Sei sulla pagina 1di 15

Business Report on “Machine Learning and AI Research”

Presented to
Prof. Sriparna Basu

Presented by
Shashwat Tiwari
133127
IMG13-E
Letter of Transmittal
Shashwat Tiwari
NVIDIA Corporation
Santa Clara, California, U.S.
Zip Code: 95050

26 November 2019

Prof. Sriparna Basu


Professor (Communication)
FORE School of Management
Adhitam Kendra,
B 18, NRPC Colony
Block B, Qutab Institutional Area, New Delhi (110016)

Ma’am,

Subject: Transmittal of NVIDIA’s Machine learning and AI research

The enclosed report is being submitted to take you through NVIDIA’s research on machine
learning and Artificial Intelligence (AI). The report describes how machine learning and AI,
especially deep learning, can be used to solve real-world problems and accelerate innovation.

If any additional information is needed, please let me know. You may reach me at
133127@fsm.ac.in . Thank you in advance for your attention.

Regards,
Shashwat Tiwari

2
Table of Contents
Executive Summary.................................................................................................................................. 4

Introduction .............................................................................................................................................. 5

Discussions of findings ............................................................................................................................. 7

Deep Learning....................................................................................................................................................... 7

AI Powered RTX technology .............................................................................................................................. 8

The State of RTX .................................................................................................................................................. 9

India’s First Deep Learning Workshop ............................................................................................................. 9

Conclusions............................................................................................................................................. 11

Recommendations .................................................................................................................................. 12

References ............................................................................................................................................... 13

Appendix ................................................................................................................................................. 14

3
Executive Summary
The impact of technology on shaping our values is well established. AI can take many forms and can be
applied in many different ways. It can be extremely advantageous as it can perform fast sharing and robust
data analysis. One of the major applications of AI is machine learning which provides system the ability
to automatically learn and improve from experience without being explicitly programmed. The report
highlights the work done by NVIDIA Corporation in the field of Machine learning and Artificial
intelligence. NVIDIA research shows how Artificial intelligence and machine learning, particularly deep
learning, can solve real-world problems and accelerate innovation. Billions of people around the world
are impacted by advances in this field of research. Recent publications include Image Inpainting for
Irregular Holes Using Partial Convolutions, High-Quality Self-Supervised Deep Image Denoising and
Video Stitching for Linear Camera Arrays.

The purpose of this report is to study the latest trends and development in the area of AI and machine
learning. The report further provides recommendations on how NVIDIA could enhance it R&D activities
on AI and Machine learning.

4
Introduction
From driverless cars to special purpose robots, it is conspicuous that AI is meant to enhance the quality of
human lives. The AI powered technologies have been transforming the landscape, providing
organizations with tremendous benefits and adding great value to their businesses. The sectors of
transportation, healthcare, logistics, finance and industrial manufacturing, among a plethora of
others, would undergo an immense transformation under the influence of AI and they would
evolve to be more productive and cheap and, more importantly, provide better services. (see
Fig1 in the appendix)

Machine Learning involves the use of algorithms to parse data and learn from it, and making a
determination or prediction as a result. Instead of hand coding software libraries with well- defined
specific instructions for a particular task, the machine gets “trained” using large amounts of data and
algorithms, and in turn gains the capability to perform specific tasks.

Deep Learning is a technique for implementing Machine Learning. Deep Learning was inspired by the
structure and function of the brain, specifically the interconnecting of many neurons. Artificial Neural
Networks (ANNs) are algorithms that are based on the biological structure of the brain. In ANNs, there
are ‘neurons’ which have discrete layers and connections to other “neurons”. Each layer picks out a
specific feature to learn. It’s this layering that gives deep learning its name, depth is created by using
multiple layers as opposed to a single layer

The artificial intelligence market was valued at USD 16.06 billion in 2017 and is expected to
reach USD 190.61 billion by 2025, at a CAGR of 36.62% during the forecast period. The base
year considered for the study is 2017, and the forecast period is between 2018 and 2025. (See
Figure 2 in the Appendix)
In India, there is a vast scope for Artificial Intelligence. India ranks among the most lucrative investment
destination for technology transactions worldwide. Various sectors of enterprises, agriculture, education,
healthcare and transportation can involve AI as transformative technology for the country.

