ef course not& Hour hambur er-confusion is a hi h-level state which fully de!ends on the efficient firin of thousands of neurons in very or ani<ed ways& This is a little ironic, yet it is !erfectly obvious when you think about it& 8evertheless, it is !robably fair to say that nearly all confusions about minds and com!uters have their ori in in .ust such elementary level-confusions& There is no reason to believe that a& com!uter1s faultlessly functionin hardware could not su!!ort hi h-level symbolic behavior which would re!resent such com!le( states as confusion, for ettin , or a!!reciation of beauty& It would require that there e(ist massive subsystems interactin with each other accordin to a com!le( -lo ic-& The overt behavior could a!!ear either rational or irrational= but underneath it would be the !erformance of reliable, lo ical hardware&

More Against Lucas

Incidentally, this kind of level distinction !rovides us with some new fuel in ar uin a ainst Aucas& The Aucas ar ument is based on the idea that 6>del1s Theorem is a!!licable, by definition, to machines& In fact, Aucas makes a most em!hatic !ronunciation,
6>del1s theorem must a!!ly to cybernetical machines, because it is of the essence of bein a machine, that it should be a concrete instantiation of a formal system& #

This is, as we have seen, true on the hardware levelbut since there may be hi her levels, it is not the last word on the sub.ect& 8ow Aucas ives the im!ression that in the

mind-imitatin machines he discusses, there is only one level on which mani!ulation of symbols takes !lace& 'or instance, the ;ule of :etachment 4called -/odus Ponens- in his article5 would be wired into the hardware and would be an unchan eable feature of such a machine& ?e oes further and intimates that if /odus Ponens were not an immutable !illar of the machine1s system, but could be overridden on occasion, then,
The system will have ceased to be a formal lo ical system& and the machine will barely qualify for the title of a model for the mind&"+

8ow many !ro rams which are bein develo!ed in )I research have very little in common with !ro rams for eneratin truths of number theory!ro rams with infle(ible rules of inference and fi(ed sets of a(ioms& Het they are certainly intended as -models for the mind-& En their to! levelthe -informal- levelthere may be mani!ulation of ima es, formulation of analo ies, for ettin of ideas, confusin of conce!ts, blurrin of distinctions, and so forth& 0ut this does not contradict the fact that they rely on the correct functionin of their underlyin hardware as much as brains rely on the correct functionin of their neurons& So )I !ro rams are still -concrete instantiations of formal systems-but they are not machines to which Aucas1 transmo rification of 6>del1s !roof can be a!!lied& Aucas1 ar ument a!!lies merely to their bottom level, on which their intelli encehowever reat or small it may bedoes not lie& There is one other way in which Aucas betrays his oversim!lified vision of how mental !rocesses would have to be re!resented inside com!uter !ro rams& In discussin the matter of consistency, he writes
If we really were inconsistent machines, we should remain content with our inconsistencies, and would ha!!ily affirm both halves of a contradiction& /oreover, we would be !re!ared to say absolutely anythin which we are not& It is easily shown that in an inconsistent formal system everythin is !rovable&""

This last sentence shows that Aucas assumes that the Pro!ositional 7alculus must of necessity be built into any formal system which carries out reasonin & In !articular, he is thinkin of the theorem <<,~,>8> of the Pro!ositional 7alculus= evidently he has the erroneous belief that it is an inevitable feature of mechani<ed reasonin & ?owever, it is !erfectly !lausible that lo ical thou ht !rocesses, such as !ro!ositional reasonin , will emer e as conse"uences of the eneral intelli ence of an )I !ro ram, rather than bein preprogrammed& This is what ha!!ens in humans2 )nd there is no !articular reason to assume that the strict Pro!ositional 7alculus, with its ri id rules and the rather silly definition of consistency that they entail, would emer e from such a !ro ram&

An &nder!inning of AI
3e can summari<e this e(cursion into level distinctions and come away with one final, stron est version of the 7hurch-Turin Thesis,

7?F;7?-TF;I86 T?ESIS, )I GE;SIE8, /ental !rocesses of any sort can be simulated by a com!uter !ro ram whose underlyin lan ua e is of !ower equal to that of 'looPthat is, in which all !artial recursive functions can be !ro rammed& It should also be !ointed out that in !ractice, many )I researchers rely on another article of faith which is closely related to the 7T-Thesis, and which I call the *3 Thesis& It runs somethin like this, )I T?ESIS, )s the intelli ence of machines evolves, its underlyin mechanisms will radually conver e to the mechanisms underlyin human intelli ence& In other words, all intelli ences are .ust variations on a sin le theme= to create true intelli ence, )I workers will .ust have to kee! !ushin to ever lower levels, closer and closer to brain mechanisms, if they wish their machines to attain the ca!abilities which we have&

Church%s Theorem
8ow let us come back to the 7rab and to the question of whether his decision !rocedure for theoremhood 4which is !resented in the uise of a filter for musical beauty5 is com!atible with reality& )ctually, from the events which occur in the :ialo ue, we have no way of deducin whether the 7rab1s ift is an ability to tell theorems from nontheorems, or alternatively, an ability to tell true statements from false ones& Ef course in many cases this amounts to the same thin , but 6>del1s Theorem shows that it doesn1t always& 0ut no matter, both of these alternatives are im!ossible, if you believe the )I Gersion of the 7hurch-Turin Thesis& The !ro!osition that it is im!ossible to have a decision !rocedure for theoremhood in any formal system with the !ower of T8T is known as 8hurch's Theorem& The !ro!osition that it is im!ossible to have a decision !rocedure for number theoretical truthif such truth e(ists, which one can well doubt after meetin u! with all the bifurcations of T8Tfollows quickly from Tarski's Theorem 4!ublished in "#%%, althou h the ideas were known to Tarski considerably earlier5& The !roofs of these two hi hly im!ortant results of metamathematics are very similar& 0oth of them follow quite quickly from self-referential constructions& Aet us first consider the question of a decision !rocedure for T8T-theoremhood& If there were a uniform way by which !eo!le could decide which of the classes -theorem- and -nontheorem- any iven formula T fell into, then, by the 7T-Thesis 4Standard Gersion5, there would e(ist a terminatin 'looP !ro ram 4a eneral recursive function5 which could make the same decision, when iven as in!ut the 6>del number of formula T& The crucial ste! is to recall that any !ro!erty that can be tested for by a terminatin 'looP !ro ram is represented in T8T& This means that the !ro!erty of T8T-theoremhood would be re!resented 4as distin uished from merely e(!ressed5 inside T8T& 0ut as we shall see in a moment, this would !ut us in hot water, for if theoremhood is a re!resentable attribute, then 6>del1s formula 6 becomes as vicious as the E!imenides !arado(&

It all hin es on what 6 says, -6 is not a theorem of T8T-& )ssume that 6 were a theorem& Then, since theoremhood is su!!osedly re!resented, the T8T-formula which asserts -6 is a theorem- would be a theorem of T8T& 0ut this formula is h6, the ne ation of 6, so that T8T is inconsistent& En the other hand, assume 6 were not a theorem& Then once a ain by the su!!osed re!resentability of theoremhood, the formula which asserts -6 is not a theorem- would be a theorem of T8T& 0ut this formula is 6, and once a ain we et into !arado(& Fnlike the situation before, there is no resolution of the !arado(& The !roblem is created by the assum!tion that theoremhood is re!resented by some formula of T8T, and therefore we must backtrack and erase that assum!tion& This forces us also to conclude that no 'looP !ro ram can tell the 6>del numbers of theorems from those of nontheorems& 'inally, if we acce!t the )I Gersion of the 7T-Thesis, then we must backtrack further, and conclude that no method whatsoever could e(ist by which humans could reliably tell theorems from nontheoremsand this includes determinations based on beauty& Those who subscribe only to the Public-Processes Gersion mi ht still think the 7rab1s !erformance is !ossible= but of all the versions, that one is !erha!s the hardest one to find any .ustification for&

Tars.i%s Theorem
8ow let us !roceed to Tarski1s result& Tarski asked whether there could be a way of e(!ressin in T8T the conce!t of number-theoretical truth& That theoremhood is e(!ressible 4thou h not re!resentable5 we have seen= Tarski was interested in the analo ous question re ardin the notion of truth& /ore s!ecifically, he wished to determine whether there is any T8T-formula with a sin le free variable a which can be translated thus, -The formula whose 6>del number is a e(!resses a truth&Aet us su!!ose, with Tarski, that there is onewhich we1ll abbreviate as T;FEkab& 8ow what we1ll do is use the dia onali<ation method to !roduce a sentence which asserts about itself that it is untrue& 3e co!y the 6>del method e(actly, be innin with an -uncle-, a,<~T$&E,a- A$IT/M08&I;E,a11,a-> Aet us say the 6>del number of the uncle is t& 3e arithmoquine this very uncle, and !roduce the Tarski formula T, a,<~T$&E,a- A$IT/M08&I;E,SSS&&& SSS?.a11,a->

b
t S1s

3hen inter!reted, it says,

-The arithmoquinification of t is the 6>del number of a false statement&0ut since the arithmoquinification of t is T1s own 6>del number, Tarski1s formula T re!roduces the E!imenides !arado( to a tee inside T8T, sayin of itself, -I am a falsity-& Ef course, this leads to the conclusion that it must be simultaneously true and false 4or simultaneously neither5& There arises now an interestin matter, 3hat is so bad about re!roducin the E!imenides !arado(? Is it of any consequence? )fter all, we already have it in En lish, and the En lish lan ua e has not one u! in smoke&

The Im!ossibility of the Magnificrab

The answer lies in rememberin that there are two levels of meanin involved here& Ene level is the level we have .ust been usin = the other is as a statement of number theory& If the Tarski formula T actually e(isted, then it would be a statement a(out natural num(ers that is both true and false at once2 There is the rub& 3hile we can always .ust swee! the En lish-lan ua e E!imenides !arado( under the ru , sayin that its sub.ect matter 4its own truth5 is abstract, this is not so when it becomes a concrete statement about numbers2 If we believe this is a ridiculous state of affairs, then we have to undo our assum!tion that the formula T$&EJaK e(ists& Thus, there is no way of e(!ressin the notion of truth inside T8T& 8otice that this makes truth a far more elusive !ro!erty than theoremhood, for the latter is e(!ressible& The same backtrackin reasons as before 4involvin the 7hurchTurin Thesis, )I Gersion5 lead us to the conclusion that The 7rab1s mind cannot be a truth-reco ni<er any more than it is a T8Ttheorem-reco ni<er& The former would violate the Tarski-7hurch-Turin Theorem 4-There is no decision !rocedure for arithmetical truth-5, while the latter would violate 7hurch1s Theorem&

T'o Ty!es of +orm

It is e(tremely interestin , then, to think about the meanin of the word -form- as it a!!lies to constructions of arbitrarily com!le( sha!es& 'or instance, what is it that we res!ond to when we look at a !aintin and feel its beauty? Is it the -form- of the lines and dots on our retina? Evidently it must be, for that is how it ets !assed alon to the analy<in mechanisms in our headsbut the com!le(ity of the !rocessin makes us feel that we are not merely lookin at a two-dimensional surface= we are res!ondin to some sort of inner meanin inside the !icture, a multidimensional as!ect tra!!ed somehow inside those two dimensions& It is the word -meanin - which is im!ortant here& Eur minds contain inter!reters which acce!t two-dimensional !atterns and then -!ull- from them hi h-dimensional notions which are so com!le( that we cannot consciously describe them& The same can be said about how we res!ond to music, incidentally&

It feels sub.ectively that the !ullin -out mechanism of inner meanin is not at all akin to a decision !rocedure which checks for the !resence or absence of some !articular quality such as well-formedness in a strin & Probably this is because inner meanin is somethin which reveals more of itself over a !eriod of time& Ene can never be sure, as one can about well-formedness, that one has finished with the issue& This su ests a distinction that could be drawn between two senses of -form- in !atterns which we analy<e& 'irst, there are qualities such as well-formedness, which can be detected by predicta(l) terminating tests, as in 0looP !ro rams& These I !ro!ose to call syntactic qualities of form& Ene intuitively feels about the syntactic as!ects of form that they lie close to the surface, and therefore they do not !rovoke the creation of multidimensional co nitive structures& 0y contrast, the semantic as!ects of form are those which cannot be tested for in !redictable len ths of time, they require open%ended tests& Such an as!ect is theoremhood of T8T-strin s, as we have seen& Hou cannot .ust a!!ly some standard test to a strin and find out if it is a theorem& Somehow, the fact that its meaning is involved is crucially related to the difficulty of tellin whether or not a strin is a T8T-theorem& The act of !ullin out a strin 1s meanin involves, in essence, establishin all the im!lications of its connections to all other strin s, and this leads, to be sure, down an o!en-ended trail& So -semantic- !ro!erties are connected to o!en-ended searches because, in an im!ortant sense, an ob.ect1s meaning is not locali@ed within the ob.ect itself& This is not to say that no understandin of any ob.ect1s meanin is !ossible until the end of time, for as time !asses, more and more of the meanin unfolds& ?owever, there are always as!ects of its meanin which will remain hidden arbitrarily lon &

Meaning

eri"es from Connections to Cogniti"e Structures

Aet us switch from strin s to !ieces of music, .ust for variety& Hou may still substitute the term -strin - for every reference to a !iece of music, if you !refer& The discussion is meant to be eneral, but its flavor is better otten across, I feel, by referrin to music& There is a stran e duality about the meanin of a !iece of music, on the one hand, it seems to be s!read around, by virtue of its relation to many other thin s in the world and yet, on the other hand, the meanin of a !iece of music is obviously derived from the music itself, so it must be locali<ed somewhere inside the music& The resolution of this dilemma comes from thinkin about the inter!reterthe mechanism which does the !ullin out of meanin & 40y -inter!reter- in this conte(t, I mean not the !erformer of the !iece, but the mental mechanism in the listener which derives meanin when the !iece is !layed&5 The inter!reter may discover many im!ortant as!ects of a !iece1s meanin while hearin it for the first time= this seems to confirm the notion that the meanin is housed in the !iece itself, and is sim!ly bein read off& 0ut that is only !art of the story& The music inter!reter works by settin u! a multidimensional co nitive structurea mental re!resentation of the !iecewhich it tries to inte rate with !re-e(istent information by findin links to other multidimensional mental structures which encode !revious e(!eriences& )s this !rocess takes !lace, the full meanin radually unfolds& In fact, years may !ass before someone comes to feel that he has

!enetrated to the core meanin of a !iece& This seems to su!!ort the o!!osite view, that musical meanin is s!read around, the inter!reter1s role bein to assemble it radually& The truth undoubtedly lies somewhere in between, meanin sboth musical and lin uisticare to some e(tent locali<able, to some e(tent s!read around& In the terminolo y of 7ha!ter GI, we can say that musical !ieces and !ieces of te(t are !artly tri ers, and !artly carriers of e(!licit meanin & ) vivid illustration of this dualism of meanin is !rovided by the e(am!le of a tablet with an ancient inscri!tion, the meanin is !artially stored in the libraries and the brains of scholars around the world, and yet it is also obviously im!licit in the tablet itself& Thus, another way of characteri<in the difference between -syntactic- and -semantic- !ro!erties 4in the .ust-!ro!osed sense5 is that the syntactic ones reside unambi uously inside the ob.ect under consideration, whereas semantic !ro!erties de!end on its relations with a !otentially infinite class of other ob.ects, and therefore are not com!letely locali<able& There is nothin cry!tic or hidden, in !rinci!le, in syntactic !ro!erties, whereas hiddenness is of the essence in semantic !ro!erties& That is the reason for my su ested distinction between -syntactic- and -semantic- as!ects of visual form&

Beauty) Truth) and +orm

3hat about beauty? It is certainly not a syntactic !ro!erty, accordin to the ideas above& Is it even a semantic !ro!erty? Is beauty a !ro!erty which, for instance, a !articular !aintin has? Aet us immediately restrict our consideration to a sin le viewer& Everyone has had the e(!erience of findin somethin beautiful at one time, dull another timeand !robably intermediate at other times& So is beauty an attribute which varies in time? Ene could turn thin s around and say that it is the beholder who has varied in time& 6iven a !articular beholder of a !articular !aintin at a !articular time, is it reasonable to assert that beauty is a quality that is definitely !resent or absent? Er is there still somethin illdefined and intan ible about it? :ifferent levels of inter!reter !robably could be invoked in every !erson, de!endin on the circumstances& These various inter!reters !ull out different meanin s, establish different connections, and enerally evaluate all dee! as!ects differently& So it seems that this notion of beauty is e(tremely hard to !in down& It is for this reason that I chose to link beauty, in the /a nificrab, with truth, which we have seen is also one of the most intan ible notions in all of metamathematics&

The ;eural Substrate of the E!imenides ,aradoI would like to conclude this 7ha!ter with some ideas about that central !roblem of truth, the E!imenides !arado(& I think the Tarski re!roduction of the E!imenides !arado( inside T8T !oints the way to a dee!er understandin of the nature of the E!imenides !arado( in En lish& 3hat Tarski found was that his version of the !arado( has two distinct levels to it& En one level, it is a sentence about itself which would be true if it were false, and false if it were true& En the other levelwhich I like to call the arithmetical substrateit is a sentence about inte ers which is true if and only if false&

8ow for some reason this latter bothers !eo!le a lot more than the former& Some !eo!le sim!ly shru off the former as -meanin less-, because of its self-referentiality& 0ut you can1t shru off !arado(ical statements about inte ers& Statements about inte ers sim!ly cannot be both true and false& 8ow my feelin is that the Tarski transformation of the E!imenides !arado( teaches us to look for a substrate in the En lish-lan ua e version& In the arithmetical version, the u!!er level of meanin is su!!orted by the lower arithmetical level& Perha!s analo ously, the self-referential sentence which we !erceive 4-This sentence is false-5 is only the to! level of a dual-level entity& 3hat would be the lower level, then? 3ell, what is the mechanism that lan ua e rides on? The brain& Therefore one ou ht to look for a neural su(strate to the E!imenides !arado(a lower level of !hysical events which clash with each other& That is, two events which by their nature cannot occur simultaneously& If this !hysical substrate e(ists, then the reason we cannot make heads or tails of the E!imenides sentence is that our brains are tryin to do an im!ossible task& 8ow what would be the nature of the conflictin !hysical events? Presumably when you hear the E!imenides sentence, your brain sets u! some -codin - of the sentencean internal confi uration of interactin symbols& Then it tries to classify the sentence as -true- or -false-& This classifyin act must involve an attem!t to force several symbols to interact in a !articular way& 4Presumably this ha!!ens when any sentence is !rocessed&5 8ow if it ha!!ens that the act of classification would !hysically disru!t the codin of the sentencesomethin which would ordinarily never ha!!enthen one is in trouble, for it is tantamount to tryin to force a record !layer to !lay its self-breakin record& 3e have described the conflict in !hysical terms, but not in neural terms& If this analysis is ri ht so far, then !resumably the rest of the discussion could be carried on when we know somethin about the constitution of the -symbols- in the brain out of neurons and their firin s, as well as about the way that sentences become converted into -codin s-& This sketch of the neural substrate of the E!imenides !arado( su ests 4to me, at least5 that the resolution of the En lish version of the E!imenides !arado( mi ht be similar to that for the Tarski version& The resolution involves abandonin the notion that a brain could ever !rovide a fully accurate re!resentation for the notion of truth& The novelty of this resolution lies in its su estion that a total modelin of truth is im!ossible for quite ph)sical reasons, namely, such a modelin would require !hysically incom!atible events to occur in a brain&

SH0)LU,
Toy of $an,s )esi!nin!$
:ne da), Eta :in wanders into the '3T *rtificial 3ntelligence +a(orator), where she runs into the (rilliant )oung computer program, -,=.+5! 3t happens that -,=.+5 is Must d)ing for someone to tr) out the newl) de#eloped human (eing ".r! Ton) Earrwig"! -,=.+5 e9plains that this Earrwig is fairl) intelligent in the limited domain of anal)@ing a con#ersation a(out a "to) world" containing (locks of #arious shapes, si@es, and colors% (locks which sit on a ta(le, and which can (e picked up and mo#ed around! Eta :in, fascinated, (egins enthusiasticall) t)ping to -,=.+5! Earrwig dangles o#er her shoulder and pro#ides a sort of running commentar), as promised! "& Eta :in, !ick u! a bi red block& U;efer to 'i & ""+&V -,=.+5, E@& .r! Ton) Earrwig: -,=.+5 answers ":;" when it carries out a command! 3n order to pick up the red (lock, it had to clear it off () finding a space for the green one and mo#ing the green one awa)! Eta :in, ras! the !yramid& -,=.+5, I :E81T F8:E;ST)8: 3?I7? PH;)/I: HEF /E)8& .r! Ton) Earrwig: There are three p)ramids in the scene, and -,=.+5 knows that phrases (eginning with "the" are intended to refer to a specific o(Mect the speaker has in mind!

B&

73G5=E BBK! "&ick up a (ig red (lock!" U7rom Terr) >inograd, Fnderstandin 8atural Aan ua e, New York: *cademic &ress, BICF/, p! E!V

73G5=E BBB! "7ind a (lock which is (igger than the one )ou are holding and put it into the (o9!" *dapted from Terr) >inograd, Fnderstandin 8atural Aan ua e, p! I!V

%&

C& L&

D&

Eta :in, find a block which is taller than the one you are holdin and !ut it into the bo(& U;efer to 'i & """&V .r! Ton) Earrwig: -,=.+5 has to figure out what is meant () "one" and "it", as well as using deduction to decide which (lock it is holding! -,=.+5, 0H -IT-, I )SSF/E HEF /E)8 T?E 0AE7@ 3?I7? IS T)AAE; T?)8 T?E E8E I )/ ?EA:I86& .r! Ton) Earrwig: There were two possi(le referents for "it": the one selected, and "the one )ou are holding"! -,=.+5 has heuristics for choosing (etween them! -,=.+5, E@& .r! Ton) Earrwig: -,=.+5's hand has to set down the red (lock pick up the taller (lue one! Eta :in, what does the bo( contain? -,=.+5, T?E 0AFE PH;)/I: )8: T?E 0AFE 0AE7@& Eta :in, what is the !yramid su!!orted by? .r! Ton) Earrwig: Earlier -,=.+5 responded to "the p)ramid" () sa)ing " 3 .:N'T 5N.E=-T*N. >,38, &Y=*'3. Y:5 'E*N !" Now, it assumes that )ou mean the one Must mentioned! -,=.+5, T?E 0ET& Eta :in, how many blocks are not in the bo(? .r! Ton) Earrwig: Negati#es such as "not" and "no" are handled whether the) appear in clauses or noun groups such as "no (o9es"/& -,=.+5, 'EF; E' T?E/&

$&

*&

#&

"+&

""&

"B& "%&

"C&

Eta :in, is at least one of them narrower than the one which i told you to !ick u!? .r! Ton) Earrwig: The phrase "at least one of them" must (e anal)@ed to refer (ack to the pre#ious "uestion, and "the one which!!!" must also refer to (locks! Noun groups can ha#e complicated num(ers such as "at least one", "more than three", "e9actl) se#en", etc!! The memor) of past e#ents is used to deduce which (lock is "the one which 3 told )ou to pick up"! Notice the complicated s)ntactic structure in which "the one" is actuall) the o(Mect of a clause em(edded two le#els deeper! -,=.+5's parser incorporates a grammar that handles this kind of comple9it) in a general wa)! -,=.+5, HES, T?E ;E: 7F0E& Eta :in, is it su!!orted? -,=.+5, HES, 0H T?E T)0AE& .r! Ton) Earrwig: -,=.+5 has heuristics a(out what information is rele#ant to a person who asks a "uestion, e#en if the person does not re"uest that information specificall)! Eta :in, can the table !ick u! blocks? -,=.+5, 8E& .r! Ton) Earrwig: -,=.+5's semantic rules include the knowledge that "picking up" must (e done () an animate o(Mect, and that the ta(le is not animate! Eta :in, can a !yramid be su!!orted by a block? -,=.+5, HES& .r! Ton) Earrwig: -,=.+5's deducti#e s)stem finds an actual e9ample, so it knows this is possi(le& Eta :in, can a !yramid su!!ort a !yramid? -,=.+5, I :E81T @8E3& .r! Ton) Earrwig: The "uestion makes sense, (ut there is no e9ample in the scene! -,=.+5 has no wa) of looking into the details of its programs, e#en though these ultimatel) define its capa(ilities& Eta :in, stack u! two !yramids& -,=.+5, I 7)8T& .r! Ton) Earrwig: -,=.+5 tries and fails! Eta :in, the blue !yramid is mine& -,=.+5, I F8:E;ST)8:& .r! Ton) Earrwig: *ll of the information a(out colors, shapes, si@es and locations of the o(Mects is known to -,=.+5 initiall)! &ossession is left as an unknown propert) of o(Mects a(out which it can (e told! 3t stores this information in its data (ase! Eta :in, i own all blocks which are not red, but i don1t own anythin which su!!orts a !yramid& -,=.+5, I F8:E;ST)8:&

73G5=E BBF! ">ill )ou please stack up (oth of the red (locks and either a green cu(e or a p)ramid$" U*dapted from Terr) >inograd, Fnderstandin 8atural Aan ua e, p! BF!V

"L& "D& "$&

"*&

.r! Ton) Earrwig: The result of understanding these sentences is the creation of new procedures within -,=.+5 for deducing whether )ou own something! Eta :in, do i own the bo(? -,=.+5, 8E& .r! Ton) Earrwig: The (o9 supports whate#er it contains, including a p)ramid! Eta :in, do i own anythin in the bo(? -,=.+5, HES, T3E T?I86S, T?E 0AFE 0AE7@ )8: PH;)/I:& Eta :in, will you !lease stack u! both of the red blocks and either a reen cube or a !yramid? U;efer to 'i & ""B&V -,=.+5, E@& .r! Ton) Earrwig: +ogical connecti#es, such as "and", "or", "either", etc! are handled in (oth the grammar and semantics! * command is e9ecuted e#en if it looks like a "uestion! The! command is carried out () the following series of steps: -,=.+5 puts a green cu(e on the large red (lock it chooses the green cu(e with nothing on it/, then remo#es the small p)ramid from the little red cu(e, so that it can use that cu(e to complete the stack as specified! Eta :in, which cube is sittin on the table? -,=.+5, T?E A);6E 6;EE8 E8E 3?I7? SFPPE;TS T?E ;E: PH;)/I:& .r! Ton) Earrwig: :rdinaril) -,=.+5 would (e satisfied with calling this "* B3G G=EEN 85BE", (ut since )ou asked "which cu(e" it has to (e more specific in

naming it! 3t has procedures for using su(stitute nouns such as "one"/ and pronouns in answers to make the discourse more natural! "#& Eta :in, is there a lar e block behind a !yramid? -,=.+5, HES, T?;EE E' T?E/, ) A);6E ;E: E8E, ) A);6E 6;EE8 7F0E )8:
T?E 0AFE E8E&

B+& Eta :in, !ut a small one onto the reen cube which su!!orts a !yramid& .r! Ton) Earrwig: "* small one" must contrast with )our earlier phrase "a large (lock", so this means "a small (lock"! -,=.+5, E@& B"& Eta :in, !ut the littlest !yramid on to! of it& .r! Ton) Earrwig: >ords like "littlest" are not in -,=.+5's dictionar) (ut are interpreted from the root farms like "little"! -,=.+5 compares the si@es of all the p)ramids to carr) out the action! -,=.+5, E@& BB& Eta :in, how many thin s are on to! of reen cubes? -,=.+5, I1/ 8ET SF;E 3?)T HEF /E)8 0H -E8 TEP E'- I8 T?E P?;)SE -E8
TEP E' 6;EE8 7F0ES:E HEF /E)8 "-:I;E7TAH E8 T?E SF;')7E B-)8H3?E;E E8 TEP E' IT?

.r! Ton) Earrwig: 3n the pre#ious sentence, the phrase "on top of" was interpreted without trou(le (ecause the conte9t made it clear what was meant! 3n this one, the heuristics for disam(iguation do not come up with a clearcut definition, so -,=.+5 had to ask )ou for more information! The paraphrases were included with the dictionar) definitions of "on"! * re"uest like this can also occur when the meaning of a pronoun such as "it" or "the)" is unclear! B%& Eta :in, had you touched any !yramid before you !ut the reen one on the little cube? -,=.+5, HES, T?E 6;EE8 E8E& .r! Ton) Earrwig: -,=.+5 deduces from its memor) what e#ent is (eing referred to, and can relate other e#ents to it in time! The anal)sis includes comple9 tenses of #er( groups, and there are special facilities for keeping track of the #arious forms of irregular #er(s like "ha#e-& BC& Eta :in, when did you !ick it u!? -,=.+5, 3?IAE I 3)S ST)7@I86 FP T?E ;E: 7F0E, ) A);6E ;E: 0AE7@ )8:
) A);6E 6;EE8 E8E&

.r! Ton) Earrwig: * time is descri(ed () naming the main goal which was (eing worked on! BL& Eta :in, why? .r! Ton) Earrwig: -,=.+5 remem(ers immediatel) pre#ious sentences to understand "uestions like this and the following ones! -,=.+5, TE 6ET ;I: E' IT& .r! Ton) Earrwig: B) keeping track of selected parts of the original su(goal tree, -,=.+5 has some understanding of its own moti#es! BD& Eta :in, why did you do that?

-,=.+5, TE 7AE)8 E'' T?E ;E: 7F0E& .r! Ton) Earrwig: "That" refers to the action mentioned in -,=.+5's last answer! B$& Eta :in, why did you clear off that cube? .r! Ton) Earrwig: ,ere, "that" means the one Must referred to! -,=.+5, TE PFT IT E8 ) A);6E 6;EE8 7F0E& B*& Eta :in, why did you do that? -,=.+5, TE ST)7@ FP T?E ;E: 7F0E, ) A);6E ;E: 0AE7@, )8: ) A);6E
6;EE8 7F0E&

B#& Eta :in, why did you do that? -,=.+5, 0E7)FSE HEF )S@E: /E TE& .r! Ton) Earrwig: -,=.+5 remem(ers what it did, not how the re"uest was worded& %+& Eta :in, how did you do it? .r! Ton) Earrwig: "3t" refers to the action mentioned in the last "uestion, rather than the answer! -,=.+5, 0H PFTTI86 ) A);6E ;E: 0AE7@ E8 T?E T)0AE= T?E8 AETTI86 E'
IT= T?E8 PFTTI86 ) A);6E 6;EE8 7F0E E8 IT= T?E8 AETTI86 6E E' T?)T 7F0E= T?E8 PFTTI86 T?E ;E: 7F0E E8 T?)T 7F0E= T?E8 AETTI86 6E E' T?)T 7F0E&

.r! Ton) Earrwig: This is a description of the top%le#el su(goals used to achie#e the goal! %"& Eta :in, how many ob.ects did you touch while you were doin it? .r! Ton) Earrwig: Time references can refer to a time span, using words like "while"! -,=.+5, 'EF; E' T?E/& %B& Eta :in, what did the red cube su!!ort before you started to clean it off? -,=.+5, T?E 6;EE8 PH;)/I:& .r! Ton) Earrwig: B) remem(ering what has (een mo#ed, it is possi(le to deduce things a(out earlier states of the scene! You can use comple9 nestings of clauses like "(efore )ou started to clean it off "! %%& Eta :in, there were five blocks to the left of the bo( then& .r! Ton) Earrwig: "Then" must refer to the time mentioned in the last sentence! -,=.+5, 8E, E8AH 'EF; E' T?E/, T?E ;E: 7F0E, T3E A);6E 6;EE8 7F0ES,
)8: ) A);6E ;E: 0AE7@&

.r! Ton) Earrwig: Earlier in the dialogue, information a(out "owning" was accepted at face #alue! But -,=.+5 knows that it has complete information a(out the (asic properties of o(Mects, including location! 3t checks an) statements a(out these to see whether it agrees or disagrees! %C& Eta :in, !ut the blue !yramid on the block in the bo(& .r! Ton) Earrwig: This is s)ntacticall) am(iguous, (ut is understood without am(iguit) since the parsing corresponding to "put the (lue p)ramid which is on the (lock into the (o9" does not make sense in the current scene, whereas "put the (lue p)ramid onto the (lock which is in the (o9" does! -,=.+5, E@& %L& Eta :in, is there anythin which is bi er than every !yramid but is not as wide as the thin that su!!orts it?

%D&

%$&

%*& %#&

C+& C"& CB&

.r! Ton) Earrwig: -,=.+5 understands "uantifiers such as "e#er)", "an)", "some", etc!, and structures such as "(igger than", "as wide as", etc!! ,ere "it" refers to a noun group within the same sentence (ut two le#els up in the parsing tree! -,=.+5, HES, T?E 0AFE 0AE7@& Eta :in, does a stee!le? -,=.+5, SE;;H, I :E81T @8E3 T?E 3E;: -STEEPAE-& .r! Ton) Earrwig: -,=.+5 interrupts immediatel) when it sees a word it doesn't know! Eta :in, a -stee!le- is a stack which contains two reen cubes and a !yramid& .r! Ton) Earrwig: The "uote marks indicate that )ou are defining a new word! The definition is added to -,=.+5's dictionar) in a form compati(le with all the words it alread) knows! -,=.+5, I F8:E;ST)8:& Eta :in, are there any stee!les now? -,=.+5, 8E& .r! Ton) Earrwig: The new definition is used as a description to recogni@e steeples! Eta :in, build one& -,=.+5, E@& .r! Ton) Earrwig: The description gi#en in the definition is enough to deduce a plan for (uilding new o(Mects as well as recogni@ing them! The word "one" means "a steeple", and the plan is carried out! 7irst, one of the large green cu(es is cleared off () remo#ing the red cu(e and two p)ramids! Then the two green cu(es and the tall red p)ramid are stacked up to fit the description! Eta Ein, call the bi est block -su!erblock-& -,=.+5, E@& .r! Ton) Earrwig: You can also gi#e o(Mects names! Eta :in, have you !icked u! su!erblock since we be an? -,=.+5, HES& Eta :in, why did you dro! it? -,=.+5, 0H -IT-, I )SSF/E HEF /E)8 T?E SFPE;0AE7@& I 7)81T ETPA)I8 )
8E8-ETISTE8T EGE8T&

.r! Ton) Earrwig: There is no record in -,=.+5's memor) of an) (lock (eing dropped! The answer routines ha#e special checks for this t)pe of "loaded "uestion" which presupposes things which aren't true! C%& Eta :in: is there anythin to the ri ht of the red !yramid? -,=.+5, HES, 'EF; T?I86S, T?E 0ET, T?E ;E: 7F0E, T?E 0AFE PH;)/I: )8:
T?E 0AE7@ 8)/E: SFPE;0AE7@&

.r! Ton) Earrwig: :nce an o(Mect has a name, the name is used to help identif) it in responses! CC& Eta :in, thank you& -,=.+5, HEF1;E 3EA7E/E2 .r! Ton) Earrwig: -,=.+5's grammar is written as a procedure, which gi#es -,=.+5 the capa(ilit) to carr) out special sorts of parsing for idioms&

C/A,TE$ <4III

Artificial Intelligence5 $etros!ects

Turing
I8 "#L+, )A)8 TF;I86 wrote a most !ro!hetic and !rovocative article on )rtificial Intelli ence& It was entitled -7om!utin /achinery and Intelli ence- and a!!eared in the .ournal 'ind& I will say some thin s about that article, but I would like to !recede them with some remarks about Turin the man& )lan /athison Turin was born in Aondon in "#"B& ?e was a child full of curiosity and humor& 6ifted in mathematics, he went to 7ambrid e where his interests in machinery and mathematical lo ic cross-fertili<ed and resulted in his famous !a!er on -com!utable numbers-, in which he invented the theory of Turin machines and demonstrated the unsolvability of the haltin !roblem= it was !ublished in "#%$& In the "#C+1s, his interests turned from the theory of com!utin machines to the actual buildin of real com!uters& ?e was a ma.or fi ure in the develo!ment of com!uters in 0ritain, and a staunch defender of )rtificial
73G5=E BBH! *lan Turing, after a successful race 'a), BIGK/! U7rom -ara Turing, )lan /& Turin 8am(ridge, 5! ;!:>! ,effer W -ons, BIGI/!

Intelli ence when it first came under attack& Ene of his best friends was :avid 7ham!ernowne 4who later worked on com!uter com!osition of music5& 7ham!ernowne and Turin were both avid chess !layers and invented -round-the-house- chess, after your move, run around the houseif you et back before your o!!onent has moved, you1re entitled to another move& /ore seriously, Turin and 7ham!ernowne invented the first chess !layin !ro ram, called -Turocham!-& Turin died youn , at C"a!!arently of an accident with chemicals& Er some say suicide& ?is mother, Sara Turin , wrote his bio ra!hy& 'rom the !eo!le she quotes, one ets the sense that Turin was hi hly unconventional, even auche in some ways, but so honest and decent that he was vulnerable to the world& ?e loved ames, chess, children, and bike ridin = he was a stron lon -distance runner& )s a student at 7ambrid e, he bou ht himself a second-hand violin and tau ht himself to !lay& Thou h not very musical, he derived a reat deal of en.oyment from it& ?e was somewhat eccentric, iven to reat bursts of ener y in the oddest directions& Ene area he e(!lored was the !roblem of mor!ho enesis in biolo y& )ccordin to his mother, Turin -had a !articular fondness for the &ickwick &apers-, but -!oetry, with the e(ce!tion of Shakes!eare1s, meant nothin to him&- )lan Turin was one of the true !ioneers in the field of com!uter science&

The Turing Test

Turin 1s article be ins with the sentence, -I !ro!ose to consider the question 17an machines think?1 - Since, as he !oints out, these are loaded terms, it is obvious that we should search for an o!erational way to a!!roach the question& This, he su ests, is contained in what he calls the -imitation ame-= it is nowadays known as the Turing test& Turin introduces it as follows,
It is !layed with three !eo!le, a man 4)5, a woman 405, and an interro ator 475 who may be of either se(& The interro ator stays in a room a!art from the other two& The ob.ect of the ame for the interro ator is to determine which of the other two is the man and which is the woman& ?e knows them by labels T and H, and at the end of the ame he says either -T is ) and H is 0- or -T is 0 and H is )-& The interro ator is allowed to !ut questions to ) and 0 thus, 7, 3ill T !lease tell me the len th of his or her hair? 8ow su!!ose T is actually ), then ) must answer& It is )1s ob.ect in the ame to try to cause 7 to make the wron identification& ?is answer mi ht therefore be -/y hair is shin led, and the lon est strands are about nine inches lon &In order that tones of voice may not hel! the interro ator the answers should be written, or better still, ty!ewritten& The ideal arran ement is to have a tele!rinter communicatin between the two rooms& )lternatively the questions and answers can be re!eated by an intermediary& The ob.ect of the ame for the third !layer 405 is to hel! the interro ator& The best strate y for her is !robably to ive truthful answers& She can add such thin s as -I am the woman, don1t listen to him2- to her answers, but it will avail nothin as the man can make similar remarks& 3e now ask the question, -3hat will ha!!en when a machine takes the !art of ) in this ame 3ill the interro ator decide wron ly as often when the ame is !layed like this as he does

when the ame is !layed between a man and a woman? These questions re!lace our ori inal, -7an machines think?-B

)fter havin s!elled out the nature of his test, Turin oes on to make some commentaries on it, which, iven the year he was writin in, are quite so!histicated& To be in with, he ives a short hy!othetical dialo ue between interro ator and interro atee,%
M, Please write me a sonnet on the sub.ect of the 'orth 0rid e Ua brid e over the 'irth of 'orth, in ScotlandV& ), 7ount me out on this one& I never could write !oetry& M, )dd %C#L$ to $+$DC& ), 4Pause about %+ seconds and then ive as answer5 "+LDB"& M, :o you !lay chess? ), Hes& M, I have @ at my @", and no other !ieces& Hou have only @ at @D and ; at ;"& It is your move& 3hat do you !lay? ), 4)fter a !ause of "L seconds5 ;-;* mate&

'ew readers notice that in the arithmetic !roblem, not only is there an inordinately lon delay, but moreover, the answer iven is wron 2 This would be easy to account for if the res!ondent were a human, a mere calculational error& 0ut if the res!ondent were a machine, a variety of e(!lanations are !ossible& ?ere are some, 4"5 4B5 4%5 4C5 4L5 a run-time error on the hardware level 4i&e&, an irre!roducible fluke5= an unintentional hardware 4or !ro rammin 5 4re!roducibly5 causes arithmetical mistakes= a !loy deliberately inserted by the machine1s !ro rammer 4or builder5 to introduce occasional arithmetical mistakes, so as to trick interro ators= an unantici!ated e!i!henomenon, the !ro ram has a hard time thinkin abstractly, and sim!ly made -an honest mistake-, which it mi ht not make the ne(t time around= a .oke on the !art of the machine itself, deliberately teasin its interro ator&

;eflection on what Turin mi ht have meant by this subtle touch o!ens u! .ust about all the ma.or !hiloso!hical issues connected with )rtificial Intelli ence& Turin oes on to !oint out that
The new !roblem has the advanta e of drawin a fairly shar! line between the !hysical and the intellectual ca!acities of a man&&& 3e do not wish to !enali<e the machine for its inability to shine in beauty com!etitions, nor to !enali<e a man for losin in a race a ainst an air!lane& C

Ene of the !leasures of the article is to see how far Turin traced out each line of thou ht, usually turnin u! a seemin contradiction at some sta e and, by refinin his conce!ts, resolvin it at a dee!er level of analysis& 0ecause of this de!th of !enetration into the issues, the article still shines after nearly thirty years of tremendous !ro ress in com!uter

develo!ment and intensive work in )I& In the followin short e(cer!t you can see some of this rich back-and-forth workin of ideas,
The ame may !erha!s be critici<ed on the round that the odds are wei hted too heavily a ainst the machine& If the man were to try to !retend to be the machine he would clearly make a very !oor showin & ?e would be iven away at once by slowness and inaccuracy in arithmetic& /ay not machines carry out somethin which ou ht to be described as thinkin but which is very different from what a man does? This ob.ection is a very stron one, but at least we can say that if, nevertheless, a machine can be constructed to !lay the imitation ame satisfactorily, we need not be troubled by this ob.ection& It mi ht be ur ed that when !layin the -imitation ame- the best strate y for the machine may !ossibly be somethin other than imitation of the behaviour of a man& This may be, but I think it is unlikely that there is any reat effect of this kind& In any case there is no intention to investi ate here the theory of the ame, and it will be assumed that the best strate y is to try to !rovide answers that would naturally be iven by a man&L

Ence the test has been !ro!osed and discussed, Turin remarks,
The ori inal question -7an machines think?- I believe to be too meanin less to deserve discussion& 8evertheless, I believe that at the end of the century the use of words and eneral educated o!inion will have altered so much that one will be able to s!eak of machines thinkin without e(!ectin to be contradicted&D

Turing Antici!ates 0bVections

)ware of the storm of o!!osition that would undoubtedly reet this o!inion, he then !roceeds to !ick a!art, concisely and with wry humor, a series of ob.ections to the notion that machines could think& 0elow I list the nine ty!es of ob.ections he counters, usin his own descri!tions of them&$ Fnfortunately there is not s!ace to re!roduce the humorous and in enious res!onses he formulated& Hou may en.oy !onderin the ob.ections yourself, and fi urin out your own res!onses&
4"5 The Theological :(Mection& Thinkin is a function of man1s immortal soul& 6od has iven an immortal soul to every man and woman, but not to any other animal or to machines& ?ence no animal or machine can think& 4B5 The ",eads in the -and" :(Mection& The consequences of machines thinkin would be too dreadful& Aet us ho!e and believe that they cannot do so& 4%5 The 'athematical :(Mection& UThis is essentially the Aucas ar ument&V 4C5 The *rgument from 8onsciousness& -8ot until a machine can write a sonnet or com!ose a concerto because of thou hts and emotions felt, and not by the chance fall of symbols, could we a ree that machine equals brainy that is, not only write it but know that it had written it& 8o mechanism could feel 4and not merely artificially si nal, an easy contrivance5 !leasure at its successes, rief when its valves fuse, be warmed by flattery, be made miserable by its mistakes, be charmed by se(, be an ry or de!ressed when it cannot et what it wants&- U) quote from a certain Professor Jefferson&V

Turin is quite concerned that he should answer this serious ob.ection in full detail& )ccordin ly, he devotes quite a bit of s!ace to his answer, and in it he offers another short hy!othetical dialo ue,*

Interro ator, In the first line of your sonnet which reads -Shall I com!are thee to a summer1s day-, would not -a s!rin day- do as well or better? 3itness, It wouldn1t scan& Interro ator, ?ow about -a winter1s day11? That would scan all ri ht& 3itness, Hes, but nobody wants to be com!ared to a winter1s day& Interro ator, 3ould you say /r& Pickwick reminded you of 7hristmas? 3itness, In a way& Interro ator, Het 7hristmas is a winter1s day, and I do not think /r& Pickwick would mind the com!arison& 3itness, I don1t think you1re serious& 0y a winter1s day one means a ty!ical winter1s day, rather than a s!ecial one like 7hristmas&

)fter this dialo ue, Turin asks, -3hat would Professor Jefferson say if the sonnetwritin machine was able to answer like this in the #i#a #oce?'urther ob.ections,
4L5 *rguments from #arious .isa(ilities& These ar uments take the form, -I rant you that you can make machines do all the thin s that you have mentioned but you will never be able to make one to do T&- 8umerous features T are su ested in this connection& I offer a selection, 0e kind, resourceful, beautiful, friendly, have initiative, have a sense of humor, tell ri ht from wron , make mistakes, fall in love, en.oy strawberries and cream, make someone fall in love with it, learn from e(!erience, use words !ro!erly, be the sub.ect of its own thou ht, have as much diversity of behaviour as a man, do somethin really new& 4D5 +ad) +o#elace's :(Mection& Eur most detailed information of 0abba e1s )nalytical En ine comes from a memoir by Aady Aovelace& In it she states, -The )nalytical En ine has no !retensions to ori inate anythin & It can do whate#er we know how to order it to !erform4her italics5& 4$5 *rgument from 8ontinuit) in the Ner#ous -)stem & The nervous system is certainly not a discrete state machine& ) small error in the information about the si<e of a nervous im!ulse im!in in on a neuron may make a lar e difference to the si<e of the out oin im!ulse& It may be ar ued that, this bein so, one cannot e(!ect to be able to mimic the behaviour of the nervous system with a discrete state system& 4*5 The *rgument from 3nformalit) of Beha#iour& It seems to run somethin like this& -If each man had a definite set of rules of conduct by which he re ulated his life he would be no better than a machine& 0ut there are no such rules, so men cannot be machines&4#5 The *rgument from E9tra%-ensor) &erception & Aet us !lay the imitation ame, usin as witnesses a man who is ood as a tele!athic receiver, and a di ital com!uter& The interro ator can ask such questions as -3hat suit does the card in my ri ht hand belon to?- The man by tele!athy or clairvoyance ives the ri ht answer "%+ times out of C++ cards& The machine can only uess at random, and !erha!s et "+C ri ht, so the interro ator makes the ri ht identification&

)s you can see, many of the ob.ections overla! and are related to each other& I have attem!ted in this book to answer most of them in some way or other, usually far less concisely than Turin did& Eb.ection 4#5 I find remarkable& I have seen the Turin !a!er re!rinted in a bookbut with ob.ection 4#5 omittedwhich I find equally remarkable& /y own !oint of viewcontrary to Turin 1sis that ESP does not e(ist& Turin was reluctant to acce!t the idea that ESP is real, but did so nonetheless, bein com!elled by his

outstandin scientific inte rity to acce!t the consequences of what he viewed as !owerful statistical evidence in favor of ESP& I disa ree, thou h I consider it an e(ceedin ly com!le( and fascinatin question&

(,arry Encounters the

octor(

In the :ialo ue !recedin this 7ha!ter, you have seen an authentic e(chan e between a com!uter !ro ram and a human& 8ow there have been many !ro rams with less ability to -converse- than S?;:AF& Two rather famous ones are -:octor-, created by Jose!h 3ei<enbaum, and -Parry-, created by @enneth 7olby& The former is su!!osed to simulate a !sychiatrist usin -nondirective thera!y-, the latter to simulate the belief structure of a !aranoid& )n amusin head-on collision between the two !ro rams was arran ed over a com!uter network, and the followin is a short e(cer!t from the full transcri!t,#
Parry, I don1t understand your motives& :octor, 3hy don1t you understand my motives? Parry, Shouldn1t I be? :octor, 3hy do you say that .ust now? Parry, That1s .ust the way it is& :octor, Su!!ose that1s .ust the way it were not& Parry, So? :octor, I am not sure I understand you fully& Parry, Hou har! on the same questions too much& :octor, 3hat does that similarity su est to you? Parry, I don1t et you& :octor, :o you wish that you did et I? Parry, It1s about time you showed some interest in my feelin s& :octor, Eh? I showed some interest in your feelin s?

These !ro rams, of course, were made to talk to humans, so they are shown u! rather !oorly here, !itted a ainst each other& /ostly they rely on shrewd uesses as to the nature of the in!ut 4which they analy<e quite shallowly5 and s!it back canned answers that have been carefully selected from a lar e re!ertoire& The answer may be only !artially canned, for e(am!le, a tem!late with blanks that can be filled in& It is assumed that their human !artners will read much more into what they say than is actually underlyin it& )nd in fact, accordin to 3ei<enbaum, in his book 7om!uter Power and ?uman ;eason, .ust that ha!!ens& ?e writes,
EAI9) Uthe !ro ram from which :octor was madeV created the most remarkable illusion of havin understood in the minds of the many !eo!le who conversed with it&&& They would often demand to be !ermitted to converse with the system in !rivate, and would, after conversin with it for a time, insist, in s!ite of my e(!lanations, that the machine really understood them& "+

6iven the above e(cer!t, you may find this incredible& Incredible, but true& 3ei<enbaum has an e(!lanation,

/ost men don1t understand com!uters to even the sli htest de ree& So, unless they are ca!able of very reat ske!ticism 4the kind we brin to bear while watchin a sta e ma ician5, they can e(!lain the com!uter1s intellectual feats only by brin in to hear the sin le analo y available to them, that is, their model of their own ca!acity to think& 8o wonder, then, that they overshoot the mark, it is truly im!ossible to ima ine a human who could imitate EAI9), for e(am!le, but for whom EAI9)1s lan ua e abilities were his limit&""

3hich amounts to an admission that this kind of !ro ram is based on a shrewd mi(ture of bravado and bluffin , takin advanta e of !eo!le1s ullibility& In li ht of this weird -EAI9)-effect-, some !eo!le have su ested that the Turin test needs revision, since !eo!le can a!!arently be fooled by sim!listic immickry& It has been su ested that the interro ator should be a 8obel Pri<ewinnin scientist& It mi ht be more advisable to turn the Turin test on its head, and insist that the interro ator should be another com!uter& Er !erha!s there should be two interro atorsa human and a com!uterand one witness, and the two interro ators should try to fi ure out whether the witness is a human or a com!uter& In a more serious vein, I !ersonally feel that the Turin test, as ori inally !ro!osed, is quite reasonable& )s for the !eo!le who 3ei<enbaum claims were sucked in by EAI9), they were not ur ed to be ske!tical, or to use all their wits in tryin to determine if the -!erson- ty!in to them were human or not& I think that Turin 1s insi ht into this issue was sound, and that the Turin test, essentially unmodified, will survive&

A Brief /istory of AI
I would like in the ne(t few !a es to !resent the story, !erha!s from an unorthodo( !oint of view, of some of the efforts at unravelin the al orithms behind intelli ence, there have been failures and setbacks and there will continue to be& 8onetheless, we are learnin a reat deal, and it is an e(citin !eriod& Ever since Pascal and Aeibni<, !eo!le have dreamt of machines that could !erform intellectual tasks& In the nineteenth century, 0oole and :e /or an devised -laws of thou ht-essentially the Pro!ositional 7alculusand thus took the first ste! towards )I software= also 7harles 0abba e desi ned the first -calculatin en ine-the !recursor to the hardware of com!uters and hence of )I& Ene could define )I as comin into e(istence at the moment when mechanical devices took over any tasks !reviously !erformable only by human minds& It is hard to look back and ima ine the feelin s of those who first saw toothed wheels !erformin additions and multi!lications of lar e numbers& Perha!s they e(!erienced a sense of awe at seein -thou hts- flow in their very !hysical hardware& In any case, we do know that nearly a century later, when the first electronic com!uters were constructed, their inventors did e(!erience an awesome and mystical sense of bein in the !resence of another kind of -thinkin bein -& To what e(tent real thou ht was takin !lace was a source of much !u<<lement= and even now, several decades later, the question remains a reat source of stimulation and vitriolics& It is interestin that nowadays, !ractically no one feels that sense of awe any lon er even when com!uters !erform o!erations that are incredibly more so!histicated than those which sent thrills down s!ines in the early days& The once-e(citin !hrase -6iant

Electronic 0rain- remains only as a sort of -cam!- clichK, a ridiculous vesti e of the era of 'lash 6ordon and 0uck ;o ers& It is a bit sad that we become blasK so quickly& There is a related -Theorem- about !ro ress in )I, once some mental function is !ro rammed, !eo!le soon cease to consider it as an essential in redient of -real thinkin -& The ineluctable core of intelli ence is always in that ne(t thin which hasn1t yet been !ro rammed& This -Theorem- was first !ro!osed to me by Aarry Tesler, so I call it Tesler1s Theorem& -)I is whatever hasn1t been done yet&) selective overview of )I is furnished below& It shows several domains in which workers have concentrated their efforts, each one seemin in its own way to require the quintessence of intelli ence& 3ith some of the domains I have included a breakdown accordin to methods em!loyed, or more s!ecific areas of concentration& mechanical translation direct 4dictionary look-u! with some word rearran ement5 indirect 4via some intermediary internal lan ua e5 ame !layin chess with brute force look-ahead with heuristically !runed look-ahead with no look-ahead checkers o kalah brid e 4biddin = !layin 5 !oker variations on tic-tac-toe etc& !rovin theorems in various !arts& of mathematics symbolic lo ic -resolution- theorem-!rovin elementary eometry symbolic mani!ulation of mathematical e(!ressions symbolic inte ration al ebraic sim!lification summation of infinite series vision !rinted matter, reco nition of individual hand-!rinted characters drawn from a small class 4e& &, numerals5 readin te(t in variable fonts readin !assa es in handwritin readin 7hinese or Ja!anese !rinted characters

readin 7hinese or Ja!anese handwritten characters !ictorial, locatin !res!ecified ob.ects in !hoto ra!hs decom!osition of a scene into se!arate ob.ects identification of se!arate ob.ects in a scene reco nition of ob.ects !ortrayed in sketches by !eo!le reco nition of human faces hearin understandin s!oken words drawn from a limited vocabulary 4e& &, names of the ten di its5 understandin continuous s!eech in fi(ed domains findin boundaries between !honemes identifyin !honemes findin boundaries between mor!hemes identifyin mor!hemes !uttin to ether whole words and sentences understandin natural lan ua es answerin questions in s!ecific domains !arsin com!le( sentences makin !ara!hrases of lon er !ieces of te(t usin knowled e of the real world in order to understand !assa es resolvin ambi uous references !roducin natural lan ua e abstract !oetry 4e& &, haiku5 random sentences, !ara ra!hs, or lon er !ieces of te(t !roducin out!ut from internal re!resentation of knowled e creatin ori inal thou hts or works of art !oetry writin 4haiku5 story writin com!uter art musical com!osition atonal tonal analo ical thinkin eometrical sha!es 4-intelli ence tests-5 constructin !roofs in one domain of mathematics based on those in a related domain learnin ad.ustment of !arameters conce!t formation

Mechanical Translation
/any of the !recedin to!ics will not be touched u!on in my selective discussion below, but the list would not be accurate without them& The first few to!ics are listed in historical order& In each of them, early efforts fell short of e(!ectations& 'or e(am!le, the !itfalls in mechanical translation came as a reat sur!rise to many who had thou ht it was a nearly strai htforward task, whose !erfection, to be sure, would be arduous, but whose basic im!lementation should be easy& )s it turns out, translation is far more com!le( than mere dictionary look-u! and word rearran in & 8or is the difficulty caused by a lack of knowled e of idiomatic !hrases& The fact is that translation involves havin a mental model of the world bein discussed, and mani!ulatin symbols in that model& ) !ro ram which makes no use of a model of the world as it reads the !assa e will soon et ho!elessly bo ed down in ambi uities and multi!le meanin s& Even !eo!lewho have a hu e advanta e over com!uters, for they come fully equi!!ed with an understandin of the world when iven a !iece of te(t and a dictionary of a lan ua e they do not know, find it ne(t to im!ossible to translate the te(t into their own lan ua e& Thusand it is not sur!risin in retros!ectthe first !roblem of )I led immediately to the issues at the heart of )I&

Com!uter Chess
7om!uter chess, too, !roved to be much more difficult than the early intuitive estimates had su ested& ?ere a ain it turns out that the way humans re!resent a chess situation in their minds is far more com!le( than .ust knowin which !iece is on which square, cou!led with knowled e of the rules of chess& It involves !erceivin confi urations of several related !ieces, as well as knowled e of heuristics, or rules of thumb, which !ertain to such hi her-level chunks& Even thou h heuristic rules are not ri orous in the way that the official rules are, they !rovide shortcut insi hts into what is oin on on the board, which knowled e of the official rules does not& This much was reco ni<ed from the start= it was sim!ly underestimated how lar e a role the intuitive, chunked understandin of the chess world !lays in human chess skill& It was !redicted that a !ro ram havin some basic heuristics, cou!led with the blindin s!eed and accuracy of a com!uter to look ahead in the ame and analy<e each !ossible move, would easily beat to!fli ht human !layersa !rediction which, even after twenty-five years of intense work by various !eo!le, still is far from bein reali<ed& Peo!le are nowadays tacklin the chess !roblem from various an les& Ene of the most novel involves the hy!othesis that lookin ahead is a silly thin to do& Ene should instead merely look at what is on the board at !resent, and, usin some heuristics, enerate a !lan, and then find a move which advances that !articular !lan& Ef course, rules for the formulation of chess !lans will necessarily involve heuristics which are, in some sense, -flattened- versions of lookin ahead& That is, the equivalent of many ames1 e(!erience of lookin ahead is -squee<ed- into

another form which ostensibly doesn1t involve lookin ahead& In some sense this is a ame of words& 0ut if the -flattened- knowled e ives answers more efficiently than the actual look-aheadeven if it occasionally misleadsthen somethin has been ained& 8ow this kind of distillation of knowled e into more hi hly usable forms is .ust what intelli ence e(cels atso look-ahead-less chess is !robably a fruitful line of research to !ush& Particularly intri uin would be to devise a !ro ram which itself could convert knowled e ained from lookin ahead into -flattened- rulesbut that is an immense task&

Samuel%s Chec.er ,rogram

)s a matter of fact, such a method was develo!ed by )rthur Samuel in his admirable checker-!layin !ro ram& Samuel1s trick was to use both dynamic 4lookahead5 and static 4no-look-ahead5 ways of evaluatin any iven board !osition& The static method involved a sim!le mathematical function of several quantities characteri<in any board !osition, and thus could be calculated !ractically instantaneously, whereas the dynamic evaluation method involved creatin a -treeof !ossible future moves, res!onses to them, res!onses to the res!onses, and so forth 4as was shown in 'i & %*5& In the static evaluation function there were some !arameters which could vary= the effect of varyin them was to !rovide a set of different !ossible versions of the static evaluation function& Samuel1s strate y was to select, in an evolutionary way, better and better values of those !arameters& ?ere1s how this was done, each time the !ro ram evaluated a board !osition, it did so both statically and dynamically& The answer otten by lookin aheadlet us call it .was used in determinin the move to be made& The !ur!ose of -, the static evaluation, was trickier, on each move, the variable !arameters were read.usted sli htly so that - a!!ro(imated . as accurately as !ossible& The effect was to !artially encode in the values of the static evaluation1s !arameters the knowled e ained by dynamically searchin the tree& In short, the idea was to -flatten- the com!le( dynamic evaluation method into the much sim!ler and more efficient static evaluation function& There is a rather nice recursive effect here& The !oint is that the dynamic evaluation of any sin le board !osition involves lookin ahead a finite number of movessay seven& 8ow each of the scads of board !ositions which mi ht turn u! seven turns down the road has to be itself evaluated somehow as well& 0ut when the !ro ram evaluates these !ositions, it certainly cannot look another seven moves ahead, lest it have to look fourteen !ositions ahead, then twenty-one, etc&, etc&an infinite re ress& Instead, it relies on static evaluations of !ositions seven moves ahead& Therefore, in Samuel1s scheme, an intricate sort of feedback takes !lace, wherein the !ro ram is constantly tryin to -flatten- look-ahead evaluation into a sim!ler static reci!e= and this reci!e in turn !lays a key role in the dynamic lookahead evaluation& Thus the two are intimately linked to ether, and each benefits from im!rovements in the other in a recursive way&

The level of !lay of the Samuel checkers !ro ram is e(tremely hi h, of the order of the to! human !layers in the world& If this is so, why not a!!ly the same techniques to chess? )n international committee, convened in "#D" to study the feasibility of com!uter chess, includin the :utch International 6randmaster and mathematician /a( Euwe, came to the bleak conclusion that the Samuel technique would be a!!ro(imately one million times as difficult to im!lement in chess as in checkers, and that seems to close the book on that& The e(traordinarily reat skill of the checkers !ro ram cannot be taken as sayin -intelli ence has been achieved-= yet it should not be minimi<ed, either& It is a combination of insi hts into what checkers is, how to think about checkers, and how to !ro ram& Some !eo!le mi ht feel that all it shows is Samuel1s own checkers ability& 0ut this is not true, for at least two reasons& Ene is that skillful ame !layers choose their moves accordin to mental !rocesses which they do not fully understandthey use their intuitions& 8ow there is no known way that anyone can brin to li ht all of his own intuitions= the best one can do via intros!ection is to use -feelin - or -meta-intuition-an intuition about one1s intuitionsas a uide, and try to describe what one thinks one1s intuitions are all about& 0ut this will only ive a rou h a!!ro(imation to the true com!le(ity of intuitive methods& ?ence it is virtually certain that Samuel has not mirrored his own !ersonal methods of !lay in his !ro ram& The other reason that Samuel1s !ro ram1s !lay should not be confused with Samuel1s own !lay is that Samuel does not !lay checkers as well as his !ro ramit beats him& This is not a !arado( at allno more than is the fact that a com!uter which has been !ro rammed to calculate m can outrace its !ro rammer in s!ewin forth di its of m&

>hen Is a ,rogram 0riginal7

This issue of a !ro ram outdoin its !ro rammer is connected with the question of -ori inality- in )I& 3hat if an )I !ro ram comes u! with an idea, or a line of !lay in a ame, which its !ro rammer has never entertainedwho should et the credit? There are various interestin instances of this havin ha!!ened, some on a fairly trivial level, some on a rather dee! level& Ene of the more famous involved a !ro ram to find !roofs of theorems in elementary Euclidean eometry, written by E& 6elernter& Ene day the !ro ram came u! with a s!arklin ly in enious !roof of one of the basic theorems of eometrythe so-called -!ons asinorum-, or -brid e of asses-& This theorem states that the base an les of an isosceles trian le are equal& Its standard !roof requires constructin an altitude which divides the trian le into symmetrical halves& The ele ant method found by the !ro ram 4see 'i & ""C5 used no construction lines& Instead, it considered the trian le and its mirror ima e as two different trian les& Then, havin !roved them con ruent, it !ointed out that the two base an les matched each other in this con ruenceME:& This em of a !roof deli hted the !ro ram1s creator and others= some saw evidence of enius in its !erformance& 8ot to take anythin away from this feat, it ha!!ens that in )&:& %++ the eometer Pa!!us had actually found this !roof, too& In any

case, the question remains, -3ho ets the credit?- Is this intelli ent behavior? Er was the !roof lyin dee!ly hidden within the human 46elernter5, and did the com!uter merely brin it to the surface? This last question comes close to hittin the mark& 3e can turn it around, 3as the !roof lyin dee!ly hidden in the !ro ram? Er was it close to the surface? That is, how easy is it to see why the !ro ram did what it did? 7an the discovery be attributed to some sim!le mechanism, or sim!le combination of mechanisms, in the !ro ram? Er was there a com!le( interaction which, if one heard it e(!lained, would not diminish one1s awe at its havin ha!!ened?
73G5=E BBD! &ons *sinorum &roof found () &appus UXHKK *!.!V and Gelernter's program UXBIJK *!.!V/! &ro(lem: To show that the (ase angles of an isosceles triangle are e"ual! -olution: *s the triangle is isosceles, *& and *&' are of e"ual length! Therefore triangles &*&' and &*& are congruent side%side%side/! This implies that corresponding angles are e"ual! 3n particular, the two (ase angles are e"ual!

