Artificial Intelligence

2020-2021

Prof. Punam Bedi

Department of Computer Science
University of Delhi
http://people.du.ac.in/~pbedi/
 Outline
 Readings
 Introduction
 Artificial Intelligence
 Weak AI vs Strong AI
 Narrow AI vs. General AI
 Superintelligence
 Characteristics of AI systems
 What is AI Today
 PhD research in AI by my group at University of Delhi
 Readings
1. Russell S, Norvig P , Artificial Intelligence : A Modern

Approach, Pearson Education.

2. Elaine Rich and Kelvin Knight : Artificial Intelligence, Tata McGraw
Hill.
3. Dan W. Patterson : Introduction to Artificial Intelligence and Expert
Systems, Prentice Hall of India.
4. Kaushik Saroj, Logic and Prolog Programming, New Age International
Publishers.
5. Konar Amit : Artificial Intelligence and Soft Computing, Behavioral and
Cognitive Modeling of the Human Brain, CRC Press.
Index
Introduction
 In recent years, artificial intelligence, or AI, has gained a
surge in attention from policy makers, universities,
researchers, corporations, media, and the public.
 Driven by advances in big data and computing power,
breakthroughs in AI research and technology seem to
happen almost daily.
 Expectations, but also fears, are mounting about the
transformational power of AI to change society.
“AI is the new electricity. Just as electricity
transformed almost everything 100 years ago,
today I have a hard time thinking of an industry
that I don’t think AI will transform.”

– Andrew Ng,
AI Expert.

Don’t get left in the dark.

Benefits of AI

 Consumers use AI on a daily basis to find their

destinations using navigation and ride-sharing apps,
as smart home devices or personal assistants, or for
streaming services.
 Businesses can use AI to assess risk and define
opportunity, cut costs, and boost research and
innovation. Cost and effort estimation, Risk
estimation of projects can utilize AI as it is done
based on historical data.
Industries Currently Using AI

 The self-driving car is probably the best-known use of

AI.
 Predictive maintenance is another part of AI,
forecasting when maintenance will be needed so that
it can be done proactively, leading to tremendous
cost savings.
 AI is used in transportation such as for train
scheduling and to help Uber drivers navigate routes.
 Smart cities use AI to be more energy efficient,
reduce crime and improve safety.
Intelligence
 The capability to form concepts and grasp their
significance.
 To judge well, to comprehend well, to reason well
 Reflected in
 mental and behavioral activities
 situation specific (acquisition of skills implies behaving in
a certain way)
e.g. Car Driving
Intelligence(contd.)

 Adaptation to the physical and social environment.

 essential abilities to
 respond to situations flexibly
 make sense out of ambiguous or contradictory messages.
 Find similarity in situations separated by differences.
 To generate new concepts and new ideas.

Index
Artificial Intelligence
 AI is the study of how to make computers do things,
which at the moment, people do better (Elaine Rich).
(This definition does not include problems that can
not be solved by either computers or people).

 AI is the science of making machines do things that

would require intelligence if done by men (Marvin
Minsky).
 AI can be defined as a subject dealing with
computational models that can think and act rationally,
i.e., plan and execute the right task at the right time.
(Amit Konar)
 Artificial Intelligence
 AI is the branch of computer science concerned
with the study and creation of computer systems
that exhibit some form of intelligence: systems that
can learn new concepts and tasks, systems that
can draw useful conclusions about the world
around us, systems that can understand a natural
language or perceive and comprehend a visual
scene and systems that can perform other types of
feats that require human type of
intelligence(Patterson).
• Humans try to understand about how to perceive,
understand, predict and manipulate
AI tries not just to understand but also to build
intelligent entities.
• AI currently encompasses a huge variety of
subfields in
• general-purpose areas such as learning and
perception
• specific tasks such as playing chess, proving
mathematical theorems, writing poetry and
diagnosing diseases.
01/22/2025 Artificial Intelligence 12
• AI systematizes and automates intellectual tasks and
is therefore relevant to any sphere of human intellectual
activity. In this sense, it is truly a universal field.
• AI is one of the newest sciences. Work started soon after
world war II and name itself was coined in 1956.
Along with molecular biology, AI is regularly cited as
the “ field I would most like to be in ” by scientists in
other disciplines.