As per some reports, within two years, AI is expected to boost the rate of innovation in India
by about 230%. Accenture in one of its reports predicts the AI technology to expedite India’s
annual growth rate by 1.3 percentage points by 2035.

5
NITI Aayog has adopted a three-pronged approach – undertaking exploratory proof-of-concept
AI projects in various areas, crafting a national strategy for building a vibrant AI ecosystem in
India and collaborating with various experts and stakeholders. Since the start of this year, NITI
Aayog has partnered with several leading AI technology players to implement AI projects in
critical areas such as agriculture and health.

This study focuses on NVIDIA’s GPUs which are used in artificial intelligence, deep learning,
and accelerated analytics. The study also analyses NVIDIA’s real-time ray tracing technique
and AI powered RTX technology.

6
Discussions of findings
Deep Learning

NVIDIA GPUs is being used in AI and deep learning. The company has developed a GPU-based deep
learning such that Artificial Intelligence can be used to approach problems like weather prediction, cancer
detection and self-driving cars. They are now also being incorporated in all Tesla vehicles. The purpose
of training the network to think is typically performed in large datacentres on GPUs, almost exclusively
provided by NVIDIA.
Deep learning is a subset of machine learning and Artificial Intelligence (AI). It uses multi-layered
artificial neural network to provide state-of-the-art-accuracy in tasks such as:
• Object detection
• Speech recognition
• Language translation
Deep learning doesn’t require human domain knowledge and unlike traditional machine learning process
they can automatically learn representations from data such as images, videos or text. It has helped in
recent breakthrough in AI such as Google DeepMind’s AlphaGo, self-driving cars, intelligent
voice assistants and many more.(See Figure 3 in the Appendix)
With NVIDIA GPU-accelerated deep learning structures, analysts and information researchers
can fundamentally accelerate deep learning training, that could somehow take days and weeks
to only hours and days. At the point when models are prepared for organization, engineers can
depend on GPU-accelerated deduction stages for the cloud, embedded device or self-driving
vehicles, to convey elite, low-dormancy surmising for the most computationally-escalated
profound neural systems.

Commonly used deep learning framework supported by NVIDIA:

• PyTorch
• MXNet
• TensorFlow
• Matlab
• NVIDIA Caffe
• Chainer
• Paddle Paddle

7
AI Powered RTX technology

The newly launched RTX 2060, RTX 2070 and RTX 2080 models has triggered excitement in the
gaming segment as it brings with it the promise of the holy grail of rendering; dubbed as real-time ray
tracing.
Ray tracing is defined as a procedure which generates an accurate representation of how light rays would
affect visual appearance of an object. The ray tracing technology has been commonly used since late
1970s and has been used quite frequently in Hollywood to render elements in most of their CGI scenes.
The procedure aims to bring realistic lighting to simulations by emulating the physical behaviour of light.
Ray Tracing is implemented by projecting rays of light from the camera, ‘tracing’ it until it interacts with
an object, and calculates where each ray intersects an object in the scene. This then allows it to compute
the way pixel is affected by the ray of light, and keeps tracing the ray as it bounces around the scene. This
requires the computer to follow some number of rays out from each intersection to model reflected light,
thus creating an exponentially increasing number of rays in the scene.
These procedures culminate in the calculation of the pixel color in the scene, resulting in a realistic look
due to light movement. These calculations are computational-intensive, resulting in the need for a
powerful ray tracing computer. Each ray-traced content frame will take hours on a machine with a mini-
supercomputer level, making it a technology that Hollywood uses to make CGI look realistic in cinemas.
Pixar's Monsters University has made extensive use of ray tracing, as the technology can reproduce many
of a camera's effects such as reflections, sub-surface scattering, and translucency.
Because of the high performance required to implement ray tracing in a scene, the implementation of real-
time ray tracing has long been considered a technology target. Video games are a particularly challenging
space for real-time ray tracing, since they generate between 30 and 60 frames in a second.
However, NVIDIA has achieved this and has provided a solution that can be implemented in consumer-
grade graphics cards with their Turing series of products, which includes the RTX cards.
The highest end card offered in the RTX line-up currently is the RTX 2080Ti, which can render 10
GigaRays/s. NVIDIA predicts that future solutions will also use similar hybrid renderers combining
rasterization and ray tracing, requiring close integration with existing APIs. This is achieved by another
bunch of corresponding applications and APIs that make it possible for developers and users to achieve
maximum acceptance.
NVIDIA has already worked with Microsoft closely in order to bring RTX to the widely used DirectX
API for creating games. They have partnered with Khronos Group to introduce this software to the soon-
to-be-coming Vulkan project. Reportedly, a “cross-vendor ray tracing capability” is possible, as NVIDIA
is contributing the design to the Khronos Group.