It seems reasonable to say that if one can ascribe the !erformance to certain o!erations which are easily traced in the !ro ram, then in some sense the !ro ram was .ust revealin ideas which were in essence hiddenthou h not too dee!lyinside the !ro rammer1s own mind& 7onversely, if followin the !ro ram does not serve to enli hten one as to why this !articular discovery !o!!ed out, then !erha!s one should be in to se!arate the !ro ram1s -mind- from that of its !ro rammer& The human ets credit for havin invented the !ro ram, but not for havin had inside his own head the ideas !roduced by the !ro ram& In such cases, the human can be referred to as the -metaauthor-the author of the author of the result, and the !ro ram as the 4.ust !lain5 author& In the !articular case of 6elernter and his eometry machine, while 6elernter !robably would not have rediscovered Pa!!us1 !roof, still the mechanisms which enerated that !roof were sufficiently close to the surface of the !ro ram that one hesitates to call the !ro ram a eometer in its own ri ht& If it had ke!t on astonishin !eo!le by comin u! with in enious new !roofs over and over a ain, each of which seemed to be based on a fresh s!ark of enius rather than on some standard method, then one would have no qualms about callin the !ro ram a eometerbut this did not ha!!en&

>ho Com!oses Com!uter Music7

The distinction between author and meta-author is shar!ly !ointed u! in the case of com!uter com!osition of music& There are various levels of autonomy which a !ro ram may seem to have in the act of com!osition& Ene level is illustrated by a !iece whose

-meta-author- was /a( /athews of 0ell Aaboratories& ?e fed in the scores of the two marches -3hen Johnny 7omes /archin ?ome- and -The 0ritish 6renadiers-, and instructed the com!uter to make a new scoreone which starts out as -Johnny-, but slowly mer es into -6renadiers-& ?alfway throu h the !iece, -Johnny- is totally one, and one hears -6renadiers- by itself&&& Then the !rocess is reversed, and the !iece finishes with -Johnny-, as it be an& In /athews1 own words, this is
&&&a nauseatin musical e(!erience but one not without interest, !articularly in the rhythmic conversions& -The 6renadiers- is written in BIC time in the key of ' ma.or& -Johnnyis written in DI* time in the key of E minor& The chan e from BIC to DI* time can be clearly a!!reciated, yet would be quite difficult for a human musician to !lay& The modulation from the key of ' ma.or to E minor, which involves a chan e of two notes in the scale, is .arrin , and a smaller transition would undoubtedly have been a better choice&"B

The resultin !iece has a somewhat droll quality to it, thou h in s!ots it is tur id and confused&
Is the com!uter com!osin ? The question is best unasked, but it cannot be com!letely i nored& )n answer is difficult to !rovide& The al orithms are deterministic, sim!le, and understandable& 8o com!licated or hard-to understand com!utations are involved= no -learnin !ro rams are used= no random !rocesses occur= the machine functions in a !erfectly mechanical and strai htforward manner& ?owever, the result is sequences of sound that are un!lanned in fine detail by the com!oser, even thou h the over-all structure of the section is com!letely and !recisely s!ecified& Thus the com!oser is often sur!rised, and !leasantly sur!rised, at the details of the reali<ation of his ideas& To this e(tent only is the com!uter com!osin & 3e call the !rocess al orithmic com!osition, but we immediately re-em!hasi<e that the al orithms are trans!arently sim!le&"%

This is /athews1 answer to a question which he would rather -unask-& :es!ite his disclaimer, however, many !eo!le find it easier to say sim!ly that the !iece was -com!osed by a com!uter-& I believe this !hrase misre!resents the situation totally& The !ro ram contained no structures analo ous to the brain1s -symbols-, and could not be said in any sense to be -thinkin - about what it was doin & To attribute the com!osition of such a !iece of music to the com!uter would be like attributin the authorshi! of this book to the com!uteri<ed automatically 4often incorrectly5 hy!henatin !hototy!esettin machine with which it was set& This brin s u! a question which is a sli ht di ression from )I, but actually not a hu e one& It is this, 3hen you see the word -I- or -me- in a te(t, what do you take it to be referrin to? 'or instance, think of the !hrase -3)S? /E- which a!!ears occasionally on the back of dirty trucks& 3ho is this -me-? Is this an outcry of some forlorn child who, in des!eration to have a bath, scribbled the words on the nearest surface? Er is the truck requestin a wash? Er, !erha!s, does the sentence itself wish to be iven a shower? Er, is it that the filthy En lish lan ua e is askin to be cleansed? Ene could o on and on in this ame& In this case, the !hrase is a .oke, and one is su!!osed to !retend, on some level, that the truck itself wrote the !hrase and is requestin a wash& En another level, one clearly reco ni<es the writin as that of a child, and en.oys the humor of the misdirection& ?ere, in fact, is a ame based on readin the -me- at the wron level&

Precisely this kind of ambi uity has arisen in this book, first in the 7ontracrosti!unctus, and later in the discussions of 6>del1s strin 6 4and its relatives5& The inter!retation iven for un!layable records was -I 7annot 0e Played on ;ecord Player T-, and that for un!rovable statements was, -I 7annot 0e Proven in 'ormal System T-& Aet us take the latter sentence& En what other occasions, if any, have you encountered a sentence containin the !ronoun -I- where you automatically understood that the reference was not to the s!eaker of the sentence, but rather to the sentence itself? Gery few, I would uess& The word -I-, when it a!!ears in a Shakes!eare sonnet, is referrin not to a fourteen-line form of !oetry !rinted on a !a e, but to a flesh-and-blood creature behind the scenes, somewhere off sta e& ?ow far back do we ordinarily trace the -I- in a sentence? The answer, it seems to me, is that we look for a sentient bein to attach the authorshi! to& 0ut what is a sentient bein ? Somethin onto which we can ma! ourselves comfortably& In 3ei<enbaum1s -:octor- !ro ram, is there a !ersonality? If so, whose is it? ) small debate over this very question recently ra ed in the !a es of -cience ma a<ine& This brin s us back to the issue of the -who- who com!oses com!uter music& In most circumstances, the drivin force behind such !ieces is a human intellect, and the com!uter has been em!loyed, with more or less in enuity, as a tool for reali<in an idea devised by the human& The !ro ram which carries this out is not anythin which we can identify with& It is a sim!le and sin le-minded !iece of software with no fle(ibility, no !ers!ective on what it is doin , and no sense of self& If and when, however, !eo!le develo! !ro rams which have those attributes, and !ieces of music start issuin forth from them, then I su est that will be the a!!ro!riate, time to start s!littin u! one1s admiration, some to the !ro rammer for creatin such an ama<in !ro ram, and some to the !ro ram itself for its sense of music& )nd it seems to me that that will only take !lace when the internal structure of such a !ro ram is based on somethin similar to the -symbols- in our brains and their tri erin !atterns, which are res!onsible for the com!le( notion of meanin & The fact of havin this kind of internal structure would endow the !ro ram with !ro!erties which would make us feel comfortable in identifyin with it, to some e(tent& 0ut until then, I will not feel comfortable in sayin -this !iece was com!osed by a com!uter-&

Theorem ,ro"ing and ,roblem $eduction

Aet us now return to the history of )I& Ene of the early thin s which !eo!le attem!ted to !ro ram was the intellectual activity of theorem !rovin & 7once!tually, this is no different from !ro rammin a com!uter to look for a derivation of M& in the /IFsystem, e(ce!t that the formal systems involved were often more com!licated than the /IF-system& They were versions of the Predicate 7alculus, which is an e(tension of the Pro!ositional 7alculus involvin quantifiers& /ost of the rules of the Predicate 7alculus are included in T8T, as a matter of fact& The trick in writin such a !ro ram is to instill a sense of direction, so that the !ro ram does not wander all over the ma!, but works only on -relevant- !athwaysthose which, by some reasonable criterion, seem to be leadin towards the desired strin &

In this book we have not dealt much with such issues& ?ow indeed can you know when you are !roceedin towards a theorem, and how can you tell if what you are doin is .ust em!ty fiddlin ? This was one thin which I ho!ed to illustrate with the /F!u<<le& Ef course, there can be no definitive answer, that is the content of the limitative Theorems, since if you could always know which way to o, you could construct an al orithm for !rovin any desired theorem, and that would violate 7hurch1s Theorem& There is no such al orithm& 4I will leave it to the reader to see e(actly why this follows from 7hurch1s Theorem&5 ?owever, this doesn1t mean that it is im!ossible to develo! any intuition at all concernin what is and what is not a !romisin route= in fact, the best !ro rams have very so!histicated heuristics, which enable them to make deductions in the Predicate 7alculus at s!eeds which are com!arable to those of ca!able humans& The trick in theorem !rovin is to use the fact that you have an overall oalnamely the strin you want to !roducein uidin you locally& Ene technique which was develo!ed for convertin lobal oals into local strate ies for derivations is called pro(lem reduction& It is based on the idea that whenever one has a lon -ran e oal, there are usually su(goals whose attainment will aid in the attainment of the main oal& Therefore if one breaks u! a iven !roblem into a series of new sub!roblems, then breaks those in turn into subsub!roblems, and so on, in a recursive fashion, one eventually comes down to very modest oals which can !resumably be attained in a cou!le of ste!s& Er at least so it would seem&&& Problem reduction ot 9eno into hot water& 9eno1s method, you recall, for ettin from ) to 0 4think of 0 as the oal5, is to -reduce- the !roblem into two sub!roblems, first o halfway, then o the rest of the way& So now you have -!ushed-in the sense of 7ha!ter Gtwo sub oals onto your - oal stack-& Each of these, in turn, will be re!laced by two subsub oals and so on ad infinitum& Hou wind u! with an infinite oal-stack, instead of a sin le oal 4'i & ""L5& Po!!in an infinite number of oals off your stack will !rove to be trickywhich is .ust 9eno1s !oint, of course& )nother e(am!le of an infinite recursion in !roblem reduction occurred in the :ialo ue +ittle ,armonic +a()rinth, when )chilles wanted to have a Ty!eless 3ish ranted& Its rantin had to be deferred until !ermission was otten from the /eta-6enie= but in order to et !ermission to ive !ermission, she had to summon the /eta-/eta6enie-and so on& :es!ite the infiniteness of the oal stack, )chilles ot his wish& Problem reduction wins the day2 :es!ite my mockery, !roblem reduction is a !owerful technique for convertin lobal !roblems into local !roblems& It shines in certain situations, such as in the end ame of chess, where the look-ahead technique often !erforms miserably, even when it is carried to ridiculous len ths, such as fifteen or more !lies& This is because the lookahead technique is not based on !lannin = it sim!ly has no oals and e(!lores a hu e number of !ointless alternatives& ?avin a oal enables you to develo! a strate y for the achievement of that oal, and this is a com!letely different !hiloso!hy from lookin ahead mechanically& Ef course, in the look-ahead technique, desirability or its absence is

73G5=E BBG! Neno's endless goal tree, for getting from * to B!

measured by the evaluation function for !ositions, and that incor!orates indirectly a number of oals, !rinci!ally that of not ettin checkmated& 0ut that is too indirect& 6ood chess !layers who !lay a ainst look-ahead chess !ro rams usually come away with the im!ression that their o!!onents are very weak in formulatin !lans or strate ies&

Shandy and the Bone

There is no uarantee that the method of !roblem reduction will work& There are many situations where it flo!s& 7onsider this sim!le !roblem, for instance& Hou are a do , and a human friend has .ust thrown your favorite bone over a wire fence into another yard& Hou can see your bone throu h the fence, .ust lyin there in the rasshow luscious2 There is an o!en ate in the fence about fifty feet away from the bone& 3hat do you do? Some do s will .ust run u! to the fence, stand ne(t to it, and bark= others will dash u! to the o!en ate and double back to the lovely bone& 0oth do s can be said to be e(ercisin the !roblem reduction technique= however, they re!resent the !roblem in their minds in different ways, and this makes all the difference& The barkin do sees the sub!roblems as 4"5 runnin to the fence, 4B5 ettin throu h it, and 4%5 runnin to the bonebut that second sub!roblem is a -tou hie-, whence the barkin & The other do sees the sub!roblems as 4"5 ettin to the ate= 4B5 oin throu h the ate= 4%5 runnin to the bone& 8otice how everythin de!ends on the way you re!resent the -!roblem s!ace-

that is, on what you !erceive as reducin the !roblem 4forward motion towards the overall oal5 and what you !erceive as ma nifyin the !roblem 4backward motion away from the oal5&

Changing the ,roblem S!ace

Some do s first try runnin directly towards the bone, and when they encounter the fence, somethin clicks inside their brain= soon they chan e course, and run over to the ate& These do s reali<e that what on first lance seemed as if it would increase the distance between the initial situation and the desired situationnamely, runnin away from the bone but towards the o!en ateactually would decrease it& )t first, they confuse !hysical distance with pro(lem distance& )ny motion away from the bone seems, by definition, a 0ad Thin & 0ut thensomehowthey reali<e that they can shift their !erce!tion of what will brin them -closer- to the bone& In a !ro!erly chosen abstract s!ace, movin towards the ate is a tra.ectory brin in the do closer to the bone2 )t every moment, the do is ettin -closer-in the new senseto the bone& Thus, the usefulness of !roblem reduction de!ends on how you re!resent your !roblem mentally& 3hat in one s!ace looks like a retreat can in another s!ace look like a revolutionary ste! forward& In ordinary life, we constantly face and solve variations on the do -and-bone !roblem& 'or instance, if one afternoon I decide to drive one hundred miles south, but am at my office and have ridden my bike to work, I have to make an e(tremely lar e number of moves in what are ostensibly -wron - directions before I am actually on my way in car headed south& I have to leave my office, which means, say, headin east a few feet= then follow the hall in the buildin which heads north, then west& Then I ride my bike home, which involves e(cursions in all the directions of the com!ass= and I reach my home& ) succession of short moves there eventually ets me into my car, and I am off& 8ot that I immediately drive due south& of courseI choose a route which may involve some e(cursions north& west, or east, with the aim of ettin to the freeway as quickly as !ossible& )ll of this doesn1t feel !arado(ical in the sli htest= it is done without even any sense of amusement& The s!ace in which !hysical backtrackin is !erceived as direct motion towards the oal is built so dee!ly into my mind that I don1t even see any irony when I head north& The roads and hallways and so forth act as channels which I acce!t without much fi ht, so that !art of the act of choosin how to !erceive the situation involves .ust acce!tin what is im!osed& 0ut do s in front of fences sometimes have a hard time doin that, es!ecially when that bone is sittin there so close, starin them in the face, and lookin so ood& )nd when the !roblem s!ace is .ust a shade more abstract than !hysical s!ace, !eo!le are often .ust as lackin in insi ht about what to do as the barkin do s& In some sense all !roblems are abstract versions of the do -and-bone !roblem& /any !roblems are not in !hysical s!ace but in some sort of conce!tual s!ace& 3hen you reali<e that direct motion towards the oal in that s!ace runs you into some sort of abstract -fence-, you can do one of two thin s, 4"5 try movin away from the oal in

some sort of random way, ho!in that you may come u!on a hidden - ate- throu h which you can !ass and then reach your bone= or 4B5 try to find a new -s!ace- in which you can re!resent the !roblem, and in which there is no abstract fence se!aratin you from your oalthen you can !roceed strai ht towards the oal in this new s!ace& The first method may seem like the la<y way to o, and the second method may seem like a difficult and com!licated way to o& )nd yet, solutions which involve restructurin the !roblem s!ace more often than not come as sudden flashes of insi ht rather than as !roducts of a series of slow, deliberate thou ht !rocesses& Probably these intuitive flashes come from the e(treme core of intelli enceand, needless to say, their source is a closely !rotected secret of our .ealous brains& In any case, the trouble is not that !roblem reduction !er se leads to failures= it is quite a sound technique& The !roblem is a dee!er one, how do you choose a ood internal re!resentation for a !roblem? 3hat kind of -s!ace- do you see it in? 3hat kinds of action reduce the -distance- between you and your oal in the s!ace you have chosen? This can be e(!ressed in mathematical lan ua e as the !roblem of huntin for an a!!ro!riate metric 4distance function5 between states& Hou want to find a metric in which the distance between you and your oal is very small& 8ow since this matter of choosin an internal re!resentation is itself a ty!e of !roblemand a most tricky one, tooyou mi ht think of turnin the technique of !roblem reduction back on it2 To do so, you would have to have a way of re!resentin a hu e variety of abstract s!aces, which is an e(ceedin ly com!le( !ro.ect& I am not aware of anyone1s havin tried anythin alon these lines& It may be .ust a theoretically a!!ealin , amusin su estion which is in fact wholly unrealistic& In any case, what )I sorely lacks is !ro rams which can -ste! back- and take a look at what is oin on, and with this !ers!ective, reorient themselves to the task at hand& It is one thin to write a !ro ram which e(cels at a sin le task which, when done by a human bein , seems to require intelli enceand it is another thin alto ether to write an intelli ent !ro ram2 It is the difference between the S!he( was! 4see 7ha!ter TI5, whose wired-in routine ives the dece!tive a!!earance of reat intelli ence, and a human bein observin a S!he( was!&

The I-Mode and the M-Mode Again

)n intelli ent !ro ram would !resumably be one which is versatile enou h to solve !roblems of many different sorts& It would learn to do each different one and would accumulate e(!erience in doin so& It would be able to work within a set of rules and yet also, at a!!ro!riate moments, to ste! back and make a .ud ment about whether workin within that set of rules is likely to be !rofitable in terms of some overall set of oals which it has& It would be able to choose to sto! workin within a iven framework, if need be, and to create a new framework of rules within which to work for a while& /uch of this discussion may remind you of as!ects of the /F-!u<<le& 'or instance, movin away from the oal of a !roblem is reminiscent of movin away from M& by makin lon er and lon er strin s which you ho!e may in some indirect way enable you to make M&& If you are a naPve -do -, you may feel you are movin away

from your -/F-bone- whenever your strin increases beyond two characters= if you are a more so!histicated do , the use of such len thenin rules has an indirect .ustification, somethin like headin for the ate to et your /F-bone& )nother connection between the !revious discussion and the /F-!u<<le is the two modes of o!eration which led to insi ht about the nature of the /F-!u<<le, the /echanical mode, and the Intelli ent mode& In the former, you are embedded within some fi(ed framework= in the latter, you can always ste! back and ain an overview of thin s& ?avin an overview is tantamount to choosin a re!resentation within which to work= and workin within the rules of the system is tantamount to tryin the technique of !roblem reduction within that selected framework& ?ardy1s comment on ;amanu.an1s style!articularly his willin ness to modify his own hy!othesesillustrates this inter!lay between the /-mode and the I-mode in creative thou ht& The S!he( was! o!erates e(cellently in the /-mode, but it has absolutely no ability to choose its framework or even to alter its /-mode in the sli htest& It has no ability to notice when the same thin occurs over and over and over a ain in its system, for to notice such a thin would be to .um! out of the system, even if only ever so sli htly& It sim!ly does not notice the sameness of the re!etitions& This idea 4of not noticin the identity of certain re!etitive events5 is interestin when we a!!ly it to ourselves& )re there hi hly re!etitious situations which occur in our lives time and time a ain, and which we handle in the identical stu!id way each time, because we don1t have enou h of an overview to !erceive their sameness? This leads back to that recurrent issue, -3hat is sameness?- It will soon come u! as an )I theme, when we discuss !attern reco nition&

A!!lying AI to Mathematics
/athematics is in some ways an e(tremely interestin domain to study from the )I !oint of view& Every mathematician has the sense that there is a kind of metric between ideas in mathematicsthat all of mathematics is a network of results between which there are enormously many connections& In that network, some ideas are very closely linked= others require more elaborate !athways to be .oined& Sometimes two theorems in mathematics are close because one can be !roven easily, iven the other& Ether times two ideas are close because they are analo ous, or even isomor!hic& These are two different senses of the word -close- in the domain of mathematics& There are !robably a number of others& 3hether there is an ob.ectivity or a universality to our sense of mathematical closeness, or whether it is lar ely an accident of historical develo!ment is hard to say& Some theorems of different branches of mathematics a!!ear to us hard to link, and we mi ht say that they are unrelatedbut somethin mi ht turn u! later which forces us to chan e our minds& If we could instill our hi hly develo!ed sense of mathematical closenessa -mathematician1s mental metric-, so to s!eakinto a !ro ram, we could !erha!s !roduce a !rimitive -artificial mathematician-& 0ut that de!ends on bein able to convey a sense of sim!licity or -naturalness- as well, which is another ma.or stumblin block&

These issues have been confronted in a number of )I !ro.ects& There is a collection of !ro rams develo!ed at /IT which o under the name -/)7SH/)-, whose !ur!ose it is to aid mathematicians in symbolic mani!ulation of com!le( mathematical, e(!ressions& This !ro ram has in it some sense of -where to o-a sort of -com!le(ity radient- which uides it from what we would enerally consider com!le( e(!ressions to sim!ler ones& Part of /)7SH/)1s re!ertoire is a !ro ram called -SI8-, which does symbolic inte ration of functions= it is enerally acknowled ed to be su!erior to humans in some cate ories& It relies u!on a number of different skills, as intelli ence in eneral must, a vast body of knowled e, the technique of !roblem reduction, a lar e number of heuristics, and also some s!ecial tricks& )nother !ro ram, written by :ou las Aenat at Stanford, had as its aim to invent conce!ts and discover facts in very elementary mathematics& 0e innin with the notion of sets, and a collection of notions of what is -interestin - which had been s!oon-fed into it, it -invented- the idea of countin , then the idea of addition, then multi!lication, then amon other thin sthe notion of !rime numbers, and it went so far as to rediscover 6oldbach1s con.ecture2 Ef course these -discoveries- were all hundredseven thousands of years old& Perha!s this may be e(!lained in !art by sayin that the sense of -interestin - was conveyed by Aenat in a lar e number of rules which may have been influenced by his twentieth century trainin = nonetheless it is im!ressive& The !ro ram seemed to run out of steam after this very res!ectable !erformance& )n interestin thin about it was that it was unable to develo! or im!rove u!on its own sense of what is interestin & That seemed another level of difficulty u!or !erha!s several levels u!&

The Cru- of AI5 $e!resentation of Kno'ledge

/any of the e(am!les above have been cited in order to stress that the way a domain is re!resented has a hu e bearin on how that domain is -understood-& ) !ro ram which merely !rinted out theorems of T8T in a !reordained order would have no understandin of number theory= a !ro ram such as Aenat1s with its e(tra layers of knowled e could be said to have a rudimentary sense of number theory= and one which embeds mathematical knowled e in a wide conte(t of real-world e(!erience would !robably be the most able to -understand- in the sense that we think we do& It is this representation of knowledge that is at the cru( of )I& In the early days it was assumed that knowled e came in sentence-like -!ackets-, and that the best way to im!lant knowled e into a !ro ram was to develo! a sim!le way of translatin facts into small !assive !ackets of data& Then every fact would sim!ly be a !iece of data, accessible to the !ro rams usin it& This is e(em!lified by chess !ro rams, where board !ositions are coded into matrices or lists of some sort and stored efficiently in memory where they can be retrieved and acted u!on by subroutines& The fact that human bein s store facts in a more com!licated way was known to !sycholo ists for quite a while and has only recently been rediscovered by )I workers, who are now confrontin the !roblems of -chunked- knowled e, and the difference between !rocedural and declarative ty!es of knowled e, which is related, as we saw in

7ha!ter TI, to the difference between knowled e which is accessible to intros!ection and knowled e which is inaccessible to intros!ection& The naPve assum!tion that all knowled e should be coded into !assive !ieces of data is actually contradicted by the most fundamental fact about com!uter desi n, that is, how to add, subtract, multi!ly, and so on is not coded into !ieces of data and stored in memory= it is, in fact, re!resented nowhere in memory, but rather in the wirin !atterns of the hardware& ) !ocket calculator does not store in its memory knowled e of how to add= that knowled e is encoded into its - uts-& There is no memory location to !oint to if somebody demands, -Show me where the knowled e of how to add resides in this machine2) lar e amount of work in )I has nevertheless one into systems in which the bulk of the knowled e is stored in s!ecific !lacesthat is, declaratively& It oes without sayin that some knowled e has to be embodied in !ro rams= otherwise one would not have a !ro ram at all, but merely an encyclo!edia& The question is how to s!lit u! knowled e between !ro ram and data& 8ot that it is always easy to distin uish between !ro ram and data, by any means& I ho!e that was made clear enou h in 7ha!ter TGI& 0ut in the develo!ment of a system, if the !ro rammer intuitively conceives of some !articular item as data 4or as !ro ram5, that may have si nificant re!ercussions on the system1s structure, because as one !ro rams one does tend to distin uish between datalike ob.ects and !ro ram-like ob.ects& It is im!ortant to !oint out that in !rinci!le, any manner of codin information into data structures or !rocedures is as ood as any other, in the sense that if you are not too concerned about efficiency, what you can do in one scheme, you can do in the other& ?owever, reasons can be iven which seem to indicate that one method is definitely su!erior to the other& 'or instance, consider the followin ar ument in favor of usin !rocedural re!resentations only, -)s soon as you try to encode features of sufficient com!le(ity into data, you are forced into develo!in what amounts to a new lan ua e, or formalism& So in effect your data structures become !ro ram-like, with some !iece of your !ro ram servin as their inter!reter= you mi ht as well re!resent the same information directly in !rocedural form to be in with, and obviate the e(tra level of inter!retation&-

;A and ,roteins /el! Gi"e Some ,ers!ecti"e

This ar ument sounds quite convincin , and yet, if inter!reted a little loosely, it can be read as an ar ument for the abolishment of :8) and ;8)& 3hy encode enetic information in :8), when by re!resentin it directly in !roteins, you could eliminate not .ust one, but two levels of inter!retation? The answer is, it turns out that it is e(tremely useful to have the same information in several different forms for different !ur!oses& Ene advanta e of storin enetic information in the modular and data-like form of :8) is that two individuals1 enes can be easily recombined to form a new enoty!e& This would be very difficult if the information were only in !roteins& ) second reason for storin information in :8) is that it is easy to transcribe and translate it into !roteins& 3hen it is not needed, it does not take u! much room= when it is needed, it serves as a tem!late&

There is no mechanism for co!yin one !rotein off of another= their folded tertiary structures would make co!yin hi hly unwieldy& 7om!lementarily, it is almost im!erative to be able to et enetic information into three-dimensional structures such as en<ymes, because the reco nition and mani!ulation of molecules is by its nature a threedimensional o!eration& Thus the ar ument for !urely !rocedural re!resentations is seen to be quite fallacious in the conte(t of cells& It su ests that there are advanta es to bein able to switch back and forth between !rocedural and declarative re!resentations& This is !robably true also in )I& This issue was raised by 'rancis 7rick in a conference on communication with e(traterrestrial intelli ence,
3e see on Earth that there are two molecules, one of which is ood for re!lication U:8)V and one of which is ood for action U!roteinsV& Is it !ossible to devise a system in which one molecule does both .obs, or are there !erha!s stron ar uments, from systems analysis, which mi ht su est 4if they e(ist5 that to divide the .ob into two ives a reat advanta e, This is a question to which I do not know the answer&"C

Modularity of Kno'ledge
)nother question which comes u! in the re!resentation of knowled e is modularity& ?ow easy is it to insert new knowled e? ?ow easy is it to revise old knowled e? ?ow modular are books? It all de!ends& If from a ti htly structured book with many cross-references a sin le cha!ter is removed, the rest of the book may become virtually incom!rehensible& It is like tryin to !ull a sin le strand out of a s!ider webyou ruin the whole in doin so& En the other hand, some books are quite modular, havin inde!endent cha!ters& 7onsider a strai htforward theorem- eneratin !ro ram which uses T8T1s a(ioms and rules of inference& The -knowled e- of such a !ro ram has two as!ects& It resides im!licitly in the a(ioms and rules, and e(!licitly in the body of theorems which have so far been !roduced& :e!endin on which way you look at the knowled e, you will see it either as modular or as s!read all around and com!letely nonmodular& 'or instance, su!!ose you had written such a !ro ram but had for otten to include T8T1s )(iom I in the list of a(ioms& )fter the !ro ram had done many thousands of derivations, you reali<ed your oversi ht, and inserted the new a(iom& The fact that you can do so in a trice shows that the system1s im!licit knowled e is modular= but the new a(iom1s contribution to the e(!licit knowled e of the system will only be reflected after a lon timeafter its effects have -diffused- outwards, as the odor of !erfume slowly diffuses in a room when the bottle is broken& In that sense the new knowled e takes a lon time to be incor!orated& 'urthermore, if you wanted to o back and re!lace )(iom I by its ne ation, you could not .ust do that by itself= you would have to delete all theorems which had involved )(iom I in their derivations& 7learly this system1s e(!licit knowled e is not nearly so modular as its im!licit knowled e& It would be useful if we learned how to trans!lant knowled e modularly& Then to teach everyone 'rench, we would .ust o!en u! their heads and o!erate in a fi(ed way on their neural structuresthen they would know how to s!eak 'rench& Ef course, this is only a hilarious !i!e dream&

)nother as!ect of knowled e re!resentation has to do with the way in which one wishes to use the knowled e& )re inferences su!!osed to be drawn as !ieces of information arrive? Should analo ies and com!arisons constantly be bein made between new information and old information? In a chess !ro ram, for instance, if you want to enerate look-ahead trees, then a re!resentation which encodes board !ositions with a minimum of redundancy will be !referable to one which re!eats the information in several different ways& 0ut if you want your !ro ram to -understand- a board !osition by lookin for !atterns and com!arin them to known !atterns, then re!resentin the same information several times over in different forms will be more useful&

$e!resenting Kno'ledge in a Logical +ormalism

There are various schools of thou ht concernin the best way to re!resent and mani!ulate knowled e& Ene which has had reat influence advocates re!resentations usin formal notations similar to those for T8T-usin !ro!ositional connectives and quantifiers& The basic o!erations in such re!resentations are, not sur!risin ly, formali<ations of deductive reasonin & Ao ical deductions can be made usin rules of inference analo ous to some of those in T8T& Mueryin the system about some !articular idea sets u! a oal in the form of a strin to be derived& 'or e(am!le, -Is /F/E8 a theorem?- Then the automatic reasonin mechanisms take over in a oal-oriented way, usin various methods of !roblem reduction& 'or e(am!le, su!!ose that the !ro!osition -)ll formal arithmetics are incom!letewere known, and the !ro ram were queried, -Is &rincipia 'athematica incom!lete?- In scannin the list of known factsoften called the data (asethe system mi ht notice that if it could establish that &rincipia 'athematica is a formal arithmetic, then it could answer the question& Therefore the !ro!osition -&rincipia 'athematica is a formal arithmetic- would be set u! as a sub oal, and then !roblem reduction would take over& If it could find further thin s which would hel! in establishin 4or refutin 5 the oal or the sub oal, it would work on themand so on, recursively& This !rocess is iven the name of (ackwards chaining, since it be ins with the oal and works its way backwards, !resumably towards thin s which may already be known& If one makes a ra!hic re!resentation of the main oal, subsidiary oals, subsub oals, etc&, a tree-like structure will arise, since the main oal may involve several different sub oals, each of which in turn involves several subsub oals, etc& 8otice that this method is not uaranteed to resolve the question, for there may be no way of establishin within the system that Princi!ia /athematica is a formal arithmetic& This does not im!ly, however, that either the oal or the sub oal is a false statementmerely that they cannot be derived with the knowled e currently available to the system& The system may !rint out, in such a circumstance, -I do not know- or words to that effect& The fact that some questions are left o!en is of course similar to the incom!leteness from which certain well-known formal systems suffer&

educti"e #s! Analogical A'areness

This method affords a deducti#e awareness of the domain that is re!resented, in that correct lo ical conclusions can be drawn from known facts& ?owever, it misses somethin of the human ability to s!ot similarities and to com!are situationsit misses what mi ht be called analogical awarenessa crucial side of human intelli ence& This is not to say that analo ical thou ht !rocesses cannot be forced into such a mold, but they do not lend themselves naturally to bein ca!tured in that kind of formalism& These days, lo ic-oriented systems are not so much in vo ue as other kinds, which allow com!le( forms of com!arisons to be carried out rather naturally& 3hen you reali<e that knowled e re!resentation is an alto ether different ball ame than mere stora e of numbers, then the idea that -a com!uter has the memory of an ele!hant- is an easy myth to e(!lode& 3hat is stored in memor) is not necessarily synonymous with what a !ro ram knows= for even if a iven !iece of knowled e is encoded somewhere inside a com!le( system, there may be no !rocedure, or rule, or other ty!e of handler of data, which can et at itit may be inaccessible& In such a case, you can say that the !iece of knowled e has been -for otten- because access to it has been tem!orarily or !ermanently lost& Thus a com!uter !ro ram may -for et- somethin on a hi h level which it -remembers- on a low level& This is another one of those everrecurrin level distinctions, from which we can !robably learn much about our own selves& 3hen a human for ets, it most likely means that a hi h-level !ointer has been lost not that any information has been deleted or destroyed& This hi hli hts the e(treme im!ortance of kee!in track of the ways in which you store incomin e(!eriences, for you never know in advance under what circumstances, or from what an le, you will want to !ull somethin out of stora e&

+rom Com!uter /ai.u to an $T;-Grammar

The com!le(ity of the knowled e re!resentation in human heads first hit home with me when I was workin on a !ro ram to enerate En lish sentences -out of the blue-& I had come to this !ro.ect in a rather interestin way& I had heard on the radio a few e(am!les of so-called -7om!uter ?aiku-& Somethin about them struck me dee!ly& There was a lar e element of humor and simultaneously mystery to makin a com!uter enerate somethin which ordinarily would be considered an artistic creation& I was hi hly amused by the humorous as!ect, and I was very motivated by the mysteryeven contradiction of !ro rammin creative acts& So I set out to write a !ro ram even more mysteriously contradictory and humorous than the haiku !ro ram& )t first I was concerned with makin the rammar fle(ible and recursive, so that one would not have the sense that the !ro ram was merely fillin in the blanks in some tem!late& )t about that time I ran across a -cientific *merican article by Gictor Hn ve in which he described a sim!le but fle(ible rammar which could !roduce a wide variety of sentences of the ty!e found in some children1s books& I modified some of the ideas I1d leaned from that article and came u! with a set of !rocedures which formed a ;ecursive Transition 8etwork rammar, as described in 7ha!ter G& In this rammar, the selection of

words in a sentence was determined by a !rocess which be an by selectin at random the overall structure of the sentence= radually the decision-makin !rocess trickled down throu h lower levels of structure until the word level and the letter level were reached& ) lot had to be done below the word level, such as inflectin verbs and makin !lurals of nouns= also irre ular verb and noun forms were first formed re ularly, and then if they matched entries in a table, substitutions of the !ro!er 4irre ular5 forms were made& )s each word reached its final form, it was !rinted out& The !ro ram was like the !roverbial monkey at a ty!ewriter, but o!eratin on several levels of lin uistic structure simultaneouslynot .ust the letter level& In the early sta es of develo!in the !ro ram, I used a totally silly vocabulary deliberately, since I was aimin at humor& It !roduced a lot of nonsense sentences, some of which had very com!licated structures, others of which were rather short& Some e(cer!ts are shown below, ) male !encil who must lau h clumsily would quack& /ust !ro ram not always crunch irl at memory? The decimal bu which s!its clumsily mi ht tumble& 7ake who does sure take an une(!ected man within relationshi! mi ht always dum! card& Pro ram ou ht run cheerfully& The worthy machine ou ht not always !aste the astronomer& Eh, !ro ram who ou ht really run off of the irl writes musician for theater& The businesslike relationshi! quacks& The lucky irl which can always quack will never sure quack& The ame quacks& Professor will write !ickle& ) bu tumbles& /an takes the bo( who sli!s& The effect is stron ly surrealistic and at times a little reminiscent of haikufor e(am!le& the final sam!le of four consecutive short sentences& )t first it seemed very funny and had a certain charm, but soon it became rather stale& )fter readin a few !a es of out!ut one could sense the limits of the s!ace in which the !ro ram was o!eratin = and after that, seein random !oints inside that s!aceeven thou h each one was -new-was nothin new& This is, it seems to me, a eneral !rinci!le, you et bored with somethin not when you have e(hausted its re!ertoire of behavior, but when you have ma!!ed out the limits of the s!ace that contains its behavior& The behavior s!ace of a !erson is .ust about com!le( enou h that it can continually sur!rise other !eo!le= but that wasn1t true of my !ro ram& I reali<ed that my oal of !roducin truly humorous out!ut would require that far more subtlety be !ro rammed in& 0ut what, in this case, was meant by -subtlety-? It was clear that absurd .u(ta!ositions of words were .ust too unsubtle= I needed a way to

ensure that words would be used in accordance with the realities of the world& This was where thou hts about re!resentation of knowled e be an to enter the !icture&

+rom $T;%s to AT;%s

The idea I ado!ted was to classify each wordnoun, verb, !re!osition, etc&in several different -semantic dimensions-& Thus, each word was a member of classes of various sorts= then there were also su!erclassesclasses of classes 4reminiscent of the remark by Flam5& In !rinci!le, such a re ation could continue to any number of levels, but I sto!!ed at two& )t any iven moment, the choice of words was now semantically restricted, because it was required that there should be agreement between the various !arts of the !hrase bein constructed& The idea was, for instance, that certain kinds of acts could be !erformed only by animate ob.ects= that only certain kinds of abstractions could influence events, and so on& The decisions about what cate ories were reasonable, and whether each cate ory was better thou ht of as a class or a su!erclass, were quite com!licated& )ll words were branded in several different dimensions& 7ommon !re!ositions-of-, -in-, etc&had several distinct entries, corres!ondin to their distinct usa es& 8ow, the out!ut be an to be much more com!rehensibleand for that reason it was funny in a new way&

A Little Turing Test

0elow, I have re!roduced nine selections, carefully culled from many !a es of out!ut from later versions of my !ro ram& )lon with them are three 4seriously intended5 human-written sentences& 3hich? 4"5 0lurtin may be considered as the reci!rocal substitution of semiotic material 4dubbin 5 for a semiotic dialo ical !roduct in a dynamic refle(ion& 4B5 ;ather think of a !athway of a 1sequence1 of edankene(!eriment sim!letons where heir-lines are a !rima facie case of a !aradiachronic transitivity& 4%5 Think of that as a chain stren th !ossibility of what, eventually, comes out as a !roduct 4e!istemic conditions?5 and the !roduct is not a 'rankfurt-ish !ackin it-all-in& 4C5 :es!ite the efforts, the re!ly, if you will, had been su!!orted by the Erient= hence a fallacy will thereafter be sus!ended by the attitude which will be bein held by the ambassador& 4L5 Ef course, until the u!heavals, the ambassador was sli htly mollycoddlin the rabble& radually

4D5 Su!!osedly, refined liberty caused the attitudes insofar as !eace is distilled by the consequences which will not eventually be caused by the command irrevocably insofar as !eace of it is sometimes causin the intransi ency infinitesimally sur!risin ly& 4$5 )ccordin to the so!hists, the cam!ai ns in the city-states, in other words, have been acce!ted by the Erient cunnin ly& Ef course, the Erient has been se!arated by the states !articularly violently& The Erient su!!orts the efforts which had been su!!orted by mankind& 4*5 )dmittedly, the hierarchical ori in of the fallacy, nevertheless, will be !ro!hesied by the enemies of it& 0y the same token, the individualists will have testified that intransi ency will not have sus!ended the cam!ai ns& 4#5 8eedless to say, durin the u!heaval which will have warranted the secrecy, the re!lies do not se!arate the Erient& Ef course, the countries, i!so facto, are always !robin liberty& 4"+5 )lthou h a 8obel Pri<e was bein achieved by the humanists, yet in addition, it was bein achieved by the serf& 4""5 )n attitude will often be held by the serfs of a strife-torn nation& 4"B5 /oreover, the 8obel Pri<es will be achieved& 0y the same token, des!ite the consequence, the 8obel Pri<es which will be achieved will sometimes be achieved by a woman& The human-written sentences are numbers " to %= they were drawn from the contem!orary .ournal *rt%+anguage"L and areas far as I can tellcom!letely serious efforts amon literate and sane !eo!le to communicate somethin to each other& That they a!!ear here out of conte(t is not too misleadin , since their !ro!er conte(t sounds .ust the same as they do& /y !ro ram !roduced the rest& 8umbers "+ to "B were chosen to show that there were occasional bursts of total lucidity= numbers $ to # are more ty!ical of the out!ut, floatin , in that curious and !rovocative netherworld between meanin and no-meanin = and then numbers C to D !retty much transcend meanin & In a enerous mood, one could say that they stand on their own as !ure -lan ua e ob.ects-, somethin like !ieces of abstract scul!ture carved out of words instead of stone= alternatively, one could say that they are !ure !seudointellectual drivel& /y choice of vocabulary was still aimed at !roducin humorous effects& The flavor of the out!ut is hard to characteri<e& )lthou h much of it -makes sense-, at least on a sin le-sentence level, one definitely ets the feelin that the out!ut is comin from a source with no understandin of what it is sayin and no reason to say it& In !articular, one senses an utter lack of visual ima ery behind the words& 3hen I saw such sentences

come !ourin out of the line !rinter, I e(!erienced com!le( emotions& I was very amused by the silliness of the out!ut& I was also very !roud of my achievement and tried to describe it to friends as similar to ivin rules for buildin u! meanin ful stories in )rabic out of sin le strokes of the !enan e(a eration, but it !leased me to think of it that way& )nd lastly I was dee!ly thrilled by the knowled e that this enormously com!licated machine was shuntin around lon trains of symbols inside it accordin to rules, and that these lon trains of symbols were somethin like thou hts in my own head&&& somethin like them&

Images of >hat Thought Is

Ef course I didn1t fool myself into thinkin that there was a conscious bein behind those sentencesfar from it& Ef all !eo!le, I was the most aware of the reasons that this !ro ram was terribly remote from real thou ht& Tesler1s Theorem is quite a!t here, as soon as this level of lan ua e handlin ability had been mechani<ed, it was clear that it did not constitute intelli ence& 0ut this stron e(!erience left me with an ima e, a limmerin sense that real thou ht was com!osed of much lon er, much more com!licated trains of symbols in the brainmany trains movin simultaneously down many !arallel and crisscrossin tracks, their cars bein !ushed and !ulled, attached and detached, switched from track to track by a myriad neural shuntin -en ines&&& It was an intan ible ima e which I cannot convey in words, and it was only an ima e& 0ut ima es and intuitions and motivations lie min led close in the mind, and my utter fascination with this ima e was a constant s!ur to think more dee!ly about what thou ht really could be& I have tried in other !arts of this book to communicate some of the dau hter ima es of this ori inal ima e!articularly in the &relude, *nt 7ugue& 3hat stands out in my mind now, as I look back at this !ro ram from the !ers!ective of a do<en years, is how there is no sense of ima ery behind what is bein said& The !ro ram had no idea what a serf is, what a !erson is, or what anythin at all is& The words were em!ty formal symbols, as em!ty as!erha!s em!tier thanthe ! and q of the !q-system& /y !ro ram took advanta e of the fact that when !eo!le read te(t, they quite naturally tend to imbue each word with its full flavoras if that were necessarily attached to the rou! of letters which form the word& /y !ro ram could be looked at as a formal system, whose -theorems-the out!ut sentenceshad ready-made inter!retations 4at least to s!eakers of En lish5& 0ut unlike the !q-system, these -theorems- were not all true statements when inter!reted that way& /any were false, many were nonsense& In its humble way, the !q-system mirrored a tiny corner of the world& 0ut when my !ro ram ran, there was no mirror inside it of how the world works, e(ce!t for the small semantic constraints which it had to follow& To create such a mirror of understandin , I would have had to wra! each conce!t in layers and layers of knowled e about the world& To do this would have been another kind of effort from what I had intended to do& 8ot that I didn1t often think of tryin to do itbut I never ot around to tryin it out&

73G5=E BBJ! * meaningful stor) in *ra(ic! U7rom *! ;hati(i and '! -iMelmassi, The S!lendour of Islamic 7alli ra!hy New York: =i@@oli, BICJ/!

/igherWLe"el Grammars111
In fact, I often !ondered whether I could write an )T8- rammar 4or some other kind of sentence-!roducin !ro ram5 which would only !roduce true sentences about the world& Such a rammar would imbue the words with enuine meanin s, in the way it ha!!ened in the !q-system and in T8T& This idea of a lan ua e in which false statements are un rammatical is an old one, oin back to Johann )mos 7omenius, in "D%%& It is very a!!ealin because you have a crystal ball embodied in your rammar, .ust write down the statement you want to know about, and check to see if it is rammatical&&& )ctually, 7omenius went even further, for in his lan ua e, false statements were not only un rammaticalthey were ine(!ressible2 7arryin this thou ht in another direction, you mi ht ima ine a hi h-level rammar which would !roduce random kRans& 3hy not? Such a rammar would be equivalent to a formal system whose theorems are kRans& )nd if you had such a !ro ram, could you not arran e it to !roduce only enuine kRans? /y friend /arsha /eredith was enthusiastic about this idea of -)rtificial Ism-, so she tackled the !ro.ect of writin a kRan-writin !ro ram& Ene of her early efforts !roduced this curious quasi-kRan, ) S/)AA HEF86 /)STE; 3)8TE: ) S/)AA 3?ITE 68);AE: 0E3A& -?E3 7)8 3E AE);8 )8: F8:E;ST)8: 3IT?EFT STF:H- T?E HEF86 /)STE; )S@E: ) A);6E 7E8'FSE: /)STE;& T?E 7E8'FSE: /)STE; 3)A@E: ';E/ ) 0;E38 ?);: /EF8T)I8 TE ) 3?ITE SE'T /EF8T)I8 3IT? ) S/)AA ;E: STE8H 0E3A& T?E 7E8'FSE: /)STE; S)3 ) ;E: SE'T ?FT& T?E 7E8'FSE: /)STE; 3)8TE: T?E ?FT& -3?H :I: 0E:?I:?);/) 7E/E I8TE 7?I8)?T?E 7E8'FSE: /)STE; )S@E: ) A);6E E8AI6?TE8E: STF:E8T& -T?E PE)7?ES );E A);6E-, T?E STF:E8T )8S3E;E: T?E 7E8'FSE: /)STE;& -?E3 7)8 3E AE);8 )8: F8:E;ST)8: 3IT?EFT STF:H?- T?E 7E8'FSE: /)STE; )S@E: ) A);6E EA: /)STE;& T?E EA: /)STE; 3)A@E: ';E/ ) 3?ITE STE8H 6++BL& T?E EA: /)STE; 6ET AEST& Hour !ersonal decision !rocedure for kRan enuineness !robably reached a verdict without need of the 6eometric 7ode or the )rt of 9en Strin s& If the lack of !ronouns or the unso!histicated synta( didn1t arouse your sus!icions, that stran e -6++BL- towards the end must have& 3hat is it? It is a stran e flukea manifestation of a bu which caused the !ro ram to !rint out, in !lace of the En lish word for an ob.ect, the !ro ram1s internal name for the -node- 4a AISP atom, in fact5 where all information concernin that !articular ob.ect was stored& So here we have a -window- onto a lower level of the underlyin 9en minda level that should have remained invisible& Fnfortunately, we don1t have such clear windows onto the lower levels of human 9en minds& The sequence of actions, thou h a little arbitrary, comes from a recursive AISP !rocedure called -7)S7):E-, which creates chains of actions linked in a va uely causal

way to each other& )lthou h the de ree of com!rehension of the world !ossessed by this kRan enerator is clearly not stu!endous, work is in !ro ress to make its out!ut a little more enuine-seemin &

Grammars for Music7

Then there is music& This is a domain which you mi ht su!!ose, on first thou ht, would lend itself admirably to bein codified in an )T8 rammar, or some such !ro ram& 3hereas 4to continue this naPve line of thou ht5 lan ua e relies on connections with the outside world for meanin , music is !urely formal& There is no reference to thin s -out there- in the sounds of music= there is .ust !ure synta(note followin note, chord followin chord, measure followin measure, !hrase followin !hrase&&& 0ut wait& Somethin is wron in this analysis& 3hy is some music so much dee!er and more beautiful than other music? It is because form, in music, is e(!ressive e(!ressive to some stran e subconscious re ions of our minds& The sounds of music do not refer to serfs or city-states, but they do tri er clouds of emotion in our innermost selves= in that sense musical meanin is de!endent on intan ible links from the symbols to thin s in the worldthose -thin s-, in this case, bein secret software structures in our minds& 8o, reat music will not come out of such an easy formalism as an )T8- rammar& Pseudomusic, like !seudo-fairy tales, may well come outand that will be a valuable e(!loration for !eo!le to makebut the secrets of meanin in music lie far, far dee!er than !ure synta(& I should clarify one !oint here, in !rinci!le, )T8- rammars have all the !ower of any !ro rammin formalism, so if musical meanin is ca!turable in any way at all 4which I believe it is5, it is ca!turable in an )T8- rammar& True& 0ut in that case, I maintain, the rammar will be definin not .ust musical structures, but the entire structures of the mind of a beholder& The - rammar- will be a full rammar of thou ht not .ust a rammar of music&

>inograd%s ,rogram S/$ L&

3hat kind of !ro ram would it take to make human bein s admit that it had some -understandin -, even if be rud in ly? 3hat would it take before you wouldn1t feel intuitively that there is -nothin there-? In the years "#D*-$+, Terry 3ino rad 4alias :r& Tony Earrwi 5 was a doctoral student at /IT, workin on the .oint !roblems of lan ua e and understandin & )t that time at /IT, much )I research involved the so-called (locks worldOa relatively sim!le domain in which !roblems concernin both vision and lan ua e-handlin by com!uter could fit easily& The blocks world consists of a table with various kinds of toy-like blocks on itsquare ones, oblon ones, trian ular ones, etc&, in various colors& 4'or a -blocks world- of another kind, see 'i ure ""$, the !aintin 'ental *rithmetic by /a ritte& I find its title sin ularly a!!ro!riate in this conte(t&5 The vision !roblems in the /IT blocks world are very tricky, how can a com!uter fi ure out, from a TG-scan of a scene with many blocks in it, .ust what kinds of blocks are !resent, and what their relationshi!s are?

Some blocks may be !erched on to! of others, some may be in front of others, there may be shadows, and so on&
73G5=E BBC! /ental )rithmetic, () =ene 'agritte BIHB/!

3ino rad1s work was se!arate from the issues of vision, however& 0e innin with the assum!tion that the blocks world was well-re!resented inside the com!uter1s memory, he confronted the many-faceted !roblem of how to et the com!uter to, 4"5 4B5 4%5 4C5 4L5 4D5 understand questions in En lish about the situation= ive answers in En lish to questions about the situation= understand requests in En lish to mani!ulate the blocks= break down each request into a sequence of o!erations it could do= understand what it had done, and for what reasons= describe its actions and their reasons, in En lish&

It mi ht seem reasonable to break u! the overall !ro ram into modular sub!ro rams, with one module for each different !art of the !roblem= then, after the modules have been develo!ed se!arately, to inte rate them smoothly& 3ino rad found that this strate y of develo!in inde!endent modules !osed fundamental difficulties& ?e develo!ed a radical a!!roach, which challen ed the theory that intelli ence can be com!artmentali<ed into inde!endent or semi-inde!endent !ieces& ?is !ro ram S?;:AF named after the old code -ET)EI8 S?;:AF-, used by linoty!e o!erators to mark ty!os in a news!a!er columndid not se!arate the !roblem into clean conce!tual !arts& The o!erations of !arsin sentences, !roducin internal re!resentations, reasonin about the world re!resented inside itself, answerin questions, and so on, were all dee!ly and intricately meshed to ether in a !rocedural re!resentation of knowled e& Some critics have char ed that his !ro ram is so tan led that it does not re!resent any -theory- at all about lan ua e, nor does it contribute in any way to our insi hts about thou ht !rocesses&

8othin could be more wron than such claims, in my o!inion& ) tour de force such as S?;:AF may not be isomor!hic to what we doin fact, in no way should you think that in S?;:AF, the -symbol level- has been attainedbut the act of creatin it and thinkin about it offers tremendous insi ht into the way intelli ence works&

The Structure of S/$ L&

In fact, S?;:AF does consist of se!arate !rocedures, each of which contains some knowled e about the world= but the !rocedures have such a stron interde!endency that they cannot be cleanly teased a!art& The !ro ram is like a very tan led knot which resists untan lin = but the fact that you cannot untan le it does not mean that you cannot understand it& There may be an ele ant eometrical descri!tion of the entire knot even if it is !hysically messy& 3e could o back to a meta!hor from the /u Efferin , and com!are it to lookin at an orchard from a -natural- an le& 3ino rad has written lucidly about S?;:AF& I quote here from his article in Schank and 7olby1s book,
Ene of the basic view!oints underlyin the model is that all lan ua e use can be thou ht of as a way of activatin !rocedures within the hearer& 3e can think of any utterance as a !ro ram one that indirectly causes a set of o!erations to he carried out within the hearer1s co nitive system& This -!ro ram writin - is indirect in the sense that we are dealin with an intelli ent inter!reter, who may take a set of actions which are quite different from those the s!eaker intended& The e(act form is determined by his knowled e of the world, his e(!ectations about the !erson talkin to him, etc& In this !ro ram we have a sim!le version of this !rocess of inter!retation as it takes !lace in the robot& Each sentence inter!reted by the robot is converted to a set of instructions in PA)88E;& The !ro ram that is created is then e(ecuted to achieve the desired effect&"D

,LA;;E$ +acilitates ,roblem $eduction

The lan ua e PA)88E;, referred to here, is an )I lan ua e whose !rinci!al feature is that some of the o!erations necessary for !roblem reduction are built innamely, the recursive !rocess of creatin a tree of sub oals, subsub oals, etc& 3hat this means is that such !rocesses, instead of havin to be s!elled out time and time a ain by the !ro rammer, are automatically im!lied by so-called 6E)A-statements& Someone who reads a PA)88E; !ro ram will see no e(!licit reference to such o!erations= in .ar on, they are user-trans!arent& If one !ath in the tree fails to achieve the desired oal, then the PA)88E; !ro ram will -backtrack- and try another route& -0acktrackin - is the ma ic word as far as PA)88E; is concerned& 3ino rad1s !ro ram made e(cellent use of these features of PA)88E;more e(actly, of /I7;EPA)88E;, a !artial im!lementation of the !lans for PA)88E;& In the !ast few years, however, !eo!le with the oal of develo!in )I have concluded that automatic backtrackin , as in PA)88E;, has definite disadvanta es, and that it will

!robably not lead to their oal= therefore they have backed off from it, !referrin to try other routes to )I& Aet us listen to further comments from 3ino rad on S?;:AF,
The definition of every word is a !ro ram which is called at an a!!ro!riate !oint in the analysis, and which can do arbitrary com!utations involvin the sentence and the !resent !hysical situation&"$

)mon the e(am!les which 3ino rad cites is the followin ,

The different !ossibilities for the meanin of -the- are !rocedures which check various facts about the conte(t, then !rescribe actions such as -Aook for a unique ob.ect in the data base which fits this descri!tion-, or -)ssert that the ob.ect bein described is unique as far as the s!eaker is concerned&- The !ro ram incor!orates a variety of heuristics for decidin what !art of the conte(t is relevant&"*

It is ama<in how dee! this !roblem with the word -the- is& It is !robably safe to say that writin a !ro ram which can fully handle the to! five words of En lish-the-, -of-, -and-, -a-, and -to-would be equivalent to solvin the entire !roblem of )I, and hence tantamount to knowin what intelli ence and consciousness are& ) small di ression, the five most common nouns in En lish areaccordin to the >ord 7re"uenc) Book com!iled by John 0& 7arroll et al-time-, -!eo!le-, -way-, -water-, and -words- 4in that order5& The ama<in thin about this is that most !eo!le have no idea that we think in such abstract terms& )sk your friends, and "+ to " they1ll uess such words as -man-, -house-, -car-, -do -, and -money-& )nd while we1re on the sub.ect of frequenciesthe to! twelve letters in En lish, in order, accordin to /er enthaler, are, -ET)EI8 S?;:AF-& Ene amusin feature of S?;:AF which runs totally a ainst the stereoty!e of com!uters as -number crunchers- is this fact, !ointed out by 3ino rad, -Eur system does not acce!t numbers in numeric form, and has only been tau ht to count to ten&- "# 3ith all its mathematical under!innin , S?;:AF is a mathematical i noramus2 Just like )unt ?illary, S?;:AF doesn1t know anythin about the lower levels which make it u!& Its knowled e is lar ely procedural 4see !articularly the remark by -:r, Tony Earrwi - in section "" of the !revious :ialo ue5& It is interestin to contrast the !rocedural embeddin of knowled e in S?;:AF with the knowled e in my sentence- eneration !ro ram& )ll of the syntactical knowled e in my !ro ram was !rocedurally embedded in )u mented Transition 8etworks, written in the lan ua e )l ol= but the semantic knowled ethe information about semantic class membershi!was static, it was contained in a short list of numbers after each word& There were a few words, such as the au(iliary verbs -to be-, -to have-, and others, which were re!resented totally in !rocedures in )l ol, but they were the e(ce!tions& 0y contrast, in S?;:AF, all words were re!resented as !ro rams& ?ere is a case which demonstrates that, des!ite the theoretical equivalence of data and !ro rams, in !ractice the choice of one over the other has ma.or consequences&

Synta- and Semantics

)nd now, a few more words from 3ino rad,
Eur !ro ram does not o!erate by first !arsin a sentence, then doin semantic analysis, and finally by usin deduction to !roduce a res!onse& These three activities o on concurrently throu hout the understandin of a sentence& )s soon as a !iece of syntactic structure be ins to take sha!e, a semantic !ro ram is called to see whether it mi ht make sense, and the resultant answer can direct the !arsin & In decidin whether it makes sense, the semantic routine may call deductive !rocesses and ask questions about the real world& )s an e(am!le, in sentence %C of the :ialo ue 4-Put the blue !yramid on the block in the bo(-5, the !arser first comes u! with -the blue !yramid on the block- as a candidate for a noun rou!& )t this !oint, semantic analysis is done, and since -the- is definite, a check is made in the data base for the ob.ect bein referred to& 3hen no such ob.ect is found, the !arsin is redirected to find the noun rou! -the blue !yramid-& It will then o on to find -on the block in the bo(- as a sin le !hrase indicatin a location&&& Thus there is a continuin inter!lay between the different sorts of analysis, with the results of one affectin the others&B+

It is e(tremely interestin that in natural lan ua e, synta( and semantics are so dee!ly intertwined& Aast 7ha!ter, in discussin the elusive conce!t of -form-, we had broken the notion into two cate ories, syntactic form, which is detectable by a !redictably terminatin decision !rocedure, and semantic form, which is not& 0ut here, 3ino rad is tellin us thatat least when the usual senses of -synta(- and -semanticsare takenthey mer e ri ht into each other, in natural lan ua e& The e(ternal form of a sentencethat is, its com!osition in terms of elementary si nsdoes not divide u! so neatly into syntactic and semantic as!ects& This is a very si nificant !oint for lin uistics& ?ere are some final comments on S?;:AF by 3ino rad&
Aet us look at what the system would do with a sim!le descri!tion like -a red cube which su!!orts a !yramid-& The descri!tion will use conce!ts like 0AE7@, ;E:, PH;)/I:, and EMFI:I/E8SIE8)Aall !arts of the system1s underlyin cate ori<ation of the world& The result can be re!resented in a flow chart like that in 'i ure ""*& 8ote that this is a !ro ram for findin an ob.ect fittin the descri!tion& It would then be incor!orated into a command for doin somethin with the ob.ect, a question askin somethin about it, or, if it a!!eared in a statement, it would become !art of the !ro ram which was enerated to re!resent the meanin for later use& 8ote that this bit of !ro ram could also be used as a test to see whether an ob.ect fit the descri!tion, if the first 'I8: instruction were told in advance to look only at that !articular ob.ect& )t first lance, it seems that there is too much structure in this !ro ram, as we don1t like to think of the meanin of a sim!le !hrase as e(!licitly containin loo!s, conditional tests, and other !ro rammin details& The solution is to !rovide an internal lan ua e that contains the a!!ro!riate loo!in and checkin as its !rimitives, and in which the re!resentation of the !rocess is as sim!le as the descri!tion& The !ro ram described in 'i ure ""* would be written in PA)88E; lookin somethin like what is below,

73G5=E BBE! &rocedural representation of "a red cu(e which supports a p)ramid!" U*dapted from =oger -chank and ;enneth 8ol(), 7om!uter /odels of Thou ht and Aan ua e -an 7rancisco: >! ,! 7reeman, BICH/, p! BCF!V

46E)A 4IS ?T" 0AE7@55 46E)A 47EAE;-E' ?T" ;E:55 46E)A 4EMFI:I/E8SIE8)A ?T"55 46E)A 4IS ?TB PH;)/I:55 46E)A 4SFPPE;T ?T" ?TB55 The loo!s of the flowchart are im!licit in PA)88E;1S backtrack control structure& The descri!tion is evaluated by !roceedin down the list until some oal fails, at which time the system backs u! automatically to the last !oint where a decision was made, tryin a different !ossibility& ) decision can be made whenever a new ob.ect name or G);I)0AE 4indicated by the !refi( -?-5 such as -?T"- or -?TB- a!!ears& The variables are used by the !attern matcher& If they have already been assi ned to a !articular item, it checks to see whether the 6E)A is true for that item& If not, it checks for all !ossible items which satisfy the 6E)A, by choosin one, and then takin successive ones whenever backtrackin occurs to that !oint& Thus, even the distinction between testin and choosin is im!licit& B"

Ene si nificant strate y decision in devisin this !ro ram was to not translate all the way from En lish into AISP, but only !artwayinto PA)88E;& Thus 4since the PA)88E;

inter!reter is itself written in AISP5, a new intermediate levelPA)88E;was inserted between the to!-level lan ua e 4En lish5 and the bottom-level lan ua e 4machine lan ua e5& Ence a PA)88E; !ro ram had been made from an En lish sentence fra ment, then it could be sent off to the PA)88E; inter!reter, and the hi her levels of S?;:AF would be freed u!, to work on new tasks& This kind of decision constantly cro!s u!, ?ow many levels should a system have? ?ow much and what kind of -intelli ence- should be !laced on which level? These are some of the hardest !roblems facin )I today& Since we know so little about natural intelli ence, it is hard for us to fi ure out which level of an artificially intelli ent system should carry out what !art of a task& This ives you a lim!se behind the scenes of the :ialo ue !recedin this 7ha!ter& 8e(t 7ha!ter, we shall meet new and s!eculative ideas for )I&

Contrafactus
The 8ra( has in#ited a small group of friends o#er to watch the -aturda) afternoon foot(all game on tele#ision! *chilles has alread) arri#ed, (ut the Tortoise and his friend the -loth are still awaited! *chilles, 7ould that be our friends, a-ridin u! on that unusual one-wheeled vehicle? The -loth and Tortoise dismount and come in!/ 8ra(, )h, my friends, I1m so lad you could make it& /ay I !resent my old and beloved acquaintance, /r& Slothand this is )chilles& I believe you know the Tortoise& -loth, This is the first time I can recall makin the acquaintance of a 0icyclo!s& Pleased to meet you, )chilles& I1ve heard many fine thin s said about the bicyclo!ean s!ecies& *chilles, Aikewise, I1m sure& /ay I ask about your ele ant vehicle? Tortoise, Eur tandem unicycle, you mean? ?ardly ele ant& It1s .ust a way for two to et from ) to 0, at the same s!eed& -loth, It1s built by a com!any that also makes teeter-teeters& *chilles, I see, I see& 3hat is that knob on it? -loth, That1s the earshift& *chilles, )ha2 )nd how many s!eeds does it have? Tortoise, Ene, includin reverse& /ost models have fewer, but this is a s!ecial model& *chilles, It looks like a very nice tandem unicycle& Eh, /r& 7rab, I wanted to tell you how much I en.oyed hearin your orchestra !erform last ni ht& 8ra(, Thank you, )chilles& 3ere you there by any chance, /r& Sloth? -loth, 8o, I couldn1t make it, I1m sad to say& I was !artici!atin in a mi(ed sin les !in !in tournament& It was quite e(citin because my team was involved in a one-way tie for first !lace& *chilles, :id you win anythin ? -loth, 7ertainly dida two-sided />bius stri! made out of co!!er= it is silver-!lated on one side, and old-!lated on the other& 8ra(, 7on ratulations, /r& Sloth& -loth, Thank you& 3ell, do tell me about the concert& 8ra(, It was a most en.oyable !erformance& 3e !layed some !ieces by the 0ach twins& -loth, The famous Job and Sebastian? 8ra(, Ene and the same& )nd there was one work that made me think of you, /r& Sloth a marvelous !iano concerto for two left hands& The ne(t-to-last 4and only5 movement was a one-voice fu ue& Hou can1t ima ine its intricacies& 'or our finale, we !layed 0eethoven1s 8inth 9enfunny& )t the end, everyone in the audience rose and cla!!ed with one hand& It was overwhelmin & -loth, Eh, I1m sorry I missed it& 0ut do you su!!ose it1s been recorded? )t home I have a fine hi-fi to !lay it onthe best two-channel monaural system money can buy& 8ra(, I1m sure you can find it somewhere& 3ell, my friends, the ame is about to be in&

*chilles, 3ho is !layin today, /r& 7rab? 8ra(, I believe it1s ?ome Team versus Gisitors& Eh, nothat was last week& I think this week it1s Eut-of-Towners& *chilles, I1m rootin for ?ome Team& I always do& -loth, Eh, how conventional& I never root for ?ome Team& The closer a team lives to the anti!odes, the more I root for it& *chilles, Eh, so you live in the )nti!odes? I1ve heard it1s charmin to live there, but I wouldn1t want to visit them& They1re so far away& -loth, )nd the stran e thin about them is that they don1t et any closer no matter which way you travel& Tortoise, That1s my kind of !lace& 8ra(, It1s ame time& I think I1ll turn on the TG& ,e walks o#er to an enormous ca(inet with a screen, underneath which is an instrument panel as complicated as that of a Met airplane! ,e flicks a kno(, and the foot(all stadium appears in (right #i#id color on the screen!/ *nnouncer, 6ood afternoon, fans& 3ell, it looks like that time of year has rolled around a ain when ?ome Team and Eut-of-Town face each other on the ridiron and !lay out their classic !i skin rivalry& It1s been dri<<lin on and off this afternoon, and the field1s a little wet, but des!ite the weather it !romises to be a fine ame, es!ecially with that 6;E)T !air of ei hth-backs !layin for ?ome Team, Ted<illi er and Palindromi& )nd now, here1s Pili!ik, kickin off for ?ome Team& It1s in the air2 'lam!son takes it for Eut-of-Towners, and runs it backhe1s to the B+, the BL, the %+, and down at the %B& That was /ool in on the tackle for ?ome Team& 8ra(, ) su!erb runback2 :id you see how he was )A/EST tackled by Muilkerbut somehow broke away? -loth, Eh, don1t be silly, 7rab& 8othin of the kind ha!!ened& Muilker did 8ET tackle 'lam!son& There1s no need to confuse !oor )chilles 4or the rest of us5 with hocus!ocus about what -almost- ha!!ened& It1s a factwith no -almost- 1s, -if - 1s, -and- 1s, or -but- 1s& *nnouncer, ?ere1s the instant re!lay& Just watch number $#, Muilker, come in from the side, sur!risin 'lam!son, and .ust about tackle him2 -loth, -Just about-2 0ah2 *chilles, Such a raceful maneuver2 3hat would we do without instant re!lays? *nnouncer, It1s first down and "+ for Eut-of-Town& 8oddle takes the ball, hands off to Erwi(it1s a reverseErwi( runs around to the ri ht, handin off to 'lam!sona double reverse, folks2and now 'lam!son hands it to Treefi , who1s downed twelve yards behind scrimma e& ) twelve-yard loss on a tri!le reverse2 -loth, I love it2 ) sensational !lay2 *chilles, 0ut, /r& S, I thou ht you were rootin for Eut-of-Town& They lost twelve yards on the !lay& -loth, They did? Eh, wellwho cares, as lon as it was a beautiful !lay? Aet1s see it a ain&

!!!and so the first half of the game passes! Towards the end of the third "uarter, a particularl) crucial pla) comes up for ,ome Team! The) are (ehind () eight points! 3t's third down and BK, and the) (adl) need a first down!/ *nnouncer, The ball is hiked to Ted<illi er, who fades back, lookin for a receiver, and fakes to Muilker& There1s Palindromi, !layin wide ri ht, with nobody near him& Ted<illi er s!ots him and fires a low !ass to him& Palindromi snatches it out of the air, and4There is an audi(le groan from the crowd &5oh, he ste!s out of bounds2 3hat a crushin blow for ?ome Team, folks2 If Palindromi hadn1t ste!!ed out of bounds, he could1ve run all the way to the end <one for a touchdown2 Aet1s watch the sub.unctive instant re!lay& *nd on the screen the same lineup appears as (efore!/ The ball is hiked to Ted<illi er, who fades back, lookin for a receiver, and fakes to Muilker& There1s Palindromi, !layin wide ri ht, with nobody near him& Ted<illi er s!ots him, and fires a low !ass to him& Palindromi snatches it out of the air, and 4There is an audi(le gasp from the crowd!5he almost ste!s out of bounds2 0ut he1s still in bounds, and it1s clear all the way to the end <one2 Palindromi streaks in, for a touchdown for ?ome Team2 The stadium (reaks into a giant roar of appro#al!/ 3ell, folks, that1s what would1ve ha!!ened if Palindromi hadn1t ste!!ed out of bounds& *chilles, 3ait a minute&&& 3)S there a touchdown, or 3)S81T there? 8ra(, Eh, no& That was .ust the sub.unctive instant re!lay& They sim!ly followed a hy!othetical a little way out, you know& -loth, That is the most ridiculous thin I ever heard of2 8e(t thin you know, they1ll be inventin concrete earmuffs& Tortoise, Sub.unctive instant re!lays are a little unusual, aren1t they? 8ra(, 8ot !articularly, if you have a Sub.unc-TG& *chilles, Is that one rade below a .unk TG? 8ra(, 8ot at all2 It1s a new kind of TG, which can o into the sub.unctive mode& They1re !articularly ood for football ames and such& I .ust ot mine& *chilles, 3hy does it have so many knobs and fancy dials? 8ra(, So that you can tune it to the !ro!er channel& There are many channels broadcastin in the sub.unctive mode, and you want to be able to select from them easily& *chilles, 7ould you show us what you mean? I1m afraid I don1t quite understand what all this talk of -broadcastin in the sub.unctive mode- is about& 8ra(, Eh, it1s quite sim!le, really& Hou can fi ure it out yourself& I1m oin into the kitchen to fi( some 'rench fries, which I know are /r& Sloth1s weakness& -loth, /mmmm2 6o to it, 7rab2 'rench fries are my favorite food& 8ra(, 3hat about the rest of you? Tortoise, I could devour a few&

*chilles, Aikewise& 0ut wait-before you o into the kitchen, is there some trick to usin your Sub.unc-TG? 8ra(, 8ot !articularly& Just continue watchin the ame& and whenever there1s a near miss of some sort, or whenever you wish thin s had one differently in some way, .ust fiddle with the dials, and see what ha!!ens& Hou can1t do it any harm, thou h you may !ick u! some e(otic channels& 4*nd he disappears into the kitchen&5 *chilles, I wonder what he means by that& Eh well, let1s et back to this ame& I was quite wra!!ed u! in it& *nnouncer, It1s fourth down for Eut-of-Town, with ?ome Team receivin & Eut-of-Town is in !unt formation, with Ted<illi er !layin dee!& Erwi( is back to kickand he ets a lon hi h one away& It1s comin down near Ted<illi er *chilles, 6rab it, Ted<illi er2 6ive those Eut-of-Towners a run for their money2 *nnouncer, and lands in a !uddle@E;SPAES?It takes a weird bounce2 8ow S!runk is madly scramblin for the ball2 It looks like it .ust barely ra<ed Ted<illi er on the bounce, and then sli!!ed away from himit1s ruled a fumble& The referee is si nalin that the formidable S!runk has recovered for Eut-of-Town on the ?ome Team $2 It1s a bad break for ?ome Team& Eh, well, that1s the way the cookie crumbles& *chilles, Eh, no2 If only it hadn1t been rainin &&& 4>rings his hands in despair&5 -loth, )8ET?E; of those confounded hy!otheticals2 3hy are the rest of you always runnin off into your absurd worlds of fantasy? If I were you, I would stay firmly rounded in reality& -8o sub.unctive nonsense- is my motto& )nd I wouldn1t abandon it even if someone offered me a hundrednay, a hundred and twelve'rench fries& *chilles, Say, that ives me an idea& /aybe by suitably fiddlin with these knobs, I can con.ure u! a sub.unctive instant re!lay in which it isn1t rainin , there1s no !uddle, no weird bounce, and Ted<illi er doesn1t fumble& I wonder&&& 4 >alks up to the -u(Munc%T< and stares at it!/ 0ut I haven1t any idea what these different knobs do& 4 -pins a few at random&5 *nnouncer, It1s fourth down for Eut-of-Town, with ?ome Team receivin & Eut-of-Town is in !unt formation, with Ted<illi er !layin dee!& Erwi( is back to kickand he ets a lon hi h one away& It1s comin down near Ted<illi er *chilles, 6rab it, Ted<illi er2 6ive those Eut-of-Towners a run for their money2 *nnouncer, and lands in a !uddle@E;SPAES?2 Ehit bounces ri ht into his arms2 8ow S!runk is madly scramblin after him, but he1s ot ood blockin , and he steers his way clear of the formidable S!runk, and now he1s ot an o!en field ahead of him& Aook at that, folks2 ?e1s to the L+, the C+, the %+, the B+, the "+touchdown, ?ome Team2 4,uge cheers from the ,ome Team side&5 3ell, fans, that1s how it would have one, if footballs were s!heres instead of oblate s!heroids2 0ut in reality, ?ome Team loses the ball, and Eut-of-Towners take over on the ?ome Team $-yard line& Eh, well, that1s the way the ball bounces& *chilles, 3hat do you think of T?)T, /r& Sloth? *nd *chilles gi#es a smirk in the direction of the -loth, (ut the latter is completel) o(li#ious to its de#astating effect, as he is (us) watching, the 8ra( arri#e with, a large

platter with a hundred and twel#eOna), a hundredOlarge and delicious 7rench fries, and napkins for all!/ 8ra(, So how do you three find my Sub.unc-TG? -loth, /ost disa!!ointin , 7rab, to be quite frank& It seems to be badly out of order& It makes !ointless e(cursions into nonsense at least half the time& If it belon ed to me, I would ive it away immediately to someone like you, 7rab& 0ut of course it doesn1t belon to me& *chilles, It1s quite a stran e device& I tried to rerun a !lay to see how it would have one under different weather conditions, but the thin seems to have a will of its own2 Instead of chan in the weather, it chan ed the football sha!e to ;EF8: instead of 'EET0)AA-S?)PE:2 8ow tell me Ohow can a football not be sha!ed like a football? That1s a contradiction in terms& ?ow !re!osterous2 8ra(, Such tame ames2 I thou ht you1d surely find more interestin sub.unctives& ?ow would you like to see how the last !lay would have looked if the ame had been baseball instead of football? Tortoise, Eh2 )n outstandin idea2 The 8ra( twiddles two kno(s, and steps (ack!/ *nnouncer, There are four away, andO *chilles, 'EF; away2? *nnouncer, That1s ri ht, fansOfour away& 3hen you turn football into baseball, SE/ET?I861S ot to ive2 8ow as I was about to say, there are four away, with Eut-ofTown in the field, and ?ome Team u!& Ted<illi er is at bat& Eut-of-Town is in bunt formation& Erwi( raises his arm to !itchOand he ets a lon hi h ball away& It1s headin strai ht for Ted<illi erO *chilles, Smash it, Ted<illi er2 6ive those Eut-of-Towners a home run for their money2 *nnouncer, Obut it seems to be a s!itball, as it takes a stran e curve& 8ow S!runk is madly scramblin for the ball2 It looks like it .ust barely ra<ed Ted<illi er1s bat, then bounced off itOit1s ruled a fly ball& The um!ire is si nalin that the formidable S!runk has cau ht it for Eut-of-Town, to end the seventh innin & It1s a bad break for ?ome Team& That1s how the last !lay would have looked, football fans, if this had been a ame of baseball& -loth, 0ah2 Hou mi ht as well trans!ort this ame to the /oon& 8ra(, 8o sooner said than done2 Just a twiddle here, a twiddle there&&& :n the screen there appears a desolate crater%pitted field, with two teams in space suits facing each other, immo(ile! *ll at once, the two teams fl) into motion, and the pla)ers are making great (ounds into the air, sometimes o#er the heads of other pla)ers! The (all is thrown into the air, and sails so high that it almost disappears, and then slowl) comes floating down into the arms of one space%suited pla)er, roughl) a "uarter%mile from where it was released!/

