Sei sulla pagina 1di 7

ASSIGNMENT 2

(PHILOSOPHY 1010)

SUBMITTED BY: REBECCA GILLINGHAM


TEACHER: LETITIA MEYNELL
DATE: DECEMBER, 2006
In this paper I am arguing that there are artificial neural networks, which are

computers that think and that with future improvements, some things like biological

computers, will be able to think better – they will be able to learn more and rely less on

programming. I am defining thinking as the ability to learn from experience, and the

ability to respond to an input that has not been directly programmed.

First I will explain what a neural network is and what it is capable of. Then, I will

compare Locke’s concept of the human mind with that network. My main focus will be

on Searle’s argument and how it relates to neural networks and in a lesser way biological

computers. Lastly I will consider Midgley’s definition of persons, if it can fit computers

and how this relates to their ability to think.

The creation of Artificial Neural Networks (ANNs) was inspired by the brain and

other nervous systems and the way they processed information through neurons. An ANN

is made of neurons, which are many elements that process information and are

interconnected. ANNs can actually learn by adjusting the connections between their

neurons, similar to the way that living creatures learn. (Stergiou and Siganos, 1996) This

allows ANNs to learn by example. An ANN can also “create its own organization or

representation of the information it receives during learning time. (Stergiou and Siganos,

1996)” An ANN can be unpredictable, because it can learn. The neurons of the ANN have

many inputs for each output, and it will either fire or not. But, an ANN can respond to an

input it has not been programmed to recognize by using the firing rule (the neuron will

provide the output for the given input that is the most similar to a taught input.) This is

what gives the ANN its ability to learn. Another ability of ANNs is their capacity for

feedback networks, which are loops in the ANNs programming that allow an output to

change an input, which means that an ANN could correct itself, and that an ANN is
2
capable of change. An ANN can also be programmed to assign more or less importance to

certain conditions. (Stergiou and Siganos, 1996)

Locke in his essay says that humans form an idea about an object based on how it

affects our senses – the “sense data” it produces. Computers can be made to perceive with

certain senses, and can respond to stimuli received by these senses, and just as an absence

of stimuli produces a positive idea in a human’s head, an absence of stimuli could

produce an effect in a computer with the ability to distinguish it. Locke states that the

images in our mind are only representations (simulations) of a perceived object. This

would indicate that thought consists of simulating objects in an informational way and

this is possible for certain computers. Locke is concerned with what can be inferred by

our perceptions of sense data, and what the actual properties are. This can be related to

how well a computer can sense data and be said to understand it, however this was not his

focus. (Locke, 1823)

In Searle’s argument, he argues that a computer cannot be programmed to think,

because that suggests that the mind is completely separate from the physical (i.e. the

brain), but this does not mean a computer with the right structure could not think. He uses

the man in a house example to show that a computer doesn’t understand the input or the

output and is simply following rules. (Searle 1980) I agree that blindly following rules

does not constitute thinking, however ANNs do not blindly follow the rules, even without

understanding their given input they need to be able to distinguish its similarity to input

they have learned. This involves thinking because the ANN must decide whether a given

input is similar enough to a taught input for it to respond or not.

Searle argues that the understanding computer must be built with the “ability to

produce intentional states (Searle 1980)” meaning that to think it must not rely only on its
3
programming. I say that ANNs are able to change their own programming, but it might be

argued that those changes would still be based on the original programming. This does

not mean that they are not thinking, only that they don’t understand the outputs they

produce.

However, there are biological computers, which have both biological and

mechanical parts and because these computers can be made with neurons, they are able to

make connections on their own, without outside programming. (Tongen 2003) With

further knowledge these biological computers could even be made to correspond to a

mapped out human brain and perhaps even become conscious. That biological computers

could have mental states is supported by Searle’s statement that animals can be assumed

to have mental states because their brains are “made out of the stuff that is like our stuff.

(Searle 1980)”

Searle says that a human is able to know, because of the specific structure of the

brain, not just its ability to run a program. This would indicate that something built like

the nervous system (ANNs are based on the nervous system) or with the same biological

structure (biological computers are exactly that) should be able to perform the same

functions as the brain. I.e. that computer would be able to think.