8
The State of RTX

RTX is supported in 10+ games currently and does not provide ray-tracing support for other games.
Nonetheless, due to the open API specifications and close integration that NVIDIA is bringing to the table,
the platform is set for potential adoption.
RTX triggers a noticeable output hit that is compensated by a DLSS technology that stands for deep
learning super sampling. To build this, NVIDIA trains a neural network at lower and higher resolutions
on pre-release game engine frames. The AI, via driver updates, provides weight data for Tensor Cores in
user GPUs.
The game is rendered at a lower resolution while DLSS is turned on, with Tensor cores working with
Deep Learning to upgrade to a higher resolution. This results in a higher frame rate at high resolution with
a slightly worse image. This can be used to provide better framerates in conjunction with ray tracing.
NVIDIA claims that users can achieve performance similar to ray tracing off with a combination of DLSS
and ray tracing on.
The introduction of this algorithm is required to have playable performance in games, as there is a need
to reduce noise in images delivered by ray tracing. This is because it is very difficult to cast an extremely
high number of rays, resulting in a lot of noise in the picture. The algorithm functions in the GPUs on the
Tensor cores, providing a quick and high-quality output without sacrificing performance. NVIDIA stated,
“These are early days for AI being applied to graphics and the above results are promising”, leading many
to believe that AI is the next step in improving graphics performance.

India’s First Deep Learning Workshop

Times Professional Learning in collaboration with NVIDIA Corporation have announced the Deep
Learning Workshop to be conducted for the first time in India. Instructor-led NVIDIA Deep Learning
Institute (DLI) Class will be a one day workshop which will help students and professionals acquainted
with Deep Learning and Artificial Intelligence (AI).
The workshop will be conducted for the first time in India in the following cities:
• Mumbai
• Hyderabad
• Pune
• New Delhi
• Chennai
• Bengaluru

9
This hands-on workshop will allow participants to understand how Deep Learning improves tasks of
perception and how to understand which framework is best suited to the needs of an organization. The
workshop will explore three approaches by using neural networks trained on NVIDIA DIGITS to identify
a specific feature within each image. Components include grading and segmentation of images using
DIGITS and deployment of neural networks using Tensor RTTM. It will also help understand the role of
batch size in inference performance; make various optimizations in the inference process and explore
inference for a variety of different DNN architectures trained in other DLI Labs.
At the end of the program, in addition to obtaining a global standard credential, students will also gain
practical experience with NVIDIA trainers, opportunities to work with the new GPU-accelerated DL
systems from NVIDIA to create neural networks, opportunities to meet and network with experts, access
to open-source frameworks and virtual laboratories, and a great way to improve careers.
Artificial Intelligence today powers businesses globally and plays an important role in every market field,
from improving healthcare to improving efficiency across all industries to developing smart cities to
revolutionizing data analytics.