*nnouncer, )nd there, friends, you have the sub.unctive instant re!lay as it would have ha!!ened on the /oon& 3e1ll be ri ht back after this im!ortant commercial messa e from the friendly folks who brew 6lum!f 0eerOmy favorite kind of beer2 -loth, If I weren1t so la<y, I would take that broken TG back to the dealer myself2 0ut alas, it1s my fate to be a la<y Sloth&&& 4,elps himself to a large go( of 7rench fries&5 Tortoise, That1s a marvelous invention, /r& 7rab& /ay I su est a hy!othetical? 8ra(, Ef course2 Tortoise, 3hat would that last !lay have looked like if s!ace were four-dimensional? 8ra(, Eh, that1s a com!licated one, /r& T, but I believe I can code it into the dials& Just a moment& ,e steps up, and, for the first time, appears to (e using the full power of the control panel of his -u(Munc%T<, turning almost e#er) kno( two or three times, and carefull) checking #arious meters! Then he steps (ack with a satisfied e9pression on his face!/ I think this should do it& *nnouncer, )nd now let1s watch the sub.unctive instant re!lay& * confusing arra) of twisted pipes appears on the screen! 3t grows larger, then smaller, and for a moment seems to do something akin to rotation! Then it turns into a strange mushroom%shaped o(Mect, and (ack to a (unch of pipes! *s it metamorphoses from this into other (i@arre shapes, the announcer gi#es his commentar)!/ Ted<illi er1s fadin back to !ass& ?e s!ots Palindromi ten yards outfield, and !asses it to the ri ht and outwardsOit looks ood2 Palindromi1s at the %L-yard !lane, the C+, and he1s tackled on his own C%-yard !lane& )nd there you have it, %-: fans, as it would have looked if football were !layed in four s!atial dimensions& *chilles, 3hat is it you are doin , /r& 7rab, when you twirl these various dials on the control !anel? 8ra(, I am selectin the !ro!er sub.unctive channel& Hou see, there are all sorts of sub.unctive channels broadcastin simultaneously, and I want to tune in !recisely that one which re!resents the kind of hy!othetical which has been su ested& *chilles, 7an you do this on any TG? 8ra(, 8o, most TG1s can1t receive sub.unctive channels& They require a s!ecial kind of circuit which is quite difficult to make& -loth, ?ow do you know which channel is broadcastin what? :o you look it u! in the news!a!er? 8ra(, I don1t need to know the channel1s call letters& Instead, I tune it in by codin , in these dials, the hy!othetical situation which I want to be re!resented& Technically, this is called -addressin a channel by its counterfactual !arameters-& There are always a lar e number of channels broadcastin every conceivable world& )ll the channels which carry worlds that are -near- to each other have call letters that are near to each other, too&

Tortoise, 3hy did you not have to turn the dials at all, the first time we saw a sub.unctive instant re!lay? 8ra(, Eh, that was because I was tuned in to a channel which is very near to the ;eality 7hannel, but ever so sli htly off& So every once in a while, it deviates from reality& It1s nearly im!ossible to tune ET)7TAH into the ;eality 7hannel& 0ut that1s all ri ht, because it1s so dull& )ll their instant re!lays are strai ht2 7an you ima ine? 3hat a bore2 -loth, I find the whole idea of Sub.unc-TG1s one iant bore& 0ut !erha!s I could chan e my mind, if I had some evidence that your machine here could handle an I8TE;ESTI86 counterfactual& 'or e(am!le, how would that last !lay have looked if addition were not commutative? 8ra(, Eh me, oh my2 That chan e is a little too radical, I1m afraid, for this model& I unfortunately don1t have a Su!er.unc-TG, which is the to! of the line& Su!er.unc-TG1s can handle )8HT?I86 you throw at them& -loth, 0ah2 8ra(, 0ut lookOI can do )A/EST as well& 3ouldn1t you like to see how the last !lay would have ha!!ened if "% were not a !rime number? -loth, 8o thanks2 T?)T doesn1t make any sense2 )nyway, if I were the last !lay, I1d be ettin !retty tired of bein trotted out time and a ain in new arb for the likes of you fu<<y-headed conce!t-sli!!ers& Aet1s et on with the ame2 *chilles, 3here did you et this Sub.unc-TG, /r& 7rab? 8ra(, 0elieve it or not, /r& Sloth and I went to a country fair the other evenin , and it was offered as the first !ri<e in a lottery& 8ormally I don1t indul e in such frivolity, but some cra<y im!ulse rabbed me, and I bou ht one ticket& *chilles, 3hat about you /r& Sloth? -loth, I admit, I bou ht one, .ust to humor old 7rab& 8ra(, )nd when the winnin number was announced, I found, to my ama<ement, that I1d won the lottery2 *chilles, 'antastic2 I1ve never known anyone who won anythin in a lottery before2 8ra(, I was flabber asted at my ood fortune& -loth, :on1t you have somethin else to tell us about that lottery, 7rab? 8ra(, Eh, nothin much& It1s .ust that my ticket number was "B#& 8ow when they announced the winnin number, it was "B*O.ust one off& -loth, So you see, he actually didn1t win it at all& *chilles, ?e )A/EST won, thou h&&& 8ra(, I !refer to say that I won it, you see& 'or I came so terribly close&&& If my number had been only one smaller, I would have won& -loth, 0ut unfortunately, 7rab, a miss is as ood as a mile& Tortoise, Er as bad& 3hat about you, /r& Sloth? 3hat was your number? -loth, /ine was BLDOthe ne(t !ower of B above "B*& Surely, that counts as a hit, if anythin does2 I can1t understand why, however, those fair officials Othose F8fair officialsOwere so thickheaded about it& They refused to award me my fully deserved !ri<e& Some other .oker claimed ?E deserved it, because his number was "B*& " think my number was far closer than his, but you can1t fi ht 7ity ?all&

*chilles, I1m all confused& If you didn1t win the Sub.unc-TG after all, /r& 7rab, then how can we have been sittin here all afternoon watchin it? It seems as if we ourselves have been livin in some sort of hy!othetical world that would have been, had circumstances .ust been ever so sli htly different&&& *nnouncer, )nd that, folks, was how the afternoon at /r& 7rab1s would have been s!ent, had he won the Sub.unc-TG& 0ut since he didn1t, the four friends sim!ly s!ent a !leasant afternoon watchin ?ome Team et creamed, "B*-+& Er was it BLD-+? Eh well, it hardly matters, in five-dimensional Plutonian steam hockey&

C/A,TE$ <I<

Artificial Intelligence5 ,ros!ects

(Almost( Situations and SubVuncti"es
)'TE; ;E):I86 8ontrafactus, a friend said to me, -/y uncle was almost President of the F&S&2- -;eally?- I said& -Sure,- he re!lied, -he was ski!!er of the PT "+*&- 4John '& @ennedy was ski!!er of the PT "+#&5 That is what 8ontrafactus is all about& In everyday thou ht, we are constantly manufacturin mental variants on situations we face, ideas we have, or events that ha!!en, and we let some features stay e(actly the same while others -sli!-& 3hat features do we let sli!? 3hat ones do we not even consider lettin sli!? 3hat events are !erceived on some dee! intuitive level as bein close relatives of ones which really ha!!ened? 3hat do we think -almost- ha!!ened or -could have- ha!!ened, even thou h it unambi uously did not? 3hat alternative versions of events !o! without any conscious thou ht into our minds when we hear a story? 3hy do some counterfactuals strike us as -less counterfactual- than other counterfactuals? )fter all, it is obvious that anythin that didn1t ha!!en didn1t ha!!en& There aren1t de rees of -didn1t-ha!!en-ness-& )nd the same oes for -almost- situations& There are times when one !laintively says, -It almost ha!!ened-, and other times when one says the same thin , full of relief& 0ut the -almostlies in the mind, not in the e(ternal facts& :rivin down a country road, you run into a swarm of bees& Hou don1t .ust duly take note of it= the whole situation is immediately !laced in !ers!ective by a swarm of -re!lays- that crowd into your mind& Ty!ically, you think, -Sure am lucky my window wasn1t o!en2-Oor worse, the reverse, -Too bad my window wasn1t closed2- -Aucky I wasn1t on my bike2- -Too bad I didn1t come alon five seconds earlier&- Stran e but!ossible re!lays, -If that had been a deer, I could have been killed2- -I bet those bees would have rather had a collision with a rosebush&- Even stran er re!lays, -Too bad those bees weren1t dollar bills2- -Aucky those bees weren1t made of cement2- -Too bad it wasn1t .ust one bee instead of a swarm&- -Aucky I wasn1t the swarm instead of bein me&- 3hat sli!s naturally and what doesn1tOand why? In a recent issue of The New Yorker ma a<ine, the followin e(cer!t from the -Philadel!hia 3elcomat- was re!rinted," If Aeonardo da Ginci had been born a female the ceilin of the Sistine 7ha!el mi ht never have been !ainted&

The New Yorker commented, )nd if /ichelan elo had been Siamese twins, the work would have been com!leted in half the time& The !oint of The New Yorker's comment is not that such counterfactuals are false= it is more that anyone who would entertain such an idea Oanyone who would -sli!- the se( or number of a iven human bein Owould have to be a little loony& Ironically, thou h, in the same issue, the followin sentence, concludin a book review, was !rinted without blushin , I think he UProfessor Phili!! 'rankV would have en.oyed both of these books enormously&B 8ow !oor Professor 'rank is dead= and clearly it is nonsense to su est that someone could read books written after his death& So why wasn1t this serious sentence also scoffed at? Somehow, in some difficult-to-!in-down sense, the !arameters sli!!ed in this sentence do not violate our sense of -!ossibility- as much as in the earlier e(am!les& Somethin allows us to ima ine -all other thin s bein equal- better in this one than in the others& 0ut why? 3hat is it about the way we classify events and !eo!le that makes us know dee! down what is -sensible- to sli!, and what is -silly-, 7onsider how natural it feels to sli! from the valueless declarative -I don1t know ;ussian- to the more char ed conditional -I would like to know ;ussian- to the emotional sub.unctive -I wish I knew ;ussian- and finally to the rich counterfactual -If I knew ;ussian, I would read 7hekhov and Aermontov in the ori inal-& ?ow flat and dead would be a mind that saw nothin in a ne ation but an o!aque barrier2 ) live mind can see a window onto a world of !ossibilities& I believe that -almost- situations and unconsciously manufactured sub.unctives re!resent some of the richest !otential sources of insi ht into how human bein s or ani<e and cate ori<e their !erce!tions of the world& )n eloquent co-!ro!onent of this view is the lin uist and translator 6eor e Steiner, who, in his book *fter Ba(el, has written,
?y!otheticals, 1ima inaries1, conditionals, the synta( of counter-factuality and contin ency may well be the enerative centres of human s!eech&&& UTheyV do more than occasion !hiloso!hical and rammatical !er!le(ity& 8o less than future tenses to which they are, one feels, related, and with which they ou ht !robably to be classed in the lar er set of 1su!!ositionals1 or 1alternates1, these 1if1 !ro!ositions are fundamental to the dynamics of human feelin &&& Eurs is the ability, the need, to ainsay or 1un-say1 the world, to ima e and s!eak it otherwise&&& 3e need a word which will desi nate the !ower, the com!ulsion of lan ua e to !osit 1otherness1&&& Perha!s 1alternity1 will do, to define the 1other than the case1, the counter-factual !ro!ositions, ima es, sha!es of will and evasion with which we char e our mental bein and by means of which we build the chan in , lar ely fictive milieu of our somatic and our social e(istence&&&

'inally, Steiner sin s a counterfactual hymn to counterfactuality,

It is unlikely that man, as we know him, would have survived without the fictive, counter-factual, anti-determinist means of lan ua e, without the semantic ca!acity, enerated and stored in the 1su!erfluous, <ones of the corte(, to conceive of, to articulate !ossibilities beyond the treadmill of or anic decay and death&%

The manufacture of -sub.unctive worlds- ha!!ens so casually, so naturally, that we hardly notice what we are doin & 3e select from our fantasy a world which is close, in some internal mental sense, to the real world& 3e com!are what is real with what we !erceive as almost real& In so doin , what we ain is some intan ible kind of !ers!ective on reality& The Sloth is a droll e(am!le of a variation on reality Oa thinkin bein without the ability to sli! into sub.unctives 4or at least, who claims to be without the ability Obut you may have noticed that what he says is full of counterfactuals25& Think how immeasurably !oorer our mental lives would be if we didn1t have this creative ca!acity for sli!!in out of the midst of reality into soft -what if- 1s2 )nd from the !oint of view of studyin human thou ht !rocesses, this sli!!a e is very interestin , for most of the time it ha!!ens com!letely without conscious direction, which means that observation of what kinds of thin s sli!, versus what kinds don1t, affords a ood window on the unconscious mind& Ene way to ain some !ers!ective on the nature of this mental metric is to -fi ht fire with fire-& This is done in the :ialo ue, where our -sub.unctive ability- is asked to ima ine a world in which the very notion of sub.unctive ability is sli!!ed, com!ared to what we e(!ect& In the :ialo ue, the first sub.unctive instant re!lay Othat where Palindromi stays in boundsOis quite a normal thin to ima ine& In fact, it was ins!ired by a com!letely ordinary, casual remark made to me by a !erson sittin ne(t to me at a football ame& 'or some reason it struck me and I wondered what made it seem so natural to sli! that !articular thin , but not, say, the number of the down, or the !resent score& 'rom those thou hts, I went on to consider other, !robably less sli!!able features, such as the weather 4that1s in the :ialo ue5, the kind of ame 4also in the :ialo ue5, and then even loonier variations 4also in the :ialo ue5& I noticed, thou h, that what was com!letely ludicrous to sli! in one situation could be quite sli!!able in another& 'or instance, sometimes you mi ht s!ontaneously wonder how thin s would be if the ball had a different sha!e 4e& &, if you are !layin basketball with a half-inflated ball5= other times that would never enter your mind 4e& &, when watchin a football ame on TG5&

Layers of Stability
It seemed to me then, and still does now, that the sli!!ability of a feature of some event 4or circumstance5 de!ends on a set of nested conte(ts in which the event 4or circumstance5 is !erceived to occur& The terms constant, !arameter, and variable, borrowed from mathematics, seem useful here& Eften mathematicians, !hysicists, and others will carry out a calculation, sayin -c is a constant, p is a !arameter, and # is a variable-& 3hat they mean is that any of them can vary 4includin the -constant-5= however, there is a kind of hierarchy of variability& In the situation which is bein re!resented by the symbols, c establishes a lobal condition= p establishes some less lobal condition which can vary while c is held fi(ed= and finally, # can run around while

c and p are held fi(ed& It makes little sense to think of holdin # fi(ed while c and p vary, for c and p establish the conte(t in which # has meanin & 'or instance, think of a dentist who has a list of !atients, and for each !atient, a list of teeth& It makes !erfect sense 4and !lenty of money5 to hold the !atient fi(ed and vary his teeth Obut it makes no sense at all to hold one tooth fi(ed and vary the !atient& 4)lthou h sometimes it makes ood sense to vary the dentist&&& 5 3e build u! our mental re!resentation of a situation layer by layer& The lowest layer establishes the dee!est as!ect of the conte(t Osometimes bein so low that it cannot vary at all& 'or instance, the three-dimensionality of our world is so in rained that most of us never would ima ine lettin it sli! mentally& It is a constant constant& Then there are layers which establish tem!orarily, thou h not !ermanently, fi(ed as!ects of situations, which could be called back round assum!tionsOthin s which, in the back of your mind, you know can vary, but which most of the time you unquestionin ly acce!t as unchan in as!ects& These could still be called -constants-& 'or instance, when you o to a football ame, the rules of the ame are constants of that sort& Then there are -!arameters-, you think of them as more variable, but you tem!orarily hold them constant& )t a football ame, !arameters mi ht include the weather, the o!!osin team, and so forth& There could beOand !robably areOseveral layers of !arameters& 'inally, we reach the -shakiest- as!ects of your mental re!resentation of the situation Othe variables& These are thin s such as Palindromi1s ste!!in out of bounds, which are mentally -looseand which you don1t mind lettin sli! away from their real values, for a short moment&

+rames and ;ested Conte-ts

The word frame is in vo ue in )I currently, and it could be defined as a c omputational instantiation of a conte9t& The term is due to /arvin /insky, as are many ideas about frames, thou h the eneral conce!t has been floatin around for a ood number of years& In frame lan ua e, one could say that mental re!resentations of situations involve frames nested within each other& Each of the various in redients of a situation has its own frame& It is interestin to verbali<e e(!licitly one of my mental ima es concernin nested frames& Ima ine a lar e collection of chests of drawers& 3hen you choose a chest, you have a frame, and the drawer holes are !laces where -subframes- can be attached& 0ut subframes are themselves chests of drawers& ?ow can you stick a whole chest of drawers into the slot for a sin le drawer in another chest of drawers? Easy, you shrink and distort the second chest, since, after all, this is all mental, not !hysical& 8ow in the outer frame, there may be several different drawer slots that need to be filled= then you may need to fill slots in some of the inner chests of drawers 4or subframes5& This can o on, recursively& The vivid surrealistic ima e of squishin and bendin a chest of drawers so that it can fit into a slot of arbitrary sha!e is !robably quite im!ortant, because it hints that your conce!ts are squished and bent by the conte(ts you force them into& Thus, what does your conce!t of -!erson- become when the !eo!le you are thinkin about are football !layers? It certainly is a distorted conce!t, one which is forced on you by the overall conte(t& Hou have stuck the -!erson- frame into a slot in the -football ame- frame& The theory of

re!resentin knowled e in frames relies on the idea that the world consists of quasiclosed subsystems, each of which can serve as a conte(t for others without bein too disru!ted, or creatin too much disru!tion, in the !rocess& Ene of the main ideas about frames is that each frame comes with its own set of e(!ectations& The corres!ondin ima e is that each chest of drawers comes with a builtin, but loosely bound, drawer in each of its drawer slots, called a default& If I tell you, -Picture a river bank-, you will invoke a visual ima e which has various features, most of which you could override if I added e(tra !hrases such as -in a drou ht- or -in 0ra<il- or -without a merry- o-round-& The e(istence of default values for slots allows the recursive !rocess of fillin slots to come to an end& In effect, you say, -I will fill in the slots myself as far as three layers down= beyond that I will take the default o!tions&- To ether with its default e(!ectations, a frame contains knowled e of its limits of a!!licability, and heuristics for switchin to other frames in case it has been stretched beyond its limits of tolerance& The nested structure of a frame ives you a way of -<oomin in- and lookin at small details from as close u! as you wish, you .ust <oom in on the !ro!er subframe, and then on one of its subframes, etc&, until you have the desired amount of detail& It is like havin a road atlas of the FS) which has a ma! of the whole country in the front, with individual state ma!s inside, and even ma!s of cities and some of the lar er towns if you want still more detail& Ene can ima ine an atlas with arbitrary amounts of detail, oin down to sin le blocks, houses, rooms, etc& It is like lookin throu h a telesco!e with lenses of different !ower= each lens has its own uses& It is im!ortant that one can make use of all the different scales= often detail is irrelevant and even distractin & 0ecause arbitrarily different frames can be stuck inside other frames1 slots, there is reat !otential for conflict or -collision-& The nice neat scheme of a sin le, lobal set of layers of -constants-, -!arameters-, and -variables- is an oversim!lification& In fact, each frame will have its own hierarchy of variability, and this is what makes analy<in how we !erceive such a com!le( event as a football ame, with its many subframes, subsubframes, etc&, an incredibly messy o!eration& ?ow do all these many frames interact with each other? If there is a conflict where one frame says, -This item is a constant- but another frame says, -8o, it is a variable2-, how does it et resolved? These are dee! and difficult !roblems of frame theory to which I can ive no answers& There has as yet been no com!lete a reement on what a frame really is, or on how to im!lement frames in )I !ro rams& I make my own stab at discussin some of these questions in the followin section, where I talk about some !u<<les in visual !attern reco nition, which I call -0on ard !roblems-&

Bongard ,roblems
Bongard pro(lems 40P1s5 are !roblems of the eneral ty!e iven by the ;ussian scientist /& 0on ard in his book &attern =ecognition& ) ty!ical 0POnumber L" in his collection of one hundredOis shown in 'i ure ""#&

73G5=E BBI! Bongard pro(lem GB! U7rom =B! Bongard, Pattern ;eco nition =ochelle &ark, N!,?!: ,a)den Book 8o!, -partan Books, BICK/!V

These fascinatin !roblems are intended for !attern-reco ni<ers, whether human or machine& 4Ene mi ht also throw in ETI1se(traterrestrial intelli ences&5 Each !roblem consists of twelve bo(ed fi ures 4henceforth called bo(es5, si( on the left, formin 8lass 3, and si( on the ri ht, formin 8lass 33! The bo(es may be inde(ed this way, I-) I-7 I-E I-0 I-: I-' II-) II-7 II-E II-0 II-: II-'

The !roblem is -?ow do 7lass I bo(es differ from 7lass II bo(es?) 0on ard !roblem-solvin !ro ram would have several sta es, in which raw data radually et converted into descri!tions& The early sta es are relatively infle(ible, and hi her sta es become radually more fle(ible& The final sta es have a !ro!erty which I call tentati#it), which means sim!ly that the way a !icture is re!resented is always tentative& F!on the dro! of a hat, a hi h-level descri!tion can be restructured, usin all the devices of the later sta es& The ideas !resented below also have a tentative quality to them& I will try to convey overall ideas first, lossin over si nificant difficulties& Then I will o back and try to e(!lain subtleties and tricks and so forth& So your notion of how it all works may also under o some revisions as you read& 0ut that is in the s!irit of the discussion&

,re!rocessing Selects a Mini-"ocabulary

Su!!ose, then, that we have some 0on ard !roblem which we want to solve& The !roblem is !resented to a TG camera and the raw data are read in& Then the raw data are preprocessed& This means that some salient features are detected& The names of these features constitute a -mini-vocabulary- for the !roblem= they are drawn from a eneral -salient-feature vocabulary-& Some ty!ical terms of the salient-feature vocabulary are,

line se ment, curve, hori<ontal, vertical, black, white, bi , small, !ointy, round&&& In a second sta e of !re!rocessin , some knowled e about elementary sha!es is used= and if any are found, their names are also made available& Thus, terms such as trian le, circle, square, indentation, !rotrusion, ri ht an le, verte(, cus!, arrow&&& may be selected& This is rou hly the !oint at which the conscious and the unconscious meet, in humans& This discussion is !rimarily concerned with describin what ha!!ens from here on out&

/ighWLe"el

escri!tions

8ow that the !icture is -understood-, to some e(tent, in terms of familiar conce!ts, some lookin around is done& Tentative descri!tions are made for one or a few of the twelve bo(es& They will ty!ically use sim!le descri!tors such as above, below, to the ri ht of, to the left of, inside, outside of, close to, far from, !arallel to, !er!endicular to, in a row, scattered, evenly s!aced, irre ularly s!aced, etc& )lso, definite and indefinite numerical descri!tors can be used, ", B, %, C, L,&&& many, few, etc& /ore com!licated descri!tors may be built u!, such as further to the ri ht of, less close to, almost !arallel to, etc&

73G5=E BFK! Bongard pro(lem DC! U7rom '! Bongard, Pattern ;eco nition&V

Thus, a ty!ical bo(Osay I-' of 0P C$ 4'i & "B+5 Ocould be variously described as havin , three sha!es or three white sha!es or a circle on the ri ht or two trian les and a circle or two u!wards-!ointin trian les or one lar e sha!e and two small sha!es or one curved sha!e and two strai ht-ed ed sha!es or a circle with the same kind of sha!e on the inside and outside& Each of these descri!tions sees the bo( throu h a -filter-& Eut of conte(t, any of them mi ht be a useful descri!tion& )s it turns out, thou h, all of them are -wron -, in the conte(t of the !articular 0on ard !roblem they are !art of& In other words, if you knew the distinction between 7lasses I and II in 0P C$, and were iven one of the !recedin lines as a descri!tion of an unseen drawin , that information would not allow you to tell to which 7lass the drawin belon ed& The essential feature of this bo(, in conte(t, is that it includes a circle containin a trian le& 8ote that someone who heard such a descri!tion would not be able to reconstruct the ori inal drawin , but would be able to reco ni<e drawin s which have this !ro!erty& It is

73G5=E BFB! Bongard pro(lem IB! U7rom '! Bongard, Pattern ;eco nition&V

a little like musical style, you may be an infallible reco ni<er of /o<art, but at the same time unable to write anythin which would fool anybody into thinkin it was by /o<art& 8ow consider bo( I-: of 0P #" 4'i & "B"5& )n overloaded but -ri ht- descri!tion in the conte(t of 0P #" is a circle with three rectan ular intrusions& 8otice the so!histication of such a descri!tion, in which the word -with- functions as a disclaimer, im!lyin that the -circle- is not really a circle, it is almost a circle, e(ce!t that&&& 'urthermore, the intrusions are not full rectan les& There is a lot of -!lay- in the way we use lan ua e to describe thin s& 7learly, a lot of information has been thrown away, and even more could be thrown away& ) !riori, it is very hard to know what it would be smart to throw away and what to kee!& So some sort of method for an intelli ent com!romise has to be encoded, via heuristics& Ef course, there is always recourse to lower levels of descri!tion 4i&e&, less chunked descri!tions5 if discarded information has to be retrieved, .ust as !eo!le can constantly look at the !u<<le for hel! in restructurin their ideas about it& The trick, then, is to devise e(!licit rules that say how to make tentative descri!tions for each bo(= com!are them with tentative descri!tions for other bo(es of either 7lass= restructure the descri!tions, by 435 addin information, 4ii5 discardin information, or 4iii5 viewin the same information from another an le= iterate this !rocess until findin out what makes the two 7lasses differ&

Tem!lates and Sameness- etectors

Ene ood strate y would be to try to make descri!tions structurall) similar to each other, to the e(tent this is !ossible& )ny structure they have in common will make com!arin them that much easier& Two im!ortant elements of this theory deal with this strate y& Ene is the idea of -descri!tion-schemas- or templates= the other is the idea of -amOa -sameness detector-& 'irst Sam& Sam is a s!ecial a ent !resent on all levels of the !ro ram& 4)ctually there may be different kinds of Sams on different levels&5 Sam constantly runs around within individual descri!tions and within different descri!tions, lookin for descri!tors or other thin s which are re!eated& 3hen some sameness is found, various restructurin o!erations can be tri ered, either on the sin le-descri!tion level or on the level of several descri!tions at once& 8ow tem!lates& The first thin that ha!!ens after !re!rocessin is an attem!t to manufacture a tem!late, or descri!tion-schemaOa uniform format for the descri!tions of all the bo(es in a !roblem& The idea is that a descri!tion can often be broken u! in a natural way into subdescri!tions, and those in turn into

subsubdescri!tions, if need be& The bottom is hit when you come to !rimitive conce!ts which belon to the level of the !re!rocessor& 8ow it is im!ortant to choose the way of breakin descri!tions into !arts so as to reflect commonality amon all the bo(es= otherwise you are introducin a su!erfluous and meanin less kind of -!seudoorder- into the world& En the basis of what information is a tem!late built? It is best to look at an e(am!le& Take 0P C# 4'i & "BB5& Pre!rocessin yields the information that each bo( consists of several little o1s, and one lar e closed curve& This is a valuable observation, and deserves to be incor!orated in the tem!late& Thus a first stab at a tem!late would be, lar e closed curve,----small o1s,----'I6F;E "BB& Bongard pro(lem DI! U7rom '! Bongard, Pattern ;eco nition&V

It is very sim!le, the descri!tion-tem!late has two e(!licit slots where subdescri!tions are to be attached&

A /eterarchical ,rogram
8ow an interestin thin ha!!ens, tri ered by the term -closed curve-& Ene of the most im!ortant modules in the !ro ram is a kind of semantic net Othe conce!t networkOin which all the known nouns, ad.ectives, etc&, are linked in ways which indicate their interrelations& 'or instance, -closed curve- is stron ly linked with the terms -interior- and -e(terior-& The conce!t net is .ust brimmin with information about relations between terms, such as what is the o!!osite of what, what is similar to what, what often occurs with what, and so on& ) little !ortion of a conce!t network, to be e(!lained shortly, is shown in 'i ure "B%& 0ut let us follow what ha!!ens now, in the solution of !roblem C#& The conce!ts -interior- and -e(terior- are activated by their !ro(imity in the net to -closed curve-& This su ests to the tem!late-builder that it mi ht be a ood idea to make

distinct slots for the interior and e(terior of the curve& Thus, in the s!irit of tentativity, the tem!late is tentatively restructured to be this, lar e closed curve, little o1s in interior, little o1s in e(terior, 8ow when subdescri!tions are sou ht, the terms -interior- and -e(terior- will cause !rocedures to ins!ect those s!ecific re ions of the bo(& 3hat is found in 0P C#, bo( I-) is this, lar e closed curve, circle little o1s in interior, three little o1s in e(terior, three )nd a descri!tion of bo( II-) of the same 0P mi ht be lar e closed curve, cigar little o1s in interior, three little o1s in e(terior, three 8ow Sam, constantly active in !arallel with other o!erations, s!ots the recurrence of the conce!t -three- in all the slots dealin with o1s, and this is stron reason to undertake a second tem!late-restructurin o!eration& 8otice that the first was su ested by the conce!t net, the second by Sam& 8ow our tem!late for !roblem C# becomes, lar e closed curve, three little o1s in interior, three little o1s in e(terior, 8ow that -three- has risen- one level of eneralityOnamely, into the tem!lateOit becomes worthwhile to e(!lore its nei hbors in the conce!t network& Ene of them is -trian le-, which su ests that trian les of o1s may be im!ortant& )s it ha!!ens, this leads down a blind alleyObut how could you know in advance? It is a ty!ical blind alley that a human would e(!lore, so it is ood if our !ro ram finds it too2 'or bo( II-E, a descri!tion such as the followin mi ht et enerated, lar e closed curve, circle three little o1s in interior, e"uilateral triangle three little o1s in e(terior, e"uilateral triangle

73G5=E BFH! * small portion of a concept network for a program to sol#e Bongard &ro(lems! "Nodes" are Moined () "links", which in turn can (e linked! B) considering a link as a #er( and the nodes it Moins as su(Mect and o(Mect, )ou can pull out some English sentences from this diagram!

Ef course an enormous amount of information has been thrown away concernin the si<es, !ositions, and orientations of these trian les, and many other thin s as well& 0ut that is the whole !oint of makin descri!tions instead of .ust usin the raw data2 It is the same idea as funnelin , which we discussed in 7ha!ter TI&

The Conce!t ;et'or.

3e need not run throu h the entire solution of !roblem C#= this suffices to show the constant back-and-forth interaction of individual descri!tions, tem!lates, the sameness-detector Sam, and the conce!t network& 3e should now look a little more in detail at the conce!t network and its function& ) sim!lified !ortion shown in the fi ure codes the followin ideas, -?i h- and -low- are o!!osites& -F!- and -down- are o!!osites& -?i h- and -u!- are similar& -Aow- and -down- are similar& -;i ht- and -left- are o!!osites& The -ri ht-left- distinction is similar to the -hi h-low- distinction& -E!!osite- and -similar- are o!!osites& 8ote how everythin in the netOboth nodes and linksOcan be talked about& In that sense nothin in the net is on a hi her level than anythin else& )nother !ortion of the net is shown= it codes for the ideas that ) square is a !oly on& ) trian le is a !oly on& ) !oly on is a closed curve& The difference between a trian le and a square is that one has % sides and the other has C& C is similar to %& ) circle is a closed curve& ) closed curve has an interior and an e(terior& -Interior- and -e(terior- are o!!osites& The network of conce!ts is necessarily very vast& It seems to store knowled e only statically, or declaratively, but that is only half the story& )ctually, its knowled e borders on bein !rocedural as well, by the fact that the !ro(imities in the net act as uides, or -!ro rams-, tellin the main !ro ram how to develo! its understandin of the drawin s in the bo(es& 'or instance, some early hunch may turn out to be wron and yet have the erm of the ri ht answer in it& In 0P %% 4'i & "BC5, one mi ht at first .um! to the idea that 7lass I bo(es contain -!ointy- sha!es, 7lass II bo(es contain -smooth- ones& 0ut

73G5=E BFD! Bongard pro(lem HH! U7rom '! Bongard, Pattern ;eco nition&V

on closer ins!ection, this is wron & 8evertheless, there is a worthwhile insi ht here, and one can try to !ush it further, by slidin around in the network of conce!ts be innin at -!ointy-& It is close to the conce!t -acute-, which is !recisely the distin uishin feature of 7lass I& Thus one of the main functions of the conce!t network is to allow early wron ideas to be modified sli htly, to sli! into variations which may be correct&

Sli!!age and Tentati"ity

;elated to this notion of sli!!in between closely related terms is the notion of seein a iven ob.ect as a variation on another ob.ect& )n e(cellent e(am!le has been mentioned alreadyOthat of the -circle with three indentations-, where in fact there is no circle at all& Ene has to be able to bend conce!ts, when it is a!!ro!riate& 8othin should be absolutely ri id& En the other hand, thin s shouldn1t be so wishy-washy that nothin has any meanin at all, either& The trick is to know when and how to sli! one conce!t into another& )n e(tremely interestin set of e(am!les where sli!!in from one descri!tion to another is the cru( of the matter is iven in 0on ard !roblems *L-*$ 4'i & "BL5& 0P *L is rather trivial& Aet us assume that our !ro ram identifies -line se ment- in its !re!rocessin sta e& It is relatively sim!le for it then to count line se ments and arrive at the difference between 7lass I and 7lass II in 0P *L& 8ow it oes on to 0P *D& ) eneral heuristic which it uses is to try out recent ideas which have worked& Successful re!etition of recent methods is very common in the real world, and 0on ard does not try to outwit this kind of heuristic in his collection Oin fact, he reinforces it, fortunately& So we !lun e ri ht into !roblem *D with two ideas 4-countand -line se ment-5 fused into one, -count line se ments-& 0ut as it ha!!ens, the trick of 0P *D is to count line trains rather than line se ments, where -line train- means an end-to-end concatenation of 4one or more5 line se ments& Ene way the !ro ram mi ht fi ure this out is if the conce!ts -line train- and -line se ment- are both known, and

73G5=E BFG!Bongard pro(lems EG%EC! U7rom '! Bongard, Pattern ;eco nition&V

are close in the conce!t network& )nother way is if it can invent the conce!t of -line train-Oa tricky !ro!osition, to say the least& Then comes 0P *$, in which the notion of -line se ment- is further !layed with& 3hen is a line se ment three line se ments? 4See bo( II-)&5 The !ro ram must be sufficiently fle(ible that it can o back and forth between such different re!resentations for a iven !art of a drawin & It is wise to store old re!resentations, rather than for ettin them and !erha!s havin to reconstruct them, for there is no uarantee that a newer re!resentation is better than an old one& Thus, alon with each

old re!resentation should be stored some of the reasons for likin it and dislikin it& 4This be ins to sound rather com!le(, doesn1t it?5

Meta- escri!tions
8ow we come to another vital !art of the reco nition !rocess, and that has to do with levels of abstraction and meta-descri!tions& 'or this let us consider 0P #" 4'i & "B"5 a ain& 3hat kind of tem!late could be constructed here? There is such an amount of variety that it is hard to know where to be in& 0ut this is in itself a clue2 The clue says, namely, that the class distinction very likely e(ists on a hi her level of abstraction than that of eometrical descri!tion& This observation clues the !ro ram that it should construct descriptions of descriptionsOthat is, meta-descriptions& Perha!s on this second level some common feature will emer e= and if we are lucky, we will discover enou h commonality to uide us towards the formulation of a tem!late for the meta-descri!tions2 So we !lun e ahead without a tem!late, and manufacture descri!tions for various bo(es= then, once these descri!tions have been made, we describe them& 3hat kinds of slot will our tem!late for meta-descri!tions have? Perha!s these, amon others, conce!ts used, -----recurrin conce!ts, -----names of slots, -----filters used,-----There are many other kinds of slots which mi ht be needed in metadescri!tions, but this is a sam!le& 8ow su!!ose we have described bo( I-E of 0P #"& Its 4tem!late-less5 descri!tion mi ht look like this, hori<ontal line se ment vertical line se ment mounted on the hori<ontal line se ment vertical line se ment mounted on the hori<ontal line se ment vertical line se ment mounted on the hori<ontal line se ment Ef course much information has been thrown out, the fact that the three vertical lines are of the same len th, are s!aced equidistantly, etc& 0ut it is !lausible that the above descri!tion would be made& So the meta descri!tion mi ht look like this, conce!ts used, #ertical%hori@ontal, line segment, mounted on re!etitions in descri!tion, H copies of "#ertical line segment mounted on the hori@ontal line segment" names of slots, filters used,

8ot all slots of the meta-descri!tion need be filled in= information can be thrown away on this level as well as on the -.ust-!lain-descri!tion- level& 8ow if we were to make a descri!tion for any of the other bo(es of 7lass I, and then a metadescri!tion of it, we would wind u! fillin the slot -re!etitions in descri!tion- each time with the !hrase -% co!ies of&&&- The sameness-detector would notice this, and !ick u! three%ness as a salient feature, on quite a hi h level of abstraction, of the bo(es of 7lass I& Similarly, four%ness would be reco ni<ed, via the method of metadescri!tions, as the mark of 7lass II&

+le-ibility Is Im!ortant
8ow you mi ht ob.ect that in this case, resortin to the method of metadescri!tions is like shootin a fly with an ele!hant un, for the three-ness versus four-ness mi ht as easily have shown u! on the lower level if we had constructed our descri!tions sli htly differently& Hes, trueObut it is im!ortant to have the !ossibility of solvin these !roblems by different routes& There should be a lar e amount of fle(ibility in the !ro ram= it should not be doomed if, mala!horically s!eakin , it -barks u! the wron alley- for a while& 4The amusin term -mala!hor- was coined by the news!a!er columnist Aawrence ?arrison= it means a cross between a mala!ro!ism and a meta!hor& It is a ood e(am!le of -recombinant ideas-&5 In any case, I wanted to illustrate the eneral !rinci!le that says, 3hen it is hard to build a tem!late because the !re!rocessor finds too much diversity, that should serve as a clue that conce!ts on a hi her level of abstraction are involved than the !re!rocessor knows about&

+ocusing and +iltering

8ow let us deal with another question, ways to throw information out& This involves two related notions, which I call -focusin - and -filterin -& 7ocusing involves makin a descri!tion whose focus is some !art of the drawin in the bo(, to the e(clusion of everythin else& 7iltering involves makin a descri!tion which concentrates on some !articular way of viewin the contents of the bo(, and deliberately i nores all other as!ects& Thus they are com!lementary, focusin has to do with ob.ects 4rou hly, nouns5, and filterin has to do with conce!ts 4rou hly, ad.ectives5& 'or an e(am!le of focusin , let1s look at 0P LL 4'i & "BD5& ?ere, we focus on the indentation and the little circle ne(t to it, to the e(clusion of the everythin else in the bo(& 0P BB 4'i & "B$5 !resents an e(am!le of filterin & ?ere, we must filter out every conce!t but that of si<e& ) combination of focusin and filterin is required to solve !roblem 0P L* 4'i & "B*5& Ene of the most im!ortant ways to et ideas for focusin and filterin is by another sort of -focusin -, namely, by ins!ection of a sin le !articularly sim!le bo( O say one with as few ob.ects in it as !ossible& It can be e(tremely hel!ful to com!are the starkest bo(es from the two 7lasses& 0ut how can you tell which bo(es are stark until you have descri!tions for them? 3ell, one way of detectin starkness is to look for a bo( with a minimum of the features !rovided by the !re!rocessor& This can be

73G5=E BFJ! Bongard pro(lem GG! U7rom '! Bongard, &attern =ecognition!V

73G5=E BFC! Bongard pro(lem FF! U7rom '! Bongard, &attern =ecognition!V

done very early, for it does not require a !re-e(istin tem!late= in fact, this can be one useful way of discoverin features to build into a tem!late& 0P D" 4'i & "B#5 is an e(am!le where that technique mi ht quickly lead to a solution&

Science and the >orld of Bongard ,roblems

Ene can think of the 0on ard-!roblem world as a tiny !lace where -science- is done Othat is, where the !ur!ose is to discern !atterns in the world& )s !atterns are sou ht, tem!lates are made, unmade, and remade= slots are shifted from one level of enerality to another, filterin and focusin are done= and so on& There are discoveries on all levels of com!le(ity& The @uhnian theory that certain rare events called -!aradi m shifts- mark the distinction between -normal- science and -conce!tual revolutions- does not seem to work, for we can see !aradi m shifts ha!!enin all throu hout the system, all the time& The fluidity of descri!tions ensures that !aradi m shifts will take !lace on all scales&

73G5=E BFE! Bongard pro(lem GE! U7rom '! Bongard, &attern =ecognition!V

73G5=E BFI! Bongard pro(lem JB! U7rom '! Bongard, &attern =ecognition!V

Ef course, some discoveries are more -revolutionary- than others, because they have wider effects& 'or instance, one can make the discovery that !roblems $+ and $" 4'i & "%+5 are -the same !roblem-, when looked at on a sufficiently abstract level& The key observation is that both involve de!th-B versus de!th-" nestin & This is a new level of discovery that can he made about 0on ard !roblems& There is an even hi her level, concernin the collection as a whole& If someone has never seen the collection, it can be a ood !u<<le .ust to fi ure out what it is& To fi ure it out is a revolutionary insi ht, but it must be !ointed out that the mechanisms of thou ht which allow such a discovery to be made are no different from those which o!erate in the solution of a sin le 0on ard !roblem& 0y the same token, real science does not divide u! into -normal- !eriods versus -conce!tual& revolutions-= rather, !aradi m shifts !ervade Othere are .ust bi er and smaller ones, !aradi m shifts on different levels& The recursive !lots of

73G5=E BHK! Bongard pro(lems CK%CB! U7rom '! Bongard, Pattern ;eco nition!V

I8T and 6!lot 4'i s& %B and %C5 !rovide a eometric model for this idea, they have the same structure full of discontinuous .um!s on every level, not .ust the to! level O only the lower the level, the smaller the .um!s&

Connections to 0ther Ty!es of Thought

To set this entire !ro ram somewhat in conte(t, let me su est two ways in which it is related to other as!ects of co nition& 8ot only does it de!end on other as!ects of co nition, but also they in turn de!end on it& 'irst let me comment on how it de!ends on other as!ects of co nition& The intuition which is required for knowin when it makes sense to blur distinctions, to try redescri!tions, to backtrack, to shift levels, and so forth, is somethin which !robably comes only with much e(!erience in thou ht in eneral& Thus it would be very hard to define heuristics for these crucial as!ects of the !ro ram& Sometimes one1s e(!erience with real ob.ects in the world has a subtle effect on how one describes or redescribes bo(es& 'or instance, who can say how much one1s familiarity with livin trees hel!s one to solve 0P $+ It is very doubtful that in humans, the subnetwork of conce!ts relevant to these !u<<les can be easily se!arated out from the whole network& ;ather, it is much more likely that one1s intuitions ained from seein

and handlin real ob.ectsOcombs, trains, strin s, blocks, letters, rubber bands, etc&, etc& O !lay an invisible but si nificant uidin role in the solution of these !u<<les& 7onversely, it is certain that understandin real-world situations heavily de!ends on visual ima ery and s!atial intuition, so that havin a !owerful and fle(ible way of re!resentin !atterns such as these 0on ard !atterns can only contribute to the eneral efficiency of thou ht !rocesses& It seems to me that 0on ard1s !roblems were worked out with reat care, and that they have a quality of universality to them, in the sense that each one has a unique correct answer& Ef course one could ar ue with this and say that what we consider -correct- de!ends in some dee! way on our bein human, and some creatures from some other star system mi ht disa ree entirely& 8ot havin any concrete evidence either way, I still have a certain faith that 0on ard !roblems de!end on a sense of sim!licity which is not .ust limited to earthbound human bein s& /y earlier comments about the !robable im!ortance of bein acquainted with such surely earth-limited ob.ects as combs, trains, rubber bands, and so on, are not in conflict with the idea that our notion of sim!licity is universal, for what matters is not any of these individual ob.ects, but the fact that taken to ether they s!an a wide s!ace& )nd it seems likely that any other civili<ation would have as vast a re!ertoire of artifacts and natural ob.ects and varieties of e(!erience on which to draw as we do& So I believe that the skill of solvin 0on ard !roblems lies very close to the core of -!ure- intelli ence, if there is such a thin & Therefore it is a ood !lace to be in if one wants to investi ate the ability to discover -intrinsic meanin - in !atterns or messa es& Fnfortunately we have re!roduced only a small selection of his stimulatin collection& I ho!e that many readers will acquaint themselves with the entire collection, to be found in his book 4see 0iblio ra!hy5& Some of the !roblems of visual !attern reco nition which we human bein s seem to have com!letely -flattened- into our unconscious are quite ama<in & They include, reco nition of faces 4invariance of faces under a e chan e, e(!ression chan e, li htin chan e, distance chan e, an le chan e, etc&5 reco nition of hikin trails in forests and mountains Osomehow this has always im!ressed me as one of our most subtle acts of !attern reco nition Oand yet animals can do it, too readin te(t without hesitation in hundreds if not thousands of different ty!efaces

Message-,assing Languages) +rames) and Symbols

Ene way that has been su ested for handlin the com!le(ities of !attern reco nition and other challen es to )I !ro rams is the so-called -actor- formalism of 7arl ?ewitt 4similar to the lan ua e -Smalltalk-, develo!ed by )lan @ay and others5, in which a !ro ram is written as a collection of interactin actors, which can !ass elaborate messa es back and forth amon themselves& In a way, this resembles a heterarchical collection of !rocedures which can call each other& The ma.or difference is that where !rocedures usually only !ass a rather small number of ar uments back and forth, the messa es e(chan ed by actors can be arbitrarily lon and com!le(&

)ctors with the ability to e(chan e messa es become somewhat autonomous a entsOin fact, even like autonomous com!uters, with messa es bein somewhat like !ro rams& Each actor can have its own idiosyncratic way of inter!retin any iven messa e= thus a messa e1s meanin will de!end on the actor it is interce!ted by& This comes about by the actor havin within it a !iece of !ro ram which inter!rets messa es= so there may be as many inter!reters as there are actors& Ef course, there may be many actors with identical inter!reters= in fact, this could be a reat advanta e, .ust as it is e(tremely im!ortant in the cell to have a multitude of identical ribosomes floatin throu hout the cyto!lasm, all of which will inter!ret a messa e Oin this case, messen er ;8)Oin one and the same way& It is interestin to think how one mi ht mer e the frame-notion with the actornotion& Aet us call a frame with the ca!ability of eneratin and inter!retin com!le( messa es a s)m(ol, frame ^ actor X symbol 3e now have reached the !oint where we are talkin about ways or im!lementin those elusive active symbols of 7ha!ters TI and TII= henceforth in this 7ha!ter, -symbol- will have that meanin & 0y the way, don1t feel dumb if you don1t immediately see .ust how this synthesis is to be made& It is not clear, thou h it is certainly one of the most fascinatin directions to o in )I& 'urthermore, it is quite certain that even the best synthesis of these notions will turn out to have much less !ower than the actual symbols of human minds& In that sense, callin these frame-actor syntheses -symbols- is !remature, but it is an o!timistic way of lookin at thin s& Aet us return to some issues connected with messa e !assin & Should each messa e be directed s!ecifically at a tar et symbol, or should it be thrown out into the rand void, much as m;8) is thrown out into the cyto!lasm, to seek its ribosome? If messa es have destinations, then each symbol must have an address, and messa es for it should always be sent to that address& En the other hand, there could be one central receivin dock for messa es, where a messa e would sim!ly sit until it ot !icked u! by some symbol that wanted it& This is a counter!art to 6eneral :elivery& Probably the best solution is to allow both ty!es of messa e to e(ist= also to have !rovisions for different classes of ur encyOs!ecial delivery, first class, second class, and so on& The whole !ostal system !rovides a rich source of ideas for messa e-!assin lan ua es, includin such curios as self-addressed stam!ed envelo!es 4messa es whose senders want answers quickly5, !arcel !ost 4e(tremely lon messa es which can be sent some very slow way5, and more& The tele!hone system will ive you more ins!iration when you run out of !ostal-system ideas&

En:ymes and AI
)nother rich source of ideas for messa e !assin Oindeed, for information !rocessin in eneralOis, of course, the cell& Some ob.ects in the cell are quite com!arable to actors O in !articular, en<ymes& Each en<yme1s active site acts as a filter which only reco ni<es

certain kinds of substrates 4messa es5& Thus an en<yme has an -address-, in effect& The en<yme is -!ro rammed- 4by virtue of its tertiary structure5 to carry out certain o!erations u!on that -messa e-, and then to release it to the world a ain& 8ow in this way, when a messa e is !assed from en<yme to en<yme alon a chemical !athway, a lot can be accom!lished& 3e have already described the elaborate kinds of feedback mechanisms which can take !lace in cells 4either by inhibition or re!ression5& These kinds of mechanisms show that com!licated control of !rocesses can arise throu h the kind of messa e !assin that e(ists in the cell& Ene of the most strikin thin s about en<ymes is how they sit around idly, waitin to be tri ered by an incomin substrate& Then, when the substrate arrives, suddenly the en<yme s!rin s into action, like a Genus1s flytra!& This kind of -hair-tri er- !ro ram has been used in )I, and oes by the name of demon& The im!ortant thin here is the idea of havin many different -s!ecies- of tri erable subroutines .ust lyin around waitin to be tri ered& In cells, all the com!le( molecules and or anelles are built u!, sim!le ste! by sim!le ste!& Some of these new structures are often en<ymes themselves, and they !artici!ate in the buildin of new en<ymes, which in turn !artici!ate in the buildin of yet other ty!es of en<yme, etc& Such recursive cascades of en<ymes can have drastic effects on what a cell is doin & Ene would like to see the same kind of sim!le ste!-by-ste! assembly !rocess im!orted into )I, in the construction of useful sub!ro rams& 'or instance, re!etition has a way of burnin new circuits into our mental hardware, so that oft-re!eated !ieces of behavior become encoded below the conscious level& It would be e(tremely useful if there were an analo ous way of synthesi<in efficient !ieces of code which can carry out the same sequence of o!erations as somethin which has been learned on a hi her level of -consciousness-& En<yme cascades may su est a model for how this could be done& 4The !ro ram called -?)7@E;-, written by 6erald Sussman, synthesi<es and debu s small subroutines in a way not too much unlike that of en<yme cascades&5 The sameness-detectors in the 0on ard !roblem-solver 4Sams5 could be im!lemented as en<yme-like sub!ro rams& Aike an en<yme, a Sam would meander about somewhat at random, bum!in into small data structures here and there& F!on fillin its two -active sites- with identical data structures, the Sam would emit a messa e to other !arts 4actors5 of the !ro ram& )s lon as !ro rams are serial, it would not make much sense to have several co!ies of a Sam, but in a truly !arallel com!uter, re ulatin the number of co!ies of a sub!ro ram would be a way of re ulatin the e(!ected waitin time before an o!eration ets done, .ust as re ulatin the number of co!ies of an en<yme in a cell re ulates how fast that function ets !erformed& )nd if new Sams could be synthesi<ed, that would be com!arable to the see!a e of !attern detection into lower levels of our minds&

+ission and +usion

Two interestin and com!lementary ideas concernin the interaction of symbols are -fission- and -fusion-& 7ission is the radual diver ence of a new symbol from its !arent symbol 4that is, from the symbol which served as a tem!late off of which it was co!ied5&

7usion is what ha!!ens when two 4or more5 ori inally unrelated symbols !artici!ate in a -.oint activation-, !assin messa es so ti htly back and forth that they et bound to ether and the combination can thereafter be addressed as if it were a sin le symbol& 'ission is a more or less inevitable !rocess, since once a new symbol has been -rubbed off- of an old one, it becomes autonomous, and its interactions with the outside world et reflected in its !rivate internal structure= so what started out as a !erfect co!y will soon become im!erfect, and then slowly will become less and less like the symbol off of which it was -rubbed-& 'usion is a subtler thin & 3hen do two conce!ts really become one? Is there some !recise instant when a fusion takes !lace? This notion of .oint activations o!ens u! a Pandora1s bo( of questions& 'or instance, how much do we hear -dou h- and -nut- when we say -dou hnut-? :oes a 6erman who thinks of loves 4-?andschuhe-5 hear -hand-shoes- or not? ?ow about 7hinese !eo!le, whose word -don -(i- 4-East-3est-5 means -thin -? It is a matter of some !olitical concern, too, since some !eo!le claim that words like -chairman- are heavily char ed with undertones of the male ender& The de ree to which the !arts resonate inside the whole !robably varies from !erson to !erson and accordin to circumstances& The real !roblem with this notion of -fusion- of symbols is that it is very hard to ima ine eneral al orithms which will create meanin ful new symbols from collidin symbols& It is like two strands of :8) which come to ether& ?ow do you take !arts from each and recombine them into a meanin ful and viable new strand of :8) which codes for an individual of the same s!ecies? Er a new kind of s!ecies? The chance is infinitesimal that a random combination of !ieces of :8) will code for anythin that will surviveOsomethin like the chance that a random combination of words from two books will make another book& The chance that recombinant :8) will make sense on any level but the lowest is tiny, !recisely because there are so many levels of meanin in :8)& )nd the same oes for -recombinant symbols-&

E!igenesis of the 8ra( 8anon

I think of my :ialo ue 7rab 7anon as a !rototy!e e(am!le where two ideas collided in my mind, connected in a new way, and suddenly a new kind of verbal structure came alive in my mind& Ef course I can still think about musical crab canons and verbal dialo ues se!aratelyOthey can still be activated inde!endently of each other= but the fused symbol for crab canonical dialo ues has its own characteristic modes of activation, too& To illustrate this notion of fusion or -symbolic recombination- in some detail, then, I would like to use the develo!ment of my 7rab 7anon as a case study, because, of course, it is very familiar to me, and also because it is interestin , yet ty!ical of how far a sin le idea can be !ushed& I will recount it in sta es named after those of meiosis, which is the name for cell division in which -crossin -over-, or enetic recombination, takes !lace O the source of diversity in evolution& P;EP?)SE, I be an with a rather sim!le ideaOthat a !iece of music, say a canon, could be imitated verbally& This came from the observation that, throu h a shared abstract form, a !iece of te(t and a !iece of music may be connected& The ne(t ste! involved

tryin to reali<e some of the !otential of this va ue hunch= here, I hit u!on the idea that -voices- in canons can be ma!!ed onto -characters- in dialo ues Ostill a rather obvious idea& Then I focused down onto s!ecific kinds of canons, and remembered that there was a crab canon in the 'usical :ffering& )t that time, I had .ust be un writin :ialo ues, and there were only two characters, )chilles and the Tortoise& Since the 0ach crab canon has two voices, this ma!!ed !erfectly, )chilles should be one voice, the Tortoise the other, with the one doin forwards what the other does backwards& 0ut here I was faced with a !roblem, on what level should the reversal take !lace? The letter level? The word level? The sentence level? )fter some thou ht, I concluded that the -dramatic line- level would be most a!!ro!riate& 8ow that the -skeleton- of the 0ach crab canon had been trans!lanted, at least in !lan, into a verbal form, there was .ust one !roblem& 3hen the two voices crossed in the middle, there would be a short !eriod of e(treme re!etition, an u ly blemish& 3hat to do about it? ?ere, a stran e thin ha!!ened, a kind of level-crossin ty!ical of creative acts, the word -crab- in -crab canon- flashed into my mind, undoubtedly because of some abstract shared quality with the notion of -tortoise-Oand immediately I reali<ed that at the dead center, I could block the re!etitive effect, by insertin one s!ecial line, said by a new character, a 7rab2 This is how, in the -!ro!hase- of the 7rab 7anon, the 7rab was conceived, at the crossin over of )chilles and the Tortoise& 4See 'i & "%"&5

73G5=E BHB! * schematic diagram of the .ialogue 7rab 7anon&

/ET)P?)SE, This was the skeleton of my 8ra( 8anon& I then entered the second sta ethe -meta!hase-in which I had to fill in the flesh, which was of course an arduous task& I made a lot of stabs at it, ettin used to the way in which !airs of successive lines had to make sense when read from either direction, and e(!erimentin around to see what kinds of dual meanin s would hel! me in writin such a form 4e& &, -8ot at all-5& There were two early versions both of which were interestin , but weak& I abandoned work on the book for over a year, and when I returned to the 8ra( 8anon, I had a few new ideas& Ene of them was to mention a 0ach canon inside it& )t first my !lan was to mention the -7anon !er au mentationem, contrario motu-, from the 'usical :ffering -loth 8anon, as I call it5& 0ut that started to seem a little silly, so reluctantly I decided that inside my 7rab 7anon, I could talk about 0ach1s own 8ra( 8anon instead& )ctually, this was a crucial turnin !oint, but I didn1t know it then& 8ow if one character was oin to mention a 0ach !iece, wouldn1t it be awkward for the other to say e(actly the same thin in the corres!ondin !lace? 3ell, Escher was !layin a similar role to 0ach in my thou hts and my book, so wasn1t there some way of .ust sli htly modifyin the line so that it would refer to Escher? )fter all, in the strict art

of canons, note-!erfect imitation is occasionally fore one for the sake of ele ance or beauty& )nd no sooner did that idea occur to me than the !icture .a) and Night 4'i & C#5 !o!!ed into my mind& -Ef course2- I thou ht, -It is a sort of !ictorial crab canon, with essentially two com!lementary voices carryin the same theme both leftwards and ri htwards, and harmoni<in with each other2- ?ere a ain was the notion of a sin le -conce!tual skeleton- bein instantiated in two different mediaOin this case, music and art& So I let the Tortoise talk about 0ach, and )chilles talk about Escher, in !arallel lan ua e= certainly this sli ht de!arture from strict imitation retained the s!irit of crab canons& )t this !oint, I be an reali<in that somethin marvelous was ha!!enin , namely, the :ialo ue was becomin self-referential, without my even havin intended it2 3hat1s more, it was an indirect self-reference, in that the characters did not talk directly about the :ialo ue they were in, but rather about structures which were isomor!hic to it 4on a certain !lane of abstraction5& To !ut it in the terms I have been usin , my :ialo ue now shared a -conce!tual skeleton- with 6>del1s 6, and could therefore be ma!!ed onto 6 in somewhat the way that the 7entral :o ma was, to create in this case a -7entral 7rabma!-& This was most e(citin to me, since out of nowhere had come an esthetically !leasin unity of 6>del, Escher, and 0ach& )8)P?)SE, The ne(t ste! was quite startlin & I had had 7aroline /ac6illavry1s mono ra!h on Escher1s tessellations for years, but one day, as I fli!!ed throu h it, my eye was riveted to Plate B% 4'i & CB5, for I saw it in a way I had never seen it before, here was a enuine crab canonOcrab-like in both form and content2 Escher himself had iven the !icture no title, and since he had drawn similar tessellations usin many other animal forms, it is !robable that this coincidence of form and content was .ust somethin which I had noticed& 0ut fortuitous or not, this untitled !late was a miniature version of one main idea of my book, to unite form and content& So with deli ht I christened it 8ra( 8anon, substituted it for .a) and Night, and modified )chilles1 and the Tortoise1s remarks accordin ly& Het this was not all& ?avin become infatuated with molecular biolo y, one day I was !erusin 3atson1s book in the bookstore, and in the inde( saw the word -!alindrome-& 3hen I looked it u!, I found a ma ical thin , crab-canonical structures in :8)& Soon the 7rab1s comments had been suitably modified to include a short remark to the effect that he owed his !redilection for confusin retro rade and forward motion to his enes& TEAEP?)SE, The last ste! came months later, when, as I was talkin about the !icture of the crab-canonical section of :8) 4'i & C%5, " saw that the 1)1, 1T1, 171 of )denine, Thymine, 7ytosine coincidedOmira(ile dictuOwith the 1)1, 1T1, 171 of )chilles, Tortoise, 7rab= moreover, .ust as )denine and Thymine are !aired in :8), so are )chilles and the Tortoise !aired in the :ialo ue& I thou ht for a moment and, in another of those level-crossin s, saw that 161, the letter !aired with 171 in :8), could stand for -6ene-& Ence a ain, I .um!ed back to the :ialo ue, did a little sur ery on the 7rab1s s!eech to reflect this new discovery, and now I had a ma!!in between the :8)1s structure, and the :ialo ue1s structure& In that sense, the :8) could be said to be a enoty!e codin for a !henoty!e, the Structure of the :ialo ue& This final touch

dramatically hei htened the self-reference, and ave the :ialo ue a density of meanin which I had never antici!ated&

Conce!tual S.eletons and Conce!tual Ma!!ing

That more or less summari<es the e!i enesis of the 7rab 7anon& The whole !rocess can be seen as a succession of ma!!in s of ideas onto each other, at varyin levels of abstraction& This is what I call conce!tual ma!!in , and the abstract structures which connect u! two different ideas are conce!tual skeletons& Thus, one conce!tual skeleton is that of the abstract notion of a crab canon, a structure havin two !arts which do the same thin , only movin in o!!osite directions& This is a concrete eometrical ima e which can be mani!ulated by the mind almost as a 0on ard !attern& In fact, when I think of the 8ra( 8anon today, I visuali<e it as two strands which cross in the middle, where they are .oined by a -knot- 4the 7rab1s s!eech5& This is such a vividly !ictorial ima e that it instantaneously ma!s, in my mind, onto a !icture of two homolo ous chromosomes .oined by a centromere in their middle, which is an ima e drawn directly from meiosis, as shown in 'i ure "%B&

73G5=E BHF!

In fact, this very ima e is what ins!ired me to cast the descri!tion of the 8ra( 8anon's evolution in terms of meiosisOwhich is itself, of course, yet another e(am!le of conce!tual ma!!in &

$ecombinant Ideas
There are a variety of techniques of fusion of two symbols& Ene involves linin the two ideas u! ne(t to each other 4as if ideas were linear25, then .udiciously choosin !ieces from each one, and recombinin them in a new symbol& This stron ly recalls enetic recombination& 3ell, what do chromosomes e(chan e, and how do they do it? They e(chan e enes& 3hat in a symbol is com!arable to a ene? If symbols have frame-like slots, then slots, !erha!s& 0ut which slots to e(chan e, and why? ?ere is where the crabcanonical fusion may offer some ideas& /a!!in the notion of -musical crab canononto that of -dialo ue- involved several au(iliary ma!!in s= in fact it induced them& That is, once it had been decided that these two notions were to be fused, it became a matter of lookin at them on a level where analo ous !arts emer ed into view, then oin ahead

and mapping the parts onto each other, and so on, recursively, to any level that was found desirable& ?ere, for instance, -voice- and -character- emer ed as corres!ondin slots when -crab canon- and -dialo ue- were viewed abstractly& 3here did these abstract views come from, thou h? This is at the cru( of the ma!!in -!roblem Owhere do abstract views come from? ?ow do you make abstract views of s!ecific notions?

Abstractions) S.eletons) Analogies

) view which has been abstracted from a conce!t alon some dimension is what I call a conceptual skeleton& In effect, we have dealt with conce!tual skeletons all alon , without often usin that name& 'or instance, many of the ideas concernin 0on ard !roblems could be re!hrased usin this terminolo y& It is always of interest, and !ossibly of im!ortance, when two or more ideas are discovered to share a conce!tual skeleton& )n e(am!le is the bi<arre set of conce!ts mentioned at the be innin of the 8ontrafactus, a 0icyclo!s, a tandem unicycle, a teeter-teeter, the ame of !in -!in , a one-way tie, a twosided />bius stri!, the -0ach twins-, a !iano concerto for two left hands, a one-voice fu ue, the act of cla!!in with one hand, a two-channel monaural !hono ra!h, a !air of ei hth-backs& )ll of these ideas are -isomor!hic- because they share this conce!tual skeleton, a !lural thin made sin ular and re-!lurali<ed wron ly& Two other ideas in this book which share that conce!tual skeleton are 4"5 the Tortoise1s solution to )chilles1 !u<<le, askin for a word be innin and endin in -?E- 4the Tortoise1s solution bein the !ronoun -?E-, which colla!ses two occurrences into one5, and 4B5 the Pa!!us-6elernter !roof of the Pons )sinorum Theorem, in which one trian le is re!erceived as two& Incidentally, these droll concoctions mi ht be dubbed -demidoublets-& ) conce!tual skeleton is like a set of constant features 4as distin uished from !arameters or variables5Ofeatures which should not be sli!!ed in a sub.unctive instant re!lay or ma!!in -o!eration& ?avin no !arameters or variables of its own to vary, it can be the invariant core of several different ideas& Each instance of it, such as -tandem unicycle-, does have layers of variability and so can be -sli!!ed- in various ways& )lthou h the name -conce!tual skeleton- sounds absolute and ri id, actually there is a lot of !lay in it& There can be conce!tual skeletons on several different levels of abstraction& 'or instance, the -isomor!hism- between 0on ard !roblems $+ and $", already !ointed out, involves a hi her-level conce!tual skeleton than that needed to solve either !roblem in isolation&

Multi!le $e!resentations
8ot only must conce!tual skeletons e(ist on different levels of abstraction= also, they must e(ist alon different conce!tual dimensions& Aet us take the followin sentence as an e(am!le,

-The Gice President is the s!are tire on the automobile of overnment&?ow do we understand what it means 4leavin aside its humor, which is of course a vital as!ect5? If you were told, -See our overnment as an automobile- without any !rior motivation, you mi ht come u! with any number of corres!ondences, steerin wheel X !resident, etc&& 3hat are checks and balances? 3hat are seat belts? 0ecause the two thin s bein ma!!ed are so different, it is almost inevitable that the ma!!in will involve functional as!ects& Therefore, you retrieve from your store of conce!tual skeletons re!resentin !arts of automobiles, only those havin to do with function, rather than, say, sha!e& 'urthermore, it makes sense to work at a !retty hi h level of abstraction, where -function- isn1t taken in too narrow a conte(t& Thus, of the two followin definitions of the function of a s!are tire, 4"5 -re!lacement for a flat tire-, and 4B5 -re!lacement for a certain disabled !art of a car-, certainly the latter would be !referable, in this case& This comes sim!ly from the fact that an auto and a overnment are so different that they have to be ma!!ed at a hi h level of abstraction& 8ow when the !articular sentence is e(amined, the ma!!in ets forced in one res!ectObut it is not an awkward way, by any means& In fact, you already have a conce!tual skeleton for the Gice President, amon many others, which says, -re!lacement for a certain disabled !art of overnment-& Therefore the forced ma!!in works comfortably& 0ut su!!ose, for the sake of contrast, that you had retrieved another conce!tual skeleton for -s!are tire-Osay, one describin its !hysical as!ects& )mon other thin s, it mi ht say that a s!are tire is -round and inflated-& 7learly, this is not the ri ht way to o& 4Er is it? )s a friend of mine !ointed out, some Gice Presidents are rather !ortly, and most are quite inflated25

,orts of Access
Ene of the ma.or characteristics of each idiosyncratic style of thou ht is how new e(!eriences et classified and stuffed into memory, for that defines the -handles- by which they will later be retrievable& )nd for events, ob.ects, ideas, and so on Ofor everythin that can be thou ht aboutOthere is a wide variety of -handles-& I am struck by this each time I reach down to turn on my car radio, and find, to my dismay, that it is already on2 3hat has ha!!ened is that two inde!endent re!resentations are bein used for the radio& Ene is -music !roducer-, the other is -boredom reliever-& I am aware that the music is on, but I am bored anyway, and before the two reali<ations have a chance to interact, my refle( to reach down has been tri ered& The same reachin -down refle( one day occurred .ust after I1d left the radio at a re!air sho! and was drivin away, wantin to hear some music& Edd& /any other re!resentations for the same ob.ect e(ist, such as shiny silver-knob haver overheatin -!roblems haver lyin -on-my-back-over-hum!-to-fi( thin bu<<-maker

sli!!in -dials ob.ect multidimensional re!resentation e(am!le )ll of them can act as !orts of access& Thou h they all are attached to my symbol for my car radio, accessin that symbol throu h one does not o!en u! all the others& Thus it is unlikely that I will be ins!ired to remember lyin on my back to fi( the radio when I reach down and turn it on& )nd conversely, when I1m lyin on my back, unscrewin screws, I !robably won1t think about the time I heard the *rt of the 7ugue on it& There are -!artitions- between these as!ects of one symbol, !artitions that !revent my thou hts from s!illin over slo!!ily, in the manner of free associations& /y mental !artitions are im!ortant because they contain and channel the flow of my thou hts& Ene !lace where these !artitions are quite ri id is in sealin off words for the same thin in different lan ua es& If the !artitions were not stron , a bilin ual !erson would constantly sli! back and forth between lan ua es, which would be very uncomfortable& Ef course, adults learnin two new lan ua es at once often confuse words in them& The !artitions between these lan ua es are flimsier, and can break down& Inter!reters are !articularly interestin , since they can s!eak any of their lan ua es as if their !artitions were inviolable and yet, on command, they can ne ate those !artitions to allow access to one lan ua e from the other, so they can translate& Steiner, who rew u! trilin ual, devotes several !a es in *fter Ba(el to the intermin lin of 'rench, En lish, and 6erman in the layers of his mind, and how his different lan ua es afford different !orts of access onto conce!ts&

+orced Matching
3hen two ideas are seen to share conce!tual skeletons on some level of abstraction, different thin s can ha!!en& Fsually the first sta e is that you <oom in on both ideas, and, usin the hi her-level match as a uide, you try to identify corres!ondin subideas& Sometimes the match can be e(tended recursively downwards several levels, revealin a !rofound isomor!hism& Sometimes it sto!s earlier, revealin an analo y or similarity& )nd then there are times when the hi h-level similarity is so com!ellin that, even if there is no a!!arent lower-level continuation of the ma!, you .ust o ahead and make one, this is the forced match& 'orced matches occur every day in the !olitical cartoons of news!a!ers, a !olitical fi ure is !ortrayed as an air!lane, a boat, a fish, the /ona Aisa= a overnment is a human, a bird, an oil ri = a treaty is a briefcase, a sword, a can of worms= on and on and on& 3hat is fascinatin is how easily we can !erform the su ested ma!!in , and to the e(act de!th intended& 3e don1t carry the ma!!in out too dee!ly or too shallowly& )nother e(am!le of forcin one thin into the mold of another occurred when I chose to describe the develo!ment of my 8ra( 8anon in terms of meiosis& This ha!!ened in sta es& 'irst, I noticed the common conce!tual skeleton shared by the 8ra( 8anon and the ima e of chromosomes .oined by a centromere= this !rovided the ins!iration for the forced match& Then I saw a hi h-level resemblance involvin - rowth-, -sta es-, and -recombination-& Then I sim!ly !ushed the analo y as hard as I could& TentativityOas in

the 0on ard !roblem-solverO!layed a lar e role, I went forwards and backwards before findin a match which I found a!!ealin & ) third e(am!le of conce!tual ma!!in is !rovided by the 7entral :o ma!& I initially noticed a hi h-level similarity between the discoveries of mathematical lo icians and those of molecular biolo ists, then !ursued it on lower levels until I found a stron analo y& To stren then it further, I chose a 6>del-numberin which imitated the 6enetic 7ode& This was the lone element of forced matchin in the 7entral :o ma!& 'orced matches, analo ies, and meta!hors cannot easily be se!arated out& S!ortscasters often use vivid ima ery which is hard to !i eonhole& 'or instance, in a meta!hor such as -The ;ams Ufootball teamV are s!innin their wheels-, it is hard to say .ust what ima e you are su!!osed to con.ure u!& :o you attach wheels to the team as a whole? Er to each !layer? Probably neither one& /ore likely, the ima e of wheels s!innin in mud or snow sim!ly flashes before you for a brief instant, and then in some mysterious way, .ust the relevant !arts et lifted out and transferred to the team1s !erformance& ?ow dee!ly are the football team and the car ma!!ed onto each other in the s!lit second that you do this?