01/22/2025 Artificial Intelligence 13

Definitions of AI, organized into four
categories
 Systems that think like humans
 Systems that act like humans

 Systems that think rationally

 Systems that act rationally
Systems that act like humans
(Turing Test approach)
• The act of creating machines that perform functions that
require intelligence when performed by people.( Kurzweil
1990)
• Study of how to make computers do things at which, at the
moment people are better(Elaine Rich).

01/22/2025 Artificial Intelligence

Turing Test

Named after Alan Turing who in 1950 proposed the following

method for determining whether the machine can think.

Interrogator
KB1 A

Typed
questions
answers Programmed to respond
C like a person
KB2 B
(multiplication, reversing
a sequence of letters etc.)
If the interrogator can not distinguish between a computer and a person, the computer
passes the Turing test for that problem.
CAPTCHA
Completely Automated Public Turing test to tell Computers and
Humans Apart

01/22/202 Artificial Intelligence 17

5
Computer would need to provide the following capabilities
(Various Disciplines of AI)
• NLP to enable it to communicate successfully in English
•KR to store about what it knows or hears
•Automated Reasoning to use the stored information to answer questions
and to draw new conclusions.
•Machine Learning to adapt to new circumstances and to detect and
extrapolate patterns
•Computer Vision to perceive objects
•Robotics to manipulate objects and move about.

01/22/2025 Artificial Intelligence 18

Systems that think like humans (The Cognitive
Modeling approach)
• “ The exciting new efforts to make computers
think… machines with minds, in full and literal
sense.” (Haugeland,1985)
•“[The automation of] activities that we associate
with human being, activities such as decision-
making, problem solving, learning…(Bellman,1978)

01/22/2025 Artificial Intelligence 19

 How humans think
One can understand actual working of Human mind
• through introspection - try to catch your own thoughts as they go by
•through psychological experiments - observing the person in action
•Through brain imaging – observing the brain in action
Once we have a precise theory of mind, it becomes possible to express the theory
as a computer program
The interdisciplinary field of cognitive science brings together computer
models from AI and experimental techniques from psychology to try to construct
precise and testable theories of the working of human mind.

01/22/2025 Artificial Intelligence 20

 Thinking rationally (The Law of Thought
approach)
• “The study of mental faculties through use of computational
models.” (Charniak and McDermott,1985)
•The study of the computations that make it possible to
perceive, reason and act.” (Winston, 1992)

These law of thought approach were supposed to govern the

operation of mind; their study initiated the field called logic.
01/22/2025
Artificial Intelligence
21
 Acting rationally (The Rational Agent
approach)
• “Computational Intelligence is the study of intelligent agents.”
(Poole et al, 1998)
•AI is concerned with intelligent behavior in artifacts.”(Nilson
1998.)

An agent is just something that acts (agent comes from the Latin
word agere, to do).
01/22/2025 Artificial Intelligence 22
Attributes that distinguish agents from programs
• operating under autonomous control
• perceiving their environment
• persisting over a prolonged time period
• adapting to change
• being capable of taking another’s goal
A rational agent is one that acts so as to achieve the best
outcome or, when there is uncertainty, the best expected
outcome.

