ARTIFICIAL INTELLIGENCE (AI)

A FORMAL INTRODUCTION
What does
“artificial” intelligence
mean?
 Programming a computer to successfully
perform tasks that are thought to require
intelligence
 Playing chess
 Proving theorems
 Translating Russian into English
 Walking across a room
 Recognizing a familiar face
 Understanding directions

2
 DEFINITIONS OF AI
 Study of how to make computers do things at
which, at the moment, people are better.
(Elain Rich) (Memorizing and intelligence)

 The science of making machines do

things that would require intelligence if
done by human. (Minsky)

 Based on programming techniques: A

branch of computer science dealing
with symbolic, non-algorithmic methods
of problem solving. (Buchanan and
Shortliffe) 3
 Some more definitions

Part of computer science that concerns with

designing intelligent computer systems, that is,
systems that exhibit the characteristics we associate
with intelligence in human behavior. (Barr and
Feigenbaum)

The branch of computer science that deals with ways

of representing knowledge using symbols rather than
numbers and with rules-of-thumb or heuristic
methods for processing information. (Buchanan)

4
 Some more definitions

 A way of making a machine think

intelligently.
 Branch of computer science that is

concerned with automation of intelligent

 A recursive definition: AI is the
behavior.
collection of problems and
methodologies studied by AI
researchers.
 HISTORY: 1930: John Dewey, ‘How
we think’ to 1990s Genetic
Algorithms and the 6th Generation
Computing
5
HISTORY OF AI
Main Events
The Dartmouth
Conference
 According to the organizers’
proposal, the conference was
intended to explore the conjecture,
”that every aspect of learning or
any other feature of intelligence can
in principle be so precisely
described that a machine can be
made to simulate it”. That
conjecture, of course, continues to
be a focus of AI research.
7
 The Conference
Organizers
 John McCarthy - Assistant Professor of
Mathematics at Dartmouth.
 Marvin Minsky - Junior Fellow in mathematics
and neurology at Harvard, both had worked
with Shannon at Bell Labs.
 Nathaniel Rochester - Manager of information
research for IBM. Interested in intelligent
machines.
 Claude Shannon - Bell Labs; Established his
reputation firmly earlier.
 The conference, funded by a $7500
Rockefeller Foundation Grant, was
organized by four scientists, 2 from
academia and 2 from industry.
8
The Dartmouth
Conference
 1956 Dartmouth Conference, Hanover,
New Hampshire:
AI revolution was launched. About a dozen
scientists representing disciplines of
 Mathematics
 Neurology
 Psychology
 Electrical Engineering among others
were there.

9
 The Dartmouth
Conference
 They all in their various fields, were using
computers to try to simulate various aspects
of human intelligence.
 A new branch of computer science
crystallized at the conference, combining
elements of several different avenues of
research into a unified field.
 There was no universal agreement about
what to call the new science.
 However, Artificial Intelligence, the name
suggested by John McCarthy, one of the
conference organizers, has come to be
associated firmly with the field.
10
Can Machines Think?
 Computers were used by Americans
and the British during World War II
to expedite complex tasks such as
numerical computations and code
breaking activities that previously
had been assumed to require
human intelligence.

11
Can Machines Think?
 Notice that shift in the frontier of
“intelligence” in the last 50 years.
We have become so accustomed to
calculating machines that we now
consider that kind of activity to be
“mechanical”.
 It was probably inevitable that
scientists working with the first
computers would speculate about
how intelligent these new electronic
marvels could become.
12
Can Machines Think?
 Alan Turing, a mathematician, was working on
Project Ultra, the successful British effort to
break the German Code during World War II.
As part if his role in that project, Turing helped
design one of the first computers ever built.
 Turing also wrote an article entitled
“Computing Machinery and Intelligence”,
which secured for him the distinction of being
generally recognized as the “father of AI”. He
proposed a question “Can machines think?”
 Turing suggested a test, in the form of game,
that could help decide the issue. He called it
imitation game. (Replicating, human behavior)

13
The Turing Test
 You are the “interrogator”; you can direct
questions to either Person A or Person B
through a keyboard to the screen, but
you do not know which is the man and
which is the woman. Only one of the
persons is obligated to reply truthfully,
the other person is actively engaged in
attempting to fool and confuse you,
using any deceitful tactics that will make
you guess incorrectly. The objective of
the game is to try to guess which person
is male and which is female solely by
analyzing the responses through a
keyboard. Screen communication.
14
The Turing Test
 Next, and this is the critical part of
the Turing test, substitute a
computer for one of the people. Now
the human is obligated to give you
truthful, human-like responses; but
the computer is trying to fool you
into thinking that it is human!