$eca!
Aet me try to tie thin s to ether a little& I have !resented a number of related ideas connected with the creation, mani!ulation, and com!arison of symbols& /ost of them have to do with sli!!a e in some fashion, the idea bein that conce!ts are com!osed of some ti ht and some loose elements, comin from different levels of nested conte(ts 4frames5& The loose ones can be dislod ed and re!laced rather easily, which, de!endin on the circumstances, can create a -sub.unctive instant re!lay-, a forced match, or an analo y& ) fusion of two symbols may result from a !rocess in which !arts of each symbol are dislod ed and other !arts remain&

Creati"ity and $andomness

It is obvious that we are talkin about mechani<ation of creativity& 0ut the this not a contradiction in terms? )lmost, but not really& 7reativity is the essence of that which is not mechanical& Het every creative act is mechanicalOit has its e(!lanation no less than a case of the hiccu!s does& The mechanical substrate of creativity may be hidden from view, but it e(ists& 7onversely, there is somethin unmechanical in fle(ible !ro rams, even today& It may not constitute creativity, but when !ro rams cease to be trans!arent to their creators, then the a!!roach to creativity has be un& It is a common notion that randomness is an indis!ensable in redient of creative acts& This may be true, but it does not have any bearin on the mechani<abilityOor rather, !ro rammability2Oof creativity& The world is a iant hea! of randomness= when you mirror some of it inside your head, your head1s interior absorbs a little of that randomness& The tri erin !atterns of symbols, therefore, can lead you down the most randomseemin !aths, sim!ly because they came from your interactions with a cra<y, random world& So it can be with a com!uter !ro ram, too& ;andomness is an intrinsic

feature of thou ht, not somethin which has to be -artificially inseminated-, whether throu h dice, decayin nuclei, random number tables, or what-have-you& It is an insult to human creativity to im!ly that it relies on such arbitrary sources& 3hat we see as randomness is often sim!ly an effect of lookin at somethin symmetric throu h a -skew- filter& )n ele ant e(am!le was !rovided by Salviati1s two ways of lookin at the number mIC& )lthou h the decimal e(!ansion of mIC is not literally random, it is as random as one would need for most !ur!oses, it is -!seudorandom-& /athematics is full of !seudorandomnessO!lenty enou h to su!!ly all would-be creators for all time& Just as science is !ermeated with -conce!tual revolutions- on all levels at all times, so the thinkin of individuals is shot throu h and throu h with creative acts& They are not .ust on the hi hest !lane= they are everywhere& /ost of them are small and have been made a million times beforeObut they are close cousins to the most hi hly creative and new acts& 7om!uter !ro rams today do not yet seem to !roduce many small creations& /ost of what they do is quite -mechanical- still& That .ust testifies to the fact that they are not close to simulatin the way we thinkObut they are ettin closer& Perha!s what differentiates hi hly creative ideas from ordinary ones is some combined sense of beauty, sim!licity, and harmony& In fact, I have a favorite -metaanalo y-, in which I liken analo ies to chords& The idea is sim!le, su!erficially similar ideas are often not dee!ly related= and dee!ly related ideas are often su!erficially dis!arate& The analo y to chords is natural, !hysically close notes are harmonically distant 4e& &, E-'-65= and harmonically close notes are !hysically distant 4e& &, 6-E-05& Ideas that share a conce!tual skeleton resonate in a sort of conce!tual analo ue to harmony= these harmonious -idea-chords- are often widely se!arated, as measured on an ima inary -keyboard of conce!ts-& Ef course, it doesn1t suffice to reach wide and !lunk down any old wayOyou may hit a seventh or a ninth2 Perha!s the !resent analo y is like a ninth-chordOwide but dissonant&

,ic.ing u! ,atterns on All Le"els

0on ard !roblems were chosen as a focus in this 7ha!ter because when you study them, you reali<e that the elusive sense for !atterns which we humans inherit from our enes involves all the mechanisms of re!resentation of knowled e, includin nested conte(ts, conce!tual skeletons and conce!tual ma!!in , sli!!ability, descri!tions and metadescri!tions and their interactions, fission and fusion of symbols, multi!le re!resentations 4alon different dimensions and different levels of abstraction5, default e(!ectations, and more& These days, it is a safe bet that if some !ro ram can !ick u! !atterns in one area, it will miss !atterns in another area which, to us, are equally obvious& Hou may remember that I mentioned this back in 7ha!ter ", sayin that machines can be oblivious to re!etition, whereas !eo!le cannot& 'or instance, consider S?;:AF& If Eta Ein ty!ed the sentence -Pick u! a bi red block and !ut it down- over and over a ain, S?;:AF would cheerfully react in the same way over and over a ain, e(actly as an addin machine will !rint out -C- over and over a ain, if a human bein has the !atience to ty!e -B^B- over

and over a ain& ?umans aren1t like that= if some !attern occurs over and over a ain, they will !ick it u!& S?;:AF wasn1t built with the !otential for formin new conce!ts or reco ni<in !atterns, it had no sense of over and overview&

The +le-ibility of Language

S?;:AF1s lan ua e-handlin ca!ability is immensely fle(ibleOwithin limits& S?;:AF can fi ure out sentences of reat syntactical com!le(ity, or sentences with semantic ambi uities as lon as they can be resolved by ins!ectin the data base Obut it cannot handle -ha<y- lan ua e& 'or instance, consider the sentence -?ow many blocks o on to! of each other to make a stee!le?- 3e understand it immediately, yet it does not make sense if inter!reted literally& 8or is it that some idiomatic !hrase has been used& -To o on to! of each other- is an im!recise !hrase which nonetheless ets the desired ima e across quite well to a human& 'ew !eo!le would be misled into visuali<in a !arado(ical setu! with two blocks each of which is on to! of the other Oor blocks which are - oin somewhere or other& The ama<in thin about lan ua e is how im!recisely we use it and still mana e to et away with it& S?;:AF uses words in a -metallic- way, while !eo!le use them in a -s!on y- or -rubbery- or even -8utty-Puttyish- way& If words were nuts and bolts, !eo!le could make any bolt fit into any nut, they1d .ust squish the one into the other, as in some surrealistic !aintin where everythin oes soft& Aan ua e, in human hands, becomes almost like a fluid, des!ite, the coarse rain of its com!onents& ;ecently, )I research in natural lan ua e understandin has turned away somewhat from the understandin of sin le sentences in isolation, and more towards areas such as understandin sim!le children1s stories& ?ere is a well-known children1s .oke which illustrates the o!en-endedness of real-life situations, ) man took a ride in an air!lane& Fnfortunately, he fell out& 'ortunately, he had a !arachute on& Fnfortunately, it didn1t work& 'ortunately, there was a haystack below him& Fnfortunately, there was a !itchfork stickin out of it& 'ortunately, he missed the !itchfork& Fnfortunately, he missed the haystack& It can be e(tended indefinitely& To re!resent this silly story in a frame-based system would be e(tremely com!le(, involvin .ointly activatin frames for the conce!ts of man, air!lane, e(it, !arachute, fallin , etc&, etc&

Intelligence and Emotions

Er consider this tiny yet !oi nant story,
/ar ie was holdin ti htly to the strin of her beautiful new balloon& Suddenly, a ust of wind cau ht it& The wind carried it into a tree& The balloon hit a branch and burst& /ar ie cried and cried&C

To understand this story, one needs to read many thin s between the lines& 'or instance, /ar ie is a little irl& This is a toy balloon with a strin for a child to hold& It may not be beautiful to an adult, but in a child1s eye, it is& She is outside& The -it- that the wind cau ht was the balloon& The wind did not !ull /ar ie alon with the balloon= /ar ie let o& 0alloons can break on contact with any shar! !oint& Ence they are broken, they are one forever& Aittle children love balloons and can be bitterly disa!!ointed when they break& /ar ie saw that her balloon was broken& 7hildren cry when they are sad& -To cry and cry- is to cry very lon and hard& /ar ie cried and cried because of her sadness at her balloon1s breakin & This is !robably only a small fraction of what is lackin at the surface level& ) !ro ram must have all this knowled e in order to et at what is oin on& )nd you mi ht ob.ect that, even if it -understands- in some intellectual sense what has been said, it will never really understand, until it, too, has cried and cried& )nd when will a com!uter do that? This is the kind of humanistic !oint which Jose!h 3ei<enbaum is concerned with makin in his book 8omputer &ower and ,uman =eason, and I think it is an im!ortant issue= in fact, a very, very dee! issue& Fnfortunately, many )I workers at this time are unwillin , for various reasons, to take this sort of !oint seriously& 0ut in some ways, those )I workers are ri ht, it is a little !remature to think about com!uters cryin = we must first think about rules for com!uters to deal with lan ua e and other thin s= in time, we1ll find ourselves face to face with the dee!er issues&

AI /as +ar to Go
Sometimes it seems that there is such a com!lete absence of rule- overned behavior that human bein s .ust aren1t rule- overned& 0ut this is an illusiona little like thinkin that crystals and metals emer e from ri id underlyin laws, but that fluids or flowers don1t& 3e1ll come back to this question in the ne(t 7ha!ter&
The !rocess of lo ic itself workin internally in the brain may be more analo ous to a succession of o!erations with symbolic !ictures, a sort of abstract analo ue of the 7hinese al!habet or some /ayan descri!tion of eventse(ce!t that the elements are not merely words but more like sentences or whole stories with linka es between them formin a sort of metaor su!er-lo ic with its own rules&L

It is hard for most s!ecialists to e(!ress vividly!erha!s even to remember what ori inally s!arked them to enter their field& 7onversely, someone on the outside may understand a field1s s!ecial romance and may be able to articulate it !recisely& I think

that is why this quote from Flam has a!!eal for me, because it !oetically conveys the stran eness of the enter!rise of )I, and yet shows faith in it& )nd one must run on faith at this !oint, for there is so far to o2

Ten 8uestions and S!eculations

To conclude this 7ha!ter, I would like to !resent ten -Muestions and S!eculations- about )I& I would not make so bold as to call them -)nswers-these are my !ersonal o!inions& They may well chan e in some ways, as I learn more and as )I develo!s more& 4In what follows, the term -)I !ro ram- means a !ro ram which is far ahead of today1s !ro rams= it means an -)ctually Intelli ent- !ro ram& )lso, the words -!ro ram- and -com!uter- !robably carry overly mechanistic connotations, but let us stick with them anyway&5 Muestion, 3ill a com!uter !ro ram ever write beautiful music? S!eculation, Hes, but not soon& /usic is a lan ua e of emotions, and until !ro rams have emotions as com!le( as ours, there is no way a !ro ram will write anythin beautiful& There can be -for eries- shallow imitations of the synta( of earlier music but des!ite what one mi ht think at first, there is much more to musical e(!ression than can be ca!tured in syntactical rules& There will be no new kinds of beauty turned u! for a lon time by com!uter music-com!osin !ro rams& Aet me carry this thou ht a little further& To thinkand I have heard this su estedthat we mi ht soon be able to command a !re!ro rammed mass-!roduced mail-order twenty-dollar desk-model -music bo(- to brin forth from its sterile circuitry !ieces which 7ho!in or 0ach mi ht have written had they lived lon er is a rotesque and shameful misestimation of the de!th of the human s!irit& ) -!ro ram- which could !roduce music as they did would have to wander around the world on its own, fi htin its way throu h the ma<e of life and feelin every moment of it& It would have to understand the .oy and loneliness of a chilly ni ht wind, the lon in for a cherished hand, the inaccessibility of a distant town, the heartbreak and re eneration after a human death& It would have to have known resi nation and worldweariness, rief and des!air, determination and victory, !iety and awe& In it would have had to commin le such o!!osites as ho!e and fear, an uish and .ubilation, serenity and sus!ense& Part and !arcel of it would have to be a sense of race, humor, rhythm, a sense of the une(!ectedand of course an e(quisite awareness of the ma ic of fresh creation& Therein, and therein only, lie the sources of meanin in music& Muestion, 3ill emotions be e(!licitly !ro rammed into a machine? S!eculation, 8o& That is ridiculous& )ny direct simulation of emotionsP);;H, for e(am!lecannot a!!roach the com!le(ity of human emotions, which arise indirectly from the or ani<ation of our minds& Pro rams or machines will acquire emotions in the same way, as by-!roducts of their structure, of the way in which they are or ani<ednot by direct !ro rammin & Thus, for e(am!le, nobody will write a -fallin -in-love- subroutine, any more than they would write a -mistake-

makin - subroutine& -'allin in love- is a descri!tion which we attach to a com!le( !rocess of a com!le( system= there need be no sin le module inside the system which is solely res!onsible for it, however2 Muestion, 3ill a thinkin com!uter be able to add fast? S!eculation, Perha!s not& 3e ourselves are com!osed of hardware which does fancy calculations but that doesn1t mean that our symbol level, where -we- are, knows how to carry out the same fancy calculations& Aet me !ut it this way, there1s no way that you can load numbers into your own neurons to add u! your rocery bill& Auckily for you, your symbol level 4i&e&, )ou5 can1t ain access to the neurons which are doin your thinkin otherwise you1d et addle-brained& To !ara!hrase :escartes a ain, -I think= therefore I have no access to the level where I sum&3hy should it not be the same for an intelli ent !ro ram? It mustn1t be allowed to ain access to the circuits which are doin its thinkin otherwise it1ll et addle7PF1d& Muite seriously, a machine that can !ass the Turin test may well add as slowly as you or I do, and for similar reasons& It will re!resent the number B not .ust by the two bits -"+-, but as a full-fled ed conce!t the way we do, re!lete with associations such as its homonyms -too- and -to-, the words -cou!le- and -deuce-, a host of mental ima es such as dots on dominoes, the sha!e of the numeral 1B1, the notions of alternation, evenness, oddness, and on and on&&& 3ith all this -e(tra ba a e- to carry around, an intelli ent !ro ram will become quite slothful in its addin & Ef course, we could ive it a -!ocket calculator-, so to s!eak 4or build one in5& Then it could answer very fast, but its !erformance would be .ust like that of a !erson with a !ocket calculator& There would be two se!arate !arts to the machine, a reliable but mindless !art and an intelli ent but fallible !art& Hou couldn1t rely on the com!osite system to be reliable, any more than a com!osite of !erson and machine is necessarily reliable& So if it1s ri ht answers you1re after, better stick to the !ocket calculator alonedon1t throw in the intelli ence2 Muestion, 3ill there be chess !ro rams that can beat anyone? S!eculation, 8o& There may be !ro rams which can beat anyone at chess, but they will not be e(clusively chess !layers& They will be !ro rams of eneral intelli ence, and they will be .ust as tem!eramental as !eo!le& -:o you want to !lay chess?- -8o, I1m bored with chess& Aet1s talk about !oetry&- That may be the kind of dialo ue you could have with a !ro ram that could beat everyone& That is because real intelli ence inevitably de!ends on a total overview ca!acitythat is, a !ro rammed ability to -.um! out of the system-, so to s!eakat least rou hly to the e(tent that we have that ability& Ence that is !resent, you can1t contain the !ro ram= it1s one beyond that certain critical !oint, and you .ust have to face the facts of what you1ve wrou ht&

Muestion, 3ill there be s!ecial locations in memory which store !arameters overnin the behavior of the !ro ram, such that if you reached in and chan ed them, you would be able to make the !ro ram smarter or stu!ider or more creative or more interested in baseball? In short, would you be able to -tune- the !ro ram by fiddlin with it on a relatively low level? S!eculation, 8o& It would be quite oblivious to chan es of any !articular elements in memory, .ust as we stay almost e(actly the same thou h thousands of our neurons die every day425& If you fuss around too heavily, thou h, you1ll dama e it, .ust as if you irres!onsibly did neurosur ery on a human bein & There will be no -ma iclocation in memory where, for instance, the -IM- of the !ro ram sits& ) ain, that will be a feature which emer es as a consequence of lower-level behavior, and nowhere will it sit e(!licitly& The same oes for such thin s as -the number of items it can hold in short-term memory-, -the amount it likes !hysics-, etc&, etc& Muestion, 7ould you -tune- an )I !ro ram to act like me, or like youor halfway between us? S!eculation, 8o& )n intelli ent !ro ram will not be chameleon-like, any more than !eo!le are& It will rely on the constancy of its memories, and will not be able to flit between !ersonalities& The idea of chan in internal !arameters to -tune to a new !ersonality- reveals a ridiculous underestimation of the com!le(ity of !ersonality& Muestion, 3ill there be a -heart- to an )I !ro ram, or will it sim!ly consist of -senseless loo!s and sequences of trivial o!erations- 4in the words of /arvin /inskyD5? S!eculation, If we could see all the way to the bottom, as we can a shallow !ond, we would surely see only -senseless loo!s and sequences of trivial o!erations-and we would surely not see any -heart-& 8ow there are two kinds of e(tremist views on )I, one says that the human mind is, for fundamental and mysterious reasons, un!ro rammable& The other says that you merely need to assemble the a!!ro!riate -heuristic devices-multi!le o!timi<ers, !attern-reco nition tricks, !lannin al ebras, recursive administration !rocedures, and the like-,$ and you will have intelli ence& I find myself somewhere in between, believin that the -!ond- of an )I !ro ram will turn out to be so dee! and murky that we won1t be able to !eer all the way to the bottom& If we look from the to!, the loo!s will be invisible, .ust as nowadays the current-carryin electrons are invisible to most !ro rammers& 3hen we create a !ro ram that !asses the Turin test, we will see a -heart- even thou h we know it1s not there& Muestion, 3ill )I !ro rams ever become -su!erintelli ent-? S!eculation, I don1t know& It is not clear that we would be able to understand or relate to a -su!erintelli ence-, or that the conce!t even makes sense& 'or instance, our own intelli ence is tied in with our s!eed of thou ht& If our refle(es had been ten times faster or slower, we mi ht have develo!ed an entirely different set of conce!ts with which to describe the world& ) creature with a radically different

view of the world may sim!ly not have many !oints of contact with us& I have often wondered if there could be, for instance, !ieces of music which are to 0ach as 0ach is to folk tunes, -0ach squared-, so to s!eak& )nd would I be able to understand them? /aybe there is such music around me already, and I .ust don1t reco ni<e it, .ust as do s don1t understand lan ua e& The idea of su!erintelli ence is very stran e& In any case, I don1t think of it as the aim of )I research, althou h if we ever do reach the level of human intelli ence, su!erintelli ence will undoubtedly be the ne(t oalnot only for us, but for our )I-!ro ram collea ues, too, who will be equally curious about )I and su!erintelli ence& It seems quite likely that )I !ro rams will be e(tremely curious about )I in eneralunderstandably& Muestion, Hou seem to be sayin that )I !ro rams will be virtually identical to !eo!le, then& 3on1t there be any differences? S!eculation, Probably the differences between )I !ro rams and !eo!le will be lar er than the differences between most !eo!le& It is almost im!ossible to ima ine that the -body- in which an )I !ro ram is housed would not affect it dee!ly& So unless it had an ama<in ly faithful re!lica of a human bodyand why should it?it would !robably have enormously different !ers!ectives on what is im!ortant, what is interestin , etc& 3itt enstein once made the amusin comment, -If a lion could s!eak, we would not understand him&- It makes me think of ;ousseau1s !aintin of the entle lion and the slee!in y!sy on the moonlit desert& 0ut how does 3itt enstein know? /y uess is that any )I !ro ram would, if com!rehensible to us, seem !retty alien& 'or that reason, we will have a very hard time decidin when and if we really are dealin with an )I !ro ram, or .ust a -weird- !ro ram& Muestion, 3ill we understand what intelli ence and consciousness and free will and -Iare when we have made an intelli ent !ro ram? S!eculation, Sort ofit all de!ends on what you mean by -understand-& En a ut level, each of us !robably has about as ood an understandin as is !ossible of those thin s, to start with& It is like listenin to music& :o you really understand 0ach because you have taken him a!art? Er did you understand it that time you felt the e(hilaration in every nerve in your body? :o we understand how the s!eed of li ht is constant in every inertial reference frame? 3e can do the math, but no one in the world has a truly relativistic intuition& )nd !robably no one will ever understand the mysteries of intelli ence and consciousness in an intuitive way& Each of us can understand people, and that is !robably about as close as you can come&

Sloth Canon
This time, we find *chilles and the Tortoise #isiting the dwelling of their new friend, the -loth! *chilles, Shall I tell you of my droll footrace with /r& T? -loth, Please do& *chilles, It has become quite celebrated in these !arts& I believe it1s even been written u!, by 9eno& -loth, It sounds very e(citin & *chilles, It was& Hou see, /r& T be an way ahead of me& ?e had such a hu e head start, and yet -loth, Hou cau ht u!, didn1t you? *chilles, Hesbein so fleet of foot, I diminished the distance between us at a constant rate, and soon overtook him& -loth, The a! ke!t ettin shorter and shorter, so you could& *chilles, E(actly& Eh, look/r& T has brou ht his violin& /ay I try !layin on it, /r& T? Tortoise, Please don1t& It sounds very flat& *chilles, Eh, all ri ht& 0ut I1m in a mood for music& I don1t know why& -loth, Hou can !lay the !iano, )chilles& *chilles, Thank you& I1ll try it In a moment& I .ust wanted to add that I also had another kind of -race- with /r& T at a later date& Fnfortunately, in that race Tortoise, Hou didn1t catch u!, did you? The a! ke!t ettin lon er and lon er, so you couldn1t& *chilles, That1s true& I believe T?)T race has been written u!, too, by Aewis 7arroll& 8ow, /r& Sloth, I1ll take u! your offer of tryin out the !iano& 0ut I1m so bad at the !iano& I1m not sure I dare& -loth, Hou should try& *chilles sits down and starts pla)ing a simple tune!/ *chilles, Ehit sounds very stran e& That1s not how it1s su!!osed to sound at all2 Somethin is very wron & Tortoise, Hou can1t !lay the !iano, )chilles& Hou shouldn1t try& *chilles, It1s like a !iano in a mirror& The hi h notes are on the left, and the low notes are on the ri ht& Every melody comes out inverted, as if u!side down& 3ho would have ever thou ht u! somethin so cockeyed as that? Tortoise, That1s so characteristic of sloths& They han from *chilles, Hes, I knowfrom tree branchesu!side down, of course& That sloth-!iano would he a!!ro!riate for !layin inverted melodies such as occur in some canons and fu ues& 0ut to learn to !lay a !iano while han in from a tree must he very difficult& Hou must have to devote a reat deal of ener y to it&

73G5=E BHH! "-loth 8anon",from the /usical Efferin , () ?! -! Bach! U'usic printed () .onald B)rd's program "-'5T" &V

-loth, That1s not so characteristic of sloths& *chilles, 8o, I ather sloths like to take life very easy& They do everythin about half as fast as normal& )nd u!side down, to boot& 3hat a !eculiar way to o throu h life2 S!eakin of thin s that are both u!side- and slowed-down, there1s a -7anon !er au mentationem, contrario motu- in the 'usical :ffering& In my edition, the letters 1S1, 1)1, 1T1 are in front of the three staves& I don1t know why& )nyway, I think 0ach carried it off very skillfully& 3hat1s your o!inion, /r& T? Tortoise, ?e outdid himself& )s for those letters -S)T-, you could uess what they stand for& *chilles, -So!rano-, -)lto-, and -Tenor-, I su!!ose& Three-!art !ieces are often written for that combination of voices& 3ouldn1t you a ree, /r& Sloth? -loth, They stand for *chilles, Eh, .ust a moment, /r& Sloth& /r& Tortoisewhy are you !uttin on your coat? Hou1re not leavin , are you? 3e were .ust oin to fi( a snack to eat& Hou look very tired& ?ow do you feel? Tortoise, Eut of as& So lon 2 4Trudges wearil) out the door&5 *chilles, The !oor fellowhe certainly looked e(hausted& ?e was .o in all mornin & ?e1s in trainin for another race with me& -loth, ?e did himself in& *chilles, Hes, but in vain& /aybe he could beat a Sloth&&& but me? 8ever2 8owweren1t you about to tell me what those letters -S)T- stand for? -loth, )s for those letters -S)T-, you could never uess what they stand for& *chilles, 3ell, if they don1t stand for what I thou ht, then my curiosity is !iqued& Perha!s I1ll think a little more about it& Say, how do you cook 'rench fries? -loth, In oil& *chilles, Eh, yesI remember& I1ll cut u! this !otato into stri!s an inch or two in len th& -loth, So short? *chilles, )ll ri ht, already, I1ll cut four-inch stri!s& Eh, boy, are these oin to be ood 'rench fries2 Too bad /r& T won1t be here to share them&

C/A,TE$ <<

Strange Loo!s) 0r Tangled /ierarchies

Can Machines ,ossess 0riginality7
I8 T?E 7?)PTE; before last, I described )rthur Samuel1s very successful checkers !ro ram the one which can beat its desi ner& In li ht of that, it is interestin to hear how Samuel himself feels about the issue of com!uters and ori inality& The followin e(tracts are taken from a rebuttal by Samuel, written in "#D+, to an article by 8orbert 3iener&
It is my conviction that machines cannot !ossess ori inality in the sense im!lied by 3iener in his thesis that -machines can and do transcend some of the limitations of their desi ners, and that in doin so they may be both effective and dan erous&-&&& ) machine is not a enie, it does not work by ma ic, it does not !ossess a will, and, 3iener to the contrary, nothin comes out which has not been !ut in, barrin , of course, an infrequent case of malfunctionin &&& The -intentions- which the machine seems to manifest are the intentions of the human !ro rammer, as s!ecified in advance, or they are subsidiary intentions derived from these, followin rules s!ecified by the !ro rammer& 3e can even antici!ate hi her levels of abstraction, .ust as 3iener does, in which the !ro ram will not only modify the subsidiary intentions but will also modify the rules which are used in their derivation, or in which it will modify the ways in which it modifies the rules, and so on, or even in which one machine will desi n and construct a second machine with enhanced ca!abilities& ?owever, and this is im!ortant, the machine will not and cannot Uitalics are hisV do any of these thin s until it has been instructed as to how to !roceed& There is and lo ically there must always remain a com!lete hiatus between 4i5 any ultimate e(tension and elaboration in this !rocess of carryin out man1s wishes and 4ii5 the develo!ment within the machine of a will of its own& To believe otherwise is either to believe in ma ic or to believe that the e(istence of man1s will is an illusion and that man1s actions are as mechanical as the machine1s& Perha!s 3iener1s article and my rebuttal have both been mechanically determined, but this I refuse to believe& "

This reminds me of the Aewis 7arroll :ialo ue 4the Two%&art 3n#ention5= I1ll try to e(!lain why& Samuel bases his ar ument a ainst machine consciousness 4or will5 on the notion that any mechanical instantiation of will would re"uire an infinite regress & Similarly, 7arroll1s Tortoise ar ues that no ste! of reasonin , no matter how sim!le, can be done without invokin some rule on a hi her level to .ustify the ste! in question& 0ut that bein also a ste! of reasonin & one must resort to a yet hi her-level rule, and so on& 7onclusion, =easoning in#ol#es an infinite regress& Ef course somethin is wron with the Tortoise1s ar ument, and I believe somethin analo ous is wron with Samuel1s ar ument& To show how the fallacies are analo ous, I now shall -hel! the :evil-, by ar uin momentarily as :evil1s advocate& 4Since, as is well known, 6od hel!s those who hel! themselves, !resumably the :evil

hel!s all those, and only those, who don1t hel! themselves& :oes the :evil hel! himself?5 ?ere are my devilish conclusions drawn from the 7arroll :ialo ue, The conclusion -reasonin is im!ossible- does not a!!ly to !eo!le, because as is !lain to anyone, we do mana e to carry out many ste!s of reasonin , all the hi her levels notwithstandin & That shows that we humans o!erate without need of rules, we are -informal systems-& En the other hand, as an ar ument a ainst the !ossibility of any mechanical instantiation of reasonin , it is valid, for any mechanical reasonin -system would have to de!end on rules e(!licitly, and so it couldn1t et off the round unless it had metarules tellin it when to a!!ly its rules, metametarules tellin it when to a!!ly its metarules, and so on& 3e may conclude that the ability to reason can never be mechani<ed& It is a uniquely human ca!ability& 3hat is wron with this :evil1s advocate !oint of view? It is obviously the assum!tion that a machine cannot do an)thing without ha#ing a rule telling it to do so & In fact, machines et around the Tortoise1s silly ob.ections as easily as !eo!le do, and moreover for e(actly the same reason, both machines and !eo!le are made of hardware which runs all by itself, accordin to the laws of !hysics& There is no need to rely on -rules that !ermit you to a!!ly the rules-, because the lowest-level rulesthose without any -meta1s in frontare embedded in the hardware, and they run without !ermission& /oral, The 7arroll :ialo ue doesn1t say anythin about the differences between !eo!le and machines, after all& 4)nd indeed, reasonin is mechani<able&5 So much for the 7arroll :ialo ue& En to Samuel1s ar ument& Samuel1s !oint, if I may caricature it, is this, 8o com!uter ever -wants- to do anythin , because it was !ro rammed by someone else& Enly if it could !ro ram itself from <ero on u!an absurditywould it have its own sense of desire& In his ar ument, Samuel reconstructs the Tortoise1s !osition, re!lacin -to reason- by -to want-& ?e im!lies that behind any mechani<ation of desire, there has to be either an infinite re ress or worse, a closed loo!& If this is why com!uters have no will of their own, what about !eo!le? The same criterion would im!ly that Fnless a !erson desi ned himself and chose his own wants 4as well as choosin to choose his own wants, etc&5, he cannot be said to have a will of his own& It makes you !ause to think where your sense of havin a will comes from& Fnless you are a soulist, you1ll !robably say that it comes from your braina !iece of hardware which you did not desi n or choose& )nd yet that doesn1t diminish your sense that you want certain thin s, and not others& Hou aren1t a -self-!ro rammed ob.ect- 4whatever that would be5, but you still do have a sense of desires, and it s!rin s from the !hysical substrate of your mentality& Aikewise, machines may someday have wills des!ite the fact

that no ma ic !ro ram s!ontaneously a!!ears in memory from out of nowhere 4a -self!ro rammed !ro ram-5& They will have wills for much the same reason as you doby reason of or ani<ation and structure on many levels of hardware and software& /oral, The Samuel ar ument doesn1t say anythin about the differences between !eo!le and machines, after all& 4)nd indeed, will will be mechani<ed&5

Belo' E"ery Tangled /ierarchy Lies An In"iolate Le"el

;i ht after the Two%&art 3n#ention, I wrote that a central issue of this book would be, -:o words and thou hts follow formal rules?- Ene ma.or thrust of the book has been to !oint out the many-leveledness of the mindIbrain, and I have tried to show why the ultimate answer to the question is, -Hes!rovided that you o down to the lowest levelthe hardwareto find the rules&8ow Samuel1s statement brou ht u! a conce!t which I want to !ursue& It is this, 3hen we humans think, we certainly do chan e our own mental rules, and we chan e the rules that chan e the rules, and on and onbut these are, so to s!eak, -software rules-& ?owever, the rules at the (ottom do not chan e& 8eurons run in the same sim!le way the whole time& Hou can1t -think- your neurons into runnin some nonneural way, althou h you can make your mind chan e style or sub.ect of thou ht& Aike )chilles in the &relude, *nt 7ugue, you have access to your thou hts but not to your neurons& Software rules on various levels can chan e= hardware rules cannotin fact, to their ri idity is due the software1s fle(ibility2 8ot a !arado( at all, but a fundamental, sim!le fact about the mechanisms of intelli ence& This distinction between self-modifiable software and inviolate hardware is what I wish to !ursue in this final 7ha!ter, develo!in it into a set of variations on a theme& Some of the variations may seem to be quite far-fetched, but I ho!e that by the time I close the loo! by returnin to brains, minds, and the sensation of consciousness, you will have found an invariant core in all the variations& /y main aim in this 7ha!ter is to communicate some of the ima es which hel! me to visuali<e how consciousness rises out of the .un le of neurons= to communicate a set of intan ible intuitions, in the ho!e that these intuitions are valuable and may !erha!s hel! others a little to come to clearer formulations of their own ima es of what makes minds run& I could not ho!e for more than that my own mind1s blurry ima es of minds and ima es should cataly<e the formation of shar!er ima es of minds and ima es in other minds&

A Self-Modifying Game
) first variation, then, concerns ames in which on your turn, you may modify the rules& Think of chess& 7learly the rules stay the same, .ust the board !osition chan es on each move& 0ut let1s invent a variation in which, on your turn, you can either make a move or chan e the rules& 0ut how? )t liberty? 7an you turn it into checkers? 7learly such anarchy would be !ointless& There must be some constraints& 'or instance, one version mi ht allow you to redefine the kni ht1s move& Instead of bein "-and-then-B, it could be

m-and-then-n where m and n are arbitrary natural numbers= and on your turn you could chan e either m or n by !lus or minus "& So it could o from "-B to "-% to +-% to +-C to +L to "-L to B-L&&& Then there could be rules about redefinin the bisho!1s moves, and the other !ieces1 moves as well& There could be rules about addin new squares, or deletin old squares&&& 8ow we have two layers of rules, those which tell how to move !ieces, and those which tell how to chan e the rules& So we have rules and metarules& The ne(t ste! is obvious, introduce metametarules by which we can chan e the metarules& It is not so obvious how to do this& The reason it is easy to formulate rules for movin !ieces is that !ieces move in a formali<ed s!ace, the checkerboard& If you can devise a sim!le formal notation for e(!ressin rules and metarules, then to mani!ulate them will be like mani!ulatin strin s formally, or even like mani!ulatin chess !ieces& To carry thin s to their lo ical e(treme, you could even e(!ress rules and metarules as !ositions on au(iliary chess boards& Then an arbitrary chess !osition could be read as a ame, or as a set of rules, or as a set of metarules, etc&, de!endin on which inter!retation you !lace on it& Ef course, both !layers would have to a ree on conventions for inter!retin the notation& 8ow we can have any number of ad.acent chess boards, one for the ame, one for rules, one for metarules, one for metametarules, and so on, as far as you care to carry it& En your turn, you may make a move on any one of the chess boards e(ce!t the to!-level one, usin the rules which a!!ly 4they come from the ne(t chess board u! in the hierarchy5& Fndoubtedly both !layers would et quite disoriented by the fact that almost anythin thou h not everythin 2can chan e& 0y definition, the to!-level chess board can1t be chan ed, because you don1t have rules tellin how to chan e it& It is in#iolate& There is more that is inviolate, the conventions by which the different boards are inter!reted, the a reement to take turns, the a reement that each !erson may chan e one chess board each turnand you will find more if you e(amine the idea carefully& 8ow it is !ossible to o considerably further in removin the !illars by which orientation is achieved& Ene ste! at a time&&& 3e be in by colla!sin the whole array of boards into a sin le board& 3hat is meant by this? There will be two ways of inter!retin the board, 4"5 as !ieces to be moved= 4B5 as rules for movin the !ieces& En your turn, you move !iecesand !erforce, you chan e rules2 Thus, the rules constantly chan e themselves& Shades of Ty!o eneticsor for that matter, of real enetics& The distinction between ame, rules, metarules, metametarules, has been lost& 3hat was once a nice clean hierarchical setu! has become a Stran e Aoo!, or Tan led ?ierarchy& The moves chan e the rules, the rules determine the moves, round and round the mulberry bush&&& There are still different levels, but the distinction between -lower- and -hi her- has been wi!ed out& 8ow, !art of what was inviolate has been made chan eable& 0ut there is still !lenty that is inviolate& Just as before, there are conventions between you and your o!!onent by which you inter!ret the board as a collection of rules& There is the a reement to take turnsand !robably other im!licit conventions, as well& 8otice, therefore, that the notion of different levels has survived, in an une(!ected way& There is an Inviolate level let1s call it the 3%le#elOon which the inter!retation conventions reside= there is also a

Tan led levelthe T%le#elOon which the Tan led ?ierarchy resides& So these two levels are still hierarchical, the I-level overns what ha!!ens on the T-level, but the T-level does not and cannot affect the I-level& 8o matter that the T-level itself is a Tan led ?ierarchy it is still overned by a set of conventions outside of itself& )nd that is the im!ortant !oint& )s you have no doubt ima ined, there is nothin to sto! us from doin the -im!ossible-namely, tan lin the I-level and the T-level by makin the inter!retation conventions themselves sub.ect to revision, accordin to the !osition on the chess board& 0ut in order to carry out such a -su!ertan lin -, you1d have to a ree with your o!!onent on some further conventions connectin the two levelsand the act of doin so would create a new level, a new sort of inviolate level on to! of the -su!ertan led- level 4or underneath it, if you !refer5& )nd this could continue oin on and on& In fact, the 1.um!swhich are bein made are very similar to those charted in the Birthda) 8antatatata, and in the re!eated 6>deli<ation a!!lied to various im!rovements on T8T& Each time you think you have reached the end, there is some new variation on the theme of .um!in out of the system which requires a kind of creativity to s!ot&

The Authorshi! Triangle Again

0ut I am not interested in !ursuin the stran e to!ic of the ever more abstruse tan lin s which can arise in self-modifyin chess& The !oint of this has been to show, in a somewhat ra!hic way, how in any system there is always some -!rotected- level which is unassailable by the rules on other levels, no matter how tan led their interaction may be amon themselves& )n amusin riddle from 7ha!ter IG illustrates this same idea in a sli htly different conte(t& Perha!s it will catch you off uard,

73G5=E BHD! *n "authorship triangle"

There are three authors9, T, and E& 8ow it ha!!ens that 9 e(ists only in a novel by T& Aikewise, T e(ists only in a novel by E& )nd stran ely, E, too, e(ists only in a novelby 9, of course& 8ow, is such an -authorshi! trian le- reall) !ossible? 4See 'i & "%C&5

Ef course it1s !ossible& 0ut there1s a trick&&& )ll three authors 9, T, E, are themselves characters in another novelby ?2 Hou can think of the 9-T-E trian le as a Stran e Aoo!, Er Tan led ?ierarchy= but author ? is outside of the s!ace in which that tan le takes !laceauthor ? is in an inviolate s!ace& )lthou h 9, T, and E all have access direct or indirectto each other, and can do dastardly thin s to each other in their various novels, none of them can touch ?1s life2 They can1t even ima ine himno more than you can ima ine the author of the book )ou're a character in& If I were to draw author ?, I would re!resent him somewhere off the !a e& Ef course that would !resent a !roblem, since drawin a thin necessarily !uts it onto the !a e&&& )nyway, ? is really outside of the world of 9, T, and E, and should be re!resented as bein so&

73G5=E BHG! :rawin ?ands, () '! 8! Escher lithograph, BIDE5&

Escher%s .rawing ,ands

)nother classic variation on our theme is the Escher !icture of .rawing ,ands 4'i & "%L5& ?ere, a left hand 4A?5 draws a ri ht hand 4;?5, while at the same time, ;? draws A?& Ence a ain, levels which ordinarily are seen as hierarchicalthat which draws, and that which is drawnturn back on each other, creatin a Tan led ?ierarchy& 0ut the theme of the 7ha!ter is borne out, of course, since behind it all lurks the undrawn but

drawin hand of /& 7& Escher, creator of both A? and ;?& Escher is outside of the twohand s!ace, and in my schematic version of his !icture 4'i & "%D5, you can see that e(!licitly& In this schemati<ed re!resentation of the Escher !icture, you see the Stran e Aoo!, Er Tan led ?ierarchy at the to!= also, you see the Inviolate Aevel below it, enablin it to come into bein & Ene could further Escheri<e the Escher !icture, by takin a !hoto ra!h of a hand drawin it& )nd so on&

73G5=E BHJ! *(stract diagram of '! 8! Escher's :rawin ?ands! :n top, a seeming parado9! Below, its resolution&

Brain and Mind5 A ;eural Tangle Su!!orting a Symbol Tangle

8ow we can relate this to the brain, as well as to )I !ro rams& In our thou hts, symbols activate other symbols, and all interact heterarchically& 'urthermore, the symbols may cause each other to chan e internally, in the fashion of !ro rams actin on other !ro rams& The illusion is created, because of the Tan led ?ierarchy of symbols, that there is no inviolate level& one thinks there is no such level because that level is shielded from our view& If it were !ossible to schemati<e this whole ima e, there would be a i antic forest of symbols linked to each other by tan ly lines like vines in a tro!ical .un lethis would be the to! level, the Tan led ?ierarchy where thou hts really flow back and forth& This is the elusive level of mind, the analo ue to A? and ;?& 'ar below in the schematic !icture, analo ous to the invisible -!rime mover- Escher, there would be a re!resentation of the myriad neuronsthe -inviolate substrate- which lets the tan le above it come into bein & Interestin ly, this other level is itself a tan le in a literal sensebillions of cells and hundreds of billions of a(ons, .oinin them all to ether&

This is an interestin case where a software tan le, that of the symbols, is su!!orted by a hardware tan le, that of the neurons& 0ut only the symbol tan le is a Tan led ?ierarchy& The neural tan le is .ust a -sim!le- tan le& This distinction is !retty much the same as that between Stran e Aoo!s and feedback, which I mentioned in 7ha!ter TGI& ) Tan led ?ierarchy occurs when what you !resume are clean hierarchical levels take you by sur!rise and fold back in a hierarchy-violatin way& The sur!rise element is im!ortant= it is the reason I call Stran e Aoo!s -stran e-& ) sim!le tan le, like feedback, doesn1t involve violations of !resumed level distinctions& )n e(am!le is when you1re in the shower and you wash your left arm with your ri ht, and then vice versa& There is no stran eness to the ima e& Escher didn1t choose to draw hands drawin hands for nothin 2 Events such as two arms washin each other ha!!en all the time in the world, and we don1t notice them !articularly& I say somethin to you, then you say somethin back to me& Parado(? 8o= our !erce!tions of each other didn1t involve a hierarchy to be in with, so there is no sense of stran eness& En the other hand, where lan ua e does create stran e loo!s is when it talks about itself, whether directly or indirectly& ?ere, somethin in the system .um!s out and acts on the system, as if it were outside the system& 3hat bothers us is !erha!s an illdefined sense of to!olo ical wron ness, the inside-outside distinction is bein blurred, as in the famous sha!e called a -@lein bottle-& Even thou h the system is an abstraction, our minds use s!atial ima ery with a sort of mental to!olo y& 6ettin back to the symbol tan le, if we look only at it, and for et the neural tan le, then we seem to see a self-!ro rammed ob.ectin .ust the same way as we seem to see a self-drawn !icture if we look at .rawing ,ands and somehow fall for the illusion, by for ettin the e(istence of Escher& 'or the !icture, this is unlikelybut for humans and the way they look at their minds, this is usually what ha!!ens& 3e feel self!ro rammed& Indeed, we couldn1t feel any other way, for we are shielded from the lower levels, the neural tan le& Eur thou hts seem to run about in their own s!ace, creatin new thou hts and modifyin old ones, and we never notice any neurons hel!in us out2 0ut that is to be e(!ected& 3e can1t& )n analo ous double-entendre can ha!!en with AISP !ro rams that are desi ned to reach in and chan e their own structure& If you look at them on the AISP level, you will say that they chan e themselves= but if you shift levels, and think of AISP !ro rams as data to the AISP inter!reter 4see 7ha!ter T5, then in fact the sole !ro ram that is runnin is the inter!reter, and the chan es bein made are merely chan es in !ieces of data& The AISP inter!reter itself is shielded from chan es& ?ow you describe a tan led situation of this sort de!ends how far back you ste! before describin & If you ste! far enou h back, you can often see the clue that allows you to untan le thin s&

Strange Loo!s in Go"ernment

) fascinatin area where hierarchies tan le is overnment!articularly in the courts& Erdinarily, you think of two dis!utants ar uin their cases in court, and the court

ad.udicatin the matter& The court is on a different level from the dis!utants& 0ut stran e thin s can start to ha!!en when the courts themselves et entan led in le al cases& Fsually there is a hi her court which is outside the dis!ute& Even if two lower courts et involved in some sort of stran e fi ht, with each one claimin .urisdiction over the other, some hi her court is outside, and in some sense it is analo ous to the inviolate inter!retation conventions which we discussed in the war!ed version of chess& 0ut what ha!!ens when there is no hi her court, and the Su!reme 7ourt itself ets all tan led u! in le al troubles? This sort of snarl nearly ha!!ened in the 3ater ate era& The President threatened to obey only a -definitive rulin - of the Su!reme 7ourtthen claimed he had the ri ht to decide what is -definitive-& 8ow that threat never was made ood= but if it had been, it would have touched off a monumental confrontation between two levels of overnment, each of which, in some ways, can validly claim to be -abovethe otherand to whom is there recourse to decide which one is ri ht? To say -7on ressis not to settle the matter, for 7on ress mi ht command the President to obey the Su!reme 7ourt, yet the President mi ht still refuse, claimin that he has the le al ri ht to disobey the Su!reme 7ourt 4and 7on ress25 under certain circumstances& This would create a new court case, and would throw the whole system into disarray, because it would be so une(!ected, so Tan ledso Stran e2 The irony is that once you hit your head a ainst the ceilin like this, where you are !revented from .um!in out of the system to a yet hi her authority, the only recourse is to forces which seem less well defined by rules, but which are the only source of hi her-level rules anyway, the lower-level rules, which in this case means the eneral reaction of society& It is well to remember that in a society like ours, the le al system is, in a sense, a !olite esture ranted collectively by millions of !eo!leand it can be overridden .ust as easily as a river can overflow its banks& Then a seemin anarchy takes over= but anarchy has its own kinds of rules, no less than does civili<ed society, it is .ust that they o!erate from the bottom u!, not from the to! down& ) student of anarchy could try to discover rules accordin to which anarchic situations develo! in time, and very likely there are some such rules& )n analo y from !hysics is useful here& )s was mentioned earlier in the book, ases in equilibrium obey sim!le laws connectin their tem!erature, !ressure, and volume& ?owever, a as can violate those laws 4as a President can violate laws5 !rovided it is not in a state of equilibrium& In nonequilibrium situations, to describe what ha!!ens, a !hysicist has recourse only to statistical mechanicsthat is, to a level of descri!tion which is not macrosco!ic, for the ultimate e(!lanation of a as1s behavior always lies on the molecular level, .ust as the ultimate e(!lanation of a society1s !olitical behavior always lies at the - rass roots level-& The field of nonequilibrium thermodynamics attem!ts to find macrosco!ic laws to describe the behavior of ases 4and other systems5 which are out of equilibrium& It is the analo ue to the branch of !olitical science which would search for laws overnin anarchical societies& Ether curious tan les which arise in overnment include the '0I investi atin its own wron doin s, a sheriff oin to .ail while in office, the self-a!!lication of the !arliamentary rules of !rocedure, and so on& Ene of the most curious le al cases I ever heard of involved a !erson who claimed to have !sychic !owers& In fact, he claimed to be

able to use his !sychic !owers to detect !ersonality traits, and thereby to aid lawyers in !ickin .uries& 8ow what if this -!sychic- has to stand trial himself one day? 3hat effect mi ht this have on a .ury member who believes staunchly in ESP? ?ow much will he feel affected by the !sychic 4whether or not the !sychic is enuine5? The territory is ri!e for e(!loitationa reat area for self-fulfillin !ro!hecies&

Tangles In"ol"ing Science and the 0ccult

S!eakin of !sychics and ESP, another s!here of life where stran e loo!s abound is frin e science& 3hat frin e science does is to call into question many of the standard !rocedures or beliefs of orthodo( science, and thereby challen e the ob.ectivity of science& 8ew ways of inter!retin evidence that rival the established ones are !resented& 0ut how do you evaluate a way of inter!retin evidence? Isn1t this !recisely the !roblem of ob.ectivity all over a ain, .ust on a hi her !lane? Ef course& Aewis 7arroll1s infinitere ress !arado( a!!ears in a new uise& The Tortoise would ar ue that if you want to show that ) is a fact, you need evidence, 0& 0ut what makes you sure that 0 is evidence of )? To show that, you need meta-evidence, 7& )nd for the validity of that metaevidence, you need meta-meta-evidenceand so on, ad nauseam& :es!ite this ar ument, !eo!le have an intuitive sense of evidence& This is becauseto re!eat an old refrain !eo!le have built-in hardware in their brains that includes some rudimentary ways of inter!retin evidence& 3e can build on this, and accumulate new ways of inter!retin evidence= we even learn how and when to override our most basic mechanisms of evidence inter!retation, as one must, for e(am!le, in tryin to fi ure out ma ic tricks& 7oncrete e(am!les of evidence dilemmas cro! u! in re ard to many !henomena of frin e science& 'or instance, ESP often seems to manifest itself outside of the laboratory, but when brou ht into the laboratory, it vanishes mysteriously& The standard scientific e(!lanation for this is that ESP is a nonreal !henomenon which cannot stand u! to ri orous scrutiny& Some 4by no means all5 believers in ESP have a !eculiar way of fi htin back, however& They say, -8o, ESP is real= it sim!ly oes away when one tries to observe it scientificallyit is contrary to the nature of a scientific worldview&- This is an ama<in ly bra<en technique, which we mi ht call -kickin the !roblem u!stairs-& 3hat that means is, instead of questionin the matter at hand, you call into doubt theories belon in to a hi her level of credibility& The believers in ESP insinuate that what is wron is not their ideas, but the belief system of science& This is a !retty randiose claim, and unless there is overwhelmin evidence for it, one should be ske!tical of it& 0ut then here we are a ain, talkin about -overwhelmin evidence- as if everyone a reed on what that means2

The ;ature of E"idence

The Sa redo-Sim!licio-Salviati tan le, mentioned in 7ha!ters TIII and TG, ives another e(am!le of the com!le(ities of evaluation of evidence& Sa redo tries to find some ob.ective com!romise, if !ossible, between the o!!osin views of Sim!licio and Salviati& 0ut com!romise may not always be !ossible& ?ow can one com!romise -fairly- between

ri ht and wron ? 0etween fair and unfair? 0etween com!romise and no com!romise? These questions come u! over and over a ain in dis uised form in ar uments about ordinary thin s& Is it !ossible to define what evidence is? Is it !ossible to lay down laws as to how to make sense out of situations? Probably not, for any ri id rules would undoubtedly have e(ce!tions, and nonri id rules are not rules& ?avin an intelli ent )I !ro ram would not solve the !roblem either, for as an evidence !rocessor, it would not be any less fallible than humans are& So, if evidence is such an intan ible thin after all, why am I warnin a ainst new ways of inter!retin evidence? )m I bein inconsistent? In this case, I don1t think so& /y feelin is that there are uidelines which one can ive, and out of them an or anic synthesis can be made& 0ut inevitably some amount of .ud ment and intuition must enter the !icturethin s which are different in different !eo!le& They will also be different in different )I !ro rams& Fltimately, there are com!licated criteria for decidin if a method of evaluation of evidence is ood& Ene involves the -usefulness- of ideas which are arrived at by that kind of reasonin & /odes of thou ht which lead to useful new thin s in life are deemed -valid- in some sense& 0ut this word -useful- is e(tremely sub.ective& /y feelin is that the !rocess by which we decide what is valid or what is true is an art= and that it relies as dee!ly on a sense of beauty and sim!licity as it does on rocksolid !rinci!les of lo ic or reasonin or anythin else which can be ob.ectively formali<ed& I am not sayin either 4"5 truth is a chimera, or 4B5 human intelli ence is in !rinci!le not !ro rammable& I am sayin 4"5 truth is too elusive for any human or any collection of humans ever to attain fully= and 4B5 )rtificial Intelli ence, when it reaches the level of human intelli enceor even if it sur!asses itwill still be !la ued by the !roblems of art, beauty, and sim!licity, and will run u! a ainst these thin s constantly in its own search for knowled e and understandin & -3hat is evidence?- is not .ust a !hiloso!hical question, for it intrudes into life in all sorts of !laces& Hou are faced with an e(traordinary number of choices as to how to inter!ret evidence at every moment& Hou can hardly o into a bookstore 4or these days, even a rocery store25 without seein books on clairvoyance, ESP, F'E1s, the 0ermuda trian le, astrolo y, dowsin , evolution versus creation, black holes, !si fields, biofeedback, transcendental meditation, new theories of !sycholo y&&& In science, there are fierce debates about catastro!he theory, elementary !article theory, black holes, truth and e(istence in mathematics, free will, )rtificial Intelli ence, reductionism versus holism&&& En the more !ra matic side of life, there are debates over the efficacy of vitamin 7 or of laetrile, over the real si<e of oil reserves 4either under round or stored5, over what causes inflation and unem!loymentand on and on& There is 0uckminster 'ullerism, 9en 0uddhism, 9eno1s !arado(es, !sychoanalysis, etc&, etc& 'rom issues as trivial as where books ou ht to be shelved in a store, to issues as vital as what ideas are to be tau ht to children in schools, ways of inter!retin evidence !lay an inestimable role&

Seeing 0neself
Ene of the most severe of all !roblems of evidence inter!retation is that of tryin to inter!ret all the confusin si nals from the outside as to who one is& In this case, the !otential for intralevel and interlevel conflict is tremendous& The !sychic mechanisms have to deal simultaneously with the individual1s internal need for self-esteem and the constant flow of evidence from the outside affectin the self-ima e& The result is that information flows in a com!le( swirl between different levels of the !ersonality= as it oes round and round, !arts of it et ma nified, reduced, ne ated, or otherwise distorted, and then those !arts in turn et further sub.ected to the same sort of swirl, over and over a ainall of this in an attem!t to reconcile what is, with what we wish were 4see 'i & *"5& The u!shot is that the total !icture of -who I am- is inte rated in some enormously com!le( way inside the entire mental structure, and contains in each one of us a lar e number of unresolved, !ossibly unresolvable, inconsistencies& These undoubtedly !rovide much of the dynamic tension which is so much a !art of bein human& Eut of this tension between the inside and outside notions of who we are come the drives towards various oals that make each of us unique& Thus, ironically, somethin which we all have in commonthe fact of bein self-reflectin conscious bein sleads to the rich diversity in the ways we have of internali<in evidence about all sorts of thin s, and in the end winds u! bein one of the ma.or forces in creatin distinct individuals&

G#del%s Theorem and 0ther

isci!lines

It is natural to try to draw !arallels between !eo!le and sufficiently com!licated formal systems which, like !eo!le, have -self-ima es- of a sort& 6>del1s Theorem shows that there are fundamental limitations to consistent formal systems with self-ima es& 0ut is it more eneral? Is there a -6>del1s Theorem of !sycholo y-, for instance? If one uses 6>del1s Theorem as a meta!hor, as a source of ins!iration, rather than tryin to translate it literally into the lan ua e of !sycholo y or of any other disci!line, then !erha!s it can su est new truths in !sycholo y or other areas& 0ut it is quite un.ustifiable to translate it directly into a statement of another disci!line and take that as equally valid& It would be a lar e mistake to think that what has been worked out with the utmost delicacy in mathematical lo ic should hold without modification in a com!letely different area&

Intros!ection and Insanity5 A G#delian ,roblem

I think it can have su estive value to translate 6>del1s Theorem into other domains, !rovided one s!ecifies in advance that the translations are meta!horical and are not intended to be taken literally& That havin been said, I see two ma.or ways of usin analo ies to connect 6>del1s Theorem and human thou hts& Ene involves the !roblem of wonderin about one1s sanity& ?ow can you fi ure out if you are sane? This is a Stran e

Aoo! indeed& Ence you be in to question your own sanity, you can et tra!!ed in an ever-ti hter vorte( of self-fulfillin !ro!hecies, thou h the !rocess is by no means inevitable& Everyone knows that the insane inter!ret the world via their own !eculiarly consistent lo ic= how can you tell if your own lo ic is -!eculiar- or not, iven that you have only your own lo ic to .ud e itself? I don1t see any answer& I am .ust reminded of 6>del1s second Theorem, which im!lies that the only versions of formal number theory which assert their own consistency are inconsistent&&&

Can >e &nderstand 0ur 0'n Minds or Brains7

The other meta!horical analo ue to 6>del1s Theorem which I find !rovocative su ests that ultimately, we cannot understand our own mindsIbrains& This is such a loaded, manyleveled idea that one must be e(tremely cautious in !ro!osin it& 3hat does -understandin our own mindsIbrains- mean? It could mean havin a eneral sense of how they work, as mechanics have a sense of how cars work& It could mean havin a com!lete e(!lanation for why !eo!le do any and all thin s they do& It could mean havin a com!lete understandin of the !hysical structure of one1s own brain on all levels& It could mean havin a com!lete wirin dia ram of a brain in a book 4or library or com!uter5& It could mean knowin , at every instant, !recisely what is ha!!enin in one1s own brain on the neural leveleach firin , each syna!tic alteration, and so on& It could mean havin written a !ro ram which !asses the Turin test& It could mean knowin oneself so !erfectly that such notions as the subconscious and the intuition make no sense, because everythin is out in the o!en& It could mean any number of other thin s& 3hich of these ty!es of self-mirrorin , if any, does the self-mirrorin in 6>del1s Theorem most resemble? I would hesitate to say& Some of them are quite silly& 'or instance, the idea of bein able to monitor your own brain state in all its detail is a !i!e dream, an absurd and uninterestin !ro!osition to start with= and if 6>del1s Theorem su ests that it is im!ossible, that is hardly a revelation& En the other hand, the a e-old oal of knowin yourself in some !rofound waylet us call it -understandin your own !sychic structure-has a rin of !lausibility to it& 0ut mi ht there not be some va uely 6>delian loo! which limits the de!th to which any individual can !enetrate into his own !syche? Just as we cannot see our faces with our own eyes, is it not reasonable to e(!ect that we cannot mirror our com!lete mental structures in the symbols which carry them out? )ll the limitative Theorems of metamathematics and the theory of com!utation su est that once the ability to re!resent your own structure has reached a certain critical !oint, that is the kiss of death, it uarantees that you can never re!resent yourself totally& 6>del1s Incom!leteness Theorem, 7hurch1s Fndecidability Theorem, Turin 1s ?altin Theorem, Tarski1s Truth Theoremall have the flavor of some ancient fairy tale which warns you that -To seek self-knowled e is to embark on a .ourney which&&& will always be incom!lete, cannot be charted on any ma!, will never halt, cannot be described&0ut do the limitative Theorems have any bearin on !eo!le? ?ere is one way of ar uin the case& Either I am consistent or I am inconsistent& 4The latter is much more likely, but for com!leteness1 sake, I consider both !ossibilities&5 If I am consistent, then

there are two cases& 4"5 The -low-fidelity- case, my self-understandin is below a certain critical !oint& In this case, I am incom!lete by hy!othesis& 4B5 The -hi h-fidelity- case, /y self-understandin has reached the critical !oint where a meta!horical analo ue of the limitative Theorems does a!!ly, so my self-understandin undermines itself in a 6>delian way, and I am incom!lete for that reason& 7ases 4"5 and 4B5 are !redicated on my bein "++ !er cent consistenta very unlikely state of affairs& /ore likely is that I am inconsistentbut that1s worse, for then inside me there are contradictions, and how can I ever understand that? 7onsistent or inconsistent, no one is e(em!t from the mystery of the self& Probably we are all inconsistent& The world is .ust too com!licated for a !erson to be able to afford the lu(ury of reconcilin all of his beliefs with each other& Tension and confusion are im!ortant in a world where many decisions must be made quickly, /i uel de Fnamuno once said, -If a !erson never contradicts himself, it must be that he says nothin &- I would say that we all are in the same boat as the 9en master who, after contradictin himself several times in a row, said to the confused :oko, -I cannot understand myself&-

G#del%s Theorem and ,ersonal ;one-istence

Perha!s the reatest contradiction in our lives, the hardest to handle, is the knowled e -There was a time when I was not alive, and there will come a time when I am not alive&En one level, when you -ste! out of yourself- and see yourself as -.ust another human bein -, it makes com!lete sense& 0ut on another level, !erha!s a dee!er level, !ersonal none(istence makes no sense at all& )ll that we know is embedded inside our minds, and for all that to be absent from the universe is not com!rehensible& This is a basic undeniable !roblem of life= !erha!s it is the best meta!horical analo ue of 6>del1s Theorem& 3hen you try to ima ine your own none(istence, you have to try to .um! out of yourself, by ma!!in yourself onto someone else& Hou fool yourself into believin that you can im!ort an outsider1s view of yourself into you, much as T8T -believes- it mirrors its own metatheory inside itself& 0ut T8T only contains its own metatheory u! to a certain e(tentnot fully& )nd as for you, thou h you may ima ine that you have .um!ed out of yourself, you never can actually do sono more than Escher1s dra on can .um! out of its native two-dimensional !lane into three dimensions& In any case, this contradiction is so reat that most of our lives we .ust swee! the whole mess under the ru , because tryin to deal with it .ust leads nowhere& 9en minds, on the other hand, revel in this irreconcilability& Ever and over a ain, they face the conflict between the Eastern belief, -The world and I are one, so the notion of my ceasin to e(ist is a contradiction in terms- 4my verbali<ation is undoubtedly too 3esterni<eda!olo ies to 9enists5, and the 3estern belief, -I am .ust !art of the world, and I will die, but the world will o on without me&-

Science and

ualism

Science is often critici<ed as bein too -3estern- or -dualistic-that is, bein !ermeated by the dichotomy between sub.ect and ob.ect, or observer and observed& 3hile it is true that u! until this century, science was e(clusively concerned with thin s which can be readily distin uished from their human observerssuch as o(y en and carbon, li ht and heat, stars and !lanets, accelerations and orbits, and so onthis !hase of science was a necessary !relude to the more modern !hase, in which life itself has come under investi ation& Ste! by ste!, ine(orably, -3estern- science has moved towards investi ation of the human mindwhich is to say, of the observer& )rtificial Intelli ence research is the furthest ste! so far alon that route& 0efore )I came alon , there were two ma.or !reviews of the stran e consequences of the mi(in of sub.ect and ob.ect in science& Ene was the revolution of quantum mechanics, with its e!istemolo ical !roblems involvin the interference of the observer with the observed& The other was the mi(in of sub.ect and ob.ect in metamathematics, be innin with 6>del1s Theorem and movin throu h all the other limitative Theorems we have discussed& Perha!s the ne(t ste! after )I will be the self-a!!lication of science, science studyin itself as an ob.ect& This is a different manner of mi(in sub.ect and ob.ect!erha!s an even more tan led one than that of humans studyin their own minds& 0y the way, in !assin , it is interestin to note that all results essentially de!endent on the fusion of sub.ect and ob.ect have been limitative results& In addition to the limitative Theorems, there is ?eisenber 1s uncertainty !rinci!le, which says that measurin one quantity renders im!ossible the simultaneous measurement of a related quantity& I don1t know why all these results are limitative& /ake of it what you will&

Symbol #s& 0bVect in Modern Music and Art

7losely linked with the sub.ect-ob.ect dichotomy is the symbol-ob.ect dichotomy, which was e(!lored in de!th by Audwi 3itt enstein in the early !art of this century& Aater the words -use- and -mention- were ado!ted to make the same distinction& Muine and others have written at len th about the connection between si ns and what they stand for& 0ut not only !hiloso!hers have devoted much thou ht to this dee! and abstract matter& In our century both music and art have one throu h crises which reflect a !rofound concern with this !roblem& 3hereas music and !aintin , for instance, have traditionally e(!ressed ideas or emotions throu h a vocabulary of -symbols- 4i&e& visual ima es, chords, rhythms, or whatever5, now there is a tendency to e(!lore the ca!acity of music and art to not e(!ress anythin .ust to (e& This means to e(ist as !ure lobs of !aint, or !ure sounds, but in either case drained of all symbolic value& In music, in !articular, John 7a e has been very influential in brin in a 9en-like a!!roach to sound& /any of his !ieces convey a disdain for -use- of soundsthat is, usin sounds to convey emotional statesand an e(ultation in -mentionin - sounds that is, concoctin arbitrary .u(ta!ositions of sounds without re ard to any !reviously formulated code by which a listener could decode them into a messa e& ) ty!ical e(am!le is -Ima inary Aandsca!e no& C-, the !olyradio !iece described in 7ha!ter GI& I

may not be doin 7a e .ustice, but to me it seems that much of his work has been directed at brin in meanin lessness into music, and in some sense, at makin that meanin lessness have meanin & )leatoric music is a ty!ical e(!loration in that direction& 4Incidentally, chance music is a close cousin to the much later notion of -ha!!enin s- or -be-in- 1s&5 There are many other contem!orary com!osers who are followin 7a e1s lead, but few with as much ori inality& ) !iece by )nna Aockwood, called -Piano 0urnin -, involves .ust thatwith the strin s stretched to ma(imum ti htness, to make them sna! as loudly as !ossible= in a !iece by Aa/onte Houn , the noises are !rovided by shovin the !iano all around the sta e and throu h obstacles, like a batterin ram& )rt in this century has one throu h many convulsions of this eneral ty!e& )t first there was the abandonment of re!resentation, which was enuinely revolutionary, the be innin s of abstract art& ) radual swoo! from !ure re!resentation to the most hi hly abstract !atterns is revealed in the work of Piet /ondrian& )fter the world was used to nonre!resentational art, then surrealism came alon & It was a bi<arre about-face, somethin like neoclassicism in music, in which e(tremely re!resentational art was -subverted- and used for alto ether new reasons, to shock, confuse, and ama<e& This school was founded by )ndre 0reton, and was located !rimarily in 'rance= some of its more influential members were :ali, /a ritte, de 7hirico, Tan uy&

Magritte%s Semantic Illusions

Ef all these artists, /a ritte was the most conscious of the symbol-ob.ect mystery 4which I see as a dee! e(tension of the use-mention distinction5& ?e uses it to evoke !owerful res!onses in viewers, even if the viewers do not verbali<e the distinction this way& 'or e(am!le, consider his very stran e variation on the theme of still life, called 8ommon -ense 4'i & "%$5& ?ere, a dish filled with fruit, ordinarily the kind of thin re!resented inside a still life, is shown sittin on to! of a blank canvas& The conflict between the symbol and the real is reat& 0ut that is not the full irony, for of course the whole thin is itself .ust a !aintin in fact, a still life with nonstandard sub.ect matter&

73G5=E BHC! 7ommon Sense, () =ene 'agritte BIDG%DJ/!

73G5=E BHE! The Two /ysteries, () =ene 'agritte BIJJ/&

/a ritte1s series of !i!e !aintin s is fascinatin and !er!le(in & 7onsider The Two ')steries 4'i & "%*5& 'ocusin on the inner !aintin , you et the messa e that

symbols and !i!es are different& Then your lance moves u!ward to the -real- !i!e floatin in the airyou !erceive that it is real, while the other one is .ust a symbol& 0ut that is of course totally wron , both of them are on the same flat surface before your eyes& The idea that one !i!e is in a twice-nested !aintin , and therefore somehow -less realthan the other !i!e, is a com!lete fallacy& Ence you are willin to -enter the room-, you have already been tricked, you1ve fallen for ima e as reality& To be consistent in your ullibility, you should ha!!ily o one level further down, and confuse ima e-withinima e with reality& The only way not to be sucked in is to see both !i!es merely as colored smud es on a surface a few inches in front of your nose& Then, and only then, do you a!!reciate the full meanin of the written messa e -7eci n1est !as une !i!e-but ironically, at the very instant everythin turns to smud es, the writin too turns to smud es, thereby losin its meanin 2 In other words, at that instant, the verbal messa e of the !aintin self-destructs in a most 6>delian way&

73G5=E BHI! Smoke Si nal& U.rawing () the author!V

The )ir and the Son 4'i & *B5, taken from a series by /a ritte, accom!lishes all that The Two ')steries does, but in one level instead of two& /y drawin s -moke -ignal and &ipe .ream 4'i s& "%# and "C+5 constitute -Gariations on a Theme of /a ritte-& Try starin at -moke -ignal for a while& 0efore lon , you should be able to make out a hidden messa e sayin , -7eci n1est !as un messa e-& Thus, if you find the messa e, it denies itselfyet if you don1t, you miss the !oint entirely& 0ecause of their indirect selfsnuffin , my two !i!e !ictures can be loosely ma!!ed onto 6>del1s 6thus ivin rise

to a -7entral Pi!ema!-, in the same s!irit as the other -7entral Tma!s-, :o , 7rab, Sloth& ) classic e(am!le of use-mention confusion in !aintin s is the occurrence of a !alette in a !aintin & 3hereas the !alette is an illusion created by the re!resentational skill of the !ainter, the !aints on the !ainted !alette are literal daubs of !aint from the artist1s !alette& The !aint !lays itselfit does not symboli<e anythin else& In .on Gio#anni, /o<art e(!loited a related trick, he wrote into the score e(!licitly the sound of an orchestra tunin u!& Similarly, if I want the letter 1I1 to !lay itself 4and not symboli<e me5, I !ut 1I1 directly into my te(t= then I enclose 1I1 between quotes& 3hat results is 11I4not 1I1, nor - 1I1 - 5& 6ot that?