10
Conclusions
NVIDIA is not just accelerating AI, it aims to reshape computing. Through its work in Deep Learning it
has been gaining massive competitive advantage. Organizations at every stage of growth—from startups
to Fortune 500s—are using deep learning and AI. Deep learning, AI's fastest growing field, empowers
tremendous progress in all kinds of emerging markets and will be instrumental in ways that we haven't
even imagined. Already, deep learning is enabling self-driving cars, smart personal assistants, and smarter
Web services. But the opportunities aren’t limited to a few business-specific areas. AI and deep learning
form industry-wide innovation. AI systems such as identification of fraud and optimization of the supply
chain are being used by the most innovative teams and organisations in the world.

AI will generate a revenue of $36.8 billion by 2020 as forecasted by Tractica research. At present, any
work the pushes the forward edge of technology — including automotive, analytics, or healthcare — has
at its core big data. The data, however, is just as valuable as the insights that we draw from it. Today, more
than 19,000 businesses use deep learning to develop their respective sectors and solve what was once
insolvent.

For image recognition, customer services such as Pinterest use deep learning to curate rich personalized
experiences for their users. Financial companies such as PayPal use GPU-accelerated deep learning to
detect fraud. Consulting firm Accenture’s R&D arm and other businesses are using deep learning to detect
Internet security threats. And healthcare is one of the most important industries for the application of AI
and deep learning. The sweeping change is taking place across all sectors in the data-driven market world.
The way farmers herd their animals is evolving and pushing us to reconsider transport. The first self-
driving public transit system is already in operation, giving us a glimpse of future-oriented cities that will
exist in our lives.

11
Recommendations

Based on this study following recommendations can be proposed for NVIDIA to further grow in the field
of Artificial Intelligence and Deep Learning:

1. Conduct workshops in India: India, with the world's second-largest population as the fastest growing
economy, has a major stake in the AI revolution. The hands-on workshop will allow participants to
understand how Deep Learning improves tasks of perception and how to understand which framework
is best suited to the needs of an organization.
2. R&D Centres: India has the building blocks needed to develop a vibrant ecosystem for AI research
and development. Highly educated talent pool, world-class educational institutions and an illustrious
list of top IT companies that dominate the global IT landscape. Given these advantages, India is
considerably lagging behind in producing world-class research and innovation in most fields of
technology, more so in AI. NVIDIA can exploit this opportunity and collaborate with Indian IT firms
and setup R&D centres in India.
3. AI for public good: Alongside educators, scientists, and humanitarians, NVIDIA can involve itself in
the global effort to democratize AI and deep learning for the common good. GPU-accelerated systems
assist in natural disasters; serve communities by improving safety, efficiency and traffic flows; and
provide critical health care assistance by helping to develop customized medicines. These
revolutionary solution can be made possible with NVIDIA technology.

12
References

• https://towardsdatascience.com/how-ai-will-affect-different-fields-of-technology-9977369f3e2b
• https://www.forbes.com/sites/jonmarkman/2016/11/22/deep-learning-cloud-power-nvidia-
future-growth/#4b130dcf39c8
• https://www.forbes.com/sites/moorinsights/2016/11/17/nvidia-is-not-just-accelerating-ai-it-
aims-to-reshape-computing/#3b4175bb3f4e
• https://en.wikipedia.org/wiki/Nvidia
• https://www.nvidia.com/en-us/research/machine-learning-artificial-intelligence/
• https://www.marketsandmarkets.com/Market-Reports/artificial-intelligence-market-
74851580.html
• https://www.analyticsinsight.net/artificial-intelligence-india-comprehensive-overview/
• https://www.accenture.com/_acnmedia/pdf-92/accenture-afs-responsible-ai.pdf
• https://niti.gov.in/writereaddata/files/document_publication/NationalStrategy-for-AI-
Discussion-Paper.pdf
• https://economictimes.indiatimes.com/industry/miscellaneous/tpl-and-nvidia-announce-indias-
first-deep-learning-workshop/articleshow/61527690.cms
• https://www.nvidia.com/en-us/deep-learning-ai/industries/

13
Appendix

Figure 1

Figure 2

14
Figure 3

15

Potrebbero piacerti anche