In fact Searle states that a man made machine and in fact a computer could think,

and even understand or become conscious, if they were built along the same lines as the

human nervous system (or another system that could produce the same effects.) He is

only arguing that a computer’s understanding would not only rely on its program,

because anything can follow instructions without understanding. (Searle 1980)

Searle may argue that ANNs don’t process information, because they don’t

understand their outputs, and are only able to perform pattern detection because of
4
programmed information. However, the ANNs’ feedback loops allow it to detect an

incorrect output. This is similar to a human solving a math problem. The human may not

know why a formula works. However they can learn to solve it and will be able to detect

if an answer is wrong, by knowing the rules of that equation. This does require thought,

but does not necessarily imply a greater understanding of the answer or formula.

Another way to discuss whether computers think is to examine why someone would

assume they couldn’t. In Midgley’s discussion of what constitutes a person she makes a

point of “finding the right drama” to determine what a person could be. She says that

people are more likely to dismiss any attempts to create new ‘characters’, because they

believe they know who (or what) counts or doesn’t in their drama. (Midgley, 1996) This

can also be applied to concepts of artificial intelligence or computers that think. People

may say that computers can’t participate because they are not similar enough to human,

but there are non-human persons already. They may also say that computers cannot be

considered persons because they can’t feel pain, can’t fall in love, or other characteristics

that don’t involve thinking and would therefore be predisposed to dismiss the idea of a

thinking computer. But according to Midgley, a person is commonly defined as a rational

being, capable of choice. (Midgley, 1996) Computers could definitely be considered as

rational; since they are only logical, and it seems more advanced computers do make

choices.

The idea of personhood raises some important ethical considerations, but even

though computers can think, they are not even advanced enough to properly participate in

society. Even modern thinking machines such as ANNs do not have the capacity to be

illogical; they cannot, for example make decisions based on morality or politics, which

would make interacting with society difficult. Also, until computers are able to feel as
5
well as think, they would be unable to mind being “slaves”. This means there would be

no reason to treat them as persons or as more than thinking tools.

Midgley does make a good point concerning the matter of intelligence in saying that

“intelligence is a matter of degree (Midgley, 1996)” and that creatures with lives that are

different from those of humans would need different kinds of intelligence, which could

also apply to non-living ‘creatures’ such as computers. If computers advanced enough to

acquire intelligence on par with humans, but remained completely logical, it would be an

alien intelligence, because humans do not have that ability. Midgley also makes the point,

while arguing against Kant’s definition of a thing, that anything, including a human, can

be used for a purpose. This means that a thinking thing can be used, which means the fact

that computers have a purpose does not exclude an ability to think. (Midgley, 1996)

In conclusion, in the future computers will be able to think more independently and

make more connections, because there are already computers that can think, by virtue of

their particular structure. These computers have functions that are similar to the human

brain and may even gain consciousness, although that is not necessary for thinking.

6
BIBLIOGRAPHY

Citation Style Reference Documents

(http://ilopa.ucis.dal.ca:8900/SCRIPT/biol1020_dal/scripts/serve_home) J. Van

Dommelen, Comp., Dalhousie University, Spons. (Mod. 2005, December 20.

Viewed 06 Nov. 2006).

Locke, J. 1823. “An Essay Concerning Human Understanding - Book II, Chapter

VIII”. In: First Philosophy, edited by A. Bailey, pp. 183-188. Broadview Press;

Peterborough, Ontario.

Midgley, M. 1996. “Is a Dolphin a Person?”. In: First Philosophy, edited by A. Bailey,

pp. 882-888. Broadview Press; Peterborough, Ontario.

Neural Networks

(http://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol4/cs11/report.html) C.

Stergiou, D. Siganos, Auth. (Mod. 1996. Viewed 06 Dec. 2006).

Searle, J. R. 1980. “Minds, Brains and Programs”. In: First Philosophy, edited by A.

Bailey, pp. 438-452. Broadview Press; Peterborough, Ontario.

Will Biological Computers Enable Artificially Intelligent Machines to Become Persons?

(http://www.cbhd.org/resources/biotech/tongen_2003-11-07_print.htm) A. Tongen,

Auth. The Center for Bioethics and Human Dignity, Spons. (Mod. 2003, November

7. Viewed 06 Dec. 2006).

Potrebbero piacerti anche