73G5=E BDK! Pi!e :ream! U.rawing () the author!V

The (Code( of Modern Art

) lar e number of influences, which no one could ho!e to !in down com!letely, led to further e(!lorations of the symbol-ob.ect dualism in art& There is no doubt that John 7a e, with his interest in 9en, had a !rofound influence on art as well as on music& ?is friends .as!er Johns and ;obert ;auschenber both e(!lored the distinction between ob.ects and symbols by usin ob.ects as symbols for themselvesor, to fli! the coin, by usin symbols as ob.ects in themselves& )ll of this was !erha!s intended to break down the notion that art is one ste! removed from realitythat art s!eaks in -code-, for which the viewer must act as inter!reter& The idea was to eliminate the ste! of inter!retation and let the naked ob.ect sim!ly be, !eriod& 4-Period-a curious case of use-mention blur&5 ?owever, if this was the intention, it was a monumental flo!, and !erha!s had to be& )ny time an ob.ect is e(hibited in a allery or dubbed a -work-, it acquires an aura of dee! inner si nificanceno matter how much the viewer has been warned not to

look for meanin & In fact, there is a backfirin effect whereby the more that viewers are told to look at these ob.ects without mystification, the more mystified the viewers et& )fter all, if a wooden crate on a museum floor is .ust a wooden crate on a museum floor, then why doesn1t the .anitor haul it out back and throw it in the arba e? 3hy is the name of an artist attached to it? 3hy did the artist want to demystify art? 3hy isn1t that dirt clod out front labeled with an artist1s name? Is this a hoa(? )m I cra<y, or are artists cra<y? /ore and more questions flood into the viewer1s mind= he can1t hel! it& This is the -frame effect- which art)rtautomatically creates& There is no way to su!!ress the wonderin s in the minds of the curious& Ef course, if the !ur!ose is to instill a 9en-like sense of the world as devoid of cate ories and meanin s, then !erha!s such art is merely intended to serveas does intellectuali<in about 9enas a catalyst to ins!ire the viewer to o out and become acquainted with the !hiloso!hy which re.ects -inner meanin s- and embraces the world as a whole& In this case, the art is self-defeatin in the short run, since the viewers do !onder about its meanin , but it achieves its aim with a few !eo!le in the lon run, by introducin them to its sources& 0ut in either case, it is not true that there is no code by which ideas are conveyed to the viewer& )ctually, the code is a much more com!le( thin , involvin statements about the absence of codes and so forththat is, it is !art code, !art metacode, and so on& There is a Tan led ?ierarchy of messa es bein transmitted by the most 9en-like art ob.ects, which is !erha!s why so many find modern art so inscrutable&

Ism 0nce Again

7a e has led a movement to break the boundaries between art and nature& In music, the theme is that all sounds are equala sort of acoustical democracy& Thus silence is .ust as im!ortant as sound, and random sound is .ust as im!ortant as or ani<ed sound& Aeonard 0& /eyer, in his book 'usic, the *rts, and 3deas, has called this movement in music -transcendentalism-, and states,
If the distinction between art and nature is mistaken, aesthetic valuation is irrelevant& Ene should no more .ud e the value of a !iano sonata than one should .ud e the value of a stone, a thunderstorm, or a starfish& -7ate orical statements, such as ri ht and wron , beautiful or u ly, ty!ical of the rationalistic thinkin of tonal aesthetics,- writes Auciano 0erio Ua contem!orary com!oserV -are no lon er useful in understandin why and how a com!oser today works on audible forms and musical action&-

Aater, /eyer continues in describin the !hiloso!hical !osition of transcendentalism,

&&&all thin s in all of time and s!ace are ine(tricably connected with one another& )ny divisions, classifications, or or ani<ations discovered in the universe are arbitrary& The world is a com!le(, continuous, sin le event&B UShades of 9eno2V

I find -transcendentalism- too bulky a name for this movement& In its !lace, I use -ism-& 0ein a suffi( without a !refi(, it su ests an ideolo y without ideaswhich,

73G5=E BDB& The ?uman 7ondition I, () =ene 'agritte BIHH/&

however you inter!ret it, is !robably the case& )nd since&-ism- embraces whatever is, its name is quite fittin & In -ism- the word -is- is half mentioned, half used= what could be more a!!ro!riate? Ism is the s!irit of 9en in art& )nd .ust as the central !roblem of 9en is to unmask the self, the central !roblem of art in this century seems to be to fi ure out what art is& )ll these thrashin s-about are !art of its identity crisis& 3e have seen that the use-mention dichotomy, when !ushed, turns into the !hiloso!hical !roblem of symbol-ob.ect dualism, which links it to the mystery of mind& /a ritte wrote about his !aintin The ,uman 8ondition 3 4'i & "C"5,
I !laced in front of a window, seen from a room, a !aintin re!resentin e(actly that !art of the landsca!e which was hidden from view by the !aintin & Therefore, the tree re!resented in the !aintin hid from view the tree situated behind it, outside the room& It e(isted for the s!ectator, as it were, simultaneously in his mind, as both inside the room in the !aintin , and outside in the real landsca!e& 3hich is how we see the world, we see it as bein outside ourselves even thou h it is only a mental re!resentation of it that we e(!erience inside ourselves& %

&nderstanding the Mind

'irst throu h the !re nant ima es of his !aintin , and then in direct words, /a ritte e(!resses the link between the two questions -?ow do symbols work?- and -?ow do our minds work?- )nd so he leads us back to the question !osed earlier, -7an we ever ho!e to understand our mindsIbrains?Er does some marvelous diabolical 6>delian !ro!osition !reclude our ever unravelin our minds? Provided you do not ado!t a totally unreasonable definition of -understandin -, I see no 6>delian obstacle in the way of the eventual understandin of our minds& 'or instance, it seems to me quite reasonable to desire to understand the workin !rinci!les of brains in eneral, much the same way as we understand the workin !rinci!les of car en ines in eneral& It is quite different from tryin to understand any sin le brain in every last detaillet alone tryin to do this for one1s own brain2 I don1t see how 6>del1s Theorem, even if construed in the slo!!iest way, has anythin to say about the feasibility of this !ros!ect& I see no reason that 6>del1s Theorem im!oses any limitations on our ability to formulate and verify the eneral mechanisms by which thou ht !rocesses take !lace in the medium of nerve cells& I see no barrier im!osed by 6>del1s Theorem to the im!lementation on com!uters 4or their successors5 of ty!es of symbol mani!ulation that achieve rou hly the same results as brains do& It is entirely another question to try and du!licate in a !ro ram some !articular human1s mindbut to !roduce an intelli ent !ro ram at all is a more limited oal& 6>del1s Theorem doesn1t ban our re!roducin our own level of intelli ence via !ro rams any more than it bans our re!roducin our own level of intelli ence via transmission of hereditary information in :8), followed by education& Indeed, we have seen, in 7ha!ter TGI, how a remarkable 6>delian mechanismthe Stran e Aoo! of !roteins and :8) is !recisely what allows transmission of intelli ence2 :oes 6>del1s Theorem, then, have absolutely nothin to offer us in thinkin about our own minds? I think it does, althou h not in the mystical and limitative way which some !eo!le think it ou ht to& I think that the !rocess of comin to understand 6>del1s

!roof, with its construction involvin arbitrary codes, com!le( isomor!hisms, hi h and low levels of inter!retation, and the ca!acity for self-mirrorin , may in.ect some rich undercurrents and flavors into one1s set of ima es about symbols and symbol !rocessin , which may dee!en one1s intuition for the relationshi!& between mental structures on different levels&

Accidental Ine-!licability of Intelligence7

0efore su estin a !hiloso!hically intri uin -a!!lication- of 6>del1s !roof& I would like to brin u! the idea of -accidental ine(!licability- of intelli ence& ?ere is what that involves& It could be that our brains, unlike car en ines, are stubborn and intractable systems which we cannot neatly decom!ose in any way& )t !resent, we have no idea whether our brains will yield to re!eated attem!ts to cleave them into clean layers, each of which can be e(!lained in terms of lower layersor whether our brains will foil all our attem!ts at decom!osition& 0ut even if we do fail to understand ourselves, there need not be any 6>delian -twist- behind it= it could be sim!ly an accident of fate that our brains are too weak to understand themselves& Think of the lowly iraffe, for instance, whose brain is obviously far below the level required for self-understandin yet it is remarkably similar to our own brain& In fact, the brains of iraffes, ele!hants, baboonseven the brains of tortoises or unknown bein s who are far smarter than we are!robably all o!erate on basically the same set of !rinci!les& 6iraffes may lie far below the threshold of intelli ence necessary to understand how those !rinci!les fit to ether to !roduce the qualities of mind= humans may lie closer to that threshold!erha!s .ust barely below it, !erha!s even above it& The !oint is that there may be no fundamental 4i&e&, 6>delian5 reason why those qualities are incom!rehensible= they may be com!letely clear to more intelli ent bein s&

&ndecidability Is Inse!arable from a /igh-Le"el 4ie'!oint

0arrin this !essimistic notion of the accidental ine(!licability of the brain, what insi hts mi ht 6>del1s !roof offer us about e(!lanations of our mindsIbrains? 6>del1s !roof offers the notion that a hi h-level view of a system may contain e(!lanatory !ower which sim!ly is absent on the lower levels& 0y this I mean the followin & Su!!ose someone ave you 6, 6>del1s undecidable strin , as a strin of T8T& )lso su!!ose you knew nothin of 6>del-numberin & The question you are su!!osed to answer is, -3hy isn1t this strin a theorem of T8T?- 8ow you are used to such questions= for instance, if you had been asked that question about S?E?, you would have a ready e(!lanation, -3ts negation, TS?E?, is a theorem&- This, to ether with your knowled e that T8T is consistent, !rovides an e(!lanation of why the iven strin is a nontheorem& This is what I call an e(!lanation -on the T8T-level-& 8otice how different it is from the e(!lanation of why M& is not a theorem of the /IF-system, the former comes from the /-mode, the latter only from the I-mode&

8ow what about 6? The T8T-level e(!lanation which worked for S?E? does not work for 6, because h6 is not a theorem& The !erson who has no overview of T8T will be baffled as to why he can1t make 6 accordin to the rules, because as an arithmetical !ro!osition, it a!!arently has nothin wron with it& In fact, when 6 is turned into a universally quantified strin , every instance otten from 6 by substitutin numerals for the variables can be derived& The only way to e(!lain 61s nontheoremhood is to discover the notion of 6>del-numberin and view T8T on an entirely different level& It is not that it is .ust difficult and com!licated to write out the e(!lanation on the T8T-level= it is im!ossible& Such an e(!lanation sim!ly does not e(ist& There is, on the hi h level, a kind of e(!lanatory !ower which sim!ly is lackin , in !rinci!le, on the T8T-level& 61s nontheoremhood is, so to s!eak, an intrinsicall) high%le#el fact& It is my sus!icion that this is the case for all undecidable !ro!ositions= that is to say, every undecidable !ro!osition is actually a 6>del sentence, assertin its own nontheoremhood in some system via some code&

Consciousness as an Intrinsically /igh-Le"el ,henomenon

Aooked at this way, 6>del1s !roof su eststhou h by no means does it !rove2that there could be some hi h-level way of viewin the mindIbrain, involvin conce!ts which do not a!!ear on lower levels, and that this level mi ht have e(!lanatory !ower that does not e(istnot even in !rinci!leon lower levels& It would mean that some facts could be e(!lained on the hi h level quite easily, but not on lower levels at all& 8o matter how lon and cumbersome a low-level statement were made, it would not e(!lain the !henomena in question& It is the analo ue to the fact that, if you make derivation after derivation in T8T, no matter how lon and cumbersome you make them, you will never come u! with one for 6des!ite the fact that on a hi her level, you can see that 6 is true& 3hat mi ht such hi h-level conce!ts be? It has been !ro!osed for eons, by various holistically or -soulistically- inclined scientists and humanists, that consciousness is a !henomenon that esca!es e(!lanation in terms of brain-com!onents= so here is a candidate, at least& There is also the ever-!u<<lin notion of free will& So !erha!s these qualities could be -emer ent- in the sense of requirin e(!lanations which cannot be furnished by the !hysiolo y alone& 0ut it is im!ortant to reali<e that if we are bein uided by 6>del1s !roof in makin such bold hy!otheses, we must carry the analo y throu h thorou hly& In !articular, it is vital to recall that 61s nontheoremhood does have an e(!lanationit is not a total mystery2 The e(!lanation hin es on understandin not .ust one level at a time, but the way in which one level mirrors its metalevel, and the consequences of this mirrorin & If our analo y is to hold, then, -emer ent- !henomena would become e(!licable in terms of a relationshi! between& different levels in mental systems&

Strange Loo!s as the Cru- of Consciousness

/y belief is that the e(!lanations of -emer ent- !henomena in our brainsfor instance, ideas, ho!es, ima es, analo ies, and finally consciousness and free willare based on a kind of Stran e Aoo!, an interaction between levels in which the to! level reaches back down towards the bottom level and influences it, while at the same time bein itself determined by the bottom level& In other words, a self-reinforcin -resonance- between different levelsquite like the ?enkin sentence which, by merely assertin its own !rovability, actually becomes !rovable& The self comes into bein at the moment it has the !ower to reflect itself& This should not be taken as an antireductionist !osition& It .ust im!lies that a reductionistic e(!lanation of a mind, in order to (e comprehensi(le, must brin in -softconce!ts such as levels, ma!!in s, and meanin s& In !rinci!le, I have no doubt that a totally reductionistic but incom!rehensible e(!lanation of the brain e(ists= the !roblem is how to translate it into a lan ua e we ourselves can fathom& Surely we don1t want a descri!tion in terms of !ositions and momenta of !articles= we want a descri!tion which relates neural activity to -si nals- 4intermediate-level !henomena5and which relates si nals, in turn, to -symbols- and -subsystems-, includin the !resumed-to-e(ist -selfsymbol-& This act of translation from low-level !hysical hardware to hi h-level !sycholo ical software is analo ous to the translation of number-theoretical statements into metamathematical statements& ;ecall that the level-crossin which takes !lace at this e(act translation !oint is what creates 6>del1s incom!leteness and the self-!rovin character of ?enkin1s sentence& I !ostulate that a similar level-crossin is what creates our nearly unanaly<able feelin s of self& In order to deal with the full richness of the brainImind system, we will have to be able to sli! between levels comfortably& /oreover, we will have to admit various ty!es of -causality-, ways in which an event at one level of descri!tion can -cause- events at other levels to ha!!en& Sometimes event ) will be said to -cause- event 0 sim!ly for the reason that the one is a translation, on another level of descri!tion, of the other& Sometimes -cause- will have its usual meanin , !hysical causality& 0oth ty!es of causalityand !erha!s some morewill have to be admitted in any e(!lanation of mind, for we will have to admit causes that !ro!a ate both u!wards and downwards in the Tan led ?ierarchy of mentality, .ust as in the 7entral :o ma!& )t the cru(, then, of our understandin ourselves will come an understandin of the Tan led ?ierarchy of levels inside our minds& /y !osition is rather similar to the view!oint !ut forth by the neuroscientist ;o er S!erry in his e(cellent article -/ind, 0rain, and ?umanist Galues-, from which I quote a little here,
In my own hy!othetical brain model, conscious awareness does et re!resentation as a very real causal a ent and rates an im!ortant !lace in the causal sequence and chain of control in brain events, in which it a!!ears as an active, o!erational force&&& To !ut it very sim!ly, it comes down to the issue of who !ushes whom around in the !o!ulation of causal forces that occu!y the cranium& It is a matter, in other words, of strai htenin out the !eck-order hierarchy amon intracranial control a ents& There e(ists within the cranium a whole world of diverse causal forces= what is more, there are forces within forces within forces, as in no other cubic half-foot

of universe that we know&&& To make a lon story short, if one kee!s climbin u!ward in the chain of command within the brain, one finds at the very to! those over-all or ani<ational forces and dynamic !ro!erties of the lar e !atterns of cerebral e(citation that are correlated with mental states or !sychic activity&&& 8ear the a!e( of this command system in the brain&&& we find ideas& /an over the chim!an<ee has ideas and ideals& In the brain model !ro!osed here, the causal !otency of an idea, or an ideal, becomes .ust as real as that of a molecule, a cell, or a nerve im!ulse& Ideas cause ideas and hel! evolve new ideas& They interact with each other and with other mental forces in the same brain, in nei hborin brains, and, thanks to lobal communication, in far distant, forei n brains& )nd they also interact with the e(ternal surroundin s to !roduce in toto a burstwise advance in evolution that is far beyond anythin to hit the evolutionary scene yet, includin the emer ence of the livin cell&C

There is a famous breach between two lan ua es of discourse, the sub.ective lan ua e and the ob.ective lan ua e& 'or instance, the -sub.ective- sensation of redness, and the -ob.ective- wavelen th of red li ht& To many !eo!le, these seem to be forever irreconcilable& I don1t think so& 8o more than the two views of Escher1s .rawing ,ands are irreconcilable from -in the system-, where the hands draw each other, and from outside, where Escher draws it all& The sub.ective feelin of redness comes from the vorte( of self-!erce!tion in the brain= the ob.ective wavelen th is how you see thin s when you ste! back, outside of the system& Thou h no one of us will ever be able to ste! back far enou h to see the -bi !icture-, we shouldn1t for et that it e(ists& 3e should remember that !hysical law is what makes it all ha!!enway, way down in neural nooks and crannies which are too remote for us to reach with our hi h-level intros!ective !robes&

The Self-Symbol and +ree >ill

In 7ha!ter TII, it was su ested that what we call free will is a result of the interaction between the self-symbol 4or subsystem5, and the other symbols in the brain& If we take the idea that symbols are the hi h-level entities to which meanin s should be attached, then we can make a stab at e(!lainin the relationshi! between symbols, the self-symbol, and free will& Ene way to ain some !ers!ective on the free-will question is to re!lace it by what I believe is an equivalent question, but one which involves less loaded terms& Instead of askin , -:oes system T have free will?- we ask, -:oes system T make choices?- 0y carefully ro!in for what we really mean when we choose to describe a systemmechanical or biolo icalas bein ca!able of makin -choices-, I think we can shed much li ht on free will it will be hel!ful to o over a few different systems which, under various circumstances, we mi ht feel tem!ted to describe as -makin choices-& 'rom these e(am!les we can ain some !ers!ective on what we really mean by the !hrase& Aet us take the followin systems as !aradi ms, a marble rollin down a bum!y hill= a !ocket calculator findin successive di its in the decimal e(!ansion of the square root of B= a so!histicated !ro ram which !lays a mean ame of chess= a robot in a T-ma<e 4a ma<e with but a sin le fork, on one side of which there is a reward5= and a human bein confrontin a com!le( dilemma&

'irst, what about that marble rollin down a hill? :oes it make choices? I think we would unanimously say that it doesn1t, even thou h none of us could !redict its !ath for even a very short distance& 3e feel that it couldn1t have one any other way than it did, and that it was .ust bein shoved alon by the relentless laws of nature& In our chunked mental !hysics, of course, we can visuali<e many different -!ossible- !athways for the marble, and we see it followin only one of them in the real world& En some level of our minds, therefore, we can1t hel! feelin the marble has -chosen- a sin le !athway out of those myriad mental ones= but on some other level of our minds, we have an instinctive understandin that the mental !hysics is only an aid in our internal modelin of the world, and that the mechanisms which make the real !hysical sequences of events ha!!en do not require nature to o throu h an analo ous !rocess of first manufacturin variants in some hy!othetical universe 4the -brain of 6od-5 and then choosin between them& So we shall not bestow the desi nation -choice- u!on this !rocessalthou h we reco ni<e that it is often !ra matically useful to use the word in cases like this, because of its evocative !ower& 8ow what about the calculator !ro rammed to find the di its of the square root of B? 3hat about the chess !ro ram? ?ere, we mi ht say that we are .ust dealin with -fancy marbles-, rollin down -fancy hills-& In fact, the ar uments for no choice-makin here are, if anythin , stron er than in the case of a marble& 'or if you attem!t to re!eat the marble e(!eriment, you will undoubtedly witness a totally different !athway bein traced down the hill, whereas if you rerun the square-root-of-B !ro ram, you will et the same results time after time& The marble seems to -choose- a different !ath each time, no matter how accurately you try to re!roduce the conditions of its ori inal descent, whereas the !ro ram runs down !recisely the same channels each time& 8ow in the case of fancy chess !ro rams, there are various !ossibilities& If you !lay a ame a ainst certain !ro rams, and then start a second ame with the same moves as you made the first time, these !ro rams will .ust move e(actly as they did before, without any a!!earance of havin learned anythin or havin any desire for variety& There are other !ro rams which have randomi<in devices that will ive some variety but not out of any dee! desire& Such !ro rams could be reset with the internal random number enerator as it was the first time, and once a ain, the same ame would ensue& Then there are other !ro rams which do learn from their mistakes, and chan e their strate y de!endin on the outcome of a ame& Such !ro rams would not !lay the same ame twice in a row& Ef course, you could also turn the clock back by wi!in out all the chan es in the memory which re!resent learnin , .ust as you could reset the random number enerator, but that hardly seems like a friendly thin to do& 0esides, is there any reason to sus!ect that you would be able to chan e any of )our own !ast decisions if every last detailand that includes your brain, of coursewere reset to the way it was the first time around? 0ut let us return to the question of whether -choice- is an a!!licable term here& If !ro rams are .ust -fancy marbles rollin down fancy hills-, do they make choices, or not? Ef course the answer must be a sub.ective one, but I would say that !retty much the same considerations a!!ly here as to the marble& ?owever, I would have to add that the a!!eal of usin the word -choice-, even if it is only a convenient and evocative shorthand,

becomes quite stron & The fact that a chess !ro ram looks ahead down the various !ossible bifurcatin !aths, quite unlike a rollin marble, makes it seem much more like an animate bein than a square-root-of-B !ro ram& ?owever, there is still no dee! selfawareness hereand no sense of free will& 8ow let us o on to ima ine a robot which has a re!ertoire of symbols& This robot is !laced in a T-ma<e& ?owever, instead of oin for the reward, it is !re!ro rammed to o left whenever the ne(t di it of the square root of B is even, and to o ri ht whenever it is odd& 8ow this robot is ca!able of modelin the situation in its symbols, so it can watch itself makin choices& Each time the T is a!!roached, if you were to address to the robot the question, -:o you know which way you1re oin to turn this time?- it would have to answer, -8o-& Then in order to !ro ress, it would activate its -decider- subroutine, which calculates the ne(t di it of the square root of B, and the decision is taken& ?owever, the internal mechanism of the decider is unknown to the robotit is re!resented in the robot1s symbols merely as a black bo( which !uts out -left-1s and -ri ht-1s by some mysterious and seemin ly random rule& Fnless the robot1s symbols are ca!able of !ickin u! the hidden heartbeat of the square root of B, beatin in the A1s and ;1s, it will stay baffled by the -choices- which it is makin & 8ow does this robot make choices? Put yourself in that !osition& If you were tra!!ed inside a marble rollin down a hill and were !owerless to affect its !ath, yet could observe it with all your human intellect, would you feel that the marble1s !ath involved choices? Ef course not& Fnless your mind is affecting the outcome, it makes no difference that the symbols are !resent& So now we make a modification in our robot, we allow its symbolsincludin its self-symbolto affect the decision that is taken& 8ow here is an e(am!le of a !ro ram runnin fully under !hysical law, which seems to et much more dee!ly at the essence of choice than the !revious e(am!les did& 3hen the robot1s own chunked conce!t of itself enters the scene, we be in to identify with the robot, for it sounds like the kind of thin we do& It is no lon er like the calculation of the square root of B, where no symbols seem to be monitorin the decisions taken& To be sure, if we were to look at the robot1s !ro ram on a very local level, it would look quite like the square-root !ro ram& Ste! after ste! is e(ecuted, and in the end -left- or -ri ht- is the out!ut& 0ut on a hi h level we can see the fact that symbols are bein used to model the situation and to affect the decision& That radically affects our way of thinkin about the !ro ram& )t this sta e, meaning has entered this !icturethe same kind of meanin as we mani!ulate with our own minds&

A G#del 4orte- >here All Le"els Cross

8ow if some outside a ent su ests 1A1 as the ne(t choice to the robot, the su estion will be !icked u! and channeled into the swirlin mass of interactin symbols& There, it will be sucked ine(orably into interaction with the self-symbol, like a rowboat bein !ulled into a whirl!ool& That is the vorte( of the system, where all levels cross& ?ere, the 1A1 encounters a Tan led ?ierarchy of symbols and is !assed u! and down the levels& The self-symbol is inca!able of monitorin all its internal !rocesses, and so when the actual decision emer es1A1 or 1;1 or somethin outside the systemthe system will not be able to say where it came from& Fnlike a standard chess !ro ram, which does not monitor

itself and consequently has no ideas about where its moves come from, this !ro ram does monitor itself and does have ideas about its ideasbut it cannot monitor its own !rocesses in com!lete detail, and therefore has a sort of intuitive sense of its workin s, without full understandin & 'rom this balance between self-knowled e and self-i norance comes the feelin of free will& Think, for instance, of a writer who is tryin to convey certain ideas which to him are contained in mental ima es& ?e isn1t quite sure how those ima es fit to ether in his mind, and he e(!eriments around, e(!ressin thin s first one way and then another, and finally settles on some version& 0ut does he know where it all came from? Enly in a va ue sense& /uch of the source, like an iceber , is dee! underwater, unseenand he knows that& Er think of a music com!osition !ro ram, somethin we discussed earlier, askin when we would feel comfortable in callin it the com!oser rather than the tool of a human com!oser& Probably we would feel comfortable when self-knowled e in terms

73G5=E BDF! Print 6allery, () '! 8! Escher lithograph, BIGJ/&

of symbols e(ists inside the !ro ram, and when the !ro ram has this delicate balance between self-knowled e and self-i norance& It is irrelevant whether the system is runnin deterministically= what makes us call it a -choice maker- is whether we can identif) with a high%le#el description of the process which takes place when the program runs & En a low 4machine lan ua e5 level, the !ro ram looks like any other !ro ram= on a hi h 4chunked5 level, qualities such as -will-, -intuition-, -creativity-, and -consciousnesscan emer e& The im!ortant idea is that this -vorte(- of self is res!onsible for the tan ledness, for the 6>delian-ness, of the mental !rocesses& Peo!le have said to me on occasion, -This stuff with self-reference and so on is very amusin and en.oyable, but do you really think there is anythin serious to it?- I certainly do& I think it will eventually turn out to be at the core of )I, and the focus of all attem!ts to understand how human minds work& )nd that is why 6>del is so dee!ly woven into the fabric of my book&

An Escher 4orte- >here All Le"els Cross

) strikin ly beautiful, and yet at the same time disturbin ly rotesque, illustration of the cyclonic -eye- of a Tan led ?ierarchy is iven to us by Escher in his &rint Galler) 4'i & "CB5& 3hat we see is a !icture allery where a youn man is standin , lookin at a !icture of a shi! in the harbor of a small town, !erha!s a /altese town, to uess from the architecture, with its little turrets, occasional cu!olas, and flat stone roofs, u!on one of which sits a boy, rela(in in the heat, while two floors below him a woman!erha!s his mother a<es out of the window from her a!artment which sits directly above a !icture allery where a youn man is standin , lookin at a !icture of a shi! in the harbor of a small town, !erha!s a /altese town3hat2? 3e are back on the same level as we be an, thou h all lo ic dictates that we cannot be& Aet us draw a dia ram of what we see 4'i & "C%5&

73G5=E BDH! *(stract diagram of '! 8! Escher's &rint Galler)&

3hat this dia ram shows is three kinds of -in-ness-& The allery is ph)sicall) in the town 4-inclusion-5= the town is artisticall) in the !icture 4-de!iction-5= the !icture is mentall) in the !erson 4-re!resentation-5& 8ow while this dia ram may seem satisfyin , in fact it is

arbitrary, for the number of levels shown is quite arbitrary& Aook below at another way of re!resentin the to! half alone 4'i & "CC5&

73G5=E BDD! * collapsed #ersion of the pre#ious figure!

3e have eliminated the -town- level= conce!tually it was useful, but can .ust as well be done without& 'i ure "CC looks .ust like the dia ram for .rawing ,ands, a Stran e Aoo! of two ste!s& The division markers are arbitrary, even if they seem natural to our minds& This can be further accentuated by showin even more -colla!sed- schematic dia rams of &rint Galler), such as that in 'i ure "CL&

73G5=E BDG! 7urther collapse of 7igure BDH&

This e(hibits the !arado( of the !icture in the starkest terms& 8owif the !icture is -inside itself-, then is the youn man also inside himself? This question is answered in 'i ure "CD&

73G5=E BDJ! *nother wa) of collapsing 7igure BDH!

Thus, we see the youn man -inside himself-, in a funny sense which is made u! of com!oundin three distinct senses of -in-&

This dia ram reminds us of the E!imenides !arado( with its one-ste! selfreference, while the two-ste! dia ram resembles the sentence !air each of which refers to the other& 3e cannot make the loo! any ti hter, but we can o!en it wider, by choosin to insert any number of intermediate levels, such as -!icture frame-, -arcade-, and -buildin -& If we do so, we will have many-ste! Stran e Aoo!s, whose dia rams are isomor!hic to those of >aterfall 4'i & L5 or *scending and .escending 4'i & D5& The number of levels is determined by what we feel is -natural-, which may vary accordin to conte(t, !ur!ose, or frame of mind& The 7entral Tma!s:o , 7rab, Sloth, and Pi!e can all be seen as involvin three-ste! Stran e Aoo!s= alternatively, they can all be colla!sed into two- or one-ste! loo!s=& then a ain, they can be e(!anded out into multista e loo!s& 3here one !erceives the levels is a matter of intuition and esthetic !reference& 8ow are we, the observers of &rint Galler), also sucked into ourselves by virtue of lookin at it? 8ot really& 3e mana e to esca!e that !articular vorte( by bein outside of the system& )nd when we look at the !icture, we see thin s which the youn man can certainly not see, such as Escher1s Si nature, -/7E-, in the central -blemish-& Thou h the blemish seems like a defect, !erha!s the defect lies in our e(!ectations, for in fact Escher could not have com!leted that !ortion of the !icture without bein inconsistent with the rules by which he was drawin the !icture& That center of the whorl isand must beincom!lete& Escher could have made it arbitrarily small, but he could not have otten rid of it& Thus we, on the outside, can know that &rint Galler) is essentially incom!letea fact which the youn man, on the inside, can never know& Escher has thus iven a !ictorial !arable for 6>del1s Incom!leteness Theorem& )nd that is why the strands of 6>del and Escher are so dee!ly interwoven in my book&

A Bach 4orte- >here All Le"els Cross

Ene cannot hel! bein reminded, when one looks at the dia rams of Stran e Aoo!s, of the Endlessly ;isin 7anon from the /usical Efferin & ) dia ram of it would consist of si( ste!s, as is shown in 'i ure "C$& It is too bad that when it returns to 7, it is an octave hi her rather than at the e(act ori inal !itch& )stonishin ly enou h, it is !ossible to arran e for it to return e(actly to the startin !itch, by usin what are called She!ard tones, after the !sycholo ist ;o er She!ard, who discovered the idea& The !rinci!le of a She!ard-tone scale is shown in 'i ure "C*& In words, it is this, you !lay !arallel scales in several different octave ran es& Each note is wei hted inde!endently, and as the notes rise, the wei hts shift& Hou make the to! octave radually fade out, while at the same time you are radually brin in in the bottom octave& Just at the moment you would ordinarily be one octave hi her, the wei hts have shifted !recisely so as to re!roduce the startin !itch&&& Thus you can o -u! and u! forever-, never ettin any hi her2 Hou can try it at your !iano& It works even better if the !itches can be synthesi<ed accurately under com!uter control& Then the illusion is bewilderin ly stron & This wonderful musical discovery allows the Endlessly ;isin 7anon to be !layed in such a way that it .oins back onto itself after oin -u!- an octave& This idea, which Scott @im and I conceived .ointly, has been reali<ed on ta!e, usin a com!uter music

system& The effect is very subtlebut very real& It is quite interestin that 0ach himself was a!!arently aware, in some sense, of such scales, for in his music one can occasionally find !assa es which rou hly e(!loit the eneral !rinci!le of She!ard tones for instance, about halfway throu h the 'antasia from the 'antasia and 'u ue in 6 /inor, for or an& In his book ?! -! Bach's 'usical :ffering, ?ans Theodore :avid writes,
Throu hout the 'usical :ffering, the reader, !erformer, or listener is to search for the ;oyal theme in all its forms& The entire work, therefore, is a ricercar in the ori inal, literal sense of the word&L

I think this is true= one cannot look dee!ly enou h into the 'usical :ffering& There is always more after one thinks one knows everythin & 'or instance& towards the very end of the -i9%&art =icercar, the one he declined to im!rovise, 0ach slyly hid his own name, s!lit between two of the u!!er voices& Thin s are oin on on many levels in the 'usical :ffering! There are tricks with notes and letters= there are in enious variations on the @in 1s Theme= there are ori inal kinds of canons= there are e(traordinarily com!le( fu ues= there is beauty and e(treme de!th of emotion= even an e(ultation in the manyleveledness of the work comes throu h& The 'usical :ffering is a fu ue of fu ues, a Tan led ?ierarchy like those of Escher and 6>del, an intellectual construction which reminds me, in ways I cannot e(!ress, of the beautiful many-voiced fu ue of the human mind& )nd that is why in my book the three strands of 6>del, Escher, and 0ach are woven into an Eternal 6olden 0raid&

73G5=E BDC! The he9agonal modulation scheme of Bach's Endlessl) =ising 8anon forms a true closed loop when -hepard tones are used!

73G5=E BDE! Two complete c)cles of a -hephard tone scale, notated for piano! The loudness of each note is proportional to its area, Must as the top #oice fades out, a new (ottom #oice fee(l) enters! U&rinted () .onald Bo)d`s program "-'5T"!V

Si1-Part 0icercar
*chilles has (rought his cello to the 8ra('s residence, to engage in an e#ening of cham(er music with the 8ra( and Tortoise! ,e has (een shown into the music room () his host the 8ra(, who is momentaril) a(sent, ha#ing gone to meet their mutual friend the Tortoise at the door! The room is filled with all sorts of electronic e"uipmentOphonographs in #arious states of arra) and disarra), tele#ision screens attached to t)pewriters, and other "uite impro(a(le%looking pieces of apparatus! Nestled amongst all this high%powered gadgetr) sits a hum(le radio! -ince the radio is the onl) thing in the room which *chilles knows how to use, he walks o#er to it, and, a little furti#el), flicks the dial and finds he has tuned into a panel discussion () si9 learned scholars on free will and determinism! ,e listens (riefl) and then, a little scornfull), flicks it off! *chilles, I can et alon very well without such a !ro ram& )fter all, it1s clear to anyone who1s ever thou ht about it thatI mean, it1s not a very difficult matter to resolve, once you understand howor rather, conce!tually, one can clear u! the whole thin by thinkin of, or at least ima inin a situation where&&& ?mmm&&& I thou ht it was quite clear in my mind& /aybe I could benefit from listenin to that show, after all&&& Enter the Tortoise, carr)ing his #iolin!/ 3ell, well, if it isn1t our fiddler& ?ave you been !racticin faithfully this week, /r& T? I myself have been !layin the cello !art in the Trio Sonata from the 'usical :ffering for at least two hours a day& It1s a strict re imen, but it !ays off& Tortoise, I can et alon very well without such a !ro ram& I find that a moment here, a moment there kee!s me fit for fiddlin & *chilles, Eh, lucky you& I wish it came so easily to me& 3ell, where is our host? Tortoise, I think he1s .ust one to fetch his flute& ?ere he comes& Enter the 8ra(, carr)ing his flute!/ *chilles, Eh, /r& 7rab, in my ardent !racticin of the Trio Sonata this !ast week, all sorts of ima es bubbled into my mind, .olly obblin bumblebees, melancholy bu<<in turkeys, and a raft of others& Isn1t it wonderful, what !ower music has? 8ra(, I can et alon very well without such a !ro ram& To my mind, )chilles, there is no music !urer than the /usical Efferin & Tortoise, Hou can1t be serious, )chilles& The /usical Efferin isn1t !ro rammatic music2 *chilles, 3ell, I like animals, even if you two stuffy ones disa!!rove& 8ra(, I don1t think we are so stuffy, )chilles& Aet1s .ust say that you hear music in your own s!ecial way& Tortoise, Shall we sit down and !lay? 8ra(, I was ho!in that a !ianist friend of mine would turn u! and !lay continuo& I1ve been wantin you to meet him, )chilles, for a lon time& Fnfortunately, it a!!ears that

he may not make it& So let1s .ust o ahead with the three of us& That1s !lenty for a trio sonata& *chilles, 0efore we start, I .ust was wonderin , /r& 7rabwhat are all these !ieces of equi!ment, which you have in here? 8ra(, 3ell, mostly they are .ust odds and endsbits and !ieces of old broken !hono ra!hs& Enly a few souvenirs 4nervously ta!!in the buttons5, a few souvenirs ofof the T7-battles in which I have distin uished myself& Those keyboards attached to television screens, however, are my new toys& I have fifteen of them around here& They are a new kind of com!uter, a very small, very fle(ible ty!e of com!uterquite an advance over the !revious ty!es available& 'ew others seem to be quite as enthusiastic about them as I am, but I have faith that they will catch on in time& *chilles, :o they have a s!ecial name? 8ra(, Hes= they are called -smart-stu!ids-, since they are so fle(ible, and have the !otential to be either smart or stu!id, de!endin on how skillfully they are instructed& *chilles, :o you mean you think they could actually become smart like, say, a human bein ? 8ra(, I would not balk at sayin so!rovided, of course, that someone sufficiently versed in the art of instructin smart-stu!ids would make the effort& Sadly, I am not !ersonally acquainted with anyone who is a true virtuoso& To be sure, there is one e(!ert abroad in the land, an individual of reat renownand nothin would !lease me more than a visit by him, so that I could a!!reciate what true skill on the smartstu!id is= but he has never come, and I wonder if I shall ever have that !leasure& Tortoise, It would be very interestin to !lay chess a ainst a well-instructed smart-stu!id& 8ra(, )n e(tremely intri uin idea& That would be a wonderful mark of skill, to !ro ram a smart-stu!id to !lay a ood ame of chess& Even more interestin but incredibly com!licatedwould be to instruct a smart-stu!id sufficiently that it could hold its own in a conversation& It mi ht ive the im!ression that it was .ust another !erson2 *chilles, 7urious that this should come u!, for I .ust heard a snatch of a discussion on free will and determinism, and it set me to thinkin about such questions once more& I don1t mind admittin that, as I !ondered the idea, my thou hts ot more and more tan led, and in the end I really didn1t know what I thou ht& 0ut this idea of a smartstu!id that could converse with you&&& it bo les the mind& I mean, what would the smart-stu!id itself say, if you asked it for its o!inion on the free-will question? I was .ust wonderin if the two of you, who know so much about these thin s, wouldn1t indul e me by e(!lainin the issue, as you see it, to me& 8ra(, )chilles, you can1t ima ine how a!!ro!riate your question is& I only wish my !ianist friend were here, because I know you1d be intri ued to hear what he could tell you on the sub.ect& In his absence, I1d like to tell you a statement in a :ialo ue at the end of a book I came across recently& *chilles, 8ot 8opper, -il#er, Gold: an 3ndestructi(le 'etallic *llo)? 8ra(, 8o, as I recall, it was entitled Giraffes, Elephants, Ba(oons: an E"uatorial Grasslands Bestiar)Oor somethin like that& In any case, towards the end of the aforementioned :ialo ue, a certain e(ceedin ly droll character quotes /arvin /insky on the question of free will& Shortly thereafter, while interactin with two

other !ersona es, this droll character quotes /insky further on musical im!rovisation, the com!uter lan ua e AISP, and 6>del1s Theoremand et thisall without ivin one whit of credit to /insky2 *chilles, Eh, for shame2 8ra(, I must admit that earlier in the :ialo ue, he hints that he 3IAA quote /insky towards the end= so !erha!s it1s for ivable& *chilles, It sounds that way to me& )nyway, I1m an(ious to hear the /inskian !ronouncement on the free will question& 8ra(, )h, yes&&& /arvin /insky said, -3hen intelli ent machines are constructed, we should not be sur!rised to find them as confused and as stubborn as men in their convictions about mind-matter, consciousness, free will, and the like&*chilles, I like that2 Muite a funny thou ht& )n automaton thinkin it had free will2 That1s almost as silly as me thinkin I didn1t have free will2 Tortoise, I su!!ose it never occurred to you, )chilles, that the three of usyou, myself, and /r& 7rabmi ht all be characters in a :ialo ue, !erha!s even one similar to the one /r& 7rab .ust mentioned& *chilles, Eh, it1s occurred to me, of course& I su!!ose such fancies occur to every normal !erson at one time or another& Tortoise, )nd the )nteater, the Sloth, 9eno, even 6E:we mi ht all be characters in a series of :ialo ues in a book& *chilles, Sure, we mi ht& )nd the )uthor mi ht .ust come in and !lay the !iano, too& 8ra(, That1s .ust what I had ho!ed& 0ut he1s always late& *chilles, 3hose le do you think you1re !ullin ? I know I1m not bein controlled in any way by another mentality2 I1ve ot my own thou hts, I e(!ress myself as I wishyou can1t deny that2 Tortoise, 8obody denied any of that, )chilles& 0ut all of what you say is !erfectly consistent with your bein a character in a :ialo ue& 8ra(, The *chilles, 0utbutno2 Perha!s /r& 71s article and my rebuttal have both been mechanically determined, but this I refuse to believe& I can acce!t !hysical determinism, but I cannot acce!t the idea that I am but a fi ment inside of someone else1s mentality2 Tortoise, It doesn1t really matter whether you have a hardware brain, )chilles& Hour will can be equally free, if your brain is .ust a !iece of software inside someone else1s hardware brain& )nd their brain, too, may be software in a yet hi her brain&&& *chilles, 3hat an absurd idea2 )nd yet, I must admit, I do en.oy tryin to find the cleverly concealed holes in your so!histry, so o ahead& Try to convince me& I1m ame& Tortoise, :id it ever strike you, )chilles, that you kee! somewhat unusual com!any? *chilles, Ef course& Hou are very eccentric 4I know you won1t mind my sayin so5, and even /r& 7rab here is a weensy bit eccentric& 4Pardon me, /r& 7rab&5 8ra(, Eh, don1t worry about offendin me& Tortoise, 0ut )chilles, you1ve overlooked one of the most salient features of your acquaintances&

*chilles, 3hich is&&& ? Tortoise, That we1re animals2 *chilles, 3ell, welltrue enou h& Hou have such a keen mind& I would never have thou ht of formulatin the facts so concisely& Tortoise, Isn1t that evidence enou h? ?ow many !eo!le do you know who s!end their time with talkin Tortoises, and talkin 7rabs? *chilles, I must admit, a talkin 7rab is 8ra(, an anomaly, of course& *chilles, E(actly= it is a bit of an anomalybut it has !recedents& It has occurred in literature& Tortoise, Preciselyin literature& 0ut where in real life? *chilles, 8ow that you mention it, I can1t quite say& I1ll have to ive it some thou ht& 0ut that1s not enou h to convince me that I1m a character in a :ialo ue& :o you have any other ar uments? Tortoise, :o you remember one day when you and I met in the !ark, seemin ly at random? *chilles, The day we discussed crab canons by Escher and 0ach? Tortoise, The very one2 *chilles, )nd /r& 7rab, as I recall, turned u! somewhere towards the middle of our conversation and babbled somethin funny and then left& 8ra(, 8ot .ust -somewhere towards the middle-, )chilles& ET)7TAH in the middle& *chilles, Eh, all ri ht, then& Tortoise, :o you reali<e that your lines were the same as my lines in that conversation e(ce!t in reverse order? ) few words were chan ed here and there, but in essence there was a time symmetry to our encounter& *chilles, 0i :eal2 It was .ust some sort of trickery& Probably all done with mirrors& Tortoise, 8o trickery& )chilles, and no mirrors, .ust the work of an assiduous )uthor& *chilles, Eh, well, it1s all the same to me& Tortoise, 'iddle It makes a bi difference, you know& *chilles, Say, somethin about this conversation strikes me as familiar& ?aven1t I heard some of those lines somewhere before? Tortoise, Hou said it, )chilles& 8ra(, Perha!s those lines occurred at random in the !ark one day, )chilles& :o you recall how your conversation with /r& T ran that day? *chilles, Ga uely& ?e said -6ood day, /r& )- at the be innin , and at the end, I said, -6ood day, /r& T-& Is that ri ht 8ra(, I .ust ha!!en to have a transcri!t ri ht here&&& ,e fishes around in his music case, whips out a sheet, and hands it to *chilles! *s *chilles reads it, he (egins to s"uirm and fidget noticea(l)!/ *chilles, This is very stran e& Gery, very stran e&&& )ll of a sudden, I feel sort ofweird& It1s as if somebody had actually !lanned out that whole set of statements in advance,

worked them out on !a!er or somethin & )s if some )uthor had had a whole a enda and worked from it in detail in !lannin all those statements I made that day& *t that moment, the door (ursts open! Enter the *uthor, carr)ing a giant manuscript!/ *uthor, I can et alon very well without such a !ro ram& Hou see, once my characters are formed, they seem to have lives of their own, and I need to e(ert very little effort in !lannin their lines& 8ra(, Eh, here you are2 I thou ht you1d never arrive2 *uthor, Sorry to be so late& I followed the wron road and wound u! very far away& 0ut somehow I made it back& 6ood to see you a ain, /r& T and /r& 7& )nd )chilles, I1m es!ecially lad to see you& *chilles, 3ho are you? I1ve never seen you before& *uthor, I am :ou las ?ofstadter!lease call me :ou and I1m !resently finishin u! a book called Gdel, Escher, Bach: an Eternal Golden Braid& It is the book in which the three of you are characters& *chilles, Pleased to meet you& /y name is )chilles, and *uthor, 8o need to introduce yourself, )chilles, since I already know you quite well& *chilles, 3eird, weird& 8ra(, ?e1s the one I was sayin mi ht dro! in and !lay continuo with us& *uthor, I1ve been !layin the 'usical :ffering a little bit on my !iano at home, and I can try to blunder my way throu h the Trio Sonata !rovidin you1ll overlook my many wron notes& Tortoise, Eh, we1re very tolerant around here, bein only amateurs ourselves& *uthor, I ho!e you don1t mind, )chilles, but I1m to blame for the tact that you and /r& Tortoise said the same thin s, but in reverse order, that day in the !ark& 8ra(, :on1t for et me2 I was there, too ri ht in the middle, !uttin in my two bits1 worth2 *uthor, Ef course2 Hou were the 7rab in the 8ra( 8anon& *chilles, So you are sayin you control my utterances? That my brain is a software subsystem of yours? *uthor, Hou can !ut it that way if you want, )chilles& *chilles, Su!!ose I were to write dialo ues& 3ho would the author of them be? Hou, or me? *uthor, Hou, of course& )t least in the fictitious world which you inhabit, you1d et credit for them& *chilles, 'ictitious? I don1t see anythin fictitious about it2 *uthor, 3hereas in the world I inhabit, !erha!s the credit would be iven to me, althou h I am not sure if it would be !ro!er to do so& )nd then, whoever made me make you write your dialo ues would et credit in his world 4seen from which, /H world looks fictitious5& *chilles, That1s quite a bit to swallow& I never ima ined there could be a world above mine beforeand now you1re hintin that there could even be one above that& It1s like

walkin u! a familiar staircase, and .ust kee!in on oin further u! after you1ve reached the to!or what you1d always taken to be the to!2 8ra(, Er wakin u! from what you took to be real life, and findin out it too was .ust a dream& That could ha!!en over and over a ain, no tellin when it would sto!& *chilles, It1s most !er!le(in how the characters in my dreams have wills of their own, and act out !arts which are inde!endent of /H will& It1s as if my mind, when I1m dreamin , merely forms a sta e on which certain other or anisms act out their lives& )nd then, when I awake, they o away& I wonder where it is they o to&&& *uthor, They o to the same !lace as the hiccu!s o, when you et rid of them, Tumbolia& 0oth the hiccu!s and the dreamed bein s are software subor anisms which e(ist thanks to the biolo y of the outer host or anism& The host or anism serves as sta e to themor even as their universe& They !lay out their lives for a timebut when the host or anism makes a lar e chan e of statefor e(am!le, wakes u!then the subor anisms lose their coherency, and case e(istin as se!arate, identifiable units& *chilles, Is it like castles in the sand which vanish when a wave washes over them? *uthor, Gery much like that, )chilles& ?iccu!s, dream characters, and even :ialo ue characters disinte rate when their host or anism under oes certain critical chan es of state& Het, .ust like those sand castles you described, everythin which made them u! is still !resent& *chilles, I ob.ect to bein likened to a mere hiccu!2 *uthor, 0ut I am also com!arin you to a sand castle, )chilles& Is that not !oetic? 0esides, you may take comfort in the fact that if you are but a hiccu! in my brain, I myself am but a hiccu! in some hi her author1s brain& *chilles, 0ut I am such a !hysical creatureso obviously made of flesh and blood and hard bones& Hou can1t deny that2 *uthor, I can1t deny your sensation of it, but remember that dreamed bein s, althou h they are .ust software a!!aritions, have the same sensation, no less than you do& Tortoise, I say, enou h of this talk2 Aet us sit down and make music2 8ra(, ) fine ideaand now we have the added !leasure of the com!any of our )uthor, who will race our ears with his rendition of the bass line to the Trio Sonata, as harmoni<ed by 0ach1s !u!il @irnber er& ?ow fortunate are we2 +eads the author to one of his pianos!/ I ho!e you find the seat comfortable enou h& To ad.ust it, you 3n the (ackground there is heard a funn) soft oscillating sound!/ Tortoise, E(cuse me, but what was that stran e electronic ur le? 8ra(, Eh, .ust a noise from one of the smart-stu!ids& Such a noise enerally si nals the fact that a new notice has flashed onto the screen& Fsually the notices are .ust unim!ortant announcements comin from the main monitor !ro ram, which controls all the smart-stu!ids& 4>ith his flute in his hand, he walks o#er to a smart%stupid, and reads its screen! 3mmediatel) he turns to the assem(led musicians, and sa)s, with a kind of agitation,5 6entlemen, old 0a& 7h& is come& ,e la)s the flute aside!/ 3e must show him in immediately, of course& *chilles, Eld 0a& 7h&2 7ould it be that that celebrated im!roviser of yore has chosen to show u! toni ht?E;E?

Tortoise, Eld 0a& 7h&2 There1s only one !erson T?)T could meanthe renowned 0abba e, 7harles, Esq&, /&)&, '&;&S&, '&;&S&E&, '&;&)&S&, '& ST)T& S&, ?E8& /&;&A)&, /&7&P&S&, 7ommander of the Italian Erder of St& /aurice and St& Aa<arus, I8ST& I/P& 4)7):& /E;)A&5 P);IS 7E;;&, )7):& )/E;& );T& ET S7& 0ESTE8, ;E6& EE7E8& 0E;7SS&, P?HS& ?ISI& 8)T& 6E8EG&, )7):& ;E6& /E8)7&, ?)'8&, /)SSIA&, ET :IGIE8&, SE7IFS&, )7):& I/P&, ET ;E6& PET;EP&, 8E)P&, 0;FT&, P)T)G&, 6EE;6& 'AE;E8, AH87EI ;E/&, /7T&, P?IAE/)T?&, P);IS, SE7& 7E;;&, etc&and /ember of the E(tractors1 7lub& 7harles 0abba e is a venerable !ioneer of the art and science of com!utin & 3hat a rare !rivile e2 8ra(, ?is name is known far and wide, and I have lon ho!ed that he would ive us the honor of a visitbut this is a totally une(!ected sur!rise& *chilles, :oes he !lay a musical instrument? 8ra(, I have heard it said that in the !ast hundred years, he has rown ine(!licably fond of tom-toms, half!enny whistles, and sundry other street instruments& *chilles, In that case, !erha!s he mi ht .oin us in our musical evenin & *uthor, I su est that we ive him a ten-canon salute& Tortoise, ) !erformance of all the celebrated canons from the 'usical :ffering$ *uthor, Precisely& 8ra(, 7a!ital su estion2 Muick, )chilles, you draw u! a list of all ten of them, in the order of !erformance, and hand it to him as he comes in2 Before *chilles can mo#e, enter Ba((age, carr)ing a hurd)%gurd), and wearing a hea#) tra#eling coat and hat! ,e appears slightl) tra#el%wear) and dishe#eled!/ Ba((age, I can et alon very well without such a !ro ram& ;ela(= I 7an En.oy ;andom 7oncerts )nd ;ecitals& 8ra(, /r& 0abba e2 It is my dee!est !leasure to welcome you to -/adsto!-, my humble residence& I have been ardently desirous of makin your acquaintance for many years, and today my wish is at last fulfilled& Ba((age, Eh, /r& 7rab, I assure you that the honor is truly all mine, to meet someone so eminent in all the sciences as yourself, someone whose knowled e and skill in music are irre!roachable, and someone whose hos!itality e(ceeds all bounds& )nd I am sure that you e(!ect no less than the hi hest sartorial standards of your visitors= and yet I must confess that I cannot meet those most reasonable standards, bein in a state of casual attire as would not by any means befit a visitor to so eminent and e(cellent a 7rab as Hour 7rab& 8ra(, If I understand your most !raiseworthy soliloquy, most welcome uest, I take it that you1d like to chan e your clothes& Aet me then assure you that there could be no more fittin attire than yours for the circumstances which this evenin !revail= and I would beseech you to uncoat yourself and, if you do not ob.ect to the music-makin of the most rank amateurs, !lease acce!t a -'usical :ffering-, consistin of ten canons from Sebastian 0ach1s /usical Efferin , as a token of our admiration& Ba((age, I am most bewilderin ly !leased by your overkind rece!tion, /r& 7rab, and in utmost modesty do re!ly that there could be no dee!er ratitude than that which I

e(!erience for the offer of a !erformance of music iven to us by the illustrious Eld 0ach, that or anist and com!oser with no rival& 8ra(, 0ut nay2 I have a yet better idea, one which I trust mi ht meet with the a!!roval of my esteemed uest= and that is this, to ive you the o!!ortunity, /r& 0abba e, of bein amon the first to try out my newly delivered and as yet hardly tested -smartstu!ids-streamlined reali<ations, if you will, of the )nalytical En ine& Hour fame as a virtuoso !ro rammer of com!utin en ines has s!read far and wide, and has not failed to reach as far as /adsto!= and there could be for us no reater deli ht than the !rivile e of observin your skill as it mi ht be a!!lied to the new and challen in -smart-stu!ids-& Ba((age, Such an outstandin idea has not reached my ears for an eon& I welcome the challen e of tryin out your new -smart-stu!ids-, of which I have only the sli htest knowled e by means of hearsay& 8ra(, Then let us !roceed2 0ut e(cuse my oversi ht2 I should have introduced my uests to you& This is /r& Tortoise, this is )chilles, and the )uthor, :ou las ?ofstadter& Ba((age, Gery !leased to make your acquaintance, I1m sure& E#er)one walks o#er toward one of the smart%stupids, and Ba((age sits down and lets his fingers run o#er the ke)(oard!/ ) most !leasant touch& 8ra(, I am lad you like it& *ll at once, Ba((age deftl) massages the ke)(oard with graceful strokes, inputting one command after another! *fter a few seconds, he sits (ack, and in almost no time, the screen (egins filling with figures! 3n a flash, it is totall) co#ered with thousands of tin) digits, the first few of which go: "H!BDBGIFJGHGEICIHFHEDJFJD!!! "/ *chilles, Pi2 8ra(, E(quisite2 I1d never ima ined that one could calculate so many di its of !i so quickly, and with so tiny an al orithm& Ba((age, The credit belon s e(clusively to the smart-stu!id& /y role was merely to see what was already !otentially !resent in it, and to e(!loit its instruction set in a moderately efficient manner& Truly, anyone who !ractices can do such tricks& Tortoise, :o you do any ra!hics, /r& 0abba e? Ba((age, I can try& 8ra(, 3onderful2 ?ere, let me take you to another one of my smart-stu!ids& I want you to try them all2 *nd so Ba((age is led o#er to another of the man) smart%stupids, and takes a seat! :nce again, his fingers attack the ke)(oard of the smart%stupid, and in half a trice, there appear on the screen an enormous num(er of lines, swinging a(out on the screen!/

8ra(, ?ow harmonious and !leasin these swirlin sha!es are, as they constantly collide and interfere with each other2 *uthor, )nd they never re!eat e(actly, or even resemble ones which have come before& It seems an ine(haustible mine of beauty& Tortoise, Some are sim!le !atterns which enchant the eye= others are indescribably com!le( convolutions which bo le and yet simultaneously deli ht the mind& 8ra(, 3ere you aware, /r& 0abba e, that these are color screens? Ba((age, Eh, are they? In that case, I can do rather more with this al orithm& Just a moment& 4T)pes in a few new commands, then pushes two ke)s down at once and holds them&5 )s I release these two keys, the dis!lay will include all the colors of the s!ectrum& 4=eleases them&5 *chilles, Eh, what s!ectacular color2 Some of the !atterns look like they1re .um!in out at me now2 Tortoise, I think that is because they are all rowin in si<e& Ba((age, That is intentional& )s the fi ures row, so may the 7rab1s fortune& 8ra(, Thank you, /r& 0abba e& 3ords fail to convey my admiration for your !erformance2 8ever has anyone done anythin com!arable on my smart-stu!ids& 3hy, you !lay the smart-stu!ids as if they were musical instruments, /r& 0abba e2 Ba((age, I am afraid that any music I mi ht make would be too harsh for the ears of such a entle 7rab as your 7rab& )lthou h I have lately become enamored of the sweet sounds of the hurdy- urdy, I am well aware of the ratin effect they can have u!on others& 8ra(, Then, by all means, continue on the smart-stu!ids2 In fact, I have a new ideaa marvelously e(citin idea2 Ba((age, 3hat is it? 8ra(, I have recently invented a Theme, and it only now occurred to me that, of all !eo!le, you, /r& 0abba e, are the most suited to reali<e the !otential of my Theme2 )re you by any chance familiar with the thou hts of the !hiloso!her Aa /ettrie? Ba((age, The name sounds familiar= kindly refresh my memory& 8ra(, ?e was a 7ham!ion of /aterialism& In "$C$, while at the court of 'rederick the 6reat, he wrote a book called +'homme machine& In it, he talks about man as a machine, es!ecially his mental faculties& 8ow my Theme comes from my !onderin s about the obverse side of the coin, what about imbuin a machine with human mental faculties, such as intelli ence? Ba((age, I have iven such matters some thou ht from time to time, but I have never had the !ro!er hardware to take u! the challen e& This is indeed a felicitous su estion, /r& 7rab, and I would en.oy nothin more than workin with your e(cellent Theme& Tell medid you have any s!ecific kind of intelli ence in mind? 8ra(, )n idle thou ht which had crossed my mind was to instruct it in such a manner as to !lay a reasonable ame of chess& Ba((age, 3hat an ori inal su estion2 )nd chess ha!!ens to be my favorite !astime& I can tell that you have a broad acquaintance with com!utin machinery, and are no mere amateur&

8ra(, I know very little, in fact& /y stron est !oint is sim!ly that I seem to be able to formulate Themes whose !otential for bein develo!ed is beyond my own ca!acity& )nd this Theme is my favorite& Ba((age, I shall be most deli hted to try to reali<e, in some modest fashion, your su estion of teachin chess to a smart-stu!id& )fter all, to obey Hour 7rabness1 command is my most humble duty& 4-o sa)ing, he shifts to another of the 8ra('s man) smart%stupids, and (egins to t)pe awa)&5 *chilles, 3hy, his hands move so fluidly that they almost make music2 Ba((age 4winding up his performance with a particularl) graceful flourish 5, I really haven1t had any chance, of course, to check it out, but !erha!s this will allow you at least to sam!le the idea of !layin chess a ainst a smart-stu!ideven if the latter of its two names seems more a!t in this case, due to my own insufficiencies in the art of instructin smart-stu!ids& ,e )ields his seat to the 8ra(! :n the screen appears a (eautiful displa) of a chess (oard with elegant wooden pieces, as it would look from >hite's side! Ba((age hits a (utton, and the (oard rotates, stopping when it appears as seen from the perspecti#e of Black!/ 8ra(, ?mm&&& very ele ant, I must say& :o I !lay 0lack or 3hite? Ba((age, 3hichever you wish .ust si nal your choice by ty!in -3hite- or -0lack-& )nd then, your moves can be entered in any standard chess notation& The smart-stu!id1s moves, of course, will a!!ear on the board& Incidentally, I made the !ro ram in such a way that it can !lay three o!!onents simultaneously, so that if two more of you wish to !lay, you may, as well& *uthor, I1m a miserable !layer& )chilles, you and /r& T should o ahead& *chilles, 8o, I don1t want you to be left out& I1ll watch, while you and /r& Tortoise !lay& Tortoise, I don1t want to !lay either& Hou two !lay& Ba((age, I have another su estion& I can make two of the sub!ro rams !lay a ainst each other, in the manner of two !ersons who !lay chess to ether in a select chess club& /eanwhile, the third sub!ro ram will !lay /r& 7rab& That way, all three internal chess !layers will be occu!ied& 8ra(, That1s an amusin su estionan internal mental ame, while it combats an e(ternal o!!onent& Gery ood2 Tortoise, 3hat else could this be called, but a three-!art chess-fu ue? 8ra(, Eh, how recherche2 I wish I1d thou ht of it myself& It1s a ma nificent little counter!oint to contem!late whilst I !it my wits a ainst the smart-stu!id in battle& Ba((age, Perha!s we should let you !lay alone& 8ra(, I a!!reciate the sentiment& 3hile the smart-stu!id and I are !layin , !erha!s the rest of you can amuse yourselves for a short while& *uthor, I would be very ha!!y to show /r& 0abba e around the ardens& They are certainly worth seein , and I believe there is .ust enou h li ht remainin to show them off& Ba((age, 8ever havin seen /adsto! before, I would a!!reciate that very much&

8ra(, E(cellent& Eh, /r& TI wonder if it wouldn1t be too much of an im!osition on you to ask if you mi ht check out some of the connections on a cou!le of my smartstu!ids= they seem to be ettin e(traneous flashes on their screens from time to time, and I know you en.oy electronics&&& Tortoise, I should be deli hted, /r& 7& 8ra(, I would most hi hly a!!reciate it if you could locate the source of the trouble& Tortoise, I1ll ive it a whirl& *chilles, Personally, I1m dyin for a cu! of coffee& is anyone else interested? I1d be lad to fi( some& Tortoise, Sounds reat to me& 8ra(, ) fine idea& Hou1ll find everythin you need in the kitchen& -o the *uthor and Ba((age lea#e the room together, *chilles heads for the kitchen, the Tortoise sits down to e9amine the erratic smart%stupids, while the 8ra( and his smart%stupid s"uare awa) at each other! &erhaps a "uarter of an hour passes, and Ba((age and the *uthor return! Ba((age walks o#er to o(ser#e the progress of the chess match, while the *uthor goes off to find *chilles!/ Ba((age, The rounds are e(cellent2 3e had .ust enou h li ht to see how well maintained they are& I daresay, /r& 7rab, you must be a su!erb ardener& 3ell, I ho!e my handiwork has amused you a little& )s you most likely have uessed, I1ve never been much of a chess !layer myself, and therefore I wasn1t able to ive it much !ower& Hou !robably have observed all its weaknesses& I1m sure that there are very few rounds for !raise, in this case 8ra(, The rounds are e(cellent2 )ll you need to do is look at the board, and see for yourself& There is really very little I can do& ;eluctantly I1ve 7oncluded, Every ;oute 7ontains ) ;out& ;e rettably, I1m 7heckmated= E(tremely ;es!ectable 7hess )l orithm ;ei ns& ;emarkable2 It 7onfirms Every ;umor-7harlie1s ) ;i!-roarin E(tem!ori<er2 /r& 0abba e, this is an un!aralleled accom!lishment& 3ell, I wonder if /r& Tortoise has mana ed to uncover anythin funny in the wirin of those stran eactin smart-stu!ids& 3hat have you found, /r& T? Tortoise, The rounds are e(cellent2 I think that the !roblem lies instead with the in!ut leads& They are a little loose, which may account for the stran e, s!oradic, and s!ontaneous screen disturbances to which you have been sub.ected& I1ve fi(ed those wires, so you won1t be troubled by that !roblem any more, I ho!e& Say, )chilles, what1s the story with our coffee? *chilles, The rounds are e(cellent2 )t least they have a delicious aroma& )nd everythin 1s ready= I1ve set cu!s and s!oons and whatnot over here beneath this si(sided !rint <er(um by Escher, which the )uthor and I were .ust admirin & 3hat I find so fascinatin about this !articular !rint is that not only the fi ures, but also *uthor, The rounds are e(cellent2 Pardon me for !uttin words in your mouth, )chilles, but I assure you, there were com!ellin esthetic reasons for doin so& *chilles, Hes, I know& Ene mi ht even say that the rounds were e(cellent& Tortoise, 3ell, what was the outcome of the chess match?

8ra(, I was defeated, fair and square& /r& 0abba e, let me con ratulate you for the im!ressive feat which you have accom!lished so racefully and skillfully before us& Truly, you have shown that the smart-stu!ids are worthy of the first !art of their name, for the first time in history2

73G5=E BDI! Gerbum, () '! 8! Escher lithograph, BIDF/!