15
The Turing Test
 It is considered to be any situation
in which a human converses with an
unseen respondent and attempts to
determine if the dialogue is being
conducted with a human or a
computer. If a computer can fool
you into believing that you are
talking to a human, the computer
can be said to be intelligent.

17
 Intelligent Test for
Computers
 Alan Turing, a pioneer in the theory of
computation, once proposed an intelligent test for
computer programs [Turing 1950].
 In one variant of the Turing Test, a human judge is
allowed to interrogate a program through some
sort of an interface such as a video terminal. If the
program can fool the human into believing that it
is another human responding rather than a
computer, then the program is judged intelligent.
 You can imagine variants of this test in which you
manipulate a robot’s environment to see how
robot responds and judge the robot as intelligent
or not, depending on whether the robot responds
in accordance with how a human might in the
same situation.
18
 What AI competitions exist?

The Loebner Prize, based on a fund of over $100,000

established by New York businessman Hugh G.
Loebner, is awarded annually for the computer
program that best emulates natural human behavior.
During the contest, a panel of independent judges
attempts to determine whether the responses on a
computer terminal are being produced by a computer
or a person, along the lines of the Turing Test. The
designers of the best program each year win a cash
award and a medal. If a program passes the test in all
its particulars, then the entire fund will be paid to the
program's designer and the fund abolished. For
further information about the Loebner Prize, see the
URL
http://www.loebner.net/Prizef/loebner-prize.html

19
HISTORY OF AI
Details Later
DISUCSSING NATURAL AND
ARTIFICIAL INTELLIGENCE
AI TREE
 Computers are
 Good at: • Bad at:
 Number crunching – Writing poetry
 Storing information – Composing music
 Airline scheduling
– Understanding
 Transmitting data
speech
 Structured data
bases – Driving cars
 Graphics – Enjoying peaches
– Learning new things

23
 Artificial Intelligence

Human Intelligence!

We need to build intelligent machines?

24
 INTELLIGENCE
Natural Artificial
 God made  Man made
 Associated with  Associated with
human machines
 Human behavior  To emulate human
 Psychology behavior in terms of
 Five senses computational
 Symbolic data processes
 Fuzzy ness
 Various sensors
 Numeric + Symbolic
 Uncertainty to be
dealt with

25
Definition of Intelligent Agent:

 A software routine that waits in the

background and performs an action when a
specified event occurs.
For example, agents could transmit a
summary file on the first day of the month
or monitor incoming data and alert the user
when certain transactions have arrived.

26
How do we build an intelligent
agent?

 Must be able to perceive its environment.

 Must be able to affect its environment.
 Must be able to reason about observations
and actions
 Must be able to learn from observations
and actions.
 Must have goals.

27
 Building Intelligent Agents:
First Challenge
 Create a representation of the world in terms
computers can deal with
 Numbers?
 Text
 Logic
 Let’s assume everything about the task can
be represented – we have complete
knowledge

28
 Building Intelligent Agents:
Second Challenge

 Extend your programs to handle situations

where knowledge isn’t complete, i.e.,
where there is uncertainty

29
 Intelligent Agents
prior knowledge
experience
goals/values
observations
 Building Intelligent Agents

prior knowledge
experience Agent actions
goals/values
observations

31
Intelligent Agent Skills include
Vision Processing
Planning
Robotics
Natural Language Understanding
Search Reasoning Machine Learning
Representation of the World
Symbols (Logic, Numbers)