01/22/2025 Artificial Intelligence 23

•In law of thought approach to AI, emphasis is on correct
inferences. Making correct inference is sometimes a part of being a
rational agent because one way to act rationally is to reason
logically to the conclusion that a given action will achieve one’s
goals and then to act on that conclusion.
On the other hand, correct inference is not all of rationality,
because there are often situations where there is no provably
correct thing to do, yet something must be done.
•All the skills needed for the Turing test are there to allow rational
actions. Thus we need the ability to represent knowledge and
reason with it because this enables us to reach decisions in a wide
variety of situations.
01/22/2025 Artificial Intelligence 24
Advantages of study of AI as rational agent approach
•It is more general than law of thought approach as correct inference
is just one of the several possible ways to achieve rationality.
•It is more amenable to scientific developments than the approaches
based on human behavior or human thought because the standard of
rationality is clearly defined and completely general. Human behavior
on the other hand is well adapted for one scientific environment.
Hence We will follow the rational agent approach to AI.
•Perfect rationality- always doing the right thing
•Limited rationality - acting appropriately when there is not enough
time to do all the computations one might like

01/22/2025 Artificial Intelligence 25

Foundations - Mathematics

 More formal logical methods

 Boolean logic (Boole, 1847)
 Analysis of limits to what can be computed
 Intractability (1965) – time required to solve problem scales exponentially
with the size of problem instance
 NP-complete (1971) – Formal classification of problems as intractable
 Uncertainty
 The basis for most modern approaches to AI
 Uncertainty can still be used in logical analyses
 Foundations - Neuroscience

 How do brains work?

 Earlystudies (1824) relied on injured and abnormal people to understand
what parts of brain work
 Morerecent studies use accurate sensors to correlate brain activity to
human thought
 By monitoring individual neurons, monkeys can now control a computer mouse
using thought alone
 Moore’s law states computers will have as many gates as humans have
neurons in 2020
 How close are we to having a mechanical brain?
 Parallel computation, remapping, interconnections, binary vs. gradient
Foundations – Control Theory
 Machinescan modify their behavior in response to the
environment (sense / action loop)
Water-flow regulator (250 B.C.E), steam engine governor,
thermostat
 The theory of stable feedback systems (1894)
Build systems that transition from initial
state to goal state with minimum energy
In 1950, control theory could only describe
linear systems and AI largely rose as a
response to this shortcoming
 Foundations - Linguistics
Speech demonstrates so much of human
intelligence
Analysis of human language reveals thought
taking place in ways not understood in other
settings
Children can create sentences they have never
heard before
Language and thought are believed to be tightly
intertwined
 The birth of AI (1943 – 1956)
 Warren McCulloch & Walter Pitts (1943): ANN with
on-off neurons
 Neurons triggered by sufficient #neighbors
 Showed that any computable function computable with some
network like this
 Logical connectives implementable this way
 Donald Hebb’s 1949 learning rule
 Turing & Shannon chess programs, 1950s
 SNARC, first ANN computer, Minsky & Edmonds, 1951

01/22/202 Artificial Intelligence 30

5
 The birth of AI (1943 – 1956)
 John McCarthy, who is the Father of Artificial
Intelligence coined the term Artificial
Intelligence in 1955 and he proposed this term
in the famous Dartmouth conference in 1956.
 Alan Newell & Herbert Simon created the
“first artificial intelligence program” Which
was named as "Logic Theorist". This program
had proved 38 of 52 Mathematics theorems, and
find new and more elegant proofs for some
theorems.
01/22/202 Artificial Intelligence 31
5
 Early enthusiasm (1952 – 1969)
 1956 Dartmouth conference
John McCarthy (Lisp);
Marvin Minsky (first neural network machine);
Alan Newell and Herbert Simon (GPS);
 Emphasis on intelligent general problem solving
Goal Stack Planning (means-ends analysis);
Lisp (AI programming language);
Resolution by John Robinson (basis for
automatic theorem proving);
heuristic search (A*, AO*, game tree search)
01/22/202 Artificial Intelligence 32
5
 Emphasis on knowledge (1966 – 1974)

 domain specific knowledge is the key to

overcome existing difficulties
 knowledge representation (KR) paradigms
 declarative vs. procedural representation

01/22/202 Artificial Intelligence 33

5
 Knowledge-based systems (1969 – 1979)

 DENDRAL: molecule structure identification [Feigenbaum et al.]