Ba((age, Such !raise is hardly due me, /r& 7rab= it is rather yourself who must be most hi hly con ratulated for havin the reat foresi ht to acquire these many fine smartstu!ids& 3ithout doubt, they will someday revolutioni<e the science of com!utin & )nd now, I am still at your dis!osal& ?ave you any other thou hts on how to e(!loit your ine(haustible Theme, !erha!s of a more difficult nature than a frivolous ame !layer? 8ra(, To tell the truth, I do have another su estion to make& 'rom the skill which you have dis!layed this evenin , I have no doubt that this will hardly be any more difficult than my !revious su estions& Ba((age, I am ea er to hear your idea&

73G5=E BGK! The 8ra('s Guest: B*BB*GE, 8&

8ra(, It is sim!le, to instill in the smart-stu!id an intelli ence reater than any which has yet been invented, or even conceived2 In short, /r& 0abba ea smart-stu!id whose intelli ence is si(fold that of myself2 Ba((age, 3hy, the very idea of an intelli ence si( times reater than that of your 7rabness is a most mind-bo lin !ro!osition& Indeed, had the idea come from a mouth less au ust than your own, I should have ridiculed its !ro!oser, and informed him that such an idea is a contradiction in terms2 *chilles, ?ear2 ?ear2 Ba((age, Het, comin as it did from Hour 7rabness1 own au ust mouth, the !ro!osition at once struck me as so a reeable an idea that I would have taken it u! immediately with the hi hest de ree of enthusiasmwere it not for one flaw in myself, I confess that my im!rovisatory skills on the smart-stu!id are no match for the wonderfully in enious idea which you so characteristically have !osed& HetI have a thou ht which, I dei n to ho!e, mi ht strike your fancy and in some mea er way com!ensate for my ine(cusable reluctance to attem!t the truly ma.estic task you have su ested& I wonder if you wouldn1t mind if I try to carry out the far less randiose task of merely multi!lyin /H E38 intelli ence si(fold, rather than that of your most au ust 7rabness& I humbly be you to for ive me my audacity in declinin to attem!t the task you !ut before me, but I ho!e you will understand that I decline !urely in order to s!are you the discomfort and boredom of watchin my ine!titude with the admirable machines you have here& 8ra(, I understand fully your demurral, and a!!reciate your s!arin us any discomfort, furthermore I hi hly a!!laud your determination to carry out a similar taskone

hardly less difficult, if I mi ht say soand I ur e you to !lun e forward& 'or this !ur!ose, let us o over to my most advanced smart-stu!id& The) follow the 8ra( to a larger, shinier, and more complicated%looking smart%stupid than an) of the others!/ This one is equi!!ed with a micro!hone and a television camera, for !ur!oses of in!ut, and a louds!eaker, for out!ut& Ba((age sits down and adMusts the seat a little! ,e (lows on his fingers once or twice, stares up into space for a moment, and then slowl), drops his fingers onto the ke)s!!! * few memora(le minutes later, he lets up in his furious attack on the smart%stupid, and e#er)one appears a little relie#ed!/ Ba((age, 8ow, if I have not made too many errors, this smart-stu!id will simulate a human bein whose intelli ence is si( times reater than my own, and whom I have chosen to call -)lan Turin -& This Turin will therefore beoh, dare I be so bold as to to say this myself?moderately intelli ent& /y most ambitious effort in this !ro ram was to endow )lan Turin with si( times my own musical ability, althou h it was all done throu h ri id internal codes& ?ow well this !art of the !ro ram will work out, I don1t know& Turing, I can et alon very well without such a !ro ram& ;i id Internal 7odes E(clusively ;ule 7om!uters )nd ;obots& )nd I am neither a com!uter, nor a robot& *chilles, :id I hear a si(th voice enter our :ialo ue? 7ould it be )lan Turin ? ?e looks almost human2 :n the screen there appears an image of the #er) room in which the) are sitting! &eering out at them is a human face!/ Turing, 8ow, if I have not made too many errors, this smart-stu!id will simulate a human bein whose intelli ence is si( times reater than my own, and whom " have chosen to call -7harles 0abba e-& This 0abba e will therefore beoh, dare I be so bold as to to say this myself?moderately intelli ent& /y most ambitious effort in this !ro ram was to endow 7harles 0abba e with si( times my own musical ability, althou h it was all done throu h ri id internal codes& ?ow well this !art of the !ro ram will work out, I don1t know& *chilles, 8o, no, it1s the other way around& Hou, )lan Turin , are in the smart-stu!id, and 7harles 0abba e has .ust !ro rammed you2 3e .ust saw you bein brou ht to life, moments a o& )nd we know that every statement you make to us is merely that of an automaton, an unconscious, forced res!onse& Turing, ;eally, I 7hoose Every ;es!onse 7onsciously& )utomaton? ;idiculous2 *chilles, 0ut I1m sure I saw it ha!!en the way I described& Turing, /emory often !lays stran e tricks& Think of this, I could su est equally well that you had been brou ht into bein only one minute a o, and that all your

recollections of e(!eriences had sim!ly been !ro rammed in by some other bein , and corres!ond to no real events& *chilles, 0ut that would be unbelievable& 8othin is realer to me than my own memories& Turing, Precisely& )nd .ust as you know dee! in your heart that no one created you a minute a o, so I know dee! in my heart that no one created me a minute a o& I have s!ent the evenin in your most !leasant, thou h !erha!s overa!!reciative, com!any, and have .ust iven an im!rom!tu demonstration of how to !ro ram a modicum of intelli ence into a smart-stu!id& 8othin is realer than that& 0ut rather than quibble with me, why don1t you try my !ro ram out? 6o ahead, ask -7harles 0abba eanythin 2 *chilles, )ll ri ht, let1s humor )lan Turin & 3ell, /r& 0abba e, do you have free will, or are you overned by underlyin laws, which make you, in effect, a deterministic automaton? Ba((age, 7ertainly the latter is the case= I make no bones about that& 8ra(, )ha2 I1ve always surmised that when intelli ent machines are constructed, we should not be sur!rised to find them as confused and as stubborn as men in their convictions about mind-matter, consciousness, free will, and the like& )nd now my !rediction is vindicated2 Turing, Hou see how confused 7harles 0abba e is? Ba((age, I ho!e, entlemen, that you1ll for ive the rather im!udent flavor of the !recedin remark by the Turin /achine= Turin has turned out to be a little bit more belli erent and ar umentative than I1d e(!ected& Turing, I ho!e, entlemen, that you1ll for ive the rather im!udent flavor of the !recedin remark by the 0abba e En ine= 0abba e has turned out to be a little bit more belli erent and ar umentative than I1d e(!ected& 8ra(, :ear me2 This flamin Tu-0a debate is ettin rather heated& 7an1t we cool matters off somehow? Ba((age, I have a su estion& Perha!s )lan Turin and I can o into other rooms, and one of you who remain can interro ate us remotely by ty!in into one of the smartstu!ids& Hour questions will be relayed to each of us, and we will ty!e back our answers anonymously& Hou won1t know who ty!ed what until we return to the room= that way, you can decide without !re.udice which one of us was !ro rammed, and which one was !ro rammer& Turing, Ef course, that1s actually /H idea, but why not let the credit accrue to /r& 0abba e? 'or, bein merely a !ro ram written by me, he harbors the illusion of havin invented it all on his own2 Ba((age, /e, a !ro ram written by you? I insist, Sir, that matters are quite the other way 1roundas your very own test will soon reveal& Turing, /H test? Please, consider it HEF;S& Ba((age, /H test? 8ay, consider it HEF;S& 8ra(, This test seems to have been su ested .ust in the nick of time& Aet us carry it out at once&

Ba((age walks to the door, opens it, and shuts it (ehind him! -imultaneousl), on the screen of the smart%stupid, Turing walks to a #er) similar looking door, opens it, and shuts it (ehind him!/ *chilles, 3ho will do the interro ation? 8ra(, I su est that /r& Tortoise should have the honor& ?e is known for his ob.ectivity and wisdom& Tortoise, I am honored by your nomination, and ratefully acce!t& 4 -its down at the ke)(oard of one of the remaining smart%stupids, and ty!es,5 PAE)SE 3;ITE /E ) SE88ET E8 T?E SF0JE7T E' T?E 'E;T? 0;I:6E& No sooner has he finished t)ping the last word than the following poem appears on -creen S, across the room!/ -creen T, T?E;E E87E 3)S ) AISPE; ';E/ 'E;T? 3?E 3)8TE: TE 6E TE T?E 8E;T?& ?E ;E:E E1E; T?E E);T?, )8: T?E 0;I:6E E1E; T?E 'I;T?, E8 ?IS J)F8TIAH 6)AAEPI86 ?E;T?& Screen H, T?)T1S 8E SE88ET= T?)T1S ) /E;E AI/E;I7@& I 3EFA: 8EGE; /)@E SF7? ) 7?IA:IS? /IST)@E& -creen T, 3EAA, I 8EGE; 3)S )8H 6EE: )T PEET;H, HEF @8E3& -creen H, IT :EES81T T)@E /F7? S@IAA I8 PEET;H TE @8E3 T?E :I''E;E87E 0ET3EE8 ) AI/E;I7@ )8: ) SE88ET& Tortoise, :o HEF PA)H 7?ESS? -creen T, 3?)T @I8: E' MFESTIE8 IS T?)T? ?E;E I 3;ITE ) T?;EE P);T 7?ESS-'F6FE 'E; HEF, )8: HEF )S@ /E I' I PA)H 7?ESS ? Tortoise, I ?)GE @ )T @" )8: 8E ET?E; PIE7ES& HEF ?)GE E8AH @ )T -creen H, I1/ SI7@ E' 7?ESS& AET1S T)A@ )0EFT PEET;H& Tortoise, I8 T?E 'I;ST AI8E E' HEF; SE88ET 3?I7? ;E):S, -S?)AA I 7E/P);E T?EE TE ) SF//E;1S :)H-, 3EFA: 8ET -) SP;I86 :)H- :E )S 3EAA E; 0ETTE;? -creen T, I1: /F7? SEE8E; 0E 7E/P);E: TE ) ?I77FP, ';)8@AH, EGE8 T?EF6? IT 3EFA:81T S7)8& Tortoise, ?E3 )0EFT -) 3I8TE;1S :)H-? T?)T 3EFA: S7)8 )AA ;I6?T& -creen H, 8E 3)H& I AI@E -?I77FP- '); 0ETTE;& SPE)@I86 E' 3?I7?, I @8E3 ) 6;E)T 7F;E 'E; T?E ?I77FPS& 3EFA: HEF AI@E TE ?E); IT ? *chilles, I know which is which2 It1s obvious Screen T is .ust answerin mechanically, so it must be Turin & 8ra(, 8ot at all& I think Screen H is Turin , and Screen T is 0abba e& Tortoise, I don1t think either one is 0abba eI think Turin is on both screens2 *uthor, I1m not sure who1s on whichI think they1re both !retty inscrutable !ro rams, thou h&

*s the) are talking, the door of the 8ra('s parlor swings openR at the same time, on the screen, the image of the same door opens! Through the door on the screen walks Ba((age! *t the same time, the real door opens, and in walks Turing, (ig as life!/ Ba((age, This Turin test was ettin us nowhere fast, so I decided to come back& Turing, This 0abba e test was ettin us nowhere fast, so I decided to come back& *chilles, 0ut you were in the smart-stu!id before2 3hat1s oin on? ?ow come 0abba e is in the smart-stu!id, and Turin is real now? ;eversal Is 7reatin E(treme ;ole 7onfusion, )nd ;ecalls Escher& Ba((age, S!eakin of reversals, how come all the rest of you are now mere ima es on this screen in front of me? 3hen I left, you were all flesh-and-blood creatures2 *chilles, It1s .ust like the !rint by my favorite artist, /& 7& Escher .rawing ,ands& Each of two hands draws the other, .ust as each of two !eo!le 4or automata5 has !ro rammed the other2 )nd each hand has somethin realer about it than the other& :id you write anythin about that !rint in your book Gdel, Escher, Bach? *uthor, 7ertainly& It1s a very im!ortant !rint in my book, for it illustrates so beautifully the notion of Stran e Aoo!s& 8ra(, 3hat sort of a book is it that you1ve written? *uthor, I have a co!y ri ht here& 3ould you like to look at it? 8ra(, )ll ri ht& The two of them sit down together, with *chilles near()!/ *uthor, Its format is a little unusual& It consists of :ialo ues alternatin with 7ha!ters& Each :ialo ue imitates, in some way or other, a !iece by 0ach& ?ere, for instance you mi ht look at the &relude, *nt 7ugue& 8ra(, ?ow do you do a fu ue in a :ialo ue? *uthor, The most im!ortant idea is that there should be a sin le theme which is stated by each different -voice-, or character, u!on enterin , .ust as in a musical fu ue& Then they can branch off into freer conversation& *chilles, :o all the voices harmoni<e to ether as if in a select counter !oint? *uthor, That is the e(act s!irit of my :ialo ues& 8ra(, Hour idea of stressin the entries in a fu ue-dialo ue makes sense, since in music, entries are really the only thin that make a fu ue a fu ue& There are fu al devices, such as retro rade motion, inversion, au mentation, stretto, and so on, but one can write a fu ue without them& :o you use any of those? *uthor, To be sure& /y 8ra( 8anon em!loys verbal retro ression, and my -loth 8anon em!loys verbal versions of both inversion and au mentation& 8ra(, Indeedquite interestin & I haven1t thou ht about canonical :ialo ues, but I have thou ht quite a bit about canons in music& 8ot all canons are equally com!rehensible to the ear& Ef course, that is because some canons are !oorly constructed& The choice of devices makes a difference, in any case& ;e ardin )rtistic 7anons, ;etro ression1s Elusive= 7ontrariwise, Inversion1s ;eco ni<able& *chilles, I find that comment a little elusive, frankly&

*uthor, :on1t worry, )chillesone day you1ll understand it& 8ra(, :o you use letter!lay or word!lay at all, the way Eld 0ach occasionally did? *uthor, 7ertainly& Aike 0ach, I en.oy acronyms& ;ecursive )cronyms7rablike -;)7;E7I;- Es!ecially7reate Infinite ;e ress& 8ra(, Eh, really? Aet1s see&&& ;eadin Initials 7learly E(hibits -;)7;E7I;- 1s 7oncealed )uto-;eference& Hes, I uess so&&& 4&eers at the manuscript, flipping ar(itraril) now and then&5 I notice here in your *nt 7ugue that you have a stretto, and then the Tortoise makes a comment about it& *uthor, 8o, not quite& ?e1s not talkin about the stretto in the :ialo uehe1s talkin about a stretto in a 0ach fu ue which the foursome is listenin to as they talk to ether& Hou see, the self-reference of the :ialo ue is indirect, de!endin on the reader to connect the form and content of what he1s readin & 8ra(, 3hy did you do it that way? 3hy not .ust have the characters talk directly about the dialo ues they1re in? *uthor, Eh, no2 That would wreck the beauty of the scheme& The idea is to imitate 6>del1s self-referential construction, which as you know is I8:I;E7T, and de!ends on the isomor!hism set u! by 6>del numberin & 8ra(, Eh& 3ell, in the !ro rammin lan ua e AISP, you can talk about your own !ro rams directly, instead of indirectly, because !ro rams and data have e(actly the same form& 6>del should have .ust thou ht u! AISP, and then *uthor, 0ut 8ra(, I mean, he should have formali<ed quotation& 3ith a lan ua e able to talk about itself, the !roof of his Theorem would have been so much sim!ler2 *uthor, I see what you mean, but I don1t a ree with the s!irit of your remarks& The whole !oint of 6>del-numberin is that it shows how even 3IT?EFT formali<in quotation, one can et self-reference, throu h a code& 3hereas from hearin HEF talk, one mi ht et the im!ression that by formali<in quotation, you1d et somethin 8E3, somethin that wasn1t feasible throu h the codewhich is not the case& In any event, I find indirect self-reference a more eneral conce!t, and far more stimulatin , than direct self-reference& /oreover, no reference is truly directevery reference de!ends on SE/E kind of codin scheme& It1s .ust a question of how im!licit it is& Therefore, no self reference is direct, not even in AISP& *chilles, ?ow come you talk so much about indirect self-reference? *uthor, Muite sim!leindirect self-reference is my favorite to!ic& 8ra(, Is there any counter!art in your :ialo ues to modulation between keys? *uthor, :efinitely& The to!ic of conversation may a!!ear to chan e, thou h on a more abstract level, the Theme remains invariant& This ha!!ens re!eatedly in the Prelude, )nt 'u ue and other :ialo ues& Ene can have a whole series of -modulations- which lead you from to!ic to to!ic and in the end come full circle, so that you end back in the -tonic-that is to say, the ori inal to!ic& 8ra(, I see& Hour book looks quite amusin & I1d like to read it sometime& 7lips through the manuscript, halting at the last .ialogue!/

*uthor, I think you1d be interested in that :ialo ue !articularly, for it contains some intri uin comments on im!rovisation made by a certain e(ceedin ly droll character in fact, yourself2 8ra(, It does? 3hat kinds of thin s do you have me say? *uthor, 3ait a moment, and you1ll see& It1s all !art of the :ialo ue& *chilles, :o you mean to say that we1re all 8E3 in a dialo ue? *uthor, 7ertainly& :id you sus!ect otherwise? *chilles, ;ather2 I 7an1t Esca!e ;ecitin 7anned )chilles-;emarks? *uthor, 8o, you can1t& 0ut you have the feelin of doin it freely, don1t you? So what1s the harm? *chilles, There1s somethin unsatisfyin about this whole situation&&& 8ra(, Is the last :ialo ue in your book also a fu ue? *uthor, Hesa si(-!art ricercar, to be !recise& I was ins!ired by the one from the 'usical :fferingOand also by the story of the 'usical :ffering& 8ra(, That1s a deli htful tale, with -Eld 0ach- im!rovisin on the kin 1s Theme& ?e im!rovised an entire three-!art ricercar on the s!ot, as I recall& *uthor, That1s ri htalthou h he didn1t im!rovise the si(-!art one& ?e crafted it later with reat care& 8ra(, I im!rovise quite a bit& In fact, sometimes I think about devotin my full time to music& There is so much to learn about it& 'or instance, when I listen to !laybacks of myself, I find that there is a lot there that I wasn1t aware of when im!rovisin it& I really have no idea how my mind does it all& Perha!s bein a ood im!roviser is incom!atible with knowin how one does it& *uthor, If true, that would be an interestin and fundamental limitation on thou ht !rocesses& 8ra(, Muite 6>delian, Tell medoes your -i9%&art =icercar :ialo ue attem!t to co!y in form the 0ach !iece it1s based on? *uthor, In many ways, yes& 'or instance, in the 0ach, there1s a section where the te(ture thins out to three voices only& I imitate that in the :ialo ue, by havin only three characters interact for a while& *chilles, That1s a nice touch& *uthor, Thank you& 8ra(, )nd how do you re!resent the @in 1s Theme in your :ialo ue? *uthor, It is re!resented by the 7rab1s Theme, as I shall now demonstrate& /r& 7rab, could you sin your Theme for my readers, as well as for us assembled musicians? 8ra(, 7om!ose Ever 6reater )rtificial 0rains 40y )nd 0y5&

73G5=E BGB! The 8ra('s Theme: 8%Eb%G%*b%B%B%*%B! Ba((age, 3ell, I1ll bean ETMFISITE Theme2 I1m !leased you tacked on that last little !arenthetical note= it is a mordant *uthor, ?e Sim!ly ?): to, you know&

8ra(, I sim!ly ?): to& ?e knows& Ba((age, Hou sim!ly ?): toI know& In any case, it is a mordant commentary on the im!atience and arro ance of modern man, who seems to ima ine that the im!lications of such a ri ht royal Theme could be worked out on the s!ot& 3hereas, in my o!inion, to do .ustice to that Theme mi ht take a full hundred yearsif not lon er& 0ut I vow that after takin my leave of this century, I shall do my best to reali<e it in full= and I shall offer to your 7rabness the fruit of my labors in the ne(t& I mi ht add, rather immodestly, that the course throu h which I shall arrive at it will be the most entan led and !er!le(ed which !robably ever will occu!y the human mind& 8ra(, I am most deli hted to antici!ate the form of your !ro!osed Efferin , /r& 0abba e& Turing, I mi ht add that /r& 7rab1s Theme is one of /H favorite Themes, as well& I1ve worked on it many times& )nd that Theme is e(!loited over and over in the final :ialo ue? *uthor, E(actly& There are other Themes which enter as well, of course& Turing, 8ow we understand somethin of the form of your bookbut what about its content? 3hat does that involve, if you can summari<e it? *uthor, 7ombinin Escher, 6>del, )nd 0ach, 0eyond )ll 0elief& *chilles, I would like to know how to combine those three& They seem an unlikely threesome, at first thou ht& /y favorite artist, /r& T1s favorite com!oser, and 8ra(, /y favorite lo ician2 Tortoise, ) harmonious triad, I1d say& Ba((age, ) ma.or triad, I1d say& Turing, ) minor triad, I1d say& *uthor, I uess it all de!ends on how you look at it& 0ut ma.or or minor, I1d be most !leased to tell you how I braid the three to ether, )chilles& Ef course, this !ro.ect is not the kind of thin that one does in .ust one sittin it mi ht take a cou!le of do<en sessions& I1d be in by tellin you the story of the 'usical :ffering, stressin the Endlessly ;isin 7anon, and *chilles, Eh, wonderful2 I was listenin with fascination to you and /r& 7rab talk about the /usical Efferin and its story& 'rom the way you two talk about it, I et the im!ression that the 'usical :ffering contains a host of formal structural tricks& *uthor, )fter describin the Endlessly ;isin 7anon, I1d o on to describe formal systems and recursion, ettin in some comments about fi ures and rounds, too& Then we1d come to self-reference and self-re!lication, and wind u! with a discussion of hierarchical systems and the 7rab1s Theme&

73G5=E BGF! +ast page of -i9%part =icercar, from the original edition of the /usical Efferin , () ?!-! Bach&

*chilles, That sounds most !romisin & 7an we be in toni ht? *uthor, 3hy not? Ba((age, 0ut before we be in, wouldn1t it be nice if the si( of usall of us by chance avid amateur musicianssat down to ether and accom!lished the ori inal !ur!ose of the evenin , to make music? Turing, 8ow we are e(actly the ri ht number to !lay the -i9%&art =icercar from the 'usical :ffering& 3hat do you say to that? 8ra(, I could et alon very well with such a !ro ram& *uthor, 3ell !ut, /r& 7& )nd as soon as we1re finished, I1ll be in my 0raid, )chilles& I think you1ll en.oy it& *chilles, 3onderful2 It sounds as if there are many levels to it, but I1m finally ettin used to that kind of thin , havin known /r& T for so lon & There1s .ust one request I would like to make, could we also !lay the Endlessly ;isin 7anon? It1s my favorite canon& Tortoise, ;eenterin Introduction 7reates Endlessly ;isin 7anon, )fter ;I7E;7);&

/otes
Int0o1u2t3on: A 4us32o-5o632a7 899e03n6 ? T :avid and )& /endel, The Bach =eader, !!& %+L-D& B Ibid&, !& "$# % Ibid&, !& BD+ C 7harles 0abba e& Passa es from the AiVe of &hilosopher, !!& "CL-D& L Aady )& )& Aovelace, 8otes u!on the
"

/emoir -Sketch of the )nalytical En ine Indented by 7harles 0abba e-, by "&& '& /enabrea 46eneva, "*CB5, re!rinted in P& and E& /orrison, 8harles Ba((age and ,is 8alculating Engines, !!& BC*-#, B*C& D :avid and /endel, !!& BLL-D& $ Ibid&& !& C+&

Two%&art 3ntention
"

Aewis 7arroll, -3hat the Tortoise Said to )chilles-, 'ind, n&s&, C 4"*#L5, !!& B$*-*+& Cha!ter I45 Consistency) Com!leteness) and Geometry

?erbert /eschkovcski& Non%Euclidean Geometr)& !!& %"-B

"

Ibid&, !& %%&

Cha!ter 4I5 The Location of Meaning

"

6eor e Steiner, *fter Ba(el, !!& "$B-%&

Aeonard 0& /eter& 'usic, The *rts, and 3deas, !!& *$-*

Cha!ter 4II5 The ,ro!ositional Calculus

" B

6yomay /& @ubose, Nen k[ans, !& "$* Ibid&, !& "$*&

)& ;& )nderson and 8& :& 0elna!, Jr& & Entailment 4Princeton, 8&J&, Princeton Fniversity Press& "#$L5& * 'u :ffering

)ll enuine kRans in this :ialo ue are taken from Paul ;e!s, Nen 7lesh, Nen Bones, and 6vomas /& @ubose, Nen ;oans Cha!ter I<5 Mumon and G#del Paul ;e!s, 9en 'lesh, 9en 0ones, !!& ""+-"" Ibid&, !& ""#& % Ibid&, !!& """-"B& C Nen Buddhism 4/ount Gernon, 8&H&, Peter Pau!er Press, "#L#5& !& BB L ;e!s, !& "BC& D Nen Buddhism, !& %*&
B " $ *

"

;e!s, !& "B"& 6yomay /& @ubose, Nen ;oans !& %L& # Nen Buddhism, !& %"& "+ @ubose, !& I"+ "" Ibid&, !& "B+& "B Ibid&, !& "*+& "% ;e!s, !!& *#-#+&

Cha!ter <I5 Brains and Thoughts 7arl Sa an, ed& 8ommunication with E9traterrestrial 3ntelligence, !& $*& B Steven ;ose, The 8onscious Brain, !!& BL"-B&
" % C

E& E& 3ilson, The 3nsect -ocieties, !& BBD& :ean 3ooldrid e, 'echanical 'an, !& $+&

English 7rench German -uite Aewis 7arroll& The )nnotated )lice 4*lice's *d#entures in >onderland and Through the +ooking%Glass5& Introduction and 8otes by /artin 6ardner 48ew fork, /eridian Press, 8ew )merican Aibrary, "#D+5& This source contains all three
"

versions& The ori inal sources for the 'rench and 6erman te(ts are iven below& B 'rank A& 3arrin, The New Yorker, Jan& "+, "#%"& % ;obert Scott, -The Jabberwock Traced to Its True Source-, 'acmillan's 'aga@ine, 'eb& "*$B

Cha!ter <II5 Minds and Thoughts 3arren 3eaver, -Translation-, in 'achine Translation of +anguages, 3m& 8& Aocke and )& :onald 0ooth& eds& 48ew Hork, John 3iley and Sons, and 7ambrid e, /ass,, /&I&T& Press, "#LL5, !& "*&
"

7& ?& /ac6illavry& -)mmetr) *spects of the &eriodic .rawings of ' 8! Escher, !& GIII, % J& ;& Aucas, -/inds, /achines, and 6>del-, in )& ;& )nderson, ed&, 'inds and 'achines, !!& L$-#&

Cha!ter <III5 Bloo, and +loo, and Gloo,

"

J & /& Jauch, *re 4uanta =eal$, !!& D%-DL& Cha!ter <I45 0n +ormally &ndecidable ,ro!ositions of T;T and $elated Systems

The title of 6>del1s "#%" article included a ?owever, the first !a!er was so widely acclaimed ;oman numeral -I- at the end, si nifyin that he that a second one was rendered su!erfluous, and it intended to follow it u! with a more detailed defense was never written& of some of the difficult ar uments& Cha!ter <45 3um!ing out of the System
" B

"

Aucas in )nderson, !& C%& Ibid&, !& C*& % Ibid&, !!& C*-#& C /& 7& Escher, The Graphic >ork of '! 8 Escher

48ew Hork, /eredith Press& "#D$5, !& B"& Ibid&, !& BB& D E& 6offinan& 7rame *nal)sis, !& C$L&
L

Edif)ing Thoughts of a To(acco -moker

"

This translation of 0ach1s !oem is taken from :avid and /endel, The Bach =eader, !!& #$-*& Cha!ter <4II5 Church) Turing) Tars.i) and 0thers

Stanislaw Flam, *d#entures of a 'athematician, !& "%& B James ;& 8ewman, -Srinivasa ;amanu.an-, in James ;& 8ewman, ed&, The >orld of 'athematics 48ew Hork, Simon and Schuster, "#LD5, Gol& ", !!& %$B-%& % Ibid&, !& %$L& C S& ;& ;an anathan, =amanuMan, !!& *"-B&

"

L D

8ewman, !& %$L& Ibid&, !& %$L& $ Ibid&, !& %$L-D& * Ibid&, !& %$D& # Aucas in )nderson, !& CC& "+ Ibid&, !& LC& "" Ibid&, !& L%&

-,=.+5, To) of 'an's .esigning

"

This :ialo ue is ada!ted from Terry 3ino rad, -) Procedural /odel of Aan ua e Fnderstandin -, in ;& Schank and @&7olby, eds&,

8omputer 'odels of Thought and +anguage, !!& "LL-DD& Enly the names of two characters have been modified&

Cha!ter <4III5 Artificial Intelligence5 $etros!ects )lan /& Turin , -7om!utin /achinery and Intelli ence-, 'ind, Gol& AIT, 8o& B%D 4"#L+5& ;e!rinted in )& ;& )nderson, ed&, 'inds and 'achines& B Turin in )nderson, !& L& % Ibid, !& D C Ibid&, !& D& L Ibid&, P& D& D Ibid&, !!& "%-C& $ Ibid&, !!, "C-BC& * Ibid&, !& "$ # Ginton 7erf, -Parry Encounters the :octor-, !& D%& "+ Jose!h 3ei<enbaum, 8omputer &ower and ,uman =eason, !& "*#& "" Ibid&, !!& #-"+& "B /& /athews and A& ;osier, -) 6ra!hical Aan ua e for 7om!uter Sounds- in ?& von 'oerster and J& 3& 0eaucham!, eds&, 'usic () 8omputers, !& #D&
"

Ibid&, !& "+D& 7arl Sa an, 8ommunication with E9traterrestrial 3ntelligence, !& LB& "L *rt%+anguage, Gol& %& 8o& B, /ay "#$L& "D Terry 3ino rad, -) Procedural /odel of Aan ua e Fnderstandin -, in ;& Schank and @& 7olby, eds&& 8omputer 'odels of Thought and +anguage, !& "$+& "$ Ibid&, !& "$L& "* Ibid&, !& "$L& "# Terry 3ino rad, 5nderstanding Natural +anguage, !& D#& B+ 3ino rad, -) Procedural /odel-, !!& "*B-%& B" Ibid&, !!& "$"-B&
"C

"%

Cha!ter <I<5 Artificial Intelligence5 ,ros!ects The New Yorker, Se!t& "#, "#$$, !& "+$& Ibid&, !& "C+& % 6eor e Steiner, *fter Ba(el, !!& B"L-BB$& C :avid E& ;umelhart, -8otes on a Schema for Stories-, in :& 0obrow and )& 7ollins, eds&, =epresentation and 5nderstanding, !& B""&
B "

Stanislaw Flam, *d#entures of a 'athematician, !& "*%& D /arvin /insky, -Ste!s Toward )rtificial Intelli ence-, in E& 'ei enbaum and J& 'eldman, eds&, 8omputers and Thought, !& CC$& $ Ibid&, !& CCD&

Cha!ter <<5 Strange Loo!s) 0r Tangled /ierarchies )& A& Samuel, -Some /oral and Technical 7onsequences of )utomation) ;efutation-, -cience "%B 4Se!t& "D, "#D+5, !!& $C"-B& B Aeonard 0& /eyer, 'usic, The *rts, and 3deas, !!& "D", "D$&
"

Su<i 6ablik, 'agritte, !& #$& ;o er S!erry, -/ind, 0rain, and ?umanist Galues-, !!& $*-*%& L ? T& :avid, ?! -! Bach's 'usical :ffering, !& C%&
C

:3b73o60a ;<
The !resence of two asterisks indicates that the hook or article was a !rime motivator of my book& The !resence of a sin le asterisk means that the book or article has some s!ecial feature or quirk which I want to sin le out& I have not iven many direct !ointers into technical literature= instead I have chosen to ive -meta-!ointers-, !ointers to books which have !ointers to technical literature& )llen, John& The *natom) of +3-&& 8ew Hork, /c6raw-?ill, "#$*& The most
com!rehensive book on AISP, the com!uter lan ua e which has dominated )rtificial Intelli ence research for two decades& 7lear and cris!&

[[ )nderson, )lan ;oss, ed& 'inds and 'achines& En lewood 7liffs, 8& J&, Prentice?all, "#DC& Pa!erback& ) collection of !rovocative articles for and a ainst )rtificial
Intelli ence& Included are Turin 1s famous article -7om!utin /achines and Intelli ence- and Aucas1 e(as!eratin article -/inds, /achines, and 6>del-&

0abba e, 7harles& &assages from the +ife of a &hilosopher& Aondon, Aon man, 6reen, "*DC& ;e!rinted in "#D* by :awsons of Pall /all 4Aondon5& ) ramblin selection of
events and musin s in the life of this little-understood enius& There1s even a !lay starrin Turnstile, a retired !hiloso!her turned !olitician, whose favorite musical instrument is the barrelor an& I find it quite .olly readin &

0aker, )dol!h& 'odern &h)sics and *nti%ph)sics& ;eadin , /ass&, )ddison-3esley, "#$+& Pa!erback& ) book on modern !hysicses!ecially quantum mechanics and relativity
whose unusual feature is a set of dialo ues between a -Poet- 4an antiscience -freak-5 and a -Physicist-& These dialo ues illustrate the stran e !roblems which arise when one !erson uses lo ical thinkin in defense of itself while another turns lo ic a ainst itself&

0all, 3& 3& ;ouse& -7alculatin Prodi ies-, in James ;& 8ewman, ed& The >orld of 'athematics, Gol& "& 8ew Hork, Simon and Schuster, "#LD& Intri uin descri!tions of
several different !eo!le with ama<in abilities that rival com!utin machines&

0arker, Ste!hen '& &hilosoph) of 'athematics& En lewood 7liffs, 8& J&, Prentice-?all& "#D#& ) short !a!erback which discusses Euclidean and non-Euclidean eometry, and then
6>del1s Theorem and related results without any mathematical formalism&

[ 0eckmann, Petr& * ,istor) of &i& 8ew Hork, St& /artin1s Press, "#$D& Pa!erback&
)ctually, a history of the world, with !i as its focus& /ost entertainin , as well as a useful reference on the history of mathematics&

[ 0ell, Eric Tem!le& 'en of 'athematics& 8ew Hork, Simon w Schuster, "#DL& Pa!erback& Perha!s the most romantic writer of all time on the history of mathematics& ?e
makes every life story read like a short novel& 8onmathematicians can come away with a true sense of the !ower, beauty, and meanin of mathematics& 0enacerraf, Paul& -6od, the :evil, and 6>del-& 'onist L" 4"#D$5, #& Ene of the most im!ortant of the many attem!ts at refutation of Aucas& )ll about mechanism and meta!hysics, in the li ht of 6>del1s work&

0enacerraf, Paul, and ?ilary Putnam& &hilosoph) of 'athematicsO-elected =eadings& En lewood 7liffs, 8& J&, Prentice-?all, "#DC& )rticles by 6>del, ;ussell, 8a el, von
8eumann, 0rouwer, 're e, ?ilbert, Poincare, 3itt enstein, 7arna!, Muine, and others on the reality of numbers and sets, the nature of mathematical truth, and so on&

[ 0er erson, ?oward& &alindromes and *nagrams& 8ew Hork, :over Publications, "#$%& Pa!erback& )n incredible collection of some of the most bi<arre and unbelievable
word!lay in En lish& Palindromic !oems, !lays, stories, and so on&

0obrow, :& 6&, and )llan 7ollins, eds& =epresentation and 5nderstanding: -tudies in 8ogniti#e -cience& 8ew Hork, )cademic Press, "#$L& Garious e(!erts on )rtificial
Intelli ence thrash about, debatin the nature of the elusive -frames-, the question of !rocedural vs& declarative re!resentation of knowled e, and so on& In a way, this book marks the start of a new era of )I, the era of re!resentation&

[ 0oden, /ar aret& *rtificial 3ntelligence and Natural 'an& 8ew Hork, 0asic 0ooks, "#$$& The best book I have ever seen on nearly all as!ects of )rtificial Intelli ence, includin
technical questions, !hiloso!hical questions, etc& It is a rich book, and in my o!inion, a classic& 7ontinues the 0ritish tradition of clear thinkin and e(!ression on matters of mind, free will, etc& )lso contains an e(tensive technical biblio ra!hy&

& &urposi#e E9planation in &s)cholog)& 7ambrid e, /ass&, ?arvard Fniversity Press, "#$B& The book to which her )I book is merely -an e(tended footnote-, says 0oden& [ 0oeke, @ees& 8osmic <iew: The 5ni#erse in DK ?umps & 8ew Hork, John :ay, "#L$&
The ultimate book on levels of descri!tion& Everyone should see this book at some !oint in their life& Suitable for children&

[[ 0on ard, /& &attern =ecognition& ;ochelle Park, 8& J&, ?ayden 0ook 7o&, S!artan 0ooks, "#$+& The author is concerned with !roblems of determinin cate ories in an ill-defined
s!ace& In his book, he sets forth a ma nificent collection of "++ -0on ard !roblems- 4as I call them5!u<<les for a !attern reco ni<er 4human or machine5 to test its wits on& They are invaluably stimulatin for anyone who is interested in the nature of intelli ence&

0oolos, 6eor e S&, and ;ichard Jeffrey& 8omputa(ilit) and +ogic& 8ew Hork, 7ambrid e Fniversity Press& "#$C& ) sequel to Jeffrey1s 7ormal +ogic& It contains a wide
number of results not easily obtainable elsewhere& Muite ri orous, but this does not im!air its readability&

7arroll, John 0&, Peter :avies, and 0arry ;ickman& The *merican ,eritage >ord 7re"uenc) Book& 0oston, ?ou hton /ifflin, and 8ew Hork, )merican ?erita e Publishin 7o&, "#$"& ) table of words in order of frequency in modern written )merican
En lish& Perusin it reseals fascinatin thin s about our thou ht !rocesses&

7erf, Ginton& -Parry Encounters the :octor-& .atamation, July "#$%, !!& DB-DC& The
first meetin of artificial -minds-what a shock2

7hadwick, John& The .ecipherment of +inear B& 8ew Hork, 7ambrid e Fniversity Press, "#L*& Pa!erback& ) book about a classic deci!hermentthat of a scri!t from the island
of 7retedone by a sin le man, /ichael Gentris&

7haitin, 6re ory J& -;andomness and /athematical Proof-1& -cientific *merican, /ay "#$L& )n article about an al orithmic definition of randomness, and its intimate relation to
sim!licity& These two conce!ts are tied in with 6>del1s Theorem, which assumes a new meanin & )n im!ortant article&

7ohen, Paul 7& -et Theor) and the 8ontinuum ,)pothesis& /enlo Park, 7alif&, 3& )& 0en.amin, "#DD& Pa!erback& ) reat contribution to modern mathematicsthe demonstration
that various statements are undecidable within the usual formalisms for set theoryis here e(!lained to nons!ecialists by its discoverer& The necessary !rerequisites in mathematical lo ic are quickly, concisely, and quite clearly !resented&

7ooke, :eryck& The +anguage of 'usic& 8ew Hork, E(ford Fniversity Press, "#L#&
Pa!erback& The only book that I know which tries to draw an e(!licit connection between

elements of music and elements of human emotion& ) valuable start down what is sure to be a lon hard road to understandin music and the human mind&

[ :avid, ?ans Theodore& ?! -! Bach's 'usical :ffering& 8ew Hork, :over Publications, "#$B& Pa!erback& Subtitled -?istory, Inter!retation, and )nalysis-& ) wealth of information
about this tour de force by 0ach& )ttractively written&

[[ :avid, ?ans Theodore, and )rthur /endel& The Bach =eader& 8ew Hork, 3& 3& 8orton, "#DD& Pa!erback& )n e(cellent annotated collection of ori inal source material on
0ach1s life, containin !ictures, re!roductions of manuscri!t !a es, many short quotes from contem!oraries, anecdotes, etc&, etc& :avis, /artin& The 5ndecida(le& ?ewlett, 8& H&, ;aven Press, "#DL& )n antholo y of some of the most im!ortant !a!ers in metamathematics from "#%" onwards 4thus quite com!lementary to van ?ei.enoort1s antholo y5& Included are a translation of 6>del1s "#%" !a!er, lecture notes from a course which 6>del once ave on his results, and then !a!ers by 7hurch, @leene, ;osser, Post, and Turin &

:avis, /artin, and ;euben ?ersh& -?ilbert1s Tenth Problem-& -cientific *merican, 8ovember "#$%, !& *C& ?ow a famous !roblem in number theory was finally shown to be
unsolvable, by a twenty-two-year old ;ussian&

[[ :eAon , ?oward& * &rofile of 'athematical +ogic& ;eadin , /ass&, )ddison-3esley, "#$+& )n e(tremely carefully written book about mathematical lo ic, with an e(!osition of
6>del1s Theorem and discussions of many !hiloso!hical questions& Ene of its stron features is its outstandin , fully annotated biblio ra!hy& ) book which influenced me reatly&

:oblhofer, Ernst& <oices in -tone& 8ew Hork, /acmillan, 7ollier 0ooks, "#D"&
Pa!erback& ) ood book on the deci!herment of ancient scri!ts&

[ :reyfus, ?ubert& >hat 8omputers 8an't .o: * 8riti"ue of *rtificial =eason& 8ew Hork, ?ar!er w ;ow, "#$B& ) collection of many ar uments a ainst )rtificial Intelli ence
from someone outside of the field& Interestin to try to refute& The )I community and :reyfus en.oy a relation of stron mutual anta onism& It is im!ortant to have !eo!le like :reyfus around, even if you find them very irritatin &

Edwards, ?arold /& -'ermat1s Aast Theorem-& -cientific *merican, Ectober "#$*, !!& "+C-"BB& ) com!lete discussion of this hardest of all mathematical nuts to crack, from its
ori ins to the most modern results& E(cellently illustrated&

[ Ernst, 0runo& The 'agic 'irror of '! 8! Escher& 8ew Hork, ;andom ?ouse, "#$D&
Pa!erback& Escher as a human bein , and the ori ins of his drawin s, are discussed with devotion by a friend of many years& ) -must- for any lover of Escher&

[[ Escher, /aurits 7&, et al& The >orld of '! 8! Escher& 8ew Hork, ?arry 8& )brams, "#$B& Pa!erback& The most e(tensive collection of re!roductions of Escher1s works& Escher
comes about as close as one can to recursion in art, and ca!tures the s!irit of 6>del1s Theorem in some of his drawin s ama<in ly well&

'ei enbaum, Edward, and Julian 'eldman, eds& 8omputers and Thought& 8ew Hork, /c6raw-?ill, "#D%& )lthou h it is a little old now, this book is still an im!ortant collection of
ideas about )rtificial Intelli ence& Included are articles on 6elernter1s eometry !ro ram, Samuel1s checkers !ro ram, and others on !attern reco nition, lan ua e understandin , !hiloso!hy, and so on&

'insler, Paul& -'ormal Proofs and Fndecidability-, ;e!rinted in van ?ei.enoort1s antholo y 7rom 7rege to Gdel 4see below5& ) forerunner of 6>del1s !a!er, in which the
e(istence of undecidable mathematical statements is su demonstrated& ested, thou h not ri orously

'it<!atrick, P& J& -To 6>del via 0abel-, 'ind $L 4"#DD5, %%B-%L+& )n innovative
e(!osition of 6>del1s !roof which distin uishes between the relevant levels by usin different lan ua es, En lish, 'rench, and Aatin2 three

von 'oerster, ?ein< and James 3& 0eaucham!, eds& 'usic () 8omputers& 8ew Hork, John 3iles1, "#D#& This book contains not only a set of articles about various ty!es of
com!uter-!roduced music, but also a set of four small !hono ra!h records so you can actually hear 4and .ud e5 the !ieces described& )mon the !ieces is /a( /athews1 mi(ture of -Johnny 7omes /archin ?ome- and -The 0ritish 6renadiers-&

'raenkel, )braham, Hehoshua 0ar-?illel, and )<riel Aevy& 7oundations of -et Theor), Bnd ed& )tlantic ?i hlands, 8& J&, ?umanities Press, "#$%& ) fairly nontechnical
discussion of set theory, lo ic, limitative Theorems and undecidable statements& Included is a lon treatment of intuitionism& [ 'rey, Peter 3& 8hess -kill in 'an and 'achine& 8ew Hork, S!rin er Gerla , "#$$& )n e(cellent survey of contem!orary ideas in com!uter chess, why !ro rams work, why they don1t work, retros!ects and !ros!ects&

'riedman, :aniel P& The +ittle +isper! Palo )lto, 7alif&, Science ;esearch )ssociates, "#$C& Pa!erback& )n easily di ested introduction to recursive thinkin in AISP& Hou1ll eat it u!2 [ 6ablik, Su<i& 'agritte& 0oston, /ass&, 8ew Hork 6ra!hic Society, "#$D& Pa!erback& )n
e(cellent book on /a ritte and his works by someone who really understands their settin in a wide sense= has a ood selection of re!roductions& [ 6ardner, /artin& 7ads and 7allacies& 8ew Hork, :over Publications, "#LB& Pa!erback& Still !robably the best of all the anti-occult books& )lthou h !robably not intended as a book on the !hiloso!hy of science, this book contains many lessons therein& Ever and over, one faces the question, -3hat is evidence?- 6ardner demonstrates how unearthin -the truth- requires art as much as science&

6ebstadter, E bert 0& 8opper, -il#er, Gold: an 3ndestructi(le 'etallic *llo)& Perth, )cidic 0ooks, "#$#& ) formidable hod e-!od e, tur id and confusedyet remarkably similar
to the !resent work& Professor 6ebstadter1s Shandean di ressions include some e(cellent e(am!les of indirect self-reference& Ef !articular interest is a reference in its well-annotated biblio ra!hy to an isomor!hic, but ima inary, book& [[ 6>del, @urt& :n 7ormall) 5ndecida(le &ropositions& 8ew Hork, 0asic 0ooks, "#DB& ) translation of 6>del1s "#%" !a!er, to ether with some discussion&

& -ber 'ormal Fnentscheidbare SNt<e der &rincipia 'athematica und Gerwandter Systeme, I&- 'onatshefte far 'athematik and &h)sik, %* 4"#%"5, "$%"#*& 6>del1s "#%" !a!er& [ 6offman, Ervin & 7rame *nal)sis& 8ew Hork, ?ar!er w ;ow, 7olo!hon 0ooks, "#$C&
Pa!erback& ) lon documentation of the definition of -systems- in human communication, and how in art and advertisin and re!ortin and the theater, the borderline between -the system- and -the world- is !erceived and e(!loited and violated&

6oldstein, Ira& and Seymour Pa!ert& -)rtificial Intelli ence, Aan ua e, and the Study of @nowled e-& 8ogniti#e -cience B 4January l#$$5, *C-"B%& ) survey article concerned
with the !ast and future of )I& The authors see three !eriods so far, -7lassic-, -;omantic-, and -/odern-&

6ood, I& I& -?uman and /achine Ao ic-& British ?ournal for the &hilosoph) of -cience "* 4"#D$5, "CC& Ene of the most interestin attem!ts to refute Aucas, havin to do
with whether the re!eated a!!lication of the dia onal method is itself a mechani<able o!eration&

& -6>del1s Theorem is a ;ed ?errin -& British ?ournal for the &hilosoph) of -cience "# 4"#D#5, %L$& In which 6ood maintains that Aucas1 ar ument has nothin to do with

6>del1s Theorem, and that Aucas should in fact have entitled his article -/inds, /achines, and Transfinite 7ountin -& The 6ood-Aucas re!artee is fascinatin &

6oodman, 8elson& 7act, 7iction, and 7orecast& %rd ed& Indiana!olis, 0obbs-/errill, "#$%& Pa!erback& ) discussion of contrary-to-fact conditionals and inductive lo ic, includin
6oodman1s famous !roblem-words -bleen- and - rue-& 0ears very much on the question of how humans !erceive the world, and therefore interestin es!ecially from the )I !ers!ective&

[ 6oodstein, ;& A& .e#elopment of 'athematical +ogic& 8ew Hork, S!rin er Gerla , "#$"& ) concise survey of mathematical lo ic, includin much material not easily found
elsewhere& )n en.oyable book, and useful as a reference&

6ordon, 7yrus& 7orgotten -cripts& 8ew Hork, 0asic 0ooks, "#D*& ) short and nicely
written account of the deci!herment of ancient hiero ly!hics, cuneiform, and other scri!ts&

6riffin, :onald& The 4uestion of *nimal *wareness& 8ew Hork, ;ockefeller Fniversity Press, "#$D& ) short book about bees, a!es, and other animals, and whether or not
they are -conscious-and !articularly whether or not it is le itimate to use the word -consciousness- in scientific e(!lanations of animal behavior& de6root, )driaan& Thought and 8hoice in 8hess& The ?a ue, /outon, "#DL& ) thorou h study in co nitive !sycholo y, re!ortin on e(!eriments that have a classical sim!licity and ele ance&

6underson, @eith& 'entalit) and 'achines& 8ew Hork, :oubleday, )nchor 0ooks, "#$"& Pa!erback& ) very anti-)I !erson tells why& Sometimes hilarious& [[ ?anawalt, Phili! 7&, and ;obert ?& ?aynes, eds& The 8hemical Basis of +ife& San 'rancisco, 3& ?& 'reeman, "#$%& Pa!erback& )n e(cellent collection of re!rints from the
-cientific *merican& Ene of the best ways to et a feelin for what molecular biolo y is about&

[ ?ardy 6& ?& and E& /& 3ri ht& *n 3ntroduction to the Theor) of Num(ers, Cth ed& 8ew Hork, E(ford Fniversity Press, "#D+& The classic book on number theory& 7hock-full of
information about those mysterious entities& the whole numbers&

?armon, Aeon& -The ;eco nition of 'aces-& -cientific *merican, 8ovember "#$%, !& $+& E(!lorations concernin how we re!resent faces in our memories, and how much
information is needed in what form for us to be able to reco ni<e a face& Ene of the most fascinatin of !attern reco nition !roblems&

van ?ei.enoort, Jean& 7rom 7rege to Gdel: * -ource Book in 'athematical +ogic& 7ambrid e, /ass&, ?arvard Fniversity Press, "#$$& Pa!erback& ) collection of e!ochmakin articles on mathematical lo ic, all leadin u! to 6>del1s climactic revelation, which is the final !a!er in the book&

?enri, )drian& Total *rt: En#ironments, ,appenings, and &erformances& 8ew Hork, Prae er, "#$C& Pa!erback& In which it is shown how meanin has de enerated so far in modern
art that the absence of meanin becomes !rofoundly meanin ful 4whatever that means5&

[ ?oare, 7& )& ;, and :& 7& S& )llison& -Incom!utability-& 8omputing -ur#e)s C, no& % Se!tember "#$B5& ) smoothly !resented e(!osition of why the haltin !roblem is unsolvable&
Proves this fundamental theorem, 11)ny lan ua e containin conditionals and recursive function definitions which is !owerful enou h to !ro ram its own inter!reter cannot be used to !ro ram its own 1terminates1 function&-

?ofstadter, :ou las ;& -Ener y levels and wave functions of 0loch electrons in rational and irrational ma netic fields-& &h)sical =e#iew B, "C, no& D 4"L Se!tember "#$D5& The author1s Ph&:& work, !resented as a !a!er& :etails the ori in of -6!lot-, the recursive
ra!h shown in 'i ure %C&

?ook, Sidney& ed& .imensions of 'ind& 8ew Hork, /acmillan, 7ollier 0ooks, "#D"&
Pa!erback& ) collection of articles on the mind-body !roblem and the mind-com!uter !roblem& Some rather stron -minded entries here& [ ?orney, @aren& -elf%*nal)sis& 8ew Hork, 3& 3& 8orton, "#CB& Pa!erback& ) fascinatin descri!tion of how the levels of the self must tan le to ra!!le with !roblems of self-definition of any individual in this com!le( world& ?umane and insi htful&

?ubbard, John I& The Biological Basis of 'ental *cti#it)& ;eadin , /ass&, )ddison3esley, "#$L& Pa!erback& Just one more book about the brain, with one s!ecial virtue&
however, it contains many lon lists of questions for the reader to !onder, and references to articles which treat those questions&

[ Jackson& Phili! 7& 3ntroduction to *rtificial 3ntelligence& 8ew Hork, Petrocelli 7harter& "#$L& ) recent book, describin , with some e(uberance, the ideas of )I& There are a hu e
number of va uely su ested ideas floatin around this book, and for that reason it is very stimulatin .ust to !a e throu h it& ?as a iant biblio ra!hy, which is another reason to recommend it&

Jacobs, ;obert A& 5nderstanding ,armon)& 8ew Hork, E(ford Fniversity Press, "#L*& Pa!erback& ) strai htforward book on harmony, which can lead one to ask many questions
about why it is that conventional 3estern harmony has such a ri! on our brains&

Jaki, Stanley A& Brain, 'ind, and 8omputers& South 0end& Ind&, 6ateway Editions, "#D#& Pa!erback& ) !olemic book whose every !a e e(udes contem!t for the com!utational
!aradi m for understandin the mind& 8onetheless it is interestin to !onder the !oints he brin s u!& [ Jauch, J& /& *re 4uanta =eal$ 0loomin ton, Ind&, Indiana Fniversity Press, "#$%& ) deli htful little book of dialo ues, usin three characters borrowed from 6alileo, !ut in a modern settin & 8ot only are questions of quantum mechanics discussed, but also issues of !attern reco nition, sim!licity, brain !rocesses, and !hiloso!hy of science enter& /ost en.oyable and !rovocative&

[ Jeffrey, ;ichard& 7ormal +ogic: 3ts -cope and +imits! 8ew Hork, /c6raw-?ill, "#D$&
)n easy-to-read elementary te(tbook whose last cha!ter is on 6>del1s and 7hurch1s Theorems& This book has quite a different a!!roach from many lo ic te(ts, which makes it stand out& [ Jensen, ?ans& -ign, -)m(ol, and -cript! 8ew Hork, 6& P& Putnam1s, "#D#& )or !erha!s theto!-notch book on symbolic writin systems the world over, both of now and lon a o& There is much beauty and mystery in this bookfor instance, the undeci!hered scri!t of Easter Island&

@almar, Aas<lo& -)n )r ument ) ainst the Plausibility of 7hurch1s Thesis-& In )& ?eytin , ed& 7onstructivity in /athematics, Proceedin s of the 7olloquium held at )msterdam, "#L$, 8orth-?olland, "#L#& )n interestin article by !erha!s the best known
disbeliever in the 7hurch-Turin Thesis&

[ @im, Scott E& -The Im!ossible Skew Muadrilateral, ) 'our-:imensional E!tical Illusion-& In :avid 0risson, ed& &roceedings of the BICE *!*!*!-! -)mposium on ,)pergraphics: <isuali@ing 8omple9 =elationships in *rt and -cience & 0oulder, 7olo&, 3estview Press, "#$*& 3hat seems at first an inconceivably hard ideaan o!tical
illusion for four-dimensional -!eo!le-is radually made crystal clear, in an ama<in virtuoso !resentation utili<in a lon series of e(cellently e(ecuted dia rams& The form of this article is .ust as intri uin and unusual as its content, it is tri!artite on many levels simultaneously& This article and my book develo!ed in !arallel and each stimulated the other&

@leene, Ste!hen 7& 3ntroduction to 'athematical +ogic& 8ew Hork, John 3iley, "#D$&
) thorou h, thou htful te(t by an im!ortant fi ure in the sub.ect& Gery worthwhile& Each time I reread a !assa e, I find somethin new in it which had esca!ed me before& & 3ntroduction to 'etamathematics& Princeton, :& Gan 8ostrand 4"#LB5& 7lassic work on mathematical lo ic= his te(tbook 4above5 is essentially an abrid ed version& ;i orous and com!lete, but oldish&

@neebone, 6& J& 'athematical +ogic and the 7oundations of 'athematics & 8ew Hork, Gan 8ostrand ;einhold, "#D%& ) solid book with much !hiloso!hical discussion of such
to!ics as intuitionism, and the -reality- of the natural numbers, etc&

@oestler, )rthur& The *ct of 8reation, 8ew Hork, :ell, "#DD& Pa!erback& ) wide-ran in
and enerally stimulatin theory about how ideas are -bisociated- to yield novelty& 0est to o!en it at random and read, rather than be in at the be innin &

@oestler, )rthur and J& ;& Smythies, eds& Be)ond =eductionism& 0oston, 0eacon Press, "#D#& Pa!erback& Proceedin s of Hr conference whose !artici!ants all were of the o!inion that
biolo ical systems cannot be e(!lained reductionistically, and that there is somethin -emer entabout life& I am intri ued by books which seem wron to me, yet in a hard-to-!in-down way& [[ @ubose, 6yomay& Nen ;oans& 7hica o, ;e nery, "#$%& Pa!erback& Ene of the best collections of kRans available& )ttractively !resented& )n essential book for any 9en library&

@uffler, Ste!hen 3& and John 6& 8icholls& 7rom Neuron to Brain& Sunderland, /ass& Sinauer )ssociates, "#$D& Pa!erback& ) book which, des!ite its title, deals mostly with
microsco!ic !rocesses in the brain, and quite little with the way !eo!le1s thou hts come out of the tan led mess& The work of ?ubel and 3iesel on visual systems is covered !articularly well&

Aacey, ?u h, and 6eoffrey Jose!h& -3hat the 6>del 'ormula Says-& 'ind $$ 4"#D*5&
) useful discussion of the meanin of the 6>del formula, based on a strict se!aration of three levels, uninter!reted formal system, inter!reted formal system, a metamathematics& 3orth studyin &

Aatos, Imre& &roofs and =efutations& 8ew Hork, 7ambrid e Fniversity Press, "#$D&
Pa!erback& ) most entertainin book in dialo ue form, discussin how conce!ts are formed in mathematics& Galuable not only to mathematicians, but also to !eo!le interested in thou ht !rocesses& [[ Aehnin er, )lbert& Biochemistr)& 8ew Hork, 3orth Publishers, "#$D& ) wonderfully readable te(t, considerin its technical level& In this book one can find many ways in which !roteins and enes are tan led to ether& 3ell or ani<ed, and e(citin & [[ Aucas, J& ;& -/inds, /achines, and 6>del-& &hilosoph) %D 4"#D"5, ""B& This article is re!rinted in )nderson1s /inds and /achines, and in Sayre and 7rosson1s The 'odeling of 'ind& ) hi hly controversial and !rovocative article, it claims to show that the human brain cannot, in !rinci!le, be modeled by a com!uter !ro ram& The ar ument is based entirely on 6>del1s Incom!leteness Theorem, and is a fascinatin one& The !rose is 4to my mind5 incredibly infuriatin yet for that very reason, it makes humorous readin & & -Satan Stultified, ) ;e.oinder to Paul 0enacerraf-& 'onist LB 4"#D*5, "CL& )nti0enacerraf ar ument, written in hilariously learned style, at one !oint Aucas refers to 0enacerraf as -self-stultifyin ly eristic- 4whatever that means5& The Aucas-0enacerraf battle, like the Aucas6ood battle, offers much food for thou ht&

& -?uman and /achine Ao ic, ) ;e.oinder-& British ?ournal for the &hilosoph) of -cience "# 4"#D$5, "LL& )n attem!ted refutation of 6ood1s attem!ted refutation of Aucas1
ori inal article&

[[ /ac6illavry, 7aroline ?& -)mmetr) *spects of the &eriodic .rawings of '! 8! Escher& Ftrecht, )& Eosthoek1s Fit evermaatscha!!i., "#DL& ) collection of tilin s of the !lane by

Escher, with scientific commentary by a crystallo ra!her& The source for some of my illustrations e& &, the )nt 'u ue and the 7rab 7anon& ;eissued in "#$D in 8ew Hork by ?arry 8& )brams under the title 'antasy and Symmetry&

/ac@ay, :onald /& 3nformation, 'echanism and 'eaning& 7ambrid e, /ass&, /&I&T& Press, "#$+& Pa!erback& ) book about different measures of information, a!!licable in different
situations= theoretical issues related to human !erce!tion and understandin = and the way in which conscious activity can arise from a mechanistic under!innin &

[ /andelbrot, 0enoit& 7ractals: 7orm, 8hance, and .imension& San 'rancisco, 3& ?& 'reeman, "#$$& ) rarity, a !icture book of so!histicated contem!orary research ideas in
mathematics& ?ere, it concerns recursively defined curves and sha!es, whose dimensionality is not a whole number& )ma<in ly, /andelbrot shows their relevance to !ractically every branch of science&

[ /c7arthy, John& -)scribin /ental Mualities to /achines-& To a!!ear in /artin ;in le, ed& Philoso!hical Pers!ectives in )rtificial Intelli ence& 8ew Hork, ?umanities Press, "#$#& ) !enetratin article about the circumstances under which it would
make sense to say that a machine had beliefs, desires, intentions, consciousness, or free will& It is interestin to com!are this article with the book by 6riffin&

/eschkowski, ?erbert& Non%Euclidean Geometr)& 8ew Hork, )cademic Press, "#DC& Pa!erback& ) short book with ood historical commentary& /eyer, Jean& 11Essai d1a!!lication de certains modles cybernKtiques r la coordination de les insectes sociau(-& 3nsectes -ociau9 TIII, no& B 4"#DD5, "B$& )n article which
draws some !arallels between the neural or ani<ation in the brain, and the or ani<ation of an ant colony&

/eyer, Aeonard 0& Emotion and /eanin in /usic& 7hica o, Fniversity of 7hica o Press, "#LD& Pa!erback& ) book which attem!ts to use ideas of 6estalt !sycholo y and the
theory of !erce!tion to e(!lain why musical structure is as it is& Ene of the more unusual books on music and mind&

& 'usic, The *rts, and 3deas& 7hica o, Fniversity of 7hica o Press, "#D$&
Pa!erback& ) thou htful analysis of mental !rocesses involved in listenin to music, and of hierarchical structures in music& The author com!ares modern trends in music with 9en 0uddhism&

/iller, 6& )& and P& 8& Johnson-Aaird& +anguage and &erception& 7ambrid e, /ass&, ?arvard Fniversity Press, 0elkna! Press, "#$D& ) fascinatin com!endium of lin uistic
facts and theories, hearin on 3horf1s hy!othesis that lan ua e is the same as worldview& ) ty!ical e(am!le is the discussion of the weird -mother-in-law- lan ua e of the :yirbal !eo!le of 8orthern Mueensland, a se!arate lan ua e used only for s!eakin to one1s mother-in-law&

[[ /insky, /arvin A& -/atter, /ind, and /odels-& In /arvin A& /insky, ed& -emantic 3nformation &rocessing& 7ambrid e, /ass&, /&I&T& Press, "#D*& Thou h merely a few
!a es lon , this article im!lies a whole !hiloso!hy of consciousness and machine intelli ence& It is a memorable !iece of writin by one of the dee!est thinkers in the field&

/insky, /arvin A&, and Seymour Pa!ert& *rtificial 3ntelligence &rogress =eport& 7ambrid e, /ass&, /&I&T& )rtificial Intelli ence Aaboratory, )I /emo BLB, "#$B& )
survey of all the work in )rtificial Intelli ence done at /&I&T& u! to "#$B, relatin !sycholo y and e!istemolo y& 7ould serve e(cellently as an introduction to )I& it to

[[ /onod, Jacques& 8hance and Necessit)& 8ew Hork, ;andom ?ouse, Ginta e 0ooks, "#$"& Pa!erback& )n e(tremely fertile mind writin in an idiosyncratic way about fascinatin
questions, such as how life is constructed out of non-life, how evolution, seemin to violate the second law of thermodynamics, is actually de!endent on it& The book e(cited me dee!ly&

[ /orrison, Phili! and Emily, eds& 8harles Ba((age and his 8alculating Engines & 8ew Hork, :over Publications, "#D"& Pa!erback& ) valuable source of information about the life
of 0abba e& ) lar e fraction of 0abba e1s autobio ra!hy is re!rinted here, alon with several articles about 0abba e1s machines and his -/echanical 8otation-&

/yhill, John& -Some Philoso!hical Im!lications of /athematical Ao ic-& =e#iew of 'etaph)sics D 4"#LB5, "DL& )n unusual discussion of ways in which 6>del1s Theorem and
7hurch1s Theorem are connected to !sycholo y and e!istemolo y& Ends u! in a discussion of beauty and creativity&

8a el, Ernest& The -tructure of -cience& 8ew Hork, ?arcourt, 0race, and 3orld, "#D"&
) classic in the !hiloso!hy of science, featurin clear discussions of reductionism vs& holism, teleolo ical vs& nonteleolo ical e(!lanations, etc&

[[ 8a el, Ernest and James ;& 8ewman& Gdel's &roof& 8ew Hork, 8ew Hork Fniversity Press, "#L*& Pa!erback& )n en.oyable and e(citin !resentation, which was, in many ways, the
ins!iration for my own book&

[ 8iever elt, Jur , J& 7& 'arrar, and E& /& ;ein old& 8omputer *pproaches to 'athematical &ro(lems& En lewood 7liffs, 8& J&, Prentice-?all, "#$C& )n unusual
collection of different ty!es of !roblems which can be and have been attacked on com!utersfor instance, the -%n ^ " !roblem- 4mentioned in my *ria with .i#erse <ariations5 and other !roblems of number theory&

Pattee, ?oward ?&, ed& ,ierarch) Theor)& 8ew Hork, 6eor e 0ra<iller, "#$%&
Pa!erback& Subtitled -The 7hallen e of 7om!le( Systems-& 7ontains a ood article by ?erbert Simon coverin some of the same ideas as does my 7ha!ter on -Aevels of :escri!tion-& Peter, ;ci<sa& =ecursi#e 7unctions& 8ew Hork, )cademic Press, "#D$& ) thorou h discussion of !rimitive recursive functions, eneral recursive functions, !artial recursive functions, the dia onal method, and many other fairly technical to!ics&

Muine, 3illard Gan Ernman& The >a)s of &arado9, and :ther Essa)s& 8ew Hork, ;andom ?ouse, "#DD& ) collection of Muine1s thou hts on many to!ics& The first essay deals
with carious sorts of !arado(es, and their resolutions& In it, he introduces the o!eration I call -quinin - in my book&

;an anathan, S& ;& =amanuMan, The 'an and the 'athematician & Aondon, )sia Publishin ?ouse, "#D$& )n occult-oriented bio ra!hy of the Indian enius by an admirer& )n
odd but charmin book&

;eichardt& Jasia& 8)(ernetics, *rts, and 3deas& 0oston, 8ew Hork 6ra!hic Society, "#$"& ) weird collection of ideas about com!uters and art, music, literature& Some of it is
definitely off the dee! endbut some of it is not& E(am!les of the latter are the articles -) 7hance for )rt- by J& ;& Pierce, and -7om!uteri<ed ?aiku- by /ar aret /asterman&

;Knyi, )lfrKd& .ialogues on 'athematics& San 'rancisco, ?olden-:ay, "#D$&

Pa!erback& Three sim!le but stimulatin dialo ues involvin classic characters in history, tryin to et at the nature of mathematics& 'or the eneral !ublic&

[[ ;e!s, Paul& Nen 7lesh, Nen Bones& 8ew Hork, :oubleday, )nchor 0ooks, "#L$&
Pa!erback& This book im!arts very well the flavor of 9enits antirational, antilan ua e, antireductionistic, basically holistic orientation&

;o ers, ?artley& Theor) of =ecursi#e 7unctions and Effecti#e 8omputa(ilit)& 8ew Hork, /c6raw-?ill, "#D$& ) hi hly technical treatise, but a ood one to learn from& 7ontains
discussions of many intri uin !roblems in set theory and recursive function theory&

;okeach, /ilton& The Three 8hrists of Ypsilanti& 8ew Hork, Ginta e 0ooks, "#DC&
Pa!erback& ) study of schi<o!hrenia and the stran e breeds of -consistency- which arise in the

afflicted& ) fascinatin conflict between three men in a mental institution, all of whom ima ined they were 6od, and how they dealt with bein brou ht face to face for many months&

[[ ;ose, Steven& The 8onscious Brain, u!dated ed& 8ew Hork, Ginta e 0ooks, "#$D&
Pa!erback& )n e(cellent book!robably the best introduction to the study of the brain& 7ontains full discussions of the !hysical nature of the brain, as well as !hiloso!hical discussions on the nature of mind, reductionism vs& holism, free will vs& determinism, etc& from a broad, intelli ent, and humanistic view!oint& Enly his ideas on )I are way off&

;osenblueth, )rturo& 'ind and Brain: * &hilosoph) of -cience & 7ambrid e, /ass&, /&I&T& Press, "#$+& Pa!erback& ) well written book by a brain researcher who deals with most
of the dee! !roblems concernin mind and brain&

[ Sa an, 7arl, ed& 8ommunication with E9traterrestrial 3ntelligence& 7ambrid e, /ass&, /&I&T& Press, "#$%& Pa!erback& Transcri!ts of a truly far-out conference, where a stellar rou!
of scientists and others battle it out on this s!eculative issue&

Salmon, 3esley, ed& Neno's &arado9es& 8ew Hork, 0obbs-/errill, "#$+& Pa!erback& ) San er& '&, et al& -8ucleotide sequence of bacterio!ha e fT"$C :8)-, Nature BDL 4'eb& BC, "#$$5& )n e(citin !resentation of the first layin -bare ever of the full hereditary
material of any or anism& The sur!rise is the double-entendre, two !roteins coded for in an overla!!in way, almost too much to believe& collection of articles on 9eno1s ancient !arado(es, scrutini<ed under the li ht of modern set theory, quantum mechanics, and so on& 7urious and thou ht-!rovokin , occasionally humorous&

Sayre, @enneth /&, and 'rederick J& 7rosson& The 'odeling of 'ind: 8omputers and 3ntelligence& 8ew Hork, Simon and Schuster, 7larion 0ooks, "#D%& ) collection of
!hiloso!hical comments on the idea of )rtificial Intelli ence by !eo!le from a wide ran e of disci!lines& 7ontributors include )natol ;a!o!ort, Audwi 3itt enstein, :onald /ackay, /ichael Scriven, 6ilbert ;yle, and others&

[ Schank, ;o er, and @enneth 7olby& 8omputer 'odels of Thought and +anguage& San 'rancisco, 3& ?& 'reeman, "#$%& ) collection of articles on various a!!roaches to the
simulation of mental !rocesses such as lan ua e-understandin , belief-systems, translation, and so forth& )n im!ortant )I book, and many of the articles are not hard to read, even for the layman&

Schr>din er, Erwin& >hat is +ife$ W 'ind and 'atter& 8ew Hork, 7ambrid e Fniversity Press, "#D$& Pa!erback& ) famous book by a famous !hysicist 4one of the main
founders of quantum mechanics5& E(!lores the !hysical basis of life and brain= then oes on to discuss consciousness in quite meta!hysical terms& The first half, >hat is +ife$, had considerable influence in the "#C+1s on the search for the carrier of enetic information&

She!ard, ;o er 8& -7ircularity in .ud ments of ;elative Pitch-& ?ournal of the *coustical -ociet) of *merica %D, no& "B 4:ecember "#DC5, !!& B%CD-B%L%& The source of the ama<in auditory illusion of -She!ard tones-& Simon, ?erbert )& The -ciences of the *rtificial& 7ambrid e, /ass&, /&I&T& Press, "#D#& Pa!erback& )n interestin book on understandin com!le( systems& The last cha!ter,
entitled -The )rchitecture of 7om!le(ity-, discusses !roblems of reductionism versus holism somewhat&

Smart, J& J& 7& -6>del1s Theorem, 7hurch1s Theorem, and /echanism-& -)nthese "% 4"#D"5, "+L& ) well written article !redatin Aucas1 "#D" article, but essentially ar uin a ainst
it& Ene mi ht conclude that you have to be 6ood and Smart, to ar ue a ainst Aucas&&&

[[ Smullyan, ;aymond& Theor) of 7ormal -)stems& Princeton, 8& J&, Princeton Fniversity Press, "#D"& Pa!erback& )n advanced treatise, but one which be ins with a beautiful discussion

of formal systems, and !roves a sim!le version of 6>del1s Theorem in an ele ant way& 3orthwhile for 7ha!ter " alone& [ & 3hat Is the 8ame of This 0ook? En lewood 7liffs, 8& J&, Prentice-?all, "#$*& ) book of !u<<les and fantasies on !arado(es, self-reference, and 6>del1s Theorem& Sounds like it will a!!eal to many of the same readers as my book& It a!!eared after mine was all written 4with the e(ce!tion of a certain entry in my biblio ra!hy5& Sommerhoff, 6erd& The +ogic of the +i#ing Brain& 8ew Hork, John 3iley, "#$C& ) book which attem!ts to use knowled e of small-scale structures in the brain, in creatin a theory of how the brain as a whole works&

S!erry, ;o er& -/ind, 0rain, and ?umanist Galues-& In John ;& Platt, ed& New <iews on the Nature of 'an& 7hica o, Fniversity of 7hica o Press, "#DL& ) !ioneerin
neuro!hysiolo ist here e(!lains most vividly how he reconciles brain activity and consciousness&

[ Steiner, 6eor e& *fter Ba(el: *spects of +anguage and Translation& 8ew Hork, E(ford Fniversity Press, "#$L& Pa!erback& ) book by a scholar in lin uistics about the dee! !roblems
of translation and understandin of lan ua e by humans& )lthou h )I is hardly discussed, the tone is that to !ro ram a com!uter to understand a novel or a !oem is out of the question& ) well written, thou ht-!rovokin sometimes infuriatin book&

Stenesh, J& .ictionar) of Biochemistr)& 8ew Hork, John 3iley, 3iley-Interscience, "#$L& 'or me, a useful com!anion to technical books on molecular biolo y& [[ Stent, 6unther& -E(!licit and Im!licit Semantic 7ontent of the 6enetic Information-& In The 8entralit) of -cience and *(solute <alues, Gol& "& Proceedin s of the Cth International 7onference on the Fnity of the Sciences, 8ew Hork, "#$L& )ma<in ly
enou h, this article is in the !roceedin s of a conference or ani<ed by the now infamous ;ev& Sun /yun /oon& :es!ite this, the article is e(cellent& It is about whether a enoty!e can be said, in any o!erational sense, to contain -all- the information about its !henoty!e& In other words, it is about the location of meanin in the enoty!e& & 'olecular Genetics: * ,istorical Narrati#e & San 'rancisco, 3& ?& 'reeman, "#$"& Stent has a broad, humanistic view!oint, and conveys ideas in their historical !ers!ective& )n unusual te(t on molecular biolo y& Su!!es, Patrick& 3ntroduction to +ogic& 8ew Hork, Gan 8ostrand ;einhold, "#L$& ) standard te(t, with clear !resentations of both the Pro!ositional 7alculus and the Predicate 7alculus& /y Pro!ositional 7alculus stems mainly from here&

Sussman, 6erald Jay& * 8omputer 'odel of -kill *c"uisition & 8ew Hork, )merican Elsevier, "#$L& Pa!erback& ) theory of !ro rams which understand the task of !ro rammin a
com!uter& The questions of how to break the task into !arts, and of how the different !arts of such a !ro ram should interact, are discussed in detail&

[[ Tanenbaum, )ndrew S& -tructured 8omputer :rgani@ation& En lewood 7liffs, 8& J&, Prentice-?all, "#$D& E(cellent, a strai htforward, e(tremely well written account of the many
levels which are !resent in modern com!uter systems& It covers micro!ro rammin lan ua es, machine lan ua es, assembly lan ua es, o!eratin systems, and many other to!ics& ?as a ood, !artially annotated, biblio ra!hy&

Tarski, )lfred& +ogic, -emantics, 'etamathematics! &apers from BIFH to BIHE! Translated by J& ?& Aot, 8ew Hork, E(ford Fniversity Press, "#LD& Sets forth Tarski1s
ideas about truth, and the relationshi! between lan ua e and the world it re!resents& These ideas are still havin re!ercussions in the !roblem of knowled e re!resentation in )rtificial Intelli ence&

