MODULE 1 other theorems.

Professor John McCarthy

INTRODUCTION TO ACADEMIC WRITING coined the term ’Artificial Intelligence’ at
the Dartmouth conference, and it was
Introduction
accepted as an academic field.
Artificial intelligence (AI) is a wide- - Golden years – early enthusiasm
ranging branch of computer science concerned (1956–1974): After the invention of
with building smart machines capable of high-level languages such as LISP,
performing tasks that typically require human COBOL, and FORTRAN, researchers got
intelligence. It is the study of ideas that enable more excited about AI and developed
computers to do the things that make people algorithms to solve complex
seem intelligent. The central principles of AI mathematical problems. Joseph
include such as reasoning, knowledge, planning, Weizenbaum, a computer scientist,
learning, communication, perception, and the created the first chatbot named ‘ELIZA’
ability to move and manipulate objects. It is the in the year 1966. A year later, Frank
science and engineering of making intelligent Rosenblatt built a computer named
machines, especially intelligent computer ‘Mark 1 Perceptron.’ This computer was
programs. Many of these artificial intelligence based on the biological neural network
systems are powered by machine learning and (BNN) and learned through the method
deep learning of trial and error that was later coined as
reinforced learning. In 1972, Japan built
In this module, we will have the the first intelligent humanoid robot
opportunity to explore more about the following named ‘WABOT-1.’ Since then, robots are
topics: history of artificial intelligence, why is constantly being developed and trained
artificial intelligence important, how Artificial to perform complex tasks in various
Intelligence works, how Artificial Intelligence is industries.
being used, and advantages, limitations, and - A boom in AI (1980–1987): The first AI
applications of Artificial Intelligence winter (1974–1980) was over, and
governments started seeing the
Artificial Intelligence History
potential of how useful AI systems could
The term ‘Artificial intelligence’ was coined by be for the economy and defense forces.
John McCarthy in the year 1956 at Dartmouth Expert systems and software were
College at the first-ever AI conference. Later programmed to simulate the decision-
that year, JC Shaw, Herbert Simon, and Allen making ability of the human brain in
Newell created the first AI software program machines. Al algorithms like
named ‘Logic Theorist.’ Although, the idea of a backpropagation, which uses neural
‘machine that thinks’ dates back to the Mayan networks to understand a problem and
civilization. In the modern era, there have been find the best possible solution, were
some important events since the advent of used.
electronic computers that played a crucial role - The emergence of intelligent agents
in the evolution of Artificial Intelligence: (1993–2011): In the year 1997, IBM
developed a chess-playing computer
- Maturation of Artificial Intelligence named ‘Deep Blue’ that outperformed
(1943–1952): Walter Pitts and Warren the world chess champion, Garry
S. McCulloch, two mathematicians, Kasparov, in a chess match, twice. In
published ‘A Logical Calculus of the Ideas 2002, Artificial intelligence for the first
Immanent in Nervous Activity’ in the time stepped into the domestics and
Journal of Mathematical Biophysics. They built a vacuum cleaner named ’Roomba.’
described the behavior of human By the year 2006, MNCs such as
neurons with the help of simple logical Facebook, Google, and Microsoft started
functions that inspired an English using AI algorithms and Data Analytics to
mathematician Alan Turing to publish understand customer behavior and
‘Computing Machinery and Intelligence’ improve their recommendation systems.
that comprised a test. This Turing Test - Deep Learning, Big Data, and
is used to check a machine’s ability to Artificial General Intelligence (2011–
exhibit intelligent behavior. Present): With computing systems
- The birth of Artificial Intelligence becoming more and more powerful, it is
(1952–1956): Logic Theorist, the first AI now possible to process large amounts of
program was created in the year 1955 data and train our machines to make
by Allen Newell and Herbert A Simon. It better decisions. Supercomputers take
proved around 52 mathematical the advantage of AI algorithms and
theorems and improved the proofs for neural networks to solve some of the