 Knowledge intensive
 MYCIN: medical diagnosis [Feigenbaum, Buchanan, Shortliffe]
 450 rules; knowledge from experts; no domain theory
 Better than junior doctors
 Certainty factors
 PROSPECTOR: drilling site choice [Duda et al]
 Domain knowledge in NLP
 Knowledge representation: logic, frames...
01/22/202 Artificial Intelligence 34
5
 AI became an industry (1980 –
1988)
Wide applications in various domains and
commercially available tools
 R1: first successful commercial expert system, configured
computer systems at DEC; saved 40M$/year
 1988: DEC had 40 expert systems, DuPont 100...
 1981: Japan’s 5th generation project
 Software tools for expert systems: Carnegie Group,
Inference, Intellicorp, Teknowledge
 LISP-specific hardware: LISP Machines Inc, TI, Symbolics,
Xerox
01/22/202 Artificial Intelligence 35
5
Return of ANNs (1986-)

 Mid-1980s, different research groups reinvented

backpropagation (originally from 1969)
 Disillusionment on expert systems
 Fear of AI winter ( AI winter is a period of reduced
funding and interest in artificial intelligence research )

01/22/202 Artificial Intelligence 36

5
 Current trends (1990 – present)
 Real-world applications rather than toy domains
 distributed AI and intelligent software agents
 resurgence of natural computation - neural networks
and emergence of genetic algorithms – many
applications
 dominance of machine learning and deep learning

01/22/202 Artificial Intelligence 37

5
 Two Views of AI Goal
 AI
is about duplicating what the (human) brain
DOES
 Cognitive Science

 AI
is about duplicating what the (human) brain
SHOULD do
 Rationality (doing things logically)
Weak AI vs. Strong AI

 Weak AI - describes "simulated" thinking., i.e., a system

which appears to behave intelligently, but doesn't have
any kind of consciousness about what it's doing and
why. For example, a chatbot.
 Strong AI - describes "actual" thinking., i.e., thinking
and acting intelligently as a human with a conscious,
subjective mind.
Index
Narrow AI vs. General AI

 Narrow AI - AI that is limited to a single task or

a set number of tasks.

 General AI - AI which can be used to complete a

wide range of tasks in a wide range of
environments.
Index
Superintelligence

 Superintelligence - general and strong AI at the

point at which it surpasses human intelligence, if
it ever does.

Index
Cool Stuff in AI
 Game playing agents
 Machine learning
 Speech
 Language
 Vision
 Data Mining
 Web agents …….
Useful Stuff

Medical Diagnosis
Fraud Detection
Object Identification
Space Shuttle Scheduling
Information Retrieval ….
AI Techniques

 Rule-based
 Fuzzy Logic
 Neural Networks
 Genetic Algorithms
 Characteristics of AI systems
• learn new concepts and tasks,
• reason and draw useful conclusions about the world
around us,
• understand a natural language or perceive and
comprehend a visual scene,
• plan sequences of actions to complete a goal,
• offer advice based on rules and situations,
• remember complicated interrelated facts, and draw conclusions
from them (inference),
• look through cameras and see what's there (vision), to move
themselves and objects around in the real world (robotics),
• may not necessarily imitate human senses and thought processes
but indeed, in performing some tasks differently, they may
actually exceed human abilities,
• capable of performing intelligent tasks effectively and efficiently,
• performing tasks that require high levels of intelligence,
 UNDERSTANDING OF AI
 AI techniques and ideas seem to be harder to understand than
most things in computer science.
 Artificialintelligence shows best on complex problems for which
general principles don't help much, though there are a few useful
general principles.
 Itoften means, nonnumeric ways of solving problems, since
people can't handle numbers well.
 Nonnumeric ways are generally "common sense" ways, not
necessarily the best ones.
 Artificial intelligence is also difficult to understand by its
content.
 Boundaries of AI are not well defined.
 Often it means the advanced software engineering, sophisticated
software techniques for hard problems that can't be solved in
any easy way.
 AI programs - like people - are usually not perfect, and even
make mistakes.
 Understanding of AI also requires an understanding of related terms
such as intelligence, knowledge, reasoning, thought, cognition,
learning, and a number of other computer related terms.
Components of AI Program

AI techniques must be independent of

the problem domain as far as possible.
AI program should have
 knowledge base,
 navigational capability which contains
control strategy,
 inferencing.
 A.I. Program

K.B.