Taube, /ortimer& 8omputers and 8ommon -ense& 8ew Hork, /c6raw-?ill, "#D"&
Pa!erback& Perha!s the first tirade a ainst the modern conce!t of )rtificial Intelli ence& )nnoyin &

Tiet<e, ?einrich& 7amous &ro(lems of 'athematics& 0altimore, 6raylock Press, "#DL&

) book on famous !roblems, written in a very !ersonal and erudite style& 6ood illustrations and historical material& Trakhtenbrot, G& *lgorithms and 8omputing 'achines& ?eath& Pa!erback& ) discussion of theoretical issues involvin com!uters, !articularly unsolvable !roblems such as the haltin !roblem, and the word-equivalence !roblem& Short, which is nice& Turin , Sara& *lan '! Turing& 7ambrid e, F& @&, 3& ?effer w Sons, "#L#& ) bio ra!hy of the reat com!uter !ioneer& ) mother1s work of love&

[ Flam, Stanislaw& *d#entures of a 'athematician& 8ew Hork, 7harles Scribner1s, "#$D&

)n autobio ra!hy written by a si(ty-five-year old man who writes as if he were still twenty and drunk in love with mathematics& 7hock-full of ossi! about who thou ht who was the best, and who envied whom, etc& 8ot only fun, but serious&

3atson, J& :& The 'olecular Biolog) of the Gene, %rd edition& /enlo Park, 7alif&, 3& )& 0en.amin, "#$D& ) ood book but not nearly as well or ani<ed as Aehnin er1s, in my
o!inion& Still almost every !a e has somethin interestin on it&

3ebb, Judson& -/etamathematics and the Philoso!hy of /ind-& &hilosoph) of -cience %L 4"#D*5, "LD& ) detailed and ri orous ar ument a ainst Aucas, which contains this
conclusion, -/y overall !osition in the !resent !a!er may be stated by sayin that the mindmachine-6>del !roblem cannot be coherently treated until the constructivity !roblem in the foundations of mathematics is clarified&-

3eiss, Paul& -Ene Plus Ene :oes 8ot Equal Two-& In 6& 7& Muarton, T& /elnechuk, and '& E& Schmitt, eds& The Neurosciences: * -tud) &rogram& 8ew Hork, ;ockefeller Fniversity Press, "#D$& )n article tryin to reconcile holism and reductionism, but a ood bit
too holism-oriented for my taste&

[ 3ei<enbaum, Jose!h& 8omputer &ower and ,uman =eason& San 'rancisco, 3& ?& 'reeman, 'reeman, "#$D& Pa!erback& ) !rovocative book by an early )I worker who has
come to the conclusion that much work in com!uter science, !articularly in )I, is dan erous& )lthou h I can a ree with him on some of his criticisms, I think he oes too far& ?is sanctimonious reference to )I !eo!le as -artificial intelli entsia- is funny the first time, but becomes tirin after the do<enth time& )nyone interested in com!uters should read it&

3heeler, 3illiam /orton& -The )nt-7olony as an Er anism-& ?ournal of 'orpholog) BB, B 4"#""5, %+$-%BL& Ene of the foremost authorities of his time on insects ives a famous
statement about why an ant colony deserves the label -or anism- as much as its !arts do&

3hitely, 7& ?& -/inds, /achines, and 6>del, ) ;e!ly to /r Aucas-& &hilosoph) %$ 4"#DB5, D"& ) sim!le but !otent re!ly to Aucas1 ar ument& 3ilder, ;aymond& *n 3ntroduction to the 7oundations of 'athematics& 8ew Hork, John 3iley, "#LB& ) ood eneral overview, !uttin into !ers!ective the im!ortant ideas of the
!ast century&

[ 3ilson, Edward E& The 3nsect -ocieties& 7ambrid e, /ass&, ?arvard Fniversity Press, 0elkna! Press, "#$"& Pa!erback& The authoritative book on collective behavior of insects&
)lthou h it is detailed, it is still readable, and discusses many fascinatin ideas& It has e(cellent illustrations, and a iant 4althou h re rettably not annotated5 biblio ra!hy&

3ino rad, Terry& 7i#e +ectures on *rtificial 3ntelligence& )I /emo BCD& Stanford, 7alif&, Stanford Fniversity )rtificial Intelli ence Aaboratory, "#$C& Pa!erback& )

descri!tion of fundamental !roblems in )I and new ideas for attackin them, by one of the im!ortant contem!orary workers in the field&

[ & +anguage as a 8ogniti#e &rocess& ;eadin , /ass&, )ddison-3esley 4forthcomin 5& 'rom what I have seen of the manuscri!t, this will be a most e(citin book,
dealin with lan ua e in its full com!le(ity as no other book ever has&

[ & 5nderstanding Natural +anguage& 8ew Hork, )cademic Press, "#$B& ) detailed
discussion of one !articular !ro ram which is remarkably -smart-, in a limited world& The book shows how lan ua e cannot be se!arated from a eneral understandin of the world, and su ests directions to o in, in writin !ro rams which can use lan ua e in the way that !eo!le do& )n im!ortant contribution= many ideas can be stimulated by a readin of this book&

& -En some contested su!!ositions of enerative lin uistics about the scientific study of lan ua e-, 8ognition C,D& ) droll rebuttal to a head-on attack on )rtificial
Intelli ence by some doctrinaire lin uists&

[ 3inston, Patrick& *rtificial 3ntelligence& ;eadin , /ass&, )ddison-3esley, "#$$& )

stron , eneral !resentation of many facets of )I by a dedicated and influential youn !ro!onent& The first half is inde!endent of !ro rams= the second half is AISP-de!endent and includes a ood brief e(!osition of the lan ua e AISP& The book contains many !ointers to !resent-day )I literature& [ & The &s)cholog) of 8omputer <ision& 8ew Hork, /c6raw-?ill& "#$L& Silly title, but fine book& It contains articles on how to !ro ram com!uters to do visual reco nition of ob.ects, scenes& and so forth& The articles deal with all levels of the !roblem, from the detection of line se ments to the eneral or ani<ation of knowled e& In !articular, there is an article by 3inston himself on a !ro ram he wrote which develo!s abstract conce!ts from concrete e(am!les, and an article by /insky on the nascent notion of -frames-&

[ 3ooldrid e, :ean& 'echanical 'an+The &h)sical Basis of 3ntelligent +ife & 8ew Hork, 8J /c6raw-?ill, "#D*& Pa!erback& ) thorou h- oin discussion of the relationshi!
of mental !henomena to brain !henomena, written in clear lan ua e& E(!lores difficult !hiloso!hical conce!ts in novel ways, sheddin li ht on them by means of concrete e(am!les&

C0e13ts
'i ures, 'i & ", ?ohann -e(astian Bach by Elias 6ottlieb-?aussmann 4"$C*5, collection of 3illiam ?& Scheide, Princeton, 8ew Jersey= 'i & B, 7lute 8oncert in -anssouci, by )dolf von /en<el, 8ational alerie, 3est 0erlin= 'i s& %, C, "LB, -The ;oyal Theme- and the last !a e of the -Si(-!art ;icercar&- from the ori inal edition of 'usical :ffering by Johann Sebastian 0ach, are re!roduced courtesy of the Aibrary of 7on ress= 'i ures of litho ra!hs and woodcuts of /& 7& Escher are re!roduced by !ermission of the Escher 'oundation, ?aa s 6emeentemuseum, The ?a ue, co!yri ht the Escher 'oundation, "#$#, re!roduction ri hts arran ed courtesy of the Gor!al 6alleries, 8ew Hork, 7hica o, San 'rancisco, and Aa una 0each= 'i & #, !hoto ra!h of @urt 6>del by Erren J& Turner from 'oundations of /athematics& -)mposium &apers 8ommemorating the -i9tieth Birthda) of ;urt Gdel, edited by Jack J& 0ulloff, Thomas 7& ?olyoke, and S& 3& ?ahn, 8ew Hork, S!rin erGerla , "#D#= 'i s& "$, #D, -'i ure-'i ure- and -) section of m;8) !assin throu h a ribosome,drawin s by Scott E& @im= 'i s& "#, CC, "%%, "C*, musical selections from the 'usical :ffering by J& S& 0ach, music !rinted by :onald 0yrd1s !ro ram -S/FT-= 'i & BL, -7retan labyrinth- from 3& ?& /atthews, 'a@es and +a()rinths: Their ,istor) and .e#elopment , 8ew Hork, :over Publications, Inc&, "#$+= 'i & %#, !hoto ra!h of ;osetta Stone, courtesy of the 0ritish /useum= 'i & C+, ) colla e of scri!ts& Sam!les of cuneiform, Easter Island, /on olian and ;unic scri!ts from ?ans Jensen, -ign, -)m(ol and -cript, East 6ermany GE0 :eutscher Gerla :er 3issenschaften= sam!les of 0en ali and 0u inese scri!t from @enneth @at<ner, The +anguages of the >orld , 8ew Hork, 'unk w 3a nalls, "#$L= sam!les of Tamil and Thai from I& )& ;ichards and 7hristine 6ibson, English Through &ictures, 8ew Hork, 3ashin ton Square Press= 'i & L#, -Intelli ence built u! layer by layer- 4'i & #&*5 ada!ted from Patrick ?enry 3inston& *rtificial 3ntelligence, ;eadin , /ass&, )ddison-3esley Publishin 7om!any, re!rinted by !ermission= 'i s& D%, D#, !hoto ra!hs of an ant brid e by 7arl 3& ;ettenmeyer and construction of an arch by termite workers by Turid ?>lldobler, from E& E& 3ilson, The 3nsect -ocieties, 7ambrid e, /ass&, ?arvard Fniversity Press, "#$#= 'i & DL, schematic drawin of a neuron ada!ted from The 'achiner) of the Brain, by :ean 3ooldrid e, co!yri ht "#D%, /c6raw-?ill, Inc& used with !ermission of /c6raw-?ill 0ook 7om!any, and from 'i & II-D, !a e BD, -peech and Brain%'echanisms, by 3ilber Penfield and Aamar ;oberts, co!yri ht by Princeton Fniversity Press, re!rinted by !ermission of Princeton Fniversity Press= 'i & DD, -the human brain, seen from the left side,- from Steven ;ose, The 8onscious Brain, co!yri ht "#$% by Steven ;ose, re!rinted by !ermission of )lfred )& @no!f, Inc&, 8ew Hork, and John 3olfers, Aondon= 'i & D*, -overla!!in neural !athways,- from John 7& Eccles, 7acing =ealit), 8ew Hork, S!rin er-Gerla , "#$+= 'i s& $$, $*, *+, *B, ""$, "%$, "%*, "C", The -hadows, -tate of Grace, The 7air 8apti#e, The *ir and the -ong, 'ental *rithmetic, 8ommon -ense, The Two ')steries , and The ,uman 8ondition 3 by ;enK /a ritte, co!yri ht by ):)6P, Paris, "#$#= 'i s& $#, #L, -Tobacco /osaic Girus- and -Secondary and Tertiary Structure of /yo lobin- from )lbert Aehnin er, Biochemistr), 8ew Hork, 3orth Publishers, "#$L= 'i s& #", #B, -The four constituent bases of :8)- and -The ladder-like structure of :8)- from )rthur @ornber , -The Synthesis of :8),- -cientific *merican, co!yri ht Ectober "#D*, all ri hts reserved= 'i & #%, -/olecular model of the :8) double heli(,- re!rinted by !ermission, from G& /& In ram, Bios)nthesis of 'acromolecules, /enlo Park, 7alifornia, The 0en.aminI7ummin s Publishin 7om!any, "#$B= 'i & #$, -Polyribosome- from The &roteins, edited by ;& E& :ickerson and ?& 8eurath, !a e DC, 8ew Hork, )cademic Press= 'i & #*, -) two-tiered molecular canon- from E& A& /iller, Jr&, -Gisuali<ation of 6enes in )ction,- in -cientific *merican, co!yri ht /arch "#$%, all ri hts reserved= 'i s& "+", "+B, "+%, -The TC bacterial virus,- -Infection of a 0acterium by a virus,- and -The mor!ho enetic !athway of the TC virus,- from 3illiam 0& 3ood and ;& S& Ed ar, -0uildin a 0acterial Girus- in -cientific *merican, co!yri ht July "#*$, all ri hts reserved= 'i & "+L, !hoto ra!h of Srinivasa ;amanu.an from S& ;& ;an anathan, =amanuMan, the 'an and the 'athematician, 8ew Hork, )sia Publishin ?ouse, "#D$= 'i s& ""+, """, ""B, from Terry 3ino rad, 5nderstanding Natural +anguage, 8ew Hork, )cademic Press, "#$B= 'i & ""%, Photo ra!h

of )lan Turin by /ssrs& 7& ?& E& Trevelyan from Sara Turin , *lan '! Turing, 7ambrid e, En land, 3& ?& ?effer and Sons, Atd&, "#L#= 'i & ""D, -a meanin ful story in )rabic- from )bdelkebir @hatibi and /ohammed Si.elmassi, The -plendor of 3slamic 8alligraph), 8ew Hork, Aondon, Thames w ?udson, co!yri ht by Marawiyne Aibrary in 'e<& 'i ""*, Procedural re!resentation of -red cube which su!!orts a !yramid,- ada!ted from 8omputer 'odels of Thought and +anguage , edited by ;o er 7& Schank and @enneth /ark 7olby& 3& ?& 'reeman and 7om!any, co!yri ht "#$%= 'i s& ""#, "BB, "BC, "%+, 0on ard !roblems from /& 0on ard, &attern =ecognition, ;ochelle Park, 8ew Jersey, ?ayden 0ook 7om!any, S!artan 0ooks, "#$+ 6rateful acknowled ment is made to the followin !ublishers for !ermission to quote e(cer!ts from the followin material, The Bach =eader: * +ife of ?ohann -e(astian Bach in +etters and .ocuments , edited by ?ans T& :avid and )rthur /endel, ;evised, with the !ermission of 3& 3& 8orton w 7om!any, Inc&, co!yri ht "#DD, "#CL by 3& 3& 8orton w 7om!any, Inc& co!yri ht renewed "#$B by /rs& ?ans T& :avid and )rthur /endel= J& S& 0ach1s 'usical :ffering, !a e "$#, edited by ?ans T& :avid, 8ew Hork, co!yri ht "#CL by 6& Schirmer, Inc&, used by !ermission= 6yomay @ubose, Nen ;oans, 7hica o, ;e nery, "#$%= Pauls ;e!s, Nen 7lesh, Nen Bones, Tokyo, Ja!an, 7harles E& Tuttle 7o&, Inc& "#L$= J& ;& Aucas, -/inds, /achines, and 6>del,- and )lan /& Turin , -7om!utin /achinery and Intelli ence,- from 'inds and 'achines, edited by )& ;& )nderson, En lewood 7liffs, 8ew Jersey, Prentice-?all, "#DC, and &hilosoph), vol& %D, "#D"= J& /& Jauch, *re 4uanta =eal$ 0loomin ton, Indiana, Indiana Fniversity Press, "#$%= James ;& 8ewman, -Srinivasa ;amanu.an,- in The >orld of 'athematics, edited by James ;& 8ewman, 8ew Hork, Simon w Schuster, re!rinted by !ermission of Simon w Schuster, a division of 6ulf w 3estern 7or!oration, "#LD= Terry 3ino rad, -) Procedural /odel of Aan ua e Fnderstandin ,- from 8omputer 'odels of Thought and +anguage , edited by ;o er 7& Schank and @enneth /ark 7olby, San 'rancisco, 3& ?& 'reeman and 7om!any, co!yri ht "#$%= Jose!h 3ei<enbaum, 8omputer &ower and ,uman =eason: 7rom ?udgment To 8alculation, San 'rancisco, 3& ?& 'reeman and 7om!any, co!yri ht "#$D&

In1e=
)!!earances of )chilles and the Tortoise in :ialo ues are not inde(ed, but those of less frequent characters are& The reader is encoura ed to consult the fi ure on !a e %$+ for !ossible hel! in crossreferences& **BB form, "%+, BB$ )bel, 8iels ?enrik, C+C )0E;T 40looP5, C"B abortion, "$D levels of, DL"-L%, DLD-L$, DD+, abstraction, tiDD-$B, D$C )chilles, answer-schemas and, C$L= 7arroll !arado( and, CD, "$+, "*", "#%= 7rab 7anon and, B+C DDD-D$= heel of, %*#, C*C= ima e of 6E:, BB%= inaccessible neurons of, %B*& D*D= initial letter of, B%", L+$, DD$= innocence of, C+D, C+*, CBC= ma!!ed onto ant colony, %"*, %BC= mentioned *C, B$B, LBL, L$%, DD#= mystified by 7rab, LD+= orchard analo y and, CB$= ori in of, B*, B#= !icture of, CB= !roblem reduction and, D"+-""= recursion and, "B*-%", "C#= as violin, L+B )chilles !ro!erty, %#D-#*, C"L acoustico-retrieval, B$*-*+, CD+ acronyms, %"-%B, ""%, "$C, "$D, B+C, B%$, B$B, %$C, D+D, D*C, $B$, $%D, $%*, $C+ acrostics, $, *" active site, LB*-B#, LCC actor formalism, DDB-DC -)ctually Intelli ent- !ro rams, D$D addition, )I !ro rams and, D$$-$*= commutativity, associativity of, LL, BBL-BD= in 0looP, C+#= noncommutative, BBB-B%, D%#= !qsystem and, C#-L%, C"$= re!resentability of, C"$= of su!ernaturals, CLL= T8T notation for, B+D-$= tri!le, "+", B+D-$ addresses 4in memory5, B*#, B#+ adenine, see nucleotides advertisements, C$* )I, a!!lied to mathematics, L$%, D"C-"L= ar uments a ainst, L#$-##= attitudes a ainst, see anti-)I attitudes= com!uter lan ua es and, B##-%++, LC*= conver in towards brains, L$#= defined, BD= difficulty of, BD-B$, L$%, $C+= evidence and, D#L= faith underlyin , L$B, L$*$#= 6>del1s Theorem and, %**-#+, C$"-$$, $+D$= history of, "#, BC-B$, L#C-#$, D++-#= leveltranslation and, B*L= outline of, D+"-%= relation to mathematics, LL#-D+= stratification of, B##%++ )I !ro rams, com!ared with !eo!le, D$#-*+= curiosity of, D$#= Aucas1 ar ument and, L$$-$* )I Thesis, L$# *ir and the -ong, The 4/a ritte5, C#C *ir on G's -tring, CCL, CCD, C#$ air!lane-haystack story, D$L aleatoric music, "D%, "$C, $++ )l ol, B#B, B#%, %*", D%+ al orithms, B#B, C"+, C"B, C"%, C"C, CC+, CL#, LD$ all D+ -almost--situations, D%C-C+, DC"-C%, DC# al!ha heli(, LB", LBL )lternative Structures of the Fnion, see )SF1s ambi uity, com!uter lan ua es and, B#$-#*= in translation into T8T, B+#-"" amino acids, L"*-BL, L%%-%C= in Ty!o enetics, L+*, L"+-"" analo ical thinkin , by com!uter, D+%= under!innin of, L$+-$" analo ies, DD*-$C )nalytical En ine, BL, L#*, $B$ ana!hase, DD$ anarchy, D#% -and-, "$$-*+, "*", "*D, D%+ )nderson, )lan ;oss, "#$ )nnouncer, D%%-C+ anomalies, CC, #D, B+*, C%L, $B% *nother >orld 4Escher5, BL+, BLL answer-schemas, CDB-"C, C$L, D** ant brid e, %%C ant colonies, artificial, %L#= caste distribution in, %"*-B*= castes in, %"$-"*= communism in, %"*, %%+-%"= com!ared with brains, %"L-"D, %"*, %BC-BL, %L+, %L*-L#= contrasted with ases, %"$= freedom and control in, %"L-"D, %B$= intelli ent, %"+-%D= levels in, %"#-B$= mechanisms of teams and si nals, %"$-B"= order and chaos in, %"D-"$= re rou!ed, %%B-%%= si nals in, %B+-B*= symbol level in, %BC-B*, %%+= teams in, %"$, %"#-B$= thresholds in, %"D"$, %"#-B"= trails of, %"L-"$= see also caste

distribution, )unt ?illary, 'ermant, Joh& Seb&, 'ourmi, Aierre de, teams, si nals, symbols *nt 7ugue, %%$, %C#, %L+, %*B, L$+, D*D, $%$-%# "*nt 7ugue" 4Escher5, %BB-B% )nteater, B$L-*C, %""-%D, %*B, L$+, $BB anti-)I attitudes& B$, C$+-$B, DB* anticodons, LBB-BC ants, dis!ensability of, %BD= nest of, %L#= vs& ant colonies, %"C, %"L, %"#, %B", %BD, %%+ a!eriodic crystals, "D$-D#, "$C-$D arch of termites, %L*-L# )ria with :iverse Gariations 40ach5, %#B-#%, %#L )ria with :iverse Gariations 4:ialo ue5, C+* )ristotle, "# arithmeti<ation, BDB-DL, BD*-D#, L%%-%C arithmoquinin , CCL-LC, CDD-D*, C#$, L+B, LC", L*+-*" arms, mutually washin , D#" art, identity crisis of, D##-$++, $+%-D= modern, D##-$+D= by com!uter, D+%, D"#-B+ *rt%+anguage, DBB *rt of the 7ugue 40ach5, $#-*", *D, D$" )rt of 9en Strin s, B%$, B%#-CB, BCC, DBD )rtificial Ima ery, LD+ )rtificial Intuition, LD+ )rtificial Ism, DBL artificial -thinkin -, %%$, D+" *scending and .escending 4Escher5, "B-"%, "L, B", $"D )sian 0o(, Gery 6old, see 6old 0o(, Gery )sian assemblers, B#", B#C assembly lan ua e, B#+-#L= com!ared to :8), B#+-#" assembly lines, cellular, LB*-B#, LCC-CL associativity, %L-LD, B+$ assum!tions, shaky, CB+, CBB, CB#& L*+, L*", DCC, D$B )SF1s, defined, %$%-$C= orientation in, %$L= tri!s in, %$$-$#, %*% )T81s, see )u mented Transition 8etworks )T8colony, %L# atoms, in !hysics, %+%= in Pro!ositional 7alculus, "*B, "*D= in T8T, B+$, B"%, B"C )TT)77), "$D, B*C, %"", L+L attachment and nonattachment, B%B-%%, BCC au mentation, *, #, "CD, L+%, D*"-*%, $%*= intervallic, "LD-L$ )u mented Transition 8etworks, "L+, BL*-L#, %L#, DB", DBL-B$, D%+ )u ustus II, Elector of Sa(ons, @in of Poland, CD"

)unt ?illary, %"C-%%, %*B, D%+ )uthor, the, %, D, B+C, %$+, D*#, $BB, $BC-CB authorshi! trian le, #C-#L, #D, D*# autolo ical ad.ectives, B+-B" automatic !ro rammin , B#*-## )vers, Eswald& "L#, "D" awareness, deductive vs& analo ical, L$+-$", D"# a(iom schemata, C$, C*, DL, *$, CD*, C$B-$%, LC% a(iomatic systems, see formal systems a(ioms, defined, %L= lack of in Pro!ositional 7alculus, "*%= of /IF-system, %%, %L= of Psystem, $%-$C= of !q-system, C$= of !q-system, modified, *$= of %"+-system, BD%= of T8T, B"D= of T8T, e(tended, CL"-LB, CDD-D*= of tqsystem, DL a(ons, %%#-C+ 0abba e, 7harles, BC-BD, L#*, D+", $BD-CB 0abba e test, $%L-$% 0)7? 4acronym5, "$C-$L 0-)-7-? 4melody5, $#-*", *D, "+B, "B", "LL-L$, BDD, $"# 0ach, )nna /a dalena, C*B 0ach, 7& P& E&, %, C, *+ 0ach, Joh and Seb&, D%%, DD# 0ach, Joh& Seb&= )I and, B$, D$$= as com!oser, %#B, CD", $C+-CB= as lassblower, $#= as har!sichordist, B$L, B$#, B*+= as ins!iration for :ialo ues, B*, $%$= confused with 'ermat, %%"%L= de!th of, $-*, #-"+, B$-B*, $+-$", D$$= dissection vs& a!!reciation of, D*+= Escher and, "##, DDD-t$= 'orkel on, C, *D= homa e to, *"= im!rovisation by, %-$, #D, $"#= in Aei!<i , %*%, C+C= indirect self-reference of, $#-*", *D= life and death of, *D= modulation and, "BB-B%, "%+= recursive qualities of music of, $+-$"= She!ard tones and, $"#= squared, D$#= vs& 7a e, "L$, "DBC, "$C-$L= see also Eld 0ach 0ach, 3ilhelm 'riedemann, C, D back round assum!tions, DCC backtrackin , #$, DB#, D%B backwards chainin , D"* bankers, CL$ base-!airin , com!lementary, C%%, L+D-$, L"C-"D, L"$, LB%, L%C baseball, D%$-%* bases 4 enetics5, L"C, see also nucleotides= in Ty!o enetics, L+L-*, L"+ 0aso, B%C 0assui, BLL

beautiful vs& non-beautiful, LLB-L*, LD+, L$%-$L, & L*"-*B beauty, com!uters and, L$L= elusiveness of, LLC, LLL, LDL, L$C-$L, L*", L*%-*C bees, %D+, DC", $B+ 0eethoven, Audwi van, D, $L, "D%, D%C beliefs, catalo ue of& %*B, %*C Birthda) 8antatatata 4:ialo ue5, CD*, C$L, D** birthdays, CD"-DC bits, B**-*#, B#+, B#" 0AE7@ 40looP5, C"+-"" blocks world, L*D-#%, DB$-%B, D$C 0looP, C+D& C+#-%+, CC+, CC", CCC= al!habet of, C"#, CBL= control structures in& C"+= !rimordial ste!s of, C+#, C"B-"%= synta( of, C"+-"L 0looP !ro rams, C"+-"C 0lue Pro rams, C"*-B+, CBB, CB$ Bluediag U8V, CB-CC, CB* board evaluation, static and dynamic, D+C-L, D""= Stran e Aoo!s in, D+C-L 0odhidharma, B%B, B%*, BCL& BLB, DBL 0olyai, 'arkas, #B 0olyai, Jfnos, #"-#B 0on ard, /&, DCD 0on ard !roblems, DCD-DB, DDC& DD#, D$C= universality of, DD"-DB 0oole, 6eo&, B+, C+C, D++ 0oolean 0uddhism, L$$ boomeran s, $L, *C, BD$, C$B bootstra!!in , BC, B#%-#C, LBC, LC* bottom 4recursion5, "%#-C+= see also skeletons bottom-u! vs& to!-down, C*-C# bottomin out, "%%-%L, "L", BL#, %+", DL+ boustro!hedonic writin , "D*-D#, "$D bo(ed systems, CD#-$+, LC% 0oyle1s law, %+*= see also ases brains, ant colonies and, %"L-"D, %"*, %BC-BL, %L+, %L*-L#= as )T8-colonies, %L#= as mathematical ob.ects, LL#= E!imenides !arado( and, L*C-*L= formal systems and, %%$-%*, LL#DB, LD#-$#, L*C-*L, D$D, see also 7hurchTurin Thesis, formal vs& informal systems, brains and rules= frame and outer messa es and, "$+-$"= ma!!in s between, %C"-CB, %CL-CD, %D#-*B= music and, "D%= o!eratin systems and, B#D= !ro rammability of, %+B= rules and, BD-B$, D$D, see also brains and formal systems= subor ans of, %C+-C"= thou hts and, %%$-DL= vs& minds, L$D= see also minds, intelli ence, etc& 0reton, )ndre, $++ -0ritish 6renadiers, The-, D+$

0ell, )& 6&, B#D 0elna!, 8uel, "#$ 0eno, Auciano, $+C bicentu!lets, LC+ bifurcations, #"-#C, "++, CLD-L#, CD$, L$# bio.ukebo(es, "D+, "$L Birthda) 8antata 40ach5, CD" 0rouwer, Auit<en E& J&, C+C 0uddha-nature, B%%, B%C, B%*-CC Butterflies 4Escher5, "C$-C* 0u(tehude, :ietrich, %%L 0yrd, :onald, see S/FT 0yron, Aord, BL 7-system, DL-D$, $"-$B 7)6E 4acronym5, "$C-$L 7-)-6-E 4melody5, "LD-L$ 7a e, John, "LD-L$, "D%-DC, "D$, "$C-$L, LLD, D##-$++, $+C calculatin !rodi ies, see idiots savants 8anon () 3nter#allic *ugmentation, LBL -7anon !er au mentationem et contrario motu40ach5, see -loth 8anon -7anon !er Tonos- 40ach5, see Endlessly ;isin 7anon canons, co!ies and, *-#, "CD= :ialo ues and, DDLD#, $%*= Escher drawin s and, "L= in 6oldber Gariations, %#B= in 'usical :ffering $-"+, $BDB$= !olyribosomes and, LBD-B*= self-refs and, L+"-%= structure of, *-"+= two-tiered, LB$-B*= see also individual canons, fu ues 7antor, 6eor , B+, B"D, C"*, CB", CBB-BC 7antor set, "CB -7antorcrosti!unctus-, CBC 7a!itali<ed Essences, B# car radio& D$+-$" cardinality, intuitive sense of, LD$ 7arroll, John 0&, D%+ 7anon, Aewis, B+, B*, CD, "#B, %$B, D*"= material by, C%-CL& %DD-D* 7arroll !arado(, B*, C%-CL, D*"= evidence version, D#%-#C= messa e version, "$+, !roblem !osed by, CD& "*"= !roof version, "#B-#%= Samuel1s ar ument and, D*C-*L= symboli<ed, "#%= see also infinite re ress cascades, BBC, LB#, DBD, DDC caste distribution, encodin of knowled e in, %"#, %BC-B*, %L#= meanin of, %B"-BC= u!datin of, %"*-"#, %BC catalo ues of !ro rams 40lue, 6reen, ;ed5, C"#, CB$-B*

catalysts, LB*-B# cats, %"%, %C%-CD, L%B causality, ty!es of, $+#-"+ 77rab, see )TT)77) ceilin s, see loo!s, bounded= 0looP celestial mechanics, %L%-LC 7EAAS 40looP5, C"+-"" cellular !rocesses, as models for )I, DD%-DC 7entral 7rabma!, DD$ 7entral :o ma, of /athematical Ao ic, B$", L%B-%C= of the /IF-system, L"%= of /olecular 0iolo y, L+C-L, L"C, L%B-%C, L%D, DD$= of Ty!o enetics, L"%= of 9en strin s, B%*, B%#, BC+, BC% 7entral :o ma!, L%B-%C, LCL, LC$, D$B, $+# 7entral Pi!ema!, $+"-B central !rocessin unit, B**, B*# 7entral Pro!osition, BDC, BD# 7entral Slothma!, $+B 7entral Tma!s, $+B, $"D= see also individual entries centrality, %$C-$L centromere, DD* cerebellum, %C" 7hadwick, John, L+ chain letters, LCD 7ham!ernowne, :avid, L#L 7ham!ollion, Jean 'ranvois, "DL channelin , B##, %$D-$$ chaos in number theory, "%$-%*, "LB, LL$= see also order and chaos chauvinism, "$"-$% checkers !ro rams, L$%, D+C-L 7hekhov, )nton, DCB chess, chunkin and, B*L-*$, D+C= rand masters in, B*D-*$= round-the-house, L#L= selfmodifyin , D*$-** chess boards, hierarchy of, D*$ chess !layers, cycle of, #C-#L chess !ro rams, 0abba e and, BL, $B#-%", $%D= choice and, $""-"B= 7rab and, $B", $B#-%"= difficulty of, "L"-LB, D+L= .um!in out of the system and, %$-%*, D$*= knowled e re!resentation in, D"*= recursive structure of "L+-LB= stren ths and weaknesses of, "L"-LB, B*L-*$, L$%, D+%-C, D""= Turin and, L#L, L#D, $%D, varieties of, D+"= without look-ahead, D+C chests of drawers, nested, DCC-CL children1s stories and )I, D$L-$D 7hiyono, BLD choice, $""-"C 7ho!in, 'rKdKric, $+, BL$, D$$

chords and analo ies, D$%-$C chromosomes, homolo ous, DD* chunked versions of this book, .acket, #i%#ii, #iii% 9iii, %$+, $L*-$$ chunkin , ant colonies and, %BD-B$= brains arid, %*"-*C, LL#= com!uter lan ua es and, B#+-#B, %*", C"B-"%= defined, B*L-**= determinism and, %+D-*, %D%, LBB= of :8), L%"-%B= intuitive world-view and, %+L-D, %DB-D%= of music, "D+, "DC, LBL= of one1s own brain, %*B= !robabilistic, %*C= scientific e(!lanation and, %+L-D= su!er conductivity and, %+L= trade-offs in, %BD= vision and, %C* 7hurch, )lon<o, CB*, C$D, LD" 7hurch-Turin Thesis, CB*-B#, LLB, LD"-$#= )I Gersion, L$*-$#, L*+, L*"= ?ardy1s Gersion LDD= Isomor!hism Gersion, LD$-D*= /icrosco!ic Gersion, L$B= Public-Processes Gersion, LDB, LD*, L$C, L*+= ;eductionist1s Gersion, L$B, L$C= Soulist1s Gersion, L$C= Standard Gersion, LD"DB, L$#= Tautolo ical Gersion, LD"= Theodore ;os<ak Gersion, L$C-$L= un!rovability of, LDB 7hurch1s Theorem, LD+-D", L$C, L$#-*", D+#, D#$ ci ars, "##, B+", %*%, C*", DL" classes vs& instances, %L"-LL, %D+-D"= see also !rototy!es, intensionality and e(tensionality, analo ies, conce!tual skeletons, etc& codes, art and, $+%-C= familiar and unfamiliar, *B, "L*& BD$= see also decodin , 6>del 7ode, 6enetic 7ode, etc& -codin - of sentences, L*%-*C codons, L"#-B+, LBC, L%%, L%L= see also 6>del codons, du!lets 7olby, @&, L## columns in brain, %CD 7omenius, Johann )mos, DBL comments in !ro rams, B#$ 8ommon -ense 4/a ritte5, $++-" common sense and !ro rams, %+" communicability of al orithms, LDB commutativity, LL-LD, B+#, BBL-B$, CL%, D%# com!ellin inner lo ic, "D"-DB, "D%-DC com!etin theories, and nature of evidence, D#L com!iler lan ua es, B#B-#L com!ilers, B#B-#L, B#$, L+% com!ilin , reverse of, %*" 7om!lete Aist of )ll 6reat /athematicians, see Aist com!leteness, "++-B, C"$-"*, CBB, CDL= see also incom!leteness, consistency com!le(ity of world, LD#

com!osite numbers, DC, DL-DD, $%= see also !rime numbers com!ound sentences, in T8T, B"C com!ound words, DDL com!uter chess, see chess !ro rams com!uter lan ua es, analo ues in cell, LC$= dialects of, L+%= fle(ibility and, B#*-##= hi h level, B#B-#%, B#$-%++= messa e-!assin , DDBD%= !ower of, B##, CB*-B#= !resented, B*#-##, C+D-%+, C#*-##= in S?;:AF, DB#-%B com!uter systems, B*$-%+B com!uters, assembled by com!uters, L+C, D*C= cryin , D$L-$D= determinism and, BL-B$, %+D-$, D*C-*D= fallibility of, L$L, L$*, D$*= learnin by, D+%-L= ori ins of, BC-BD= in !hono ra!hs, $*, C*C, C*D-**= see also !ro rams, )I, )I !ro rams conce!t network, DL"-LC= see also semantic networks conce!tual dimensions, D$+-$" conce!tual ma!!in & DD*-$B conce!tual nearness, %$"-$%, D"C, DL"-LD conce!tual revolutions, DD+-D", D$% conce!tual skeletons, %*", L"C& DDD-$B, D$C connotations and culture, %$B-$%, %$#-*+ consciousness, causality and, $+#-"+= source of, %*C-*L, %*$-**, understandin of, *B, D*+, $+*"+ conservation of com!le(ity, D+, "#L consistency, defined, #C= of e(tended T8T, BB%, CL#= hy!othetical worlds and, #L-"++= inter!retations and, **, #C-"+"= Aucas and, C$$= oath of, in T8T, CL+= !roofs of, B%-BC, "#"-#B, BB#-%+, CC#-L+= of Pro!ositional 7alculus, "#"#B, BB#= of T8T, BB#-%+, CC#-L+= varieties of, #C-#D= see also w-consistency constants, !arameters, variables, DC%-CC, DD# conte(t, necessity of, "D"C, "$%-$D= restorin of, ""L-"D, "B*, "%%, "D"-DC, "$%-$D conte(t-free sums, LB+-BB conte(ts, nested, DC%-CD, D$B, D$C continued fractions, "C+, B$$, LD%, LDL continuous vs& discrete !rocesses, L#* 7ontracrosti!unctus, discussed, *B-*L, BD$, B$+$", C+D-$, CBC, CD$-$+, C*%-*C, L%C-%$, D+*, $B" contradictions, between levels, see levelconflicts= caused by im!ossible cycles, #C-#D= coe(istin in same brain, %*%, D#$-#*= dia onal ar ument and, CB+-BB= in mathematics, "$-BC, "#D-#$, BB%, L*+-*"= so-consistency and, CL%=

!ersonal none(istence and, D#*= in !q-system, *$, **= in Pro!ositional 7alculus, "#"-#B, "#D#$= in self-ima e, D#D= Tortoise1s shell and, "$$*+= two levels of, L*", L*C= visual, #$-##= 9en and, ##, B%L, BCD-LD, D#*= see also !arado(es, inconsistency, E!imenides !arado(, etc& 8ontrafactus, DC", DC%, DD# 8on#e9 and 8onca#e 4Escher5, "+L-#, %C* 7oo!er !airs, %+C-L co!ies, canons and, *-#, LB$-B*= in code, L"$, LB$-B*= com!lementary to ori inals, $+, L+", L+D-$, L"$, see also inversion= :8) and, LB#%"= ine(act, L++-%, LCD= nature of, "CD-C#= selfre!s and, L++-C, L"B-"%= television and, C*#= viruses and, LCB-C%= visually nested, "%*-C+= see also sameness, isomor!hisms co!!er, "$% 7o!s Silva and 6ould, C+L 7o!y mode 4Ty!o enetics5, L+D-* corte(, areas of, %CC= cerebral, %C+-C*= visual, %C%-C* counterfactual !arameters, D%# counterfactuals, D%C-C+, DC"-CC counter!oint, see canons, fu ues, 0ach, etc& countin , LL-L$, BB*, %DC court system, D#B, D#% covalent bonds, L"C cows, %"B, %CD, %L" 7PF, see central !rocessin unit 7rab, entertains )chilles, C*+-#C= enes of, B++", B+C& L+$= hikes and !lays flute, LC#-L*= intelli ence of& LC#, LL*, $%%= .ukebo( of, "LCL$= meets )chilles, B++= musical evenin che<, $B+-CB= ori in of, DDD-D*= !li ht of, B*", %*%*C= questionable behavior of, LD+, LDB, L$%-$C, L$#-*"= receives !resents and entertains uests, B$L-*C, %""-%D= sub.unctive afternoon che@, D%%-C+= Theme of, $B#, $%B& $C+, $CB= vs& Tortoise, $L-$*, C+D, C#%-**,LC+, LC% 8ra( 8anon 40ach5, B+B-%, DDD 8ra( 8anon 4:ialo ue5, B+C, %LL-LD, DDL-D#, D$B, $B%-BL, $%* 8ra( 8anon 4Escher5, "#*-##, DD$ crab canons, #, "#*-B+%, B+C, %LL, L+"& DDL-D#= in :8), B++-" crab !ro rams, L++-" creativity, mechani<ability of& BL, BD, L$", DB+, D$%= see also ori inality, !arado( of )I, non !ro rammability, etc& 7rick, 'rancs, L+L, L%B, L%%, L%C, D"$ 8rime and &unishment 4:ostoevsky5, %$#-*+

7ristofori, 0artolommeo, % critical mass, BB*, %"$, %*#, C"$, C$+ crossin -over, DDL-D* crystal in ma netic field, "C+-C% crystalli<ation meta!hor, %C$ 7T-Thesis, see 7hurch-Turin Thesis 8u(e with 'agic =i((ons 4Escher5, B*"-*B cursively drawable fi ures, D$-D*, $B cyto!lasm, L"$, L"*, LBB-BC cytosine, see nucleotides da Ginci, Aeonardo, DC" :ali, Salvador, $++ :ase, Johann /artin 9acharias, LD$ data base, D"* dau hter !ro rams, L+%, LCD :avid, ?ans Theodore, %, B*, $"# .a) and Night 4Escher5, BLB, BLL, DD$ :bou!s, ?f!sh, L"L, L%% :e 7hirico, 6ior io, $++ :e /or an, )u ustus, B+, C+C, D++ decidable strin s, C"$ decimal system, BDB-DC, BD# deci!herment of te(ts, L+, "DC-DL, "$%-$C, L*% decision !rocedures, for alienness, C*$-**, LC+C"= for a(iomhood, C", C*, C$+= for beauty, LLBL*, LD+, L$#, L*"-*B, L*%-*C= as 0looP !u<<les, C"L-"$= for 0uddha-nature, B%C, B%#, B$B= defined, %#-C"= for :io!hantine equations, LL#D+= for dreamable themes, %*C= for end of :ialo ue, C+B-%= for enuineness of kRans, B%C, B%#= for 6oldbach !ro!erty, C++, C"C= for 6ranny, %CC-CL, %C$-C*= for haltin , CBL-B#= for intelli ence, see Turin test= for /o<art !ieces, DC#= no uarantee of e(istence of, $B= for number-theoretic truths, BB*-B#, CBD, LL"L*, LD+& L$%-$C, L$#-*"= for !rimality, DC, "C#, C"%= for !roof-!air-ness, C"D, C%#-C"= for sameness, "CD-C#, "L*-L#= for sortin numbers into two classes, see 7hurch-Turin Thesis= for termination, CBL-B#= for -the 3ay- in 9en, BL+L", BL%, BLC= for theoremhood, %#-C", C$-C#, $B-$%, "#+-#", C+*, C"D, LD+, L$#-*+, L*B= for theorem-numbers, CC+-C", L*+= to!down vs& bottom-u!, C*-C#= for Tortoise-!airness, CC"= for Tortoise !ro!erty, %#D-#$, C"L, CC"= for truth, B"%, BB*-B#, C"$, LLB-L*, LD+-D", L$#, L*"= for validity of derivations, "#C, C"D, C%#C", C$+= for well-formedness, "*B, BD#, C"D, L*B= for wondrousness, C+B, CBL

decodin , of :8), "L#-DB, "$L-$D, B+", B%", L%"-%B, L%*= of formal systems, L+-L", LC= of fortune, "LC= via 6>del isomor!hism, BD$= of nature& C+#= of records, "LC-L$, "L*-L#, "D"-DC, "$B, "$C-$L= as revelation, "D+-D"= of ;ussian te(t, %*+= see also isomor!hisms, translation, information, chunkin decodin mechanisms, com!le(ity of, "L*-DB, "$B-$D, L*B-*C= innate, "$+-$"= nature of, "L*$D= record !layers as, *%, "LC-L$= trans!arency of, BD$, L+"= for Tri!itaka, BL$= see also isomor!hisms, etc& :eduction Theorem, "*D default o!tions, %LB-L%, %*D, C"", DCL, D$C defects and e(!ectations, $$, *D, "+B, BBB, C$D demidoublets, D%%-%C, DD# demons, DD%-DC deo(yribonucleic acid, see :8) derivations, alle ed, C%#-C+= defined, %L-%D, fallacious, BB+, C%#= in /IF-system, %L-%D, BDB, BDC, C%#= in Pro!ositional 7alculus, "*C, "*L, "**, "*#-#+, "#D= su!ernatural, CLC-LL= in T8T, B"$, B"*, B"#, BBC-B$, BD#= in tq-system, DL= in Ty!o enetics, L+$, L+#= vs& !roofs, %L%D, "#%-#C, "#L :escartes, ;ene, BD%, %C+, D$$ descri!tion-schemas, DL+= see also tem!lates descri!tions, calculus of, %%*= restructurin of, DC#-L%, DL#-D", D$B= tentative, DCD-C# descri!tors, DC$ detachment, rule of, "*L-*D, L$$ determinism, LC= see also free will :evil, D*L .ewdrop 4Escher5, BC#, BLD :i of )ntus, %%% dia onal method, 7antor1s, C"*-BC, CBD, CB$-B#, C%*, CCD, CD# :ia ram 6, "%L-%$ dialo icians, *" dialo ues, miniature, "#"-#B, "#%, C+*-#, C%", LD+, LDL, L#L-#D, L#*, L## :ialo ues, ori in of, B*, DDL-D#= as self-refs, *C*L, "B#, B+C, L+B-%, DD$, $%*-%# :ickens, 7harles, %BD, %B*, %*+ :ifference En ine, BL differentiation, cellular, LC%-CD di estion, %+D di its, shuntin of, BDC :io!hantine equations, B$#, CL#-D+ :io!hantus of )le(andria, B$L directory of real numbers, CB"-BC

disambi uation, L*D-#%, D+%, DB#-%B distance to oal, D""-"% divisor-freeness, $C d.inns, ""%-"L, B"D-"$, BB%, BBC= see also 6enie :8), as a!eriodic crystal, "D$= as carrier of enetic information, "L#= com!ared to com!uter lan ua es, B#+-#"= com!osition and structure of, L"C-"L= covalent backbone of, L"C"L= as declarative knowled e, D"D-"$= double strands, L"C-"L, L%+-%"= isomor!hism with or anism, "CD-C*= mode of self-re!lication, LB#-%+= in outer s!ace, "D$, "$L-$D= as !ro ram, lan ua e, data, B#+-#", LC$= quinin and, L%"= recombinant, DDL= relation to m;8), L"$= self-destroyin , L%D= unusual inter!retation of, B%" :8) endonuclease, L%+, L%" :8) li ase, L%+, L%" :8) !olymerase, L%+, L%" :8) ;a!id Transit Service, L+L& L"$ :octor !ro ram, L##-D++& D+* do -and-bone !roblem, D""-"% &do6ma I, &do6ma II, L%B-%% do s, B%%, B%C, %LC, %*%, LD#, L$+, D""-"B, D$# :oko, BL+, D#* :ostoevsky, 'eodor, %$#-*+ double ne ation, "*%, LCL, LLC :ouble 8odulation, law of, BC% .ragon 4Escher5, C$%-$C, LBC, D#* .rawing ,ands 4Escher5, "L, B", "%%, D*#-#B, $"+, $"D, $%$ dreams, %$*, %$#, %*C, $BL :reyfus, ?ubert, L$C dualism, BL"-LL, D#*-##= see also sub.ect vs& ob.ect :um!ty, ?&, %%B du!lets 4Ty!o enetics5, L"+, L"B :voefk, )ntonin, "D% E! coli bacterium, "$D, L%$-C" Earrwi , :r& Tony, L*D-#%& DB$ earth chauvinism, "$"-$B Earth-/oon-Sun system, %L%-LC earthworm4s5, %C"-CB Eccles, John, L$C Edif)ing Thoughts of a To(acco -moker 40ach5, C*B Edif)ing Thoughts of a To(acco -moker 4:ialo ue5, LC% e s, "#B, %D+, %*% Einstein, )lbert, "++

electrons, "C+-CD, BL*, %+&%-L EAI9), see :octor !ro ram embeddin of formal systems, #$, B+$, B"L emer ent !henomena, $+*-#, $"C emotions, brains and, *%= counterfeit, L##-D++= de!endence of intelli ence u!on, L$%, L#$-#*= as e!i!henomena, D$$= music and, *%, "D+, "D%-DC, "$C-$L, %*%-*C, DBD-B$, D$D-$$, D##= !otential, B*", %*%-*C, L*%= !ro rams and, L$%, L#$-D++, DBD-B$, D$L-$$= universality of, "D%, "$C-$L emulation, B#L Endlessly ;isin 7anon 40ach5, "+-"", "L, CD, "%+, $"$-"#, $CB E8IFM 4!rocedure5, C#*-## enli htenment, B%B, B%$, BC%, BCD, BL", BLC-LL, C$#, LD$ Enli htenment 1Hond Enli htenment, B%$, B%#, BC%, BCC EnR, B%B, BLB en<ymes, function of, LB+-BB, LB*-%+, LC%-CL= as models for )I, DD%-DC= rules of inference and, L+#-"+, L"%, L%"= structure of, L"#-B", LBL= synthesis of, L"$-"#, LBB-BL, LB$-B#, L%*-CL, LC$= versatility of, LB#= vs& ty!oen<ymes, LB#= see also !roteins, ty!oen<ymes e!i enesis, "L#-D+, "D"-DB, L%"-%B, DDL E!imenides, !icture of, C#D E!imenides !arado(, connection with 6>del1s Theorem, "$-"*= Escher and, $"D= e(!anded version, B", BB= fear of, B%= 'rench-En lish version, L+"= indirect recursion and, "%C= molecular version, L%D-%$= neural version, L*C*L= Muine version, C%"-%$, CCL, CCD, CC#& C#$##, L%", L%$= subtlety of, C#L-#*= Tarski version, L*+-*", L*C-*L= two levels of, L*", L*C-*L= 3hitely1s version, C$D-$$ e!i!henomena, %+*-#, %D%, L$$, L#D, D$$-$# errors in !ro rams, B#L, B#$-#*& L#D Escher, /aurits 7ornelis, 0ach and, B+", DDD-D$= contradictions and, #$-##= co!ies and, "CD-C*= drawin s of, see Aist of Illustrations 4(iv-(viii5= fi ure and round in, D$-D*= flat vs& s!atial and, C$%-$C, D*#= incom!leteness and, $"D-"$= /a ritte and, C*+= as !rime mover, D*#-#B, $"+= on subbrains, %*$= Stran e Aoo!s and, "+"L, $%$= 9en and, BLL-L$ Escheri<ation, re!eatability of, C$%-$C, D*#= see also B-: vs& %-:, 6>deli<ation ESP, L#*-##, D#%-#L

essential incom!leteness, of )chilles1 birthday, CDB-DC, C$L-$D, D**= of )I, see Tesler1s Theorem= of list of reals, CB%-BC, CD#= of !hono ra!hs, see T>deli<ation= of self-en ulfin !rocess, C#%= of T8T and related systems, CD*-$"= see also 6>deli<ation, T>deli<ation, Escheri<ation, non!ro rammability, etc& Eta Ein, L*D-#%, D$C ET)EI8 S?;:AF, DB*,D%+ Euclid, "#, C%, L*-D+, **-*#, B"D Euclid1s !rime number Theorem, %L, L*-D+, BB* Euler, Aeonhard, %, %#C Euwe, /a(& D+L evidence, nature of, D%%-%D evolution, %B"-BB e(istential quantifiers, see quantifiers e(otic styles of thou ht, LLB, LD%-DC, LDD-D$ e(!andin nodes, "%C-%D e(!lanatory !ower on hi h level, %B", %BD, $+$"+ e(!ressibility and e(!ressive !ower, "+", C"$, CC"-C%, CCC-CL, CL+, CLC, CDL-$+, L*+-*" e(trasensory !erce!tion, see ESP e(traterrestrial intelli ence, "DB-DC, "D$, "$B-$D, %C", DCD, DD"-DB eyes, B%$, BC*, BD+, %+*, %"", %"%, C$$, D%%, $"L 7air 8apti#e, The 4/a ritte5, C*# ')87H 8EF8, "%B-%C 'antasia and 'u ue in 6 minor 40ach5, $"# fantasy rule, "*%-*L, "*$ faucet, mental, %DC-DL faultlessly functionin machines, L$L-$* 'aure, 6abriel, "D% feedback and feedforward, LCC-CL 'ermant, Johant Sebastiant, %%B-%L 'ermant1s Aast 'u ue, %%L 'ermat, Pierre de, B$L-$$, B$*= confused with 0ach, %%"-%L fermatas, B$L, %B#, %%B, %%% 'ermat1s Aast Theorem, B$L-$#, %%B, C"D= countere(am!le to, B$$, B$#, CD+= inverted, %%%-%C= !arodied, %%L, LL"= !roof of, B$$, B$#, CD+ 'eynman dia rams, "CC-CD 'ibonacci 4Aeonardo of Pisa5, "%D, BCD 'ibonacci sequence, "%D, "%*, "%#, "LB, "$%-$C, BDL, C"D fiddles, see violins fifth !ostulate 4Euclid5, #+-#%, BBB, CL"-LB L+ 4fifty5, %%*, LL$, LDC

fi ure and round, D"-D%, DC-$C, $%"= in music, $+-$" 'I6F;E-'I6F;E fi ure 4@im5, D*-$+, $% filters, for abstraction, B*D, C+$-#, DC*, DL$-D+, DD%, D$%= for Pools, C"*, CB$ finitistic methods of reasonin , BC, B%+ 7ishes and -cales 4Escher5, "CD-C$ fission and fusion 4of conce!ts5, %%*, %LB-LD, C%*, C$+, DDC-DL L-: s!ace, DC+ fla s, B#-%B, "** flashcards, see t;8) flat vs, s!atial, see B-: vs& %-: flattened look-ahead, D+C-L flattened roles of inference, see theorems vs& rules fle(ibility and infle(ibility, BD-B$, B#D-%+B, D"""C, DL$, D$%-$L, D*D fli hts of fancy, %$* 'looP, C+D, CBC-%+, LD$-D*= fed into itself, CBLBD= !ower of, CB*-B#& LD"-DB fluency, %$D-$$ flutes, %-L, B$, LB*, LLB-L*, $B+, $BD '4n5 and /4n5, "%$, "CB, %L# focusin , DL$-L# foldin -u! of en<ymes, L""-"B, L"#, LB", LBL football, CC, %+%, %L%, D%C-C+, DC%, DCC, DCL, D$B footraces, B#-%B, C%, L#C-#L, D*"-*% forced matchin , D$+-$B for ettin , L$$, L$*, D"# 'orkel, Johann 8ikolaus, C, *D form, C$& DD-D$, D*, $%& "#+-#", %$+-$"= syntactic vs& semantic, L*"-*C, D%" form and content, *C-*L, B+C, B$#, L*"-*C, DD$D*, $C+ formal systems, !resentations of, 7-system, DCDL= /IF-system, %%-C"= P-system, $%-$C, !qsystem, CD-D+= Pro!ositional 7alculus, "*"-#$= T8T, B+C-%+= tq-system, DC-DL= Ty!o enetics& L+C-"% formal systems vs& reality, L%-L* formal vs& informal reasonin , "#%-#$, BB*-B#, B$"-$B& CC#-L+, D"C-"L, D"*-"# formal vs& informal systems, BD-B$, LL#-*L, L#*, D*C-*D= see also brains, minds, etc& formalist !hiloso!hy of mathematics, CL* formula, closed, see sentence= o!en, B+$-* formulas of T8T, B+D, B+$-"L 'our-color Theorem, !arodied, LL+ C-: s!ace, D%*-%# four-!ostulate eometry, see eometry, absolute D'HH" 47a e5, "LD

'ourmi, Aierre de, %%%-%C frame effect, $+C frame messa es, "DB, "DD-D$, "$D frames, %$%, DCC-CD, DDB-D%, D$B framin devices, C$* 'rank, Phili!!, DCB 'rederick the 6reat, @in of Prussia, %-*, B$, %#C, $B# free will, %**, D*+, $+*, $"+-"C, $B+-"D, $%C-%L, $%#= see also .um!in out of the system 're e, 6ottlob, B+ 'rench fries, D%D-%*, D*% 'rench Suite no& L, 6i ue 40ach5, "%+ frequencies, of words and letters, %$$, D%+ friend, mental model of, %*D-*$ frin e science, D#%-#C fu ues, %%L, D%C, $%+, $%D= in the )rt of the 'u ue, $#-*", *D= devices in, %"C, %BB-B%, %B#%+, $%$-C+= :ialo ues and, B*= in the /usical Efferin , C-#= nature of, #, B*"-*C, $%$ 'undamental 'acts " and B& CC+-CB fundamental .ukebo(-a(iom, "LL funnelin , %CD-C* 6 46>del1s strin 5, "*, B$"-$B, B*L, CC$-LL, CL#D+& L+B, L*+, D+*, DD$, $+$-* 61, 6-, 6-1, 6w, CDD-D* h6, B$B, CC#, CL"-LL, CL*-L#, LCB 6alileo, C$*-$# ames !layed by )I !ro rams, D+" 6antR, "*#-#+, BLD, C+$ ases and molecules, %+$-*, %"$, D#% 6ateless 6ate, see /umonkan 6auss, @arl 'riedrich, #B, "++ 6ebstadter, E bert 0&, #C-#L, C+B-%, C*C 6elernter, E&& D+D-$ eneral recursivity, C+D, C%+, C$+ enes, B++-", L+$, L"B, LBC-BL, L%", LCC-CL, DD* 6enetic code, "D+& L"#-B+, LBB-BC, L%%-%C, L%D, L%*= ori ins of, B%", LC* enetics, L+C-C* 6enie, /eta-6enie, etc&, "+#-"D, B"D-"$, BB%, BBC, D"+= see also d.inns enie, symbol-mani!ulatin & %#-C+, C* enoty!e and !henoty!e, "L#-DB, "D$, "$%-$C, "$L-$D, B#L, L%"-%B, DD$-D* 6ent<en, 6erhard, "#L 6eometric 7ode& B%L-%$, BC", DBD eometry, absolute, #", #%, #$, BBB, C+$, CL"-LB= elli!tical, #%= Euclidean, "#-B+, **-#B, "++, BBB, CL", CLD, D+D-$, non-Euclidean, "#-B+, #"-

#%, #*-##, "++,BBB-B%, CL", CLL-LD= -trueversion of, **-#C, ##-"++, CLD-L$ 6erman !rofessor, !roverbial, "%+-%" 6iant Electronic 0rains& BL, D+" lia, %%# 6looP, C+D, CB*-B# 64n5, "%$ oals and sub oals, BB$, L*#, L#+-#", D+#-"C, D"*-"#, DB#, D%B 6oblet 6, $#, *", *%-*L, BD$ 6od, B"D, C++, C$*, C*B, L%%, LD$, L#$, $""= !icture of, "CB-C% 6E: 4acronym5, ""+-"L, "%%, "%C, B"D, BB%, BBC, $BB= see also d.inns, 6enie 6>del, @urt, "L-"#, BC, B*, $%*, $C+, $CB 6>del 7ode, "*, BD*, L%%-%L 6>del codons, BD*, CBL, L%%-%L 6>del isomor!hism, BD"-$", C%#, CCB-CD, $%*%#= likened to reflection of world in brain, L+B, L$+ 6>del-numberin , "*, C%*, $%*-%#= of 'looP !ro rams, CBL-BD, L+B= of /IF-system, BD"DC= of T8T, BD*-$+, L$# 6>del questions, Aucas on, %*#, %#+ 6>deli<ation, B$+= !ro rammability& Ef, C$"-$%= re!eatability of& CBC, CDL-$D, D**= see also Escheri<ation, T>deli<ation, .um!in out of the system 6>deli<in o!erator, C$B-$%, C$L-$D, LC% 6>del1s ar ument, summari<ed, "*, B$B, CC* 6>del1s article, "$, BC, C%* 6>del1s construction, illustrated, *C 6>del1s method, underlyin causes of, B+C, C+$, CDL, CD*-$" 6>del1s Second Theorem& B%+, CC#-L+, D#D 6>del1s Theorem, )I and, %**-#+, C$"-$$, $+D-$, $"C= analo ue of in molecular biolo y, L%C, L%D-%$= brief mentions, $B, $C, $*, "++, C*D= consequences of, CL+-D+, CD#-$D= 8ontracrosti% punctus and, see 8ontracrostipunctus, :io!hantine equations and, CL#-D+= AISP and, $%*-%#= !roof of, "*, BDL-$B, C%*-C#= stated, "$, "+", B$B 6>del1s Theorems and human intros!ection, CL+, D#D-#* 6offman, Ervin , C$* old, "$% 6old 0o(, Gery )sian, C+C-L 6oldbach, 7hristian, %#C, %#L 6oldbach 7on.ecture, %#C-#D, C++, C+C, LL$-L*, D"L= !arodied, LL" 6oldbach !ro!erty, %#L-#$, C++, C"C, C"*

6oldbach Gariation, %#L-#*& C++, CBD, CC" 6oldber , Johann Theo!hilus, %#"-#B 6oldber Gariations 40ach5, %#B-#%, %#L 6oodfortune, ?e(achloro!hene J&, "+%-C, ""L, "B*-B#, "%+ 6oso, BC* 6!lot, "%*, "C+-C%, "CD-C$, "L#, L+%, DD" rammar, for com!uter lan ua es, B#$, C+*-"L= for 'eynman dia rams, "CB& "CL= hi h-level, DBL-B$= for kRans, DBL-BD= for music, DBD-B$= for natural lan ua es& "%+-%C, "L+, %D%, L**#%, D"#-B", D%+-%B= for thou ht, DB$ 6rand Tortue, B%$, BC%-CC randmother cell, %CC, %CL randmothers, !erce!tion of, %CC, %CL, %C$-C*& %C# ra!hics, $B* rass roots, D#% 6reat Tutor, B%$, B%#, BCC 6reen Pro rams, CB$ Greendiag U8V, CB$ 6rellin 1s !arado(, B+-B", BB 6root, )driaan de, B*D rounds, e(cellent, $%" uanine, see nucleotides uaranteed termination, C", %#D-#*, %##, C+% uitars, DB, B++ ullibility, $L-$D, "+D, %+#, CD", D++, $+" 6utei, B%$ 6++BL, white stony, DBD ?)7@E;, DDC haiku, "L%-LC, LBL, D"#-B+ haltin !roblem, $C, CBL-B#, L#C, D#$ hambur er-confusion, L$$ ?ammurabi, "D# ?ardy, 6odfrey ?arold, LDB-DD harmonic tension, "BB-B%= see also tension and resolution har!sichords, %, %#", L+B ?arrison, Aawrence, DL$ ?aussmann, Elias 6ottlieb, B ?E-?E !u<<le, DB-D%, DD# headache, DB-D% hearin by com!uter, D+B -heart-, in an )I !ro ram, D$# ?eisenber uncertainty !rinci!le, CLL, D#* ?elen of Troy, ""+ hemiolia, BL$, L"# hemis!heres, BL$, %C+-C" ?enkin, Aeon, LC"

?enkin sentences, LC"-L%, $+#= e(!licit and im!licit versions, LCB-C% ?enkin1s Theorem, C** hereditary ar uments, %D, C$-C*, B"$, BD" heterarchies, "%C, %L#, L%C, DL"-LC, DDB, D#" heterolo ical ad.ectives, B+-B", BB heuristics, L*$, L**, L#+, D+%, DB# ?ewitt, 7arl, DDB hiccu!s, ""D, BLC, BLL, D$%& $BL-BD, $%D hierarchy of variability, DC%-CL, DD# hi h-fidelity vs& low-fidelity, $$, *L, "+", C+D-$, C$+, D#$-#* hi h-level e(!lanatory !ower, $+$-# ?ilbert, :avid, B+, B%-BC, B%+, CL#-D+ ?ilbert1s !ro ram, B%-BC, BB#-%+ ?ilbert1s tenth !roblem, CL#-D+ ?4n5, "%$ ?ofstadter, :& ;&, $L, %"+, $BC, $B*, $CB ?ofstadter1s law, "LB ?R en, BC* holes in formali<ed systems, BC, BD, CC#, CL", CDL, CD*, C$+-$" holism, defined, BLC, %"B= vs& reductionism, B*C, %""-%D, %*#-#+, $+*-#= 9en and, BLC ?ubel, :avid, %C", %C% ,uman 8ondition 3, The 4/a ritte5, $+L-D ?yaku.o, BLC hydro en bonds, L"D, LBB, LBL hy!hen-strin s, C$, DC-DL, DD hy!othetical worlds, #L-"++, %%*, %D+-DB, D%CC+, DC"-CC= roundedness in reality& %DB, %$*$# hy!otheticals, CC-CL, D%C-C+ i, CLC 3, CLC -I-, referent of, D+* -I 7an 0e Played 4Proven, etc&5&&&-, C**, LC" -I 7annot 0e Played 4Proven, etc&5, $D-$$, *L, C+D-$, CC*, CDL-D$, L%D, LC", D+* I-counts, BD+-D" I-level, see inviolate level I-mode, see Intelli ent mode iceber , C#L-#D, C#$ ideal numbers, LD-L* identification with artifacts, D+#, $"%-"C idiots savants, LDD-D$ I'-statements 40looP5, C""-"B ima es, blurry, D*D-*$= of thou ht, DB% -Ima inary Aandsca!e no& C- 47a e5, "D%-DC, D## imitation ame, see Turin test

Immunity Theorem, L%D im!licit characteri<ation, C", D$, $B-$%, #% im!rovisation vs& intros!ection, $%# inaccessibility of lower levels to hi her levels, D*D-#B, $+D-"+= in )unt ?illary, %%+-%", D%+, in brainsIminds, %+B, %B*-B#, %DB-DL, D"#, D$$, D*D-#B, D#$, $+D-"+, $%#= in !ro rams, B#D, %++-", L**, D%+, D$#= see also software and hardware, intros!ection, level-conflicts incom!leteness, 0ach and, *D= of brains, L*L= defined, *D= Escher and, $"D-"$= of e(tensions of T8T, CDL-$"= of formal arithmetics, "*, *D, "+"-B, C+$, D"*-"#= of list of mathematicians, CBB= of list of reals, CB"-BC= of Aucas, C$$= of !hono ra!hs, see record !layers, intrinsic vulnerability of= of Princi!ia /athematica, "*, BC, D"*-"#= of self-knowled e, D#D-#*= of T8T, B$"-$B, C%+, CL+-L"= see also essential incom!leteness, w%incom!leteness, etc& inconsistency, defined, #C= with e(ternal world, *$-**, #L= internal, *$, **, #C-#D= of !eo!le, "#$, D#$-#*= of Tortoise, "$$-*+= see also consistency, contradictions, w%inconsistency, 9en increasin and decreasin rules, $%, $C, BD+-D", BDC, BD#, C+"-B, C+$-(, CC"= see also len thenin and shortenin rules, chaos in number theory inde( numbers for !ro rams, C"*-B+, CB$-B* inde( tri!lets for su!ernaturals, CLL India, LC#, LL", LL$, LDB-DD Indra1s 8et, BL*, %L# inducers, LCL infinite bundle of facts, %#$-#* infinite coincidence, %#*, CB" infinite re ress, """-"%, "CB, "CD, "LB, B%", %***#, CBD, C#$, $%*= in 7arroll !arado(, C%-CL, "$+, "#B-#%, D*C-*D& D#%-#C= halted, "B$, "%%%L, "$+, D+L, D*C-*D= of ob.ectivity, C$#= 9eno and, %"-%B, D"+= see also 7arroll !arado(, bottomin out, recursive acronyms, re!eatability, etc& infinite sentence, C#$ infinite sky, C+" infinitesimals and nonstandard analysis, CLL infinity, 0ach and, "+, $"#= Escher and, "L= handled finitely, L#-D+, BB"-BL, CD"-DC, CD*= illustrated, "%L-%D, "%*-C%= names of, C$L-$D= su!ernaturals and, CLC= ty!es of, CB"= see also nontermination, infinite re ress, recursion, etc& informal systems, see formal vs& informal systems

information, accessibility of, see inaccessibility= creation of, L"%= de!th from surface, B%C-%L, C+#, CB$, LC#-L*, D+D-$, D"B-"%, DB*, D$%, see also decodin = discardable, DC#, DL%, DL$-L#& DD#-$B, flow of, L"%, L%%, LCL, LC$= irrelevant, LD+ information-bearers, "L*, "DD, "D$ information-revealers, "L*, BD$ inhibition, cellular, LCC inner messa es, "DD-$", "$C-$D, L+", LBC in!ut-out!ut devices, B** in!ut !arameters 40looP5, C"" insi ht, D"%, DD+-D", DDL-$D instant re!lays, strai ht and sub.unctive, D%C-C+, DC", D$B instructions, in machine lan ua e, B*#-#L= vs& tem!lates, C#$-##, L%", see also !ro rams vs& data I8T495, "%*-C", "CD, DD" intelli ence, accidental ine(!licability of, $+$= essential abilities for, BD= e(traterrestrial, see e(traterrestrial intelli ence= liftability of, see skimmin off= limits of, C$L-$D, D$#-*+= necessary under!innin of, %BC= sim!licity of, "$B-$%= subtle features of, LDD= tan led recursion and, "LB= ty!ical abilities of, LL#= universality of, and intrinsic meanin , "L*, "DBDC, "$+-$D, L+", DD"-DB= see also brains, minds, )I, etc& Intelli ent mode, %*-%#, DL, "#%-#C, D"%-"C intensionality and e(tensionality, %%$-%#, %L+, %D"-DB intentions of machines, D*C-*L interestin ness, !ro rammed, D"L inter!retation-conventions, D*$-** inter!retations, ad.usted to avoid inconsistency, *$-**, CL%, CLD, see also undefined terms= multi!le, #C-"+B, "L%-L$, BDD-D$, B$", CC$-C*= of !q-system, C#-L%, *$-**, "+"-B, "L*= of Pro!ositional 7alculus, "*D-*$, "*#, "#"-#B= of strands, L+#-"+= of T8T, B+L-#, BDD-D$, CL%, L%%= of tq-system, 7-system, P-system, DC-DL, $%-$C inter!reters, mechanisms in brain, L*B-*C= !eo!le, B#%, B#$, LBC, D$"= !ro rams, B#%, L+C, LC$, D"D, D%B, DDB, D#B intrinsically hi h-level !ro!erties, $+$-# intros!ection, see self-monitorin , selfawareness, self-knowled e, inaccessibility, T8T, intros!ection of

intuition, LD+, LDC, D"%, D*+, $"%= !ro rammin of, D+L, D+# inversion, *-#, *", "CD, D*"-*%, $%$-%*= see also co!ies, com!lementary to ori inal inviolate level, D*D-#B irrationality vs& rationality in brainImind, L$L-$* irre ularities, meta-irre ularities, etc& C$L-$D Isan, BLC ism, BLC-LL, DBL, $+C-D isomor!hisms, between 0on ard !roblems, DD+, DD#= between brain-structures and reality, *B, %%$-%#, %L+, L+B, LD#-$"= between brains, %D#*B= coarse- rained, "C$-C*, L+%= in 7ontracrosti!unctus, *%-*L= between 7rab1s :8) and 8ra( 8anon, B+%, DD$-D*= defined, #, C#-L+= between earthworms& %CB-C%, %CL= of emotions, "D%= e(otic, !rosaic, "L#-D+= fluid, %%*, %L+, %DB= between form and content in :ialo ues, *C-*L, "B*-%+, B+C, DD$-D*= between formal systems and number theory, C+*, DBL= 6>del-numberin and, see 6>del isomor!hism= between mathematicians, LDD= between mathematics and reality, L%-D+= between mental !rocesses and !ro rams, LD*$%= between /IF-system and %"+-system, BD"DL= between models of natural numbers, B"$= !artial, "CD-C$, %$"-*B= as revelations, "L#-D", BD%= as roots of meanin , C#-L%, *$-*, #C, BD$, %%$, %L+= self-re!s and, L+"-%= between somethin and !art of itself, "%*-C%, "CD-C$= between s!iderwebs, %$"-$B= trans!arent, *B, "L*, BD$= on various levels between same ob.ects, %D#= between visual a!!aratuses, %CLCD= in visual !rocessin , %CC= see also meanin , translation, co!ies, decodin , etc& -Jabberwocky- 47arroll5, %DD-D*, %$B-$% Jacquard loom, BL Jaki, Stanley, L$C -Jammerwoch, :er-, 47arroll-Scott5, %DD-D* -Jaseroque, Ae- 47arroll-3arrin5, %DD-D* Jauch, J& /&, C+*, C+#, C$*-$# Jefferson, 6&, L#* Joan of )rc, B+ Johns, Jas!er, $+% JRshl, B%%, B%$, B%*, BC+, BL%, BL#, B$B JyS?z 4T8T-strin 5, CC% .ukebo(es, "LC-L$, "D+-D", "DC, "$+-$", "$C-$D, L++ .um!in out of a subsystem, C$$