1|Computer Science
 most complex problems of the modern data. You just have to apply AI to find
world. Recently, Neuralink, a company them. Since the role of the data is now
owned by Elon Musk, successfully more important than ever, it can create
demonstrated a brainmachine interface a competitive advantage. If you have the
where a monkey played the ping pong best data in a competitive industry, even
ball video game from his mind if everyone is applying similar
techniques, the best data will win.
Importance
How It Works
- AI automates repetitive learning
and discovery through data. Instead AI works by combining large amounts of
of automating manual tasks, AI performs data with fast, iterative processing and
frequent, high-volume, computerized intelligent algorithms, allowing the software to
tasks. And it does so reliably and without learn automatically from patterns or features in
fatigue. Of course, humans are still the data. AI is a broad field of study that
essential to set up the system and ask includes many theories, methods and
the right questions. technologies, as well as the following major
- AI adds intelligence to existing subfields:
products. Many products you already use
will be improved with AI capabilities,
much like Siri was added as a feature to
a new generation of Apple products.
Automation, conversational platforms,
bots and smart machines can be
combined with large amounts of data to
improve many technologies. Upgrades at - Machine learning automates analytical
home and in the workplace, range from model building. It uses methods from
security intelligence and smart cams to neural networks, statistics, operations
investment analysis. research and physics to find hidden
- AI adapts through progressive insights in data without explicitly being
learning algorithms to let the data do programmed for where to look or what to
the programming. AI finds structure and conclude.
regularities in data so that algorithms - A neural network is a type of machine
can acquire skills. Just as an algorithm learning that is made up of
can teach itself to play chess, it can interconnected units (like neurons) that
teach itself what product to recommend processes information by responding to
next online. And the models adapt when external inputs, relaying information
given new data. between each unit. The process requires
- AI analyzes more and deeper data multiple passes at the data to find
using neural networks that have many connections and derive meaning from
hidden layers. Building a fraud detection undefined data.
system with five hidden layers used to - Deep learning uses huge neural
be impossible. All that has changed with networks with many layers of processing
incredible computer power and big data. units, taking advantage of advances in
You need lots of data to train deep computing power and improved training
learning models because they learn techniques to learn complex patterns in
directly from the data. large amounts of data. Common
- AI achieves incredible accuracy applications include image and speech
through deep neural networks. For recognition.
example, your interactions with Alexa - Cognitive Computing. The ultimate
and Google are all based on deep goal of cognitive computing is to imitate
learning. And these products keep the human thought process in a
getting more accurate the more you use computer model. How can this be
them. In the medical field, AI techniques achieved? Using self-learning algorithms,
from deep learning and object pattern recognition by neural networks,
recognition can now be used to pinpoint and natural language processing, a
cancer on medical images with improved computer can mimic the human way of
accuracy. thinking. Here, computerized models are
- AI gets the most out of data. When deployed to simulate the human
algorithms are self-learning, the data cognition process.
itself is an asset. The answers are in the - Computer vision relies on pattern
recognition and deep learning to

2|Computer Science
 recognize what’s in a picture or video. such as coal mines, exploring the
When machines can process, analyze deepest parts of the ocean, sewage
and understand images, they can treatment, and nuclear power plants to
capture images or videos in real time avoid any disaster.
and interpret their surroundings. - Replacing repetitive jobs: Our day-to-
- Natural language processing (NLP) is day work includes many repetitive tasks
the ability of computers to analyze, that we have to do every day without
understand and generate human any change. For example, washing your
language, including speech. The next clothes or mopping the floor doesn’t
stage of NLP is natural language require you to be creative and find new
interaction, which allows humans to easy to do it every day. Even big
communicate with computers using industries have production lines where
normal, everyday language to perform the same number of tasks has to be
tasks. done in an exact sequence. Now,
machines have replaced these tasks so
Additionally, several technologies enable and that humans can spend this time doing
support AI: creative things.
- Graphical processing units are key to - Digital assistance: With digital
AI because they provide the heavy assistants to interact with users 24/7,
compute power that’s required for organizations can save the need for
iterative processing. Training neural human resources and deliver faster
networks requires big data plus compute service to customers. It is a win-win
power. situation for both the organization and
- The Internet of Things generates the customers. In most cases, it is really
massive amounts of data from hard to determine whether a customer is
connected devices, most of it chatting with the chatbot or a human
unanalyzed. Automating models with AI being.
will allow us to use more of it.
Limitations
- Advanced algorithms are being
developed and combined in new ways to - High cost of creation: It may sound a
analyze more data faster and at multiple little spooky, but the rate at which
levels. This intelligent processing is key computational devices are upgraded is
to identifying and predicting rare events, phenomenal. Machines need to be
understanding complex systems and repaired and maintained with time to
optimizing unique scenarios. keep the latest requirements in check,
- APIs, or application programming which needs a lot of resources.
interfaces, are portable packages of - No emotions: There is no doubt that
code that make it possible to add AI machines are much powerful and faster
functionality to existing products and than human beings. They can perform
software packages. They can add image multiple tasks simultaneously and
recognition capabilities to home security produce results in a split second. AI-
systems and Q&A capabilities that powered robots can also lift more weight,
describe data, create captions and thereby increasing the production cycle.
headlines, or call out interesting patterns However, machines cannot build an
and insights in data. emotional
- Box thinking: Machines can perfectly
Advantage execute the preassigned tasks or
- Reduced human error: With humans operations with a definite range of
involved in the tasks where precision is constraints. However, they start
required, there will always be a chance producing ambiguous results if they get
of error. However, if programmed anything out of the trend.
properly, machines do not make - Can’t think for Itself: Artificial
mistakes and easily perform repetitive Intelligence aims to process data and
tasks without making many errors, if not make a conscious decision as we
at all. humans do. But, at present, it can only
- Risk avoidance: Replacing humans do the tasks it is programmed for. AI
with intelligent robots is one of the systems cannot make decisions based
biggest advantages of Artificial on emotions, compassion, and empathy.
Intelligence. AI robots are now doing For example, if a self-driving car is not
risky things replacing humans in places programmed to consider animals like