Control
Result
Query (Gate Keeper)

(Inference
Engine)
Control (Gate Keeper) explores the knowledge and draws the inference.
 Knowledge Base
 AI
programs should be learning in nature and
update its knowledge accordingly.
 Knowledge base consists of facts and rules.
 Characteristics of Knowledge:
 It is voluminous in nature and requires proper
structuring.
 It may be incomplete and imprecise.
 It may keep on changing (dynamic).
 new facts might emerge
* soil on Mars means evidence of life
* absence of ozone layer means no evidence of life
However, an AI technique which is a method that
exploits knowledge and attempts

•to capture generalization as far as possible

- helps to conserve storage
- enhances clarity
- one may need exception handling

•to have flexibility of representation

- for modification
- for extensibility of approach
Control strategy
 determines the rule to be applied
 some heuristics (thumb rule) may be
applied
Inferencing
 requires search through knowledge
base and derive new knowledge
 Sub-areas
 Natural language understanding,
 Computer vision,
 Understanding spoken utterances,
 Intelligent tutoring systems,
 Robotics,
 Machine translation systems,
 Expert problem solving,
 Neural networks
 Deep learning,
 AI tools etc
Applications
 Business : Financial strategies, give advice
 Engineering: check design, offer suggestions to create
new page
 Manufacturing: Assembly, inspection & maintenance
 Mining: used when conditions are dangerous
 Hospital : monitoring, diagnosing & prescribing
 Education : In teaching
 Household : Advice on cooking, shopping etc.
 Farming : prune trees & selectively harvest mixed
Index
crops.
What is AI today

Index

01/22/202 Artificial Intelligence 56

5
Recent AI

Recent AI is about designing good models and

apllications
using

 probability theory
 optimization
 decision theory
 statistics
 logic (fuzzy, modal, temporal)

to develop

 efficient algorithms using (lots of complex) data

 PhD research in AI by my group
 Harmeet Kaur, Trust based decentralized recommender systems (2007).
 Sudeep Marwaha, Temporal extension to ontologies for semantic web
enabled systems (2008).
 Suruchi Chawla, Information Retrieval using Information Scent and
Agents (2011).
 Ravish Sharma, Recommending Solutions for Complex Systems based on
Ant Colony Metaphor (2012).
 Anjali Thukral, Retrieving and Organizing Web Resources Semantically
for Informal e-Mentoring (2013).
 Roli Bansal, Computationally Intelligent Watermarking for Securing
Fingerprint Images (2013).
PhD research in AI by my group
 Richa Sharma, Adaptive Content Sequencing Incorporating Social Opinion in
an e-Learning Environment (2013).
 Aakanksha Vats, Mobile Process Groups based Trustworthy Coordination in
Ad-hoc Network (2014).
 Bhavna Gupta, Trust-based Multiagent Service Recommender System (2014).
 Vashisth, Pooja. Trust and Argumentation based Recommender Systems
(2014).
 Sumit Agrawal, Context-Aware Trust Based Mobile Recommender System
(2014).
 Anjali Gautam, Matrix Factorization Based Recommender System for Large-
Scale Data (2018).
PhD research in AI by my group

 Vinita Jindal, Traffic Signals and Route Optimization for Congestion

Control in Vehicular Ad-hoc Networks (2018).
 Veenu. Bhasin, Feature Selection Based Multi-Class Image
Steganalysis Using Soft Computing Techniques (2018).
 Richa, Cross Domain Context-aware Recommender System using
Trust and Distrust (2019).
 Chhavi Sharma, Community Detection Based Recommender Systems
(2019).
Index