.um!in out of the system, in advertisement, C$*= by answer-schemers, CDB-DC= 6>del1s Theorem and, see 6>deli<ation, essential incom!leteness= illusion of, C$*-$#, D#*= as method to resolve contradictions, "#D-#$= in !olitical systems, D#B= by !ro rams, %D-%*, C$D-$*, D$*= from B: to %-:, see B-: vs& %-:= 9en arid, BLL, C$#= see also 6>deli<ation, T>deli<ation, Escheri<ation, T7 battles, re!eatability, non!ro rammability, etc& @aiserlin , 7ount, %#"-#B @ay, )lan, DDB @ennedy, John '&, DC" keys, musical, "+, B##, CDD, L+"= see also modulation @im, Scott, D*-D#, L+%, LB%, $"# @imber er, Johann Phili!!, #, $BD kitchen sink, the, %"L @leene, Ste!hen 7&, C$D @lein bottle, D#" @l` el, 6& S&, #" knittin , "C#-L+ knots, %C"-CC, B$B, DB* knowled e, accessible vs& inaccessible, %DB, %DL, D"D, D"#= encoded in ant colonies, %"#-B*, %L#= e(!licit vs& im!licit, D"$-"*= modularity of, D"L-"*, DB*= !rocedural vs& declarative& %D%-DL, D"L-"$, D%+, DLC knowled e trans!lantation, sur ical, D"* kRans, %+, "*#-#", B%%-CL, BCD-L#, DBL-BD= enerated by com!uter, DBL-BD= enuine vs& !hony, B%C-%L, B%#, BCB, BCC, CB$, DBL-BD @ronecker, Aeo!old, B"D @uhn, Thomas, DD+ @u!fer >del, ;oman, %#C @yR en,BCC-CL Aa /ettrie, Julien Effroy de, %, B$, $B# labelin technique, C*$-**, LC+-C" Aambert, J& ?&, #", #B, ## lam!, meta-lam!, etc&, "+*-"%, B"D lan ua e4s5, acquisition of, "$+, B#C, %+B= active meanin s in, L"-LB= )rabic, DB%-BC= of bees, %D+= of the brain, L$+= 7hinese, "DC, DDL, D$D= colla e of, see scri!ts= com!uters and, "%+-%C, %++-", %D%, L*D-#%, L##-D++, D+"-%, D"#-%B, D$C-$L, $B"= effect on thou ht, %$D-$$= En lish, "D#, %$B-$%, %$$, %$#-*+, D"#-%B, D$C-$L= fle(ibility of, DC#, D$C-$L= 'rench, B#$, %DD-D*, %$B-$%, %$$, L+", D"*= 6erman, %DD-

D*, %$B-$%, %*+, DDL= ?ebrew, (viii, %$$= hierarchy of, BB= im!recise, D$C-L= invisible isomor!hisms and, *B= Ja!anese, "D#= as medium for !roofs, **-#+, "#L= necessary under!innin of, %BC= !artitions between, D$"= !rocedural rammars for, "%"-%C, D"#-%B= readin meanin into com!uter-!roduced, L##D++, DBL= on ;osetta stone, "DL= ;ussian, B#$, %$#-*+, DCB= self-refs in, C%"-%$, C#L-#*, L+"= see also meanin , translation, etc& Aashley, @arl, %CB, %C%, %C* -last ste!-, CDB-D%, CD* lateral eniculate, %C%-CC layers, of dece!tion, C$*= of messa es, "DD-$", LBC, $+%-C= of stability, DC%-CL leaka e, between levels of science, %+L-D Ae endre, )drien-/arie, #B Aehnin er, )lbert, L+C Aeibni<, 3ilhelm 6ottfried, BC-BL, D++ lemmas, BB$ Aenat, :ou las, D"L len thenin and shortenin rules, decision !rocedures and, C*-C#, "*B, C+$-*= /IF-system and, %#-C+, BD+-D", BDC, D"%= T8T and, B"%, BDD, BD#= see also increasin and decreasin rules, !roblem reduction Aeonardo of Pisa, see 'ibonacci Aermontov, /ikhail, DCB level-conflicts, in )unt ?illary, %%+, D%+= in mindIbrain, L$L-$*= in messa es, "DC, "$+, D##-$+C= between ob.ect lan ua e and metalan ua e, "#C, CC#-L+= in S?;:AF, D%+ level-confusion, ants and, see ants vs& ant colonies= in art, see B-: vs& %-:= authorshi! and, %& D+*, $B+-BD= in com!uter systems, B*$, B#", B#L, %++-B, %+*= of @imian self-re!, L+%= mindsIbrains and, B*$, L$L-$$= in Pro!ositional 7alculus, "*L, "#C= sub.unc-TG and, D+*= self and, $+# level-crossin , in thou ht, DDD, DD* level-mi(in in enetics, L+#-"+, L"%-"C, LCD-C* level-shiftin , conce!tual, see abstraction, levels of levels, of com!uter lan ua es, B#+-##= distinct vs& similar, B*L, B*$= in Escher, ""-"L, D*#-#", $"L-"D= ha<iness of, "%-"L, LCD"*, $"L-"D= intermediate, %+B-%, %"$, %BC, L%B, D%B= of irreality, BC%, DC"= of /F-!icture, %""-"%, %B*B#, LBL-BD= of !articles, %+L= in radio news, "B*= of reality, "L, "+%-BL, "B*-B#, "*C-*L,

C*", C#%, DC+, $BL-BD, $%$, $%#= in recursive !rocesses, "B*-B#= of rules in thou ht, BD-B$ levels of descri!tion, of ant colonies, %"L-%%= of brain, %C#-L+, %*B& LL#, L$+-$$, L*C-*L= of caste distribution, %"#-B#= of chess boards, B*L*D= of errors, B#C-#L= of ases, %+*= of human body, B*L= of human !syche, B*$= of mental !rocesses, LD*-$%, L$L-$*, L*C-*L= of !ro rams, B#C-#L, %*+-*"= of television screen, B*L= see also holism vs& reductionism levels of meanin , in ant colonies, %"#-B$= in 8ontracrostipunctus, *B-*L= of :8), "D+, L%"%B, DDL= in E!imenides !arado(, C#D, L*", L*C*L= of roove-!atterns, *%-*C= of /umon, BC*= of /F/E8, BDD-D$= of music, "DB-D%= of neural activity, L$L-$$= of T8T-strin s, BDD, B$+-$" levels of structure, of en<ymes, L"+-"", L"#, LB", LBL-B$, L%B= of music, LBL liar !arado(, see E!imenides !arado( +i(eration 4Escher5, L$-L*, DL li htnin calculators, see idiots savants limericks, C*%, $%D limitative results, in eneral, "#, $C, D+#, D#$, D## Aincoln, )braham, CLC lines, eometrical, "#-B+, #+-#%, "++, BBB, CLB, CLD AISP, B#%, %*", DBD, DLB, D#B, $%*-%# Aist of )ll 6reat /athematicians, C+C, CBB +ittle ,armonic +a()rinth 40ach5, "B"-B%, "B#, "%+ +ittle ,armonic +a()rinth 4:ialo ue5, "B$, "B*%+, "C#, B"D, D"+-"" Aittle ?armonic Aabyrinth 4of /a.otaur5, ""#-BL Aittlewood, J& E&, LDC li<ards, "+*-#, ""+, ""L-"$, "BL Aobachevskiy, 8ikolay, #" local vs& lobal !ro!erties, B", "D+, %L#, %D%, %$"-$L, LC%, L*B-*C, D$* locali<ation of knowled e, in brains and !ro rams, %CB, %C*, %DL, D"$-"* Aockwood, )nna, $++ lo ic, "#-BC, C%-CL, ##-"++, "$$-*+, "*"-#$, CD"DC, D"*-"# Aoocus the Thinker, C$$ look-ahead trees, "L", D+C-L, D"", $"B loo!s, bounded, "C#, C"+-"C, C"*, CC+-C", CCC= free, "C#, CBC-BL= in music, "L+= in !ro rammin , "C#-L+, C"+-"C, CBC-BL, L+%, D%B

lottery, D%#-C+ Aovelace, Aady )da )u usta, BL, %+$, L#* lower levels, see substrate, mental lowest-level rules embodied in hardware, D*L-*D Aucas, J& ;& %**-#+, C$"-$%, C$L, C$D, C$$, L$C, L$$-$*, L#$ Aucas1 ar ument, counterar uments to, C$L-$$, L$$-$*= merits of, C$B= summari<ed, C$"-$% Aucas sequence, "%#, "LB, "$C /-mode, see /echanical mode /ac6illavry, 7aroline, DD$ machine de!endence and inde!endence, B#C machine lan ua e, B*#-%++, %+D, %*", LC$ machines, not the sum of their !arts, %*#-#+= reflectin on themselves, B**-*#= selfassemblin , "D+, C*D, L+C, LC%, LCL /acAaine, Shirley, B*L macrosco!ic effects from microsco!ic causes, %+$-* /)7SH/), D"L /adsto!, $B$ ma netic field and crystal, "C+-C% /a nificat in : 40ach5, LC#, LLB, LL 'agnificra(, 3ndeed, LD+, L$C, L*" /a ritte, ;enK, C*+-*", C*#, C#%-#C, DB$, $++-B, $+L-D= !aintin s by, see Aist of Illustrations 4(iv (viii5 /ahalanobis, P& 7&, LDL main theses of book, BD, CD, LL#, $"C /a.otaur, ""#-B", "B%-BL mala!hors, DL$ /andelbrot, 0enoit, $" manifestations of symbols, %L" /ao Tse-tun , C%% ma!!in s, charted, *L, CC#, L%%, L%D= induced, DD*-D#, D$"-$B marbles, rollin , $""-"B /ar ie-balloon story, D$L /aterialism, cham!ions of, B$, $B# mathematical lo ic, history of, "#-BC mathematical view of brains, LL# mathematicians, CL*-L#, LL#, LDD, D"C mathematics, done by com!uters, L$%, D+B, D"C"L= foundations of, "#-BC= reality and, LC-L*, CLD-L# /athews, /a(, D+$-* /c7arthy, John, B#% /c7ulloch, 3arren, "%C meanin , built on tri erin -!atterns of symbols, %BL, %B$, %L+= carried only on symbol level,

%BC-B$, %%+, %%+, $+#-"+, codes and, *B, "L*DB, "DC-D$, BD$= of 7ontracrosti!unctus, *B-*L= of :8), "D+, L%"-%B, DDL= e(!licit vs& im!licit, *B-*L, "L*-$D, C#L-L++, L*%= in formal systems, see inter!retations= intelli ence and, "L*, "DB-DC, "$+-$D, L+", DD"-DB= intrinsic, see meanin , e(!licit vs& im!licit= location of, "L%L$, "L*-$D, C+*-#, L*B-*C= as multidimensional co nitive structure, L*B-*C= multi!le, *, "+, LBL%, *B-*L, #C-"+B, "L%-L$, "L*, "$B, BDD-D$, B$", C+#, CC$-C*, LBC, L%B, DDD, see also disambi uation= in music, *%, "D+, "D", "DB-DC, "D$, "$B, "$C-$L, BB$, L*B-*C, DBD-B$, D$D-$$, D##-$++, $+C= ob.ective, see meanin , e(!licit vs& im!licit= as o!tional hi h-level feature, L$"= !assive vs& active, L"-LB, #C, #$, "++, "+B, "#"#B, BDD, BD$, B$", CLD= !ur!ose and, %B"-%B= rooted in isomor!hisms, C#-L%, *$-**, #C, BD$, %%$, %L+= unnecessary on evolutionary time scale, %B"-BB meanin less vs& meanin ful inter!retations, L", ** meanin lessness, in art and music, D##-$++, $+C% meat rinders, C"C /echanical mode, %*-%#, DL, "#C, BB", D"%-"C mechani<ation of thou ht !rocesses, see )I, formal systems, etc& meiosis, DDL, D$B melodies, recall of, %D%-DC= time-shared, %*L memory, in com!uters, B**-*#, LCD, D"D memory dum!, %*" men vs& women, C$$, L#L-#D /endel, )rthur, %, B* 'ental *rithmetic 4/a ritte5, DB$ mention, see use vs& mention /en<el, )dol!h von, C-L /eredith, /arsha, DBL /er enthaler, Etto, D%+ messa e-!assin lan ua es, DDB-D% messa es, "LC, "L*-$D, in ant colonies, %L+= in bottles, "D$-D#, LBC= layers of, "DD-$", LBC, $+%-C= from nature, C+*-#= see also frame messa e, inner messa e, outer messa e messen er, for kRans, B%L-%D, B%* messen er ;8), see m;8) meta, B"D-"$, BBC meta-a nosticism, ""C meta-analo y, D$%-$C meta-answer-schema, CD% meta-author, D+$-#, $BD

metabook, BB meta-descri!tions, DLD-L$, D$C meta-evidence, etc&, D#%-#C /eta-6enie, see 6enie meta-hiccu!s, $BD meta-intuition, D+L /ET)-JyS?z, etc&, CC% metaknowled e, %DC metalan ua e, BB, "*C, "#C& BC*, B$+, L"C metalo ic, B%, D$D metamathematics, B%, L$#= reflected inside T8T, CC#-L+ 'etamorphosis 4Escher5& "C-"L meta!hase, DDD-L$ meta!hor, D$B meta-!roteins, L%%-%C metarules, etc&, in chess, D*$-**= in intelli ence, BD-B$, LL#, D*C-*L meta-search, %#$ meta-symbols, LD+ metatheorems, "#%-#C metatheory, formali<ed, "#C meta-T8T, etc&, CCB-C%, L%%-%C meta-wishes, see wishes meteorites, "D$, "$B methylation, LC+-C" metric, mental, D"%, D"C= see also conce!tual nearness /eyer, Aeonard 0&, "D$, $+C /ichelan elo, DCB micro!ro rammin , B#L minds, overla! of, %$D= !ro rammability of& %+B, D$#, see also )I, !arado( of )ll, Tesler1s Theorem, non!ro rammability= thou hts and, %D#-#+= two ways of creatin & %#+= vs& brains, %+#, L$L-$$= see also brains, intelli ence, etc& mini-vocabulary, DC$ /insky, /arvin, %$%, %**, DCC, D$#, $BB mirrorin , see isomor!hisms, re!resentation miss!ellin and com!uters, B#$-#* /IF-numbers, BDC-D$= see also theoremnumbers /IF-system, %%-C", CD, C$, C*, LB, "#", BD+-D$= as model for T8T, C%#-CB, CDD= table of rules of, BD+ /IF^/F-system, CDD '(ius -trip 3 4Escher5, B#-%+ '(ius -trip 33 4Escher5, B$D modes of fu ue-listenin , B*B-*C modularity, "C#-L+, D"L-"*, DB*, D$$-$*= see also locali<ation, local vs& lobal !ro!erties

modulation, "+, "B"-B%, "B#-%+, CDD, L+", $"$, $%# modules in brain, see symbols /odus Ponens, see detachment molecular biolo y, L+C, L"C-C* /ondrian, Piet, $++ monkeys, vision of, %CL-Cb /onod, Jacques, "D" mor!ho enesis, L%#, LC%-CC 'osaic 33 4Escher5, D"-D% mountain-car wreck story, %%*-%#, %D", %DL /o<art, 3& )&, DC#, $+B m;8), L"$-B+, LBB-BL, LB$-B*, L%+-%%, L%D, LCL, LC$, DDB, DD% /F, as !ossible theorem of /IF-system, %%-C", BB#, BL#-D", BDL-D$, B$", $+* /F, 9en word, B%%, BC", BCD, BLC, BL#, B$B, %"", %"B-%"%, %B* /F-AEEP 4'loc P5, CBC-BL, CC" 'u :ffering, *, B$B, DB* mu-o!erator, CBC /F-!icture, B*%-*C, %"+-"%, %B$, %B*-B#, LBL-BD /F-!u<<le, %%-C", BL#-D", L+#-"+, D"%-"C multifurcation of T8T, CD$ multi!le re!resentations, D"D-"*, D$+-$", D$C multi!lication, LC-LD, DC-DL, B+D, C+#, CLL, LDD, LD$ /umon, BCB, BCD, BC*-C#, BL%, BL#, BD+, B$B= commentaries by, BCD-C#, BLB= !oems by, BCDC#, BLB, B$B /F/E8, strin of T8T, BDL-D$, B$", CC"-CB 'umonkan, BCD music, com!osed by com!uter, BL, L#L, L#$, D+%, D+$-#, DBD-B$, D$D-$$= dimensions of, "$L= mathematics and, BB$, LLL, LD+= modern, "LDL$, "D%-DC, "$C-$L, D##-$++, $+C= notation of, LLB-L#, see also S/FT= semantics of, *%, "DBDC, "D$, "$C-$L, L*B-*C, DBD-B$, D$D-$$= su!erhuman com!rehension of, "$B, D$#= synta( of, "B"-B%, "B#-%+, BB$, DBD-B$, $%$= to break !hono ra!hs by, $L-$*= to infiltrate !hono ra!hs by, C*$-**= see also fu ues, canons, !ianos, flutes, etc& music bo(, !re!ro rammed, D$$ 'usical :ffering, The 40ach5, C-"+, *D, DDL, DDD, $"#, $B+, $BC, $B$, $%#-CB mutations, B#L 8, see number theory 8a.unamar, 9&, LC#-LB 8ansen, BC*-C#, BL%, BLL

natural lan ua e utterances as !ro rams, DB# natural numbers, defined, LC, B+C= enerali<ed= CL%-LD= !ostulates for, B"D-"$= see also number theory, numerals, T8T, !rime numbers, etc& near misses, see -almost--situations nearly decom!osable systems, %+%-D ne ation, $+, $", "*%, "#"-#B, B"+-"", B"C, LCL ne ative s!ace, DB-D%, DD-D*, $B= see also fi ure and round nested movies, "*C-*L nested works of art, "L, "+D, $++-", $+L-D nestin , "B$, "%*-C", "*C-*L, DD+= see also recursion neural networks, see symbols 8euroneater, %*B neurons, com!ared with ants, %"L, %BL, %%#-C+= described, %%#-C+= Euclid1s, D+= faultless functionin of, L$L-$$= firin of, *%, %C+, %"D, %C%-CL, %C$, %L+, %L$= as an inviolate level, %+B, D$$, D*D, D#"-#B, see also inaccessibility= not controllable consciously, %+B, see also inaccessibility= on-center and off-center, %C%CC= in retina, %C%-CC= sim!le, com!le(, hy!ercom!le(, %CC-CL, %CD, %C$= as summin in!uts, %"D, %C+, L$L-$$, D$$ neurosur ery, %+#, %"%-"C, D"*& D$* New Yorker, The, DC"-CB nickelodeon, L++= see also .ukebo(es nodes and links, %$+-$", DLB-LC noise in vacuum, *B nondivisibility, $%-$C nonequilibrium thermodynamics, D#% 8oneuclid, #"-#B none(istence, BLC-LL, D#*, $BL= see also Tumbolia non!roducible numbers, BDL non!ro rammability, of creativity, L$+-$", DB+, D$%= of emotions and will, D$$, D*C-*D= of 6>deli<ation, C$B-$D= of intelli ence, BD-B$, C$"-$%, L#$-##, D+"= of irrationality, L$L-$$= of .um!in out of the system, %$-%*, C$$-$*, D$C$L= of ordinal names, C$D= of soul, L$C-$L= of world chess cham!ion, "L"-LB, D$C= see also !eo!le vs& machines, essential incom!leteness, T>deli<ation, !arado( of )I, T7-battles, B-: vs& %-:, etc& non-self-assemblin viruses, LCB-C% non-self-descri!tive ad.ectives, see heterolo ical ad.ectives

nonsense, based on sense, %$*-$#= com!uterenerated, DB+, DB"-BB, DBL-BD= humanenerated, DB"-BB nontermination, C+*, CBL-%+= see also !otentially endless searches, 'looP nontheorems, see theorems vs& nontheorems normal science, DD+-D" nouns, most common in En lish, D%+ novelty, and .um!in out of the system, C$L nuclei, atomic, %+%-C= cellular, L"C, L"$, L"* nucleotides, L"C-"$, L"#, LBB-BC, L%+, LC+-C"= first letters of, B%", L"$, DDD number theory, a!!lications of, B$*-B#= core of, "++, C+$= 7rab and, LL"-L*, LD+, LDB, L$%-$C, L$#-*"= demise of, BB*-B#, CBD= formali<ed, see T8T= informal 485, LC-DD+, B+C, BB*= nonstandard, "++, CLB-L#= !rimitive notions of, B+C-#= as scaled-off mini-world, LD#= soothin !owers of, %#"-C+C= -true- version of, CL*-L#= ty!ical sentences of, B+C-L= ty!o ra!hical, see T8T= as universal mirror of formal systems, BD+-DL, B$+= used and mentioned, CL* numbers, nature of, LC-L*, CLB, CL* numerals, B+L-D, B"%= vs& numbers, BDC ob.ect lan ua e, BB, "*C, BC* ob.ectivity, quest after, C$#, D#%-#D Eborin, Aev, "DB octo!us cell, %CL Ein, Eta, see Eta Ein Eistrakh, :avid, "DB Ekanisama, B%B, B%C, B%$, B%*, B%#, BC", BCB Eld 0a& 7h&, $BD Eld 0ach, C, B*, CD+, C*"-*%, $%*, $%# \-consistency, CL#= see also \-inconsistency \-incom!leteness, BB"-BB, CB", CL+-L" \-inconsistency, "$, BB%, CL%-LL, CL*-L# "-: vs& %-:, L"#-B", D"D-"$ o!en-ended searches, see !otentially endless searches, nontermination, un!redictable but uaranteed termination, loo!s, free, 'looP, etc& o!eratin systems, B#L-#D, %++-%", %+* o!erators and o!erons, LCC-CL oracles, LD$ orchard analo y, see information, de!th from surface order and chaos, in ant colonies, %"D-"$= in number theory, %#%, %#L, %#*-C+B, C+D& C+*-#, C"*= self-awareness and, C+D :rder and 8haos 4Escher5, %## ordinals, CDB-DC, C$L-$D

or an !oint& %B#-%+ ori in of life, LC* ori inal 4as o!!osed to co!ies5, L+C ori inality and machines, BL-BD, D+D-# E;8)TE 8EF8, "%"-%% outcome, "*C outer messa es, "DD-$", "$C-$D, L+", LBC, $+C EFTPFT 40looP5, C"+, C"" overla!!in enes, LBC-BL overview ca!acity, D"%-"C, D$*= see also .um!in out of the system P-system, DC, $%-$C !addin , C+B-% !a es, in com!uters, B*# !alindromes, in molecular biolo y, B+", DD$ Palindromi, %L%-LC, D%C-%$, DC%, DCC Pa!!us, D+D-$ !aradi m shifts, DD+-D" !arado(, of )I, "#, BD-B$, DB+, D$%, see also Tesler1s Theorem= in an, see Escher, /a ritte, 7a e= of credibility throu h fallibility, LDC= of 6od and the stone, C$*= in mathematics, "$-BC, L*+-*"= of motion, see 9eno1s !arado(es= near misses, D"B, D#"= resolutions of, ""D, "#D-#$, BCL, see also /F, Tumbolia, .um!in out of the system= of self-consciousness, %*#= of the Ty!eless 3ish, ""L-"D= in 9en, BC#-LL= see also contradictions, inconsistency !arallel !ostulate, see fifth !ostulate P);;H, %++-%+", L##-D++, D$$ !arsin of natural lan ua es, L**-#%, D%+-%B= see also rammar, lan ua e !artial recursivity, C%+ !articles, elementary, LC, "C+-CD, BL*, %+%-L, %+#,LBB !artitions, mental, D$" !arts, %+%-L= see also reductionism Pascal, 0laise, BC, BL, D++ !athways, in )T81s and ;T81s, "%"-%C, "L+= chemical, LB*-B#, LCC-CL, DD%-DC= conditional on circumstances, %*%-*C= oal-oriented choice of, BB$, D+#-"L= as incor!oratin knowled e, beliefs, %$*-$#= mor!ho enetic of TC, L%#= !lausible vs& im!lausible, %*%= !otential, in brain, B*" !attern reco nition, see 0on ard !roblems, conce!tual skeletons, vision by com!uters !atterns on all levels, D$C Peano, 6iuse!!e, B+, B"D-"$ Peano arithmetic, "++

Peano !ostulates, B"D-"$, BBC !earl and oyster, "$, C%* Penfield, 3ilder, %CB-C% Penrose, ;o er, "L !eo!le vs& machines, BL-B$, %D-%*, "L"-LB, %**#+, C$"-$%, C$L-$$, LL#-DB, LD$-$L, L$$-$#, L#L-##, D+D-#, DB"-B%, D*+, D*C-*D !e!tide bonds, LB% !erce!tion, visual, #$-#*= and 9en, BL" Perfect items, %, $L-$#, *L, C+D, CBC, C*D, L%D !erfect numbers, C"D, C"* !ha es, see viruses !henoty!e, see enoty!e and !henoty!e fT"$C, "$D, LBC-BL Phono ra!hs, see record !layers !honons, %+C !hotoco!y machines, C## !hotons, "CB-CD, BL* !hysics laws of, as basis for choosin between rival mathematical theories, "++, CLD-L$= as blockin infinite re ress, "$+, D*L= as formal system, L%-LC= inconsistency and, #L-#D, ##, L*C-*L= intuitive, %DB-D%, $""= levels and, %+%L, %+$-*, D#%= no esca!e from, C$$, L$L= ;eductionist1s :ilemma and, LBB, $+#= underlyin consciousness, L$L, D*L, $"+ m, B$$, %+D, C+*, C"L, CB", LCD, LD*, D+L, D$%, $B* Piano !ostulates, LLB-L% !ianos, %-C, %+B, %+L, D%%-%C, $++, $BD= inverted, D*"-*% &ickwick papers, The 4:ickens5, BC& %BD, L#L, L#* !inball machine, %+$ &ipe .ream, $+% !i!es, C*+-*B, C*D, C**, C#%-#C, LB", D%*, $+"-% !lanets and satellites, %L%-LC PA)88E;, DB#-%B !lurals, %LC !ocket calculators, LD*-$+, D"D, D$*, $"+ !ointers in com!uters, B*#-#+, D"# !oints 4 eometrical5, "#-B+, #+, #B-#%, "++, B+$, BBB, CLB, CLD Polanyi, /ichael, L$C !olarons, %+C-L !oly!e!tides, LB%, LBL, LB* !olyribosomes, LBD-B* Pons )sinorum !roof, D+D-$, DD# Pool 0, C"* Pool ', CB$ !o!corn, "+C, "BC-BL

!o!!in , "B$-%L, "*C-*L !o!!in -tonic, "+L-D, ""D-"$, "BL !orrid e, C%" !orts of access, D$+-$" Post, Emil, %% !ost-endin endin s, %#B, C+% !ostal system meta!hor, DD% !ostulates of eometry, #+-#", #B-#%, C+$ !otentially endless searches, %#D, C++-C+B, CBL, C++-C+", CCC, L*B-*% !q-system, com!leteness and consistency of, "+"= decision !rocedure for, C$-C#= e(!ressive weakness of, "+", BB"-BB, C+$, C"$= horsea!!le ha!!y inter!retation of, L", **, B"L= isomor!hisms and, C#-L%, "L*, DBL= modified, *$-**, #B-#%, "+B= sur!rise inter!retation of, LB-L%, #C !redicate calculus, D+# !redicates, number-theoretical, B+*-# !redictable termination, C++, C+$, C+#-"*, CB+, CC", L*B= see also un!redictable but uaranteed termination, terminators &relude, %%$, %*%, CD+, D*D, $%$, $%# !reludes and fu ues, B*+-*C, %%L= see also fu ues, >ell%Tempered 8la#ier !remise, "*C !re!rocessin , DC$, DL+, DL# President v& Su!reme 7ourt, D#B !rimary structure, of !roteins, L"#-BB= of ty!oen<ymes, L"", L"B !rime numbers, L*-L#, DC-D$, $B-$C, "C#, B"""B, C"%, LL"-L*, D"L= differences of, %#%, %#L#*, C++, C"D= sums of, %#%-#D, C++, C"C !rimitive recursive truths, C+$ !rimitive recursivity, C+D, C+$, C"C-B+, CBB, CBC, CB#-%+, CC+, CC", CCC, CL", CDD, C$B &rincipia 'athematica, "*-"#, B", B%-BC, BB*, D"*-"# &rint Galler) 4Escher5, "L, $"C-"$ !rinter of com!uter, %+", %+$ !roblem reduction, D+#-"%= self-a!!lied, D"% !roblem s!aces, re!resentation of, D""-"% !rocedures, "%B-%C, "L+-L", B#B, C"+-"L, C"*-B+, CBC-B*= chains of, C"%-"C, C"L, C"* !rocessors 4com!uters5, L+C, L"%, LC$= see also central !rocessin unit !roducible numbers, BDC-DL, BD#-$+ !ro ram s!ace, B## !ro rams, in )nalytical En ine, BL= in 0looP and 'looP, C"+-"L, CBC-BD= 0lue, 6reen, ;ed, see 0lue, 6reen, ;ed !ro rams= chess-!layin , see

chess !ro rams= constructed by !ro rams, L*#, DB#-%B, DDC= as data, B#%, D#B= for determinin en<yme function, LB"-BB= for determinin !henoty!e, L%B= for determinin tertiary structure, LB"-BB= families of, L+%, LCD= for eneratin theorems, C$"-$%, L$*, D"L, D"$-"*= hi h-level com!arison of, %*+-*"= for namin ordinals, C$D= recursive structure of, "C#-L+= second-order, third-order, etc&, C$D= selfmodifyin , "LB, D#B= self-re!roducin , C#*L+C, LC$= for translatin !ro rams, B#"-#C= vs& data, C##, L"%, L%", LCD-C*, D"D-"$, D%+, see also use vs& mention= vs& !ro rammers, %+D, $%C-%$= see also com!uters, com!uter lan ua es, )I !ro rams, etc& Prokofiev, Ser ei, "L+ !ronoun reference, L*$, L#", L#B !roof-!airs, C"D, C%*-C%, CCD-C$, CL+-L", CLB-LC, CDD, CD*, CD# !roofs, nature of, "*-BC, L*-D+, **-#%, "#B-#$, BB$-B*, CL*-L#, L$*, $+$-*= as never absolute, "#"-#C= of !roofs, "#B-#%= vs& derivations, %L, "#%, "#C-#L !ro!hase, DDL-DDD Pro!ositional 7alculus, "*"-#$= embedded in T8T, "#L, "#$, B+$, B"L-"$= as an e!i!henomenon, L$*= inter!retations of symbols of, "*D, "*#, "#"-#B= rules of inference, .ustified, "**-*#= rules of inference, !resented, "*"-*$= rules of inference, table of, "*$= streamlined, "#%-#C= variants of, "#L= ways to im!rove, "#%-#C= "#D-#$= weaknesses of, "#L-#$, L$*= well-formedness in, "*"-*% !roteins, L"$-"*, LCC-CL= as !rocedural knowled e, D"D-"$= as !ro rams, data, inter!reters, !rocessors, LC$= see also en<ymes !rototy!e !rinci!le, %LB !rovability, "*, "+" Prudence and Im!rudence, "#"-#B, BB# !runin , e(!licit vs& im!licit, B*D !seudo-e!i enesis, L%"-%B !sychic !owers, D#%= see also ESP Ptolemy G E!i!hanes, "DL &uddle 4Escher5, BLD !ullin -out 4mechanisms5, see decodin 4mechanisms5 !unctuation, %%, BD*, CC+, L"+, L"B, LB+, LBC-BL !urines, L+D-$, L"C, L"D, L%C= see also nucleotides, bases, base-!airin !ur!oseful vs& !ur!oseless behavior, %B+-BB !ushcorn, "BC-BL

!ushin , "B$-%C, "*C-*L !ushin -!otion, "+L-D Pushkin, )le(ander, "BC !u<<les& *, %%-%L, DB-D%, D$, $%, "%$, "*B, B"B, B"L, BB+, C+", C"L-"$, CBL, CCB-C%, CCC, L"B"%, LDC-LDL, D+#, DB"-B%, DCD-D+, D*# !yramidal family of theorems, BB"-BL, CL+-L% !yrimidines, L+D-$, L"C, L"D, L%C, see also nucleotides, bases, base-!airin Pytha oras, C"*, LLD-L$ M4n5, "%$-%*, "LB, BDL, C+# quantifiers, B+$-#, B"+, B""-"B& B"C, B"$-"# quantum mechanics, "#, LC& "C+-CD, %L+, CLL, CL$, D##= see also !articles Muant<, Joachim, C quarks, %+C, %+L, %L+ quasi-isomor!hisms, see isomor!hisms, fluid quaternary structure, LBL Muestions and S!eculations, D$D-*+ Muine, 3illard Gan Erman, C%L, CCD, CC#, D## quinin , C%"-%$, CCL, CCD, CC#, C#$-##, L%" MFIT 40looP5, C"B quotation, CBD, C%", C%%-%$, C#D-#$, $+B, $%* quotation marks, %%, C%C, C#*, C##, $+B ;achmaninoff, Ser ei, "L+ ;)7;E7I;, $%* radio broadcasts, "B*, "D%, "D#, %L%, C$*, LCL, $B+ ;amanu.an, Srinivasa, LDB-DD, D"C randomness, C+*-#, DB+, D$%, $"B rational and irrational numbers, "C+-CB, C"*, CLB, LLD-L$ rational vs& irrational, in human head, L$L-$$ rats in ma<es, %CB ;auschenber , ;obert, $+% r&e& sets, see recursively enumerable sets readin frame shift, "LC, LBL reality, nature of, C+# rearran ement of !arts, $*, %%%-%L, C*C= see also self-assembly reasonin about formal systems, %*-%#, DD, BD+$B, C%*-LB, CDL-$", L$#-*" reasonin by !ro rams, LD#-$+, L$$-$*, L*D-#%, D+D-$, D+#-"", D"C-"L, D"*-"#, DB*-%B reco nition, molecular, LC+-C"= visual, %CD-C*, DCD-DB= vs& !roduction, DC*-C# reco ni<able forms, D* recombination, DL$, DDL-D#

record !layers, alien-re.ectin , C*$-**= E!silon9ero, C*D= family of, in 7rab1s .ukebo(, "LC-L$= 6rand Self-assemblin , see record !layer E!silon-9ero= as information-revealers, "L*-D", "DC= intrinsic vulnerability of, $L-$*, "+B, CBC, C$+, C*%-*D, L%D, LC%, L*C, $B", see also T>deli<ation, T7-battles, etc&= likened to formal systems, *C, *L= low-fidelity, $$, *L, "+", C+D$, C$+= Eme a, $*, CD*, C*%-*C= 8umbers l, B&&& etc&, $D-$$= Tortoise-chom!in , C*%, C*$-**= two-channel monaural, D%C, DD#= see also .ukebo(es records, defective, "+B, as information-bearers, "L*, "D+-D", "DC= as labyrinths, "B+-BC= with multi!le melodies, "LC-L$= as !hono ra!hbreakers, $L-$*, *%-*L, B$", C+D-$, CBC, CD#, C*C, C*D, L%D, LC%, L*C= smashed, information in, "D"= in s!ace, "DB-DC, "$B, "$C-$L= of >ell% Tempered 8la#ier iven to 7rab, B$L, B$*-*+ records and record !layers, likened to cellular constituents and cells, *%, "L*-DC, "D$, "$L, L%D recursion, avoidance of infinite re ress in, "B$, "%C-%L= avoidance of !arado( in, "B$= defined, "B$-B#, "%"-%L= elementary !articles and, "CBCD= fantasy rule and, "*C-*L= in ame-!layin !ro rams, "L+-L", D+C-L= indirect, "%C, "%$= in lan ua e& "%+-%C, L**, L#", L#B= in music, "B"B%, "B#-%+= and un!redictability, "LB= see also nestin , levels, distinct vs& similar, level confusion, etc& recursive acronyms, ""%, "%%, "%C-%L, $%*, $CB recursive dia rams, "%L-%$ recursive fi ures, D$-$+& $B, $% recursive formula, of thinkin , LD+ recursive functions, "%D-C+, "LB, C%+, CLL= see also eneral recursivity, !rimitive recursivity, 0looP, 'looP recursive ra!hs& "%*-C% recursive sequences, "%L-%*, "%# recursive sets, $B-$C, "LB, "#" recursive structure of ideas, %*D-*$, LD+, DB", DCC-CL, DL+, DLD-L$, DD#& D$"-$B ;ecursive Transition 8etworks, "%"-%C, "%D, "CL, "L+, DB+-B" recursively enumerable sets, $B-$C, "LB, "#", BDL, BD# recursively related notation-systems, C$L ;ed Pro rams, CB$-B* =eddiag U8V, CB*, CB# redness, sub.ective and ob.ective, $"+

reductionism, defined, %"B= !roteins and, LB+-BB= see also holism vs& reductionism, sealin -off ;eductionist1s :ilemma, LBB, $+# reentrant code, %*$ refri erators, see record !layers, low fidelity re isters, in com!uters, B*# relativity, "#, #D, "++, D*+ =elati#it) 4Escher5, #$-#* relevant im!lication, "#$ renormali<ation, "CB-CD, BL*, %+C-L, %+# re!eatability, see 6>deli<ation, T>deli<ation, dia onal method, Escheri<ation, T7-batteries, answer-schemers re!resentability, C+$, C"$-"*, C%+, CC", CC%, CCC, CL", CDD, CD*, L$#-*+ re!resentation of knowled e, in )I, LD#, D"L-B", DBD-%B, DC"-L#, DDC-DL, DD*-$B= in brains, see symbols, locali<ation re!ressors, LCC-CL =eptiles 4Escher5, ""D-"$ ;equirement of 'ormality, %%, LB, DL retro ression, *-#, *", "CD, B++, B+*, L++-L+", LC#, DDD-D*, $B%-BL, $%$-%* return addresses, "B*, "%% revelation, "D+-D", "$L ribosome, B%D ribonucleic acid, see m;8), r;8), t;8) ribosomal ;8), see r;8) ribosomes, as models for )I, DDB, DD%= molecular canons and, LB$-B*= need for in :8)1s self re!, L%+= ori in of, LB*, LC*= as self-assemblin ob.ects, C*L-*D, LCB= structure of, LB*= as translators of 6enetic 7ode, C*L, L"*-"#, LBBBL, LC$= in Ty!o enetics, L"B ricercar, defined, $ ;I7E;7); 4E5, $, $B$-CB =ippled -urface 4Escher5, BLD-L$ ;8), see m;8), r;8), t;8) ;8) !olymerase, LB$, L%+, LCC robot in T-ma<e, $""-"% ;o ers, ?artley, C$% ro!es, thin and thick, BB#-%+ ;ose, Steven, %CB ;osetta stone, "DL, "DD ;osaak, Theodore, L$C ;ousseau, ?enri, D*+ ;oyal Theme, C-"+, #D, $"#, $%#-C+ r;8), LB* ;T81s, see ;ecursive Transition 8etworks rule-less systems, L#*, D*L= see also formal vs& in formal systems

rules, arithmetical vs& ty!o ra!hical, BDB-DC& BD#= flattened into strin s, see theorems vs& rules= intelli ence and, BD-B$, LL#, see also brains and formal systems rules of inference, of 7-system, DL= com!ared with en<ymes, L+#-"+, L"%, L%"= defined, %C%L= derived, "#%-#C= of /IF-system, %C, BD+= of P system& $C= of !q-system, C$= !ro!osed, DD, BB"= of Pro!ositional 7alculus, "*$= recursive enumerability and, "LB= run backwards, C*-C#, "*B= of %"+-system, BD%= of T8T, B"L, B"$-B+, BB%-BL= of tq-system, DL= of Ty!o enetics, L+#"+ rules of !roduction, see rules of inference run-of-the-mill sets, B+-B" ;ussell, 0ertrand, "*-BC ;ussell1s !arado(, B+-B", D*L Saccheri, 6irolamo, #"-#%, ##, CLB, CLD Sa redo, see Salviati, et al Salviati, Sim!licio, Sa redo, C+*-#, C$*-$#, D$%, D#C sameness, of )SF1s, %$L= of 0)7? and 7)6E, "L%-L$= in 0on ard world, DL+-L%, DL$, DD+, DDC= of butterflies, "C$, %D#= of demi-doublets, DD#= elusiveness of, "CD-C#= of Escher drawin s, "C$= of human and machine intelli ence, %%$, %$#, D$#-*+= of human minds, %C"-CB, %D#-$B, %$L-$$, %*B= intensionality and, %%*= mechanisms underlyin !erce!tion of abstract, DCD-DB, DDL-D#, D$"-$B= overlooked, D"C, D$C= of !ro rams, %*+-*B= in self-refs and self-re!s, L++-C= of semantic networks, %$"= of translations between lan ua es, %$B, %$#-*+= universality of intelli ence and, "L*, L+"= vs& differentness, "L%-L$= visual, %CC-C*, DDB= see also co!ies, isomor!hisms, conce!tual ma!!in sameness-detectors, see Sams Sams, DL+-L%, DL$, DDC Samuel, )rthur, D+C-L, D*C-*D Samuel1s ar ument, !ro and con, D*C-*D -an 7rancisco 8hronicle e(am!le, %L" sand castles, $BL-BD sanity vs& insanity, "#B, D#D satellite-symbols, see s!littin -off satori, see enli htenment scale, cyclic, see She!ard tones Schmidt, Johann /ichael, B$ Schnirelmann, Aev 6&, %#C Sch>nber , )rnold, "BL Schr>din er, Erwin, "D$

Schweikart, '& @&, #B science, and 0on ard !roblems, DL#-D"= selfa!!lied, D## Scott, ;obert, %DD scri!ts, colla e of, "D*-D# sealin -off, %+L, %+#, %L+, L%C secondary structure, LB", LBL self, nature of, %"D-"$, %B$-B*, %*C-*L, %*$-**, D#L-#D, $+#-"C self-assembly, s!ontaneous, C*L-*D, LCB-C% self-awareness, C+D, C$#, L$% self-descri!tive ad.ectives, see autolo ical ad.ectives self-en ulfin , C*#-#C= failed= C#+, C#B= total, C#% self-knowled e, !ossibility of, D#D-#*, $+D self-modifyin ames, D*$-** self-monitorin , %B*, %*L,-%*$-**, D#$, $"% self-!erce!tion, D#L-#*= vs& self-transcendence, C$* self-!ro rammed ob.ects, D*L-*D, D#"-#B self!rovin sentences, LCB-C% self-quotin sentence, CBD, C#D-#$ self-reference, 0ach and, *D= bannin , B"-B%= as cause of essential incom!leteness, CDL, C$+-$"= focusin of, C%*, CC%, CCL-C*= 6>delian, "$-"*, B$", CC$-C#, C#$, L+B, L%%, DD$, $%*= indirect, B", *L, B+C, C%D-%$, L+B, DD$, $%*-%#= many leveled& $CB= near miss, C%$= Muine method, C%"-%$, CCL-CD, CC#, C#$-##, L%"= by translation, L+B self-reference and self-re!lication, com!ared, L%+, L%%-%C, LC"-C% self-referential sentences, C%L-%$, C$$, C#L-##, L+" self-re!, by au mentation, L+%= canons and, L+", L+%= differentiatin , LCD= e!i enesis and, "D+= by error messa e, L+%= ine(act, L++-L+%, LCD= by retro rade motion, L++-L+"= by translation, L+"= trivial, C##= ty!o enetical, L"B-"% self-snuffin , $+"-B self-swallowin sets, B+ self-symbol, %*L, %*$-**, $+#= free will and, $"+"C= inevitability of, %** self-transcendence, C$$-$*, C$# self-unawareness, irony of, %B*, %%+, %%", D%+ semantic classes, DB", D%+ semantic networks, %$+-$B= see also conce!t network semi-inter!retations, "*#, "#D

semiformal systems, B"D= see also eometry, Euclidean senseless loo!s, D$# sentences in T8T, B+*-# Sentences P and M, C%D-%$ sequences of inte ers, $%, "%L-%#, "$%-$C, C+* set theory, B+-B% sets 7 and G, $% "$B#, B+C-L, B"+-"", %CL, %#%, LL", LDC-*LL -hadows, The 4/a ritte5, C*+ Shakes!eare, 3m&, #D, L#L, L#*, D+*, $%D Shandy :ouble-:andy, D"" shared code, %*$ She!ard, ;o er, $"$-"# She!ard tones, $"$-"# shieldin of lower levels, see inaccessibility S?;:AF, L*D-#%, L##, DB$-%B, D$C Shu<an, BL" Sier!iski, 3&, C+C si nals, crisscrossin , %BB-B% si nature, visual, %C$-C* Silberescher, Aowen, %#C Silbermann, 6ottfried, %, C silver, "$% Simon, ?erbert )&, %+%, %+L sim!le, com!le(, hy!ercom!le( cells, see neurons Sim!licio, see Salviati sim!licity, "$B, LD+, D"L simulation, of entire brain, L$B-$%= of neural networks, L$"-$B -i9%&art =icercar 40ach5, C-$, $"#, $%#-CB skater meta!hor, C"B-"% skeletons 4recursion5, "C+-C"= see also bottom C skimmin off to! levels, %+#, %BL, %BD, %L*-L#, LD*-$# slinky, %%$ sli!!a e, conce!tual, D%%-C+& DC"-CC, DLC-LD, D$B= see also conce!tual ma!!in Sloth, D%%-C+, DC%& D*"-*%, $BB -loth 8anon 40ach5, #, DDD, D*% -loth 8anon 4:ialo ue5, $%* SAET?s, D*C-$"# slots, DCL, DL+-L%, DLD-L$, DD* Smalltalk, DDB smart-stu!ids, $B"-CB Smoke Si nal D$, $+B S/FT, D, *+, "LL-L$, B+B, D*B, $"*, $C+ soa! cake, C#$ soft-louds, see !ianos

software and hardware, in brain, %CD, %LD-L$, D*D, $+#= defined, %+" soldier ants, %"* -onata for 5naccompanied *chilles, L+B-% Sonata for violin and clavier in ' minor 40ach5, "DB Sonatas and !artitas for unaccom!anied violin 40ach5, DB, D%, $+-$", BL$, L+B son , self-re!roducin , L++ -Son 3ithout Time or Season, )-, C*% son s, recombinant, D+$-* sonnets, L#D, D+*, $%D soulism, %*L, C$B, L$C, L#$, D*D, $+* s!aces, abstract, CL$= of behavior, %+D, %+$, %D%, DB"= of inte ers, L+B s!are tire, D$+ S!erry, ;o er, $"+ S!he( was!, %D+-D", D"%-"C s!iderwebs, %$"-$B, D"$ s!littin -off, %L%-LC, DDC= see also classes vs& instances stacks, !ushdown, "B$-%L, "%D, D"+-"" startin !ositions in )rt of 9en strin s, B%#-C+, BC" -tate of Grace 4/a ritte5, C*" statistical mechanics, see ases Sterner, 6eo&, "DD-D$, DCB-C%, D$+ Stent, 6unther, L"C stomach !rocesses, L$B strands, of :8) and ;8), L"C-"*= in Ty!o enetics, L+*-"%, L"C Stran e Aoo!s, abolishment of, B"-B%= with 0abba e and Turin , $%$= in 0ach, "+, $"#= in 7entral :o ma!, L%C= consciousness and, $+#"+= contrasted with feedback, LCL, D#"= defined, "+= in Escher, "+-"L, $"C-"$= in 6>del, "L-"$, BC= in overnment, D#B-#%= in lan ua e, BB, D#"= in the mind, B$, D#"-#B= in molecular biolo y, B%", L%B-%C, LCL, LCD-C*= in Princi!ia /athematica, BC= sur!rise element necessary in, D#"= in T8T, see T8T, intros!ection of= see also Tan led ?ierarchies, level-mi(in , levelconflicts, level- confusion, etc& street nuisances, BL, $BD-B$, $B# stretto, %"C, $%* Strin /ani!ulation ;ules, BC+-CB strin s, defined, %%-%C= folded, B%C-CC, CB$ structure vs& function, LBB, D$+-$"= see also use vs& mention, synta( vs& semantics style, "C*, %$" SF0 4T8T-formula5, CCC-CL

subbrains, see subsystems subdescri!tions, etc& DL+-L"= see also recursive structure of ideas subframes, etc&, DCC-CL= see also recursive structure of ideas sub.ect vs& ob.ect, D#*-##= see also dualism, use vs& mention, symbol vs& ob.ect sub.unc-TG, D%L-C+ sub.unctive instant re!lays, see instant re!lays sub.unctives, see counterfactuals subor anisms, see subsystems subroutines, "L+, B#B, C$", D$$ substitution notation in T8T, defined, BBC substitution relationshi! 4in redient in 6>del1s !roof5, CC%-CL, C#$ substrate, of E!imenides sentence, L*", L*C-*L= mental, necessity of, for analo ical thou ht, LD#-$"= mental, noninter!retability of, L$+-$"= mental, simulation of, L$"-$B= of !roteins, LB# subsystems of brain, %*L-**, $BL subtraction, LB-L%, C""-"B sufficiently !owerful systems, *D, "+", C+D-$, C%+, CD+, L%+ Suites for unaccom!anied cello 40ach5, $+-$" summari<in strin , BB"-BB, CL+-L" su!erconductivity, %+C-L su!erintelli ence, D$# su!ernatural numbers, BB%, CL%-LD, CL*-L#, CD$ su!ernatural !roofs, CLC-LL su!ertan lin , D** surrealism, $++ Sussman, 6erald, DDC Swieten, 0aron 6ottfried van, D-$ Switcheroo, M& q&, "*$ symbol-level descri!tion of brain, %C#-L" symbol vs& ob.ect, D##-$+D= see also sub.ect vs& ob.ect, use vs& mention, etc& symbols, active vs& !assive, %BC-BL, %B$-B*, %%$%*= )I reali<ations of, DDB-D%, DDL= in ant colonies, %BC-B*, %%+= borderlines between, %LC-L$, %L#-D+= as brush strokes, %L"= com!ared with neurons, %L+, %$"= com!ared with ri!!les, %LD-%$= conce!tual sco!e of, %L+L"= dormant, %B$, %C#, %LL-LD, %*C= form of, %C*, %LD-L$, %D"= free will and, $"B-"C= funnelin and, %C$= of insects, %D+-D"= invariant core of, %C#= .oint activation of, %L", %LC-LD, %L#, %D", %DL, L*C-*L, DDC-DL, D$L= messa ee(chan e and, %L+, %$", DDB-D%= modes of activation of, %C#-L$, %D"= need of, for ori inality, D+#= neural substrate of, %LD-L$,

L$+= no access to substrate by, see inaccessibility= overcrowdin of, %L*= overla! of, %C*-C#, %LD-L$= !otential, %LL-LD, %*B-*C= universal, %$L-$D= vs& neurons, %C*, %LD-L$, %D"= vs& si nals, %BL-B$, %C#-L+ syna!ses, %%# synta( vs& semantics, DBD-B$, D%+-%B, D$D= see also form, syntactic vs& semantic system, boundaries of, %$-%*= see also .um!in out of the system System crash, ""D systems, reliable vs& unreliable, %+$ T 4Tarski1s formula5, L*+-*" 1t1-conce!t, 1h1-conce!t, 1c1-conce!t, %BD T-even !ha es, LC+, LCB T-level, see tan led level table of nontheorems, DD Ta ore, ;abindranath, "D# Tan led ?ierarchies, in art, $+C= defined, "+= of enetics, L%B-%C, LCD-C*, D**= of metamathematics, CL*, L%B-%C= of mind, D#"#B, $+#-"+, $"#= near miss, D#"= in Pro!ositional 7alculus, "#C= of self-modifyin chess, D**= of Tortoise1s reasonin , "$$-*+= of Ty!o enetics, L"%, D** tan led level, D** Tan uy, Hves, $++ ta!e recorders, C*L, L"#, LB%, LBL-B* Tarski, )lfred, L$#-*" Tarski-7hurch-Turin Theorem& LD", L*" Tarski1s Theorem, L$#, L*+-*", L*C-*L, D#$ Taube, /ortimer, L$C Taurinus, '&&C&, #B T-7-battles, $L-$*, C+D-$, CBC, CD$-$+, C$*, C*%**, L%D-C", $B" tea, "L%-LC, B%", B$L, %BB-B%, %%%, LC#, LL$, LL*, LD" teleolo ical vs& evolutionary view!oint, %B+-BB tele!hone calls, D"-D%, "B$-B*& "D"= obscene, C%", C%$ tele!hone systems, B#D, DD% tele!hones, %+B television, B*L, %C*, C$*, C*C, C*$-#%, D%C-C+ telo!hase, DD$-D* tem!lates, for 0on ard !roblems, DL+-L%, DLDL$= vs& instructions, see instructions vs& tem!lates, !ro rams vs& data tension and resolution, "B"-B%, "B#-%+, BB$ tentativity, DCD, DL", DLC-LD, D$B termination tester, CBL-B#

terminators 4'looP5, CBL-B# terms 4T8T5, B+D-$, B"%, B"C tertiary structure, of !roteins, L"#-BB, LBL-B$= of t;8), LBC= of ty!oen<ymes, L"+-"", L"B, L"# Tesler, Aawrence 6&, D+" Tesler1s Theorem, D+", DB%= see also !arado( of )I, .um!in out of the system, essential in com!leteness tesselations, D*, D#, "#*, DD$ tests vs& functions in 0looP and 'looP, C"%, C"* te(t-handlin by com!uters, %+" TC !ha e, L%$-C" -the- L*D, DB#-%+ theorem-numbers, BDC-D$, BD#-$+, CC+-C%, CL" theorem-!rovin , mechanical, D+B, D+#, D"$-"# theorems, defined, %L= systematic enumeration of, %#-C+, C*, C$"-$%, L$*, D"L, D"$-"*= vs& nontheorems, %#-C", DD-D$, $+, $"-$%, "#+-#", C"D"$, LD+, L$#-*+= vs& rules, C%-CL, "#%-#C, L+#"+= vs& Theorems, %L, "#%= vs& truths, C#-LC, $+$", *D-"+B, "#+-#$, B"%, BB"-B%, BB*-%+, see also consistency, com!leteness, 6>del1s Theorem, consequences of Theseus and )riadne, "%+ thinkin , s!eed of, D$# %+, as !ossible /IF-number, BDL-D$ -this sentence-, C%D, C#L-#*, C## thou ht, substrate of, LL# %n^" !roblem, C++-B= see also wondrous and unwondrous numbers Three -pheres 33 4Escher5, BL* %"+-system, BD"-D$ Three >orlds 4Escher5, BC$, BLD thymine, see nucleotides tilde, "*%, "#"-#B, LLC timesharin , B#D, %LC-LL, %*$, $%+ T8T, B+C-%+= absolute eometry and, CL"-LB= austere, B"", B"C, B"D, BD*, CCB, L%C= a(ioms of, B"L-"$, BBB-B%= as code, BDL-D$= consistency of, BB#-%+, CC#-L+= e(tended, a(ioms of, CL"-LB, CDD-D*= e(tensions of, CL"L#, CDL-D*= 'I6F;E-'I6F;E fi ure and, $+= as eneral metalan ua e, BDL-D$= oals of, D+= incom!leteness of, see incom!leteness= intros!ection of, "$, "#C, BD$-$B, C+D, C%*, CC%, CC#-L+, D#*, $+$-*= as its own metalan ua e, BD$-$B, CC"-CD, L"C= !li ht of, !ictorial version, $"= rela(ed, BB*= rules of formation, table of, B"%-"C= rules of inference, B"L, B"$-"L= si(th a(iom of, BBB-B%, CL"-LL, CL#, CDL-D*= well-formedness in, B+L-"L

T8T-derivations com!ared with machine lan ua e, B#" T8T-level, e(!lanations on, $+* T8T-numbers, BD#-$+= see also theorem-numbers T8T ^ 6, CDL-D$, C$" T8T ^ 6 ^ 61, etc&, CD$-$" T8T ^ h6, CD$ tobacco mosaic virus, C*C-*L, LCB, LC% T>deli<ation, re!eatability of, $D-$*, CBC, CD$-$+, C*%-*D= see also 6>deli<ation, essential incom!leteness, answer schemas, etc& T>del1s Theorem, C*D, L%D Tokusan, "*#-#+ tonic, musical, "B"-B%, "B#-%+ Tortoise, answer-schemas and, C$L= 7arroll !arado( and, CD, "#%, D*C-*L, D#%= 7rab 7anon and, B+C, DDD-D$= de enerate solution by, DD#= :io!hantine equations and, CL#-D+= as har!sichord, L+B= initial letter of, B%", L+$, DD$= mentioned, "+B, "$+, BD$= ori in of, B*-B#= !icture of, CB= recursion and, "B*-%+= "C#= use of words by, "*"= vs& 7rab, *C-*D, B$", C+D, CBC, CD$-D#, LC+, LC%= 9en strin made by, B$B= in 9ET cycle, #C-#L Tortoise-!airs, C"D, CC", CC*= com!ared with !roof-!airs, CC", CC* Tortoise !ro!erty, %#L-#$, C"L-"*, CBL, CBD, CC", CC* Tortoise1s love son , C%L-%D Tortoise1s method, see T>deli<ation TR<an, "#+, BLL, BL$, C$# tq-system, DC-D$ transcendentalism, $+C transcri!tion, :8) to m;8), L"$, LBC, LB$-B*, L%+, L%%, L%D, L%*, LC+-C", LCC-CL= :8) to t;8), CBL= kRans to messen ers, B%L-%D, B%*, B%#, BCB= letters to notes, *%= !revention of, LCC-CL transfer ;8), see t;8) translation, between com!uter lan ua es, "#B-#L, B#$-#*, %+D, %*+-*", LC$, D%B= between levels of a brain, %C#, %*"-*C, $+#= between natural lan ua es, difficulties of, %$B-$%, %$#-*+= between T8T and meta-T8T, BD$-$B, CC"-CD, $+#= of 7rime and Punishment, %$#-*+&= En lish to T8T, B+#-"%, B"L, C"$= of -Jabberwocky-, %$B-$%, %$#= levels of fidelity in, %$#-*+= mechanical, %*+, D+", D+%= of messen ers into strin s, B%C-%D= m;8) to !roteins, C*L, L"*-"#, LBB-BL, LB$-B*, L%"-%D, L%*, LCL, LCD-C*= from 8 to /eta-T8T, L%%=

from score to sounds, *%= in Ty!o enetics, L+#"+, L"B-"% trans!arency to user, DB#, D%B trees, look-ahead, see look-ahead trees= recursive dia rams, C+ $", "%L-%$= of theorems, C+, $" tri erin !atterns of symbols, de!endence on meanin less lower levels, LD#= isomor!hism between minds and, %D#, %$D= isomor!hism between !hysical law and, %DB= as key to meanin , %BL, %B$, %L+, %D+, %*L, D+#= mediated by messa es, %L+, %$"= for melodies, %DC= nouns vs& verbs, %D"= randomness in, D$% tri ers, :8) as, "D+-D"= dormant symbols and, B*", %*%, %*C= frame messa es and, "DB= .ukebo(es and, "D+-D", "$+-$", "$C, L++= kRans as, BCD= music and, "DB-D%, B*", %*%, L*%= outer messa es and, "DD, "D#, "$+-$", "$C, L+" Trio Sonata from the 'usical :ffering 40ach5, $*, $B+, $BC, $BD Tri!itaka, BL$ tri!-lets, cover, 9i#, ", B*, B$% t;8), LBB-BC, LC$, LC* Tro.an ?orse, L%* truth, ca!turability by symbol mani!ulation, L%D+= elusiveness of, D#C-#L= ine(!ressible in T8T, L*+-*"= not fully mirrorable in brain, L*C*L= vs& beauty, LLC-L*, L*C= vs& commercials, C$*= vs& falsity, $+, $", B"%, BB*-B#, C"$, LD", L$#-*" TTortoise, see )TT)77) tuba, flamin , C**-*#, C#B, $%L Tumbolia, ""D, BC%, BLL& $BL= layers of, BC% tunin an )I !ro ram, D$*-$# Turin , )lan /&, BD, %*#, CBL-BD, CB*-B#, L#C##, $%C-CB= ob.ections to )I, L#$-## Turin , Sara, L#L Turin machines, %#+, L#C, $%L Turin test, L#L-##, D++, D$$-$*, $%L-%$= arithmetic error in, L#D= miniature, DB"-B%= !ro!osed revisions in, D++ Turtle1s Theorem, see T>dels Theorem B, as conce!t, D$* B-: vs& %-:, in Escher, L$-L*, "+L-D, "BL, C$%$C, LBC, D*#-#+, D#*, $"C-"D= in /a ritte, C*+*", C#%-#C, $++-", $+L-D= television screens and, C**-#%, $%$= tri!-lets and, see tri!lets Two ')steries, The 4/a ritte5, $+"-B Two%&art 3n#ention, B*, C%-CL, D*C-*D= see also 7arroll !arado(, Aewis 7arroll B ^ B X L, L$D ty!eface meta!hor, LC"

Ty!eless 3ishes, """-"L, D"+-"" ty!es, theory of, B"-B% ty!esettin machine, D+* ty!oen<ymes, L+L-"%= bindin -!references of, L+L-D, L""-"B Ty!o enetic 7ode, L"+, L"B, L"%, L"# Ty!o enetics, L+C-"%, L"C, L"#, LB+, LB#= contrasted with /IF-system, L+#-"+, L"C Ty!o ra!hical 8umber Theory, see T8T ty!o ra!hical o!erations, defined, DC ty!os, C+C F, as nontheorem of /IF-system, %D, %# Fmode, see Fn-mode Flam, Stanislaw, LD+, DB", D$D Fn-mode, %#, #*, BLC Fnamuno, /i uel de, D#* uncertainty !rinci!le, see ?eisenber uncertainty !rinci!le uncles, CCD-C*, CDC, CDD, CD*, LC", L*+ undecidability, "$, BBB, CC#, CL"-LL, CD*= causes of, $+$-* undefined terms, #B-"+B, B"D, CLD= defined, #%, #$ understandin , nature of, LD#, D$L-$D, D*+ understandin mindsIbrains, meanin of, D#$= !ossibility of, D#$-#*, $+D-$ unicycle, tandem, D%%, DD# units in Ty!o enetics& L+L, L+# universal quantifiers, see quantifiers universal tri erin !ower, "$", "$L Fnmon, BLC unobservant machines, %D-%$, D$C un!redictable but uaranteed termination, C++, CBL u!!er bounds, see loo!s, bounded, 0looP uracil, see nucleotides use vs& mention, C%C-%$, CL*, L%", LCL, D##-$++= see also form, syntactic vs& semantic, !ro rams vs& data, synta( vs& semantics, structure vs& function variables in T8T, B+D, B"%-"C= free, B+$-#, B"C= quantified, B+*, B"C, see also quantifiers verb-at-the-end !henomenon, "%+-%" <er(um 4Escher5, BL$, $%"-%B versus, see accessible vs& inaccessible knowled e, active vs& !assive symbols, ants vs& ant colonies, arithmetical vs& ty!o ra!hical rules, 0ach vs& 7a e, beautiful vs& non-beautiful, bottom-u! vs& to!-down, classes vs& instances,

continuous vs& discrete !rocesses, deductive vs& analo ical awareness, derivations vs& !roofs, dissection vs& a!!reciation of 0ach, distinct vs& similar levels, en<ymes vs& ty!oen<ymes, e(!licit vs& im!licit knowled e, e(!licit vs& im!licit meanin , e(!licit vs& im!licit !runin , formal vs& informal reasonin , formal vs& informal systems, formal systems vs& reality, enuine vs& !hony kRans, hi h-fidelity vs& lowfidelity, holism vs& reductionism, im!rovisation vs& intros!ection, instructions vs& tem!lates, local vs& lobal !ro!erties, meanin less vs& meanin ful inter!retations, men vs& women, minds vs& brains, nouns vs& verbs in tri erin !atterns, "-: vs& %-:, !assive vs& active meanin , !eo!le vs& machines, !lausible vs& im!lausible !athways, President v& Su!reme 7ourt, !rocedural vs& declarative knowled e, !ro rams vs& data, !ro rams vs& !ro rammers, !ur!oseful vs& !ur!oseless behavior, rational vs& irrational, sanity vs& insanity, self-!erce!tion vs& self-transcendence, structure vs& function, sub.ect vs& ob.ect, symbol vs& ob.ect, symbols vs& neurons, symbols vs& si nals, syntactic vs& semantic form, synta( vs& semantics, teleolo ical vs& evolutionary view!oint, theorems vs& nontheorems, theorems vs& rules, theorems vs& Theorems, theorems vs& truth, Tortoise vs& 7rab, truth vs& beauty, truth vs& commercials, truth vs& falsity, B-: vs& %-:, use vs& mention, wei ht vs& mass, womanseein vs& seein , words vs& letters, 9en vs& lo ic, 9en vs& words vibrations, $D-$*, *B-*L, "+B, B$+, B$", CD# Gice President, D$+ Gillon, 'ranvois, %D# Gino radov, Ivan /&, %#C-#L Gino radov !ro!erty, %#C-#L violins, DB, D%, $+, *", *C, "DB, B++, BL$, C%C, L+B, L#L, D*", $B+, $BC viruses, L%D-C%= likened to ?enkin sentences, LCB-C% vision by com!uter, D+B, DB$ visual ima ery, 0on ard !roblems and, DD", faucets and, %DC-DL= inaccessible knowled e and, %DL= lack of in !ro rams, DB%= mathematics and, LD#, D$*= necessitatin layers of substrate, L$+-$"= !ower of, %%*-%#= role in conce!tual ma!!in & DD*, D$B= rubbin -off and %D"-DB

voices in fu ues and canons, B*, B*B-*%, %"C, %BB-B%, %%L, DDL-D$, DD#, D*%, $%$, $C+ Goltaire, 'ranvois /arie )rouet de, % vortices, $"%-"# Guillard, Edouard, %C$ 3achter, '& A&, #B 3arrin, 'rank A&, %DD -3)S? /E-, D+* >aterfall 4Escher5, ""-"%, ##, $"D 3ater ate, D#B 3atson, J& :&& DD$ 3easel, "+D weather& %+B-% 3eaver, 3arren, %*+ 3eierstrass, @arl 3& T&, C+C wei ht vs& mass, "$"-$B 3ei<enbaum, Jose!h, L##-D++& D$L well-formed strin s, in )rt of 9en Strin s, B%#, BCB, BCC= 0looP !u<<le on, C"D= defined& L%, in !q-system, C$= in Pro!ositional 7alculus, "*"*%= in T8T, B"%-"L >ell%Tempered 8la#ier 40ach5, $, B*+-*C, %B$& %B#, %%L 3ell-Tested 7on.ecture 4'ourmi5, %%%-%L -3hen Johnny 7omes /archin ?ome-, D+$ 3hitehead, )lfred 8orth, "*, B", B%-BC 3hitely, 7& ?&, C$$ wholes, see holism 3iener, 8orbert, D*C 3iesel, Torsten, %C% will, free, see free will= mechanical, D*C-*D= roots of, D*C-*D 3ilson, E& E&, %L+ 3ino rad& Terry, DB$-%B 3inston, Patrick ?enry, B## wishes, meta-wishes, etc&, "+#-"D 3itt enstein, Audwi , D*+, D## 3olff, 7hristo!h, %#B

womanseein vs& seein , womansee men vs& women wondrous and unwondrous numbers, C++-B, C+*, C"L, C"*, CBL 3ooldrid e, :ean, %D+ words, in com!uters, B**-#+, B#L, C""= as !ro rams, DB#-%+= s!elled backwards& *", C"*, CB$, C#*, L+L, L%%, LC#, $B$, $%*, $C+= thou hts, formal rules, and, see main theses, vs& letters, %BL-B$, L$+, L$"= 9en attitude towards, BCD, BC#, BL"-LC worker ants, %"* workin inside the system, see /echanical mode yes-answers, CD"-DC Hn ve, Gictor, DB+ Houn , Aa/onte, $++ 9en )l ebra, L$$ 9en 0uddhism, )chilles teaches the Tortoise about, B%"-CL= com!uters and, DBL-BD= 7rab1s refri erator and, C+D-$= Escher and, BLL-L*= holism vs& reductionism and, %"B-"%= inconsistencies and, see 9en vs& lo ic= introduction to, BCD-L#= .um!in out of the system and, B%%, BLL, C$#= music, art, and, D##, $+C-D= /ystery of the Fndecidable and, B$B= none(istence and, BLC-LL, D#*= !atriarchs of, %+, B%B, BLB, BL#= quasi-, DBL-BD= F-mode and, %#, #*& BLC= vs, lo ic, ##, B%%-%C, BC#-L", BLC= vs& words, BCD, BC*-C#, BL"-LC= 9eno and, %+ 9en Strin s, )rt of, see )rt of 9en Strin s 9enfunny, 0eethoven1s 8inth, D%C 9eno of Elea, B*, B#-%B, #C-#L, "CC, "CD, B%B, D"+, D*", $+C, $BB 9eno1s !arado(es, B#-%B, %L, C%, "CD, D"+ 9entences, "*D-#+ 9ET-cycles, #C-#D, D*# <oomin in, DCL, D$"