3|Computer Science
 deer as living organism, it will not stop Artificial Intelligence enhances the
even if it hits a deer and knock it off. speed, precision and effectiveness of
human efforts. In financial institutions, AI
Application techniques can be used to identify which
transactions are likely to be fraudulent,
adopt fast and accurate credit scoring,
as well as automate manually intense
data management tasks.

Public sector

Artificial Intelligence can make smart

cities smarter. It can support national
defense with mission readiness and
predictive maintenance. Across the
board, AI can improve program efficiency
and effectiveness.

Game playing.

You can buy machines that can play

master level chess for a few hundred
dollars. There is some AI in them, but
Every industry has a high demand for AI they play well against people mainly
capabilities – including systems that can be through brute force computation—
used for automation, learning, legal assistance, looking at hundreds of thousands of
risk notification and research. Specific uses of AI positions. To beat a world champion by
in industry include: brute force and known reliable heuristics
requires being able to look at 200 million
Health Care positions per second.
AI applications can provide personalized Speech recognition.
medicine and X-ray readings. Personal
health care assistants can act as life In the 1990s, computer speech
coaches, reminding you to take your recognition reached a practical level for
pills, exercise or eat healthier. limited purposes. Thus, United Airlines
has replaced its keyboard tree for flight
Retail information by a system using speech
AI provides virtual shopping capabilities recognition of flight numbers and city
that offer personalized recommendations names. It is quite convenient. On the the
and discuss purchase options with the other hand, while it is possible to instruct
consumer. Stock management and site some computers using speech, most
layout technologies will also be improved users have gone back to the keyboard
with AI. and the mouse as still more convenient.

Manufacturing Understanding natural language.

AI can analyze factory IoT data as it Just getting a sequence of words into a
streams from connected equipment to computer is not enough. Parsing
forecast expected load and demand sentences is not enough either. The
using recurrent networks, a specific type computer has to be provided with an
of deep learning network used with understanding of the domain the text is
sequence data. about, and this is presently possible only
for very limited domains.

Computer Vision.
Life sciences
The world is composed of three-
From ensuring drug safety to getting dimensional objects, but the inputs to
new therapies to market faster, AI the human eye and computers’ TV
technologies can unleash the full cameras are two dimensional. Some
potential of data to solve some of our useful programs can work solely in two
greatest health challenges. dimensions, but full computer vision
requires partial three-dimensional
Banking
information that is not just a set of two-

4|Computer Science
 dimensional views. At present there are
only limited ways of representing three-
dimensional information directly, and
they are not as good as what humans
evidently use.

Expert Systems.

A “knowledge engineer” interviews

experts in a certain domain and tries to
embody their knowledge in a computer
program for carrying out some task. How
well this works depends on whether the
intellectual mechanisms required for the
task are within the present state of AI.
When this turned out not to be so, there
were many disappointing results. One of
the first expert systems was MYCIN in
1974, which diagnosed bacterial
infections of the blood and suggested
treatments. It did better than medical
students or practicing doctors, provided
its limitations were observed. Namely, its
ontology included bacteria, symptoms,
and treatments and did not include
patients, doctors, hospitals, death,
recovery, and events occurring in time.
Its interactions depended on a single
patient being considered. Since the
experts consulted by the knowledge
engineers knew about patients, doctors,
death, recovery, etc., it is clear that the
knowledge engineers forced what the
experts told them into a predetermined
framework. In the present state of AI,
this has to be true. The usefulness of
current expert systems depends on their
users having common sense.

Heuristic Classification.

One of the most feasible kinds of expert

system given the present knowledge of
AI is to put some information in one of a
fixed set of categories using several
sources of information. An example is
advising whether to accept a proposed
credit card purchase. Information is
available about the owner of the credit
card, his record of payment and also
about the item he is buying and about
the establishment from which he is
buying it.

5|Computer Science