32
 AI programming Vs.
Conventional programming

 AI  CONVENTIONAL
 Primarily symbolic
 Primarily numeric
 Heuristic, Solution
 Algorithmic,
steps implicit Solution step
explicit
 Control structure  Information and
separate from
control integrated
knowledge base
 Difficult to modify
 Easy to modify
 Correct answers
 Satisfactory answers
required
acceptable

33
 EL
EM
EN
NLP
A TS
I
O
F
Heuristic Knowledge
Search Rep.
T.Prov. ROBOTS
& MV
Languages Reasoning
& Tools & Logic

ES
34
 AI: CENTRAL THEMES
THEMES
 K. representation
 Inference and control
 Learn and adapt
 Handling uncertainty
 Reasoning
 Knowledge/search
tradeoff,
Combinatorial
explosion
 Problem
decomposition
35
 WHAT IS INTELLIGENCE AND HOW IT
WORKS?
 Intelligence is ability to
 meet situations successfully
 perceive inter relationship of facts
 learn and understand from
 Goals
experience  We think because
 acquire and retain knowledge there are things we
 respond quickly and successfully
have to do
to a new situation
 To respond to a situation very
 Intelligence
flexibly  Facts and rules
 To make sense out of ambiguous
or contradictory messages  Control
 To recognize the relative  Pruning
importance of different elements
of a situation  Inferencing
 To find similarities B/W situations
despite differences which may
separate them
 To draw distinctions B/W situations
despite similarities which may link
them
36
 Facts And
Rules
Intelligence

1) Collection of facts.

2) Means of utilizing these facts to reach

goals.

3) Done by formulating set of rules relating

to all facts stored in the brain.
Fact/Rule Set1:
Fact 1: A burning stove is hot.
Rule 1:If I put my hand on a burning stove,
Then it hurt.
Fact/Rule Set 2:
Fact 2: During rush hours, streets are crowded with cars.
Rule 2: If I try to cross a major highway on foot during rush
hours,
Then I may get hit by a car.
 How Does Human Intelligence Work?

Goals:
Example
Wake up in the morning.
To go to office.
Reach in time.
Goals ----- ultimate goals.

 Thoughts are not random or arbitrary.

 Pressed into service because of goal(s).
 No thoughts without goal.
 We do not do things because we think, we
think because there are things we have to do.
Example; Wakeup in the morning.
Reach office in time.
Brain is bombarded with all type of data .
(directly/indirectly , related/unrelated).
If we had to deal with this multi data,
we might stand across the road for years.
How mind extracts the right set of rules is pruning.
Eliminates irrelevant pathways.
Reach goal immediately.
Focus on rules pertinent to solving the immediate
problem. 40
 Interference Mechanism

Reaching goal; Problem solved + New knowledge acquired.

Fact 1: Jim’s parents are John & Mary.
Fact 2: Jane’s parents are John & Mary.
• Rule: If a male person & female person have same parents.
Then they are brother and sister.
• Goals: Relation between Jim & Jane.
• Inference: New facts produced;
Jim & Jane are brother and sister.
The component of intelligence that helps us arrive at new facts
is called the “Inference Engine”, “Inference Mechanism”.
Problem may have not a vast store of information. From this we
have to determine the proper course of action. Avoid irrelevant
knowledge components. This is what we called Pruning
41
 WHAT IS INTELLIGENCE AND HOW IT
 Intelligence is ability to
WORKS?
 meet situations successfully  Goals
 perceive inter relationship of facts
 learn and understand from
 We think because

experience
acquire and retain knowledge
there are things
 respond quickly and successfully to a we have to do
new situation  Intelligence
 To respond to a situation very flexibly
 To make sense out of ambiguous or  Facts and rules
contradictory messages
 To recognize the relative importance
 Control
of different elements of a situation
 To find similarities B/W situations
 Pruning
despite differences which may
separate them
 Inferencing
 To draw distinctions B/W situations
despite similarities which may link
them

42
Strong AI
It makes the bold claim that computers can be made to think on
a level (at least) equal to humans.

Weak AI
It simply states that some “thinking like” features can be added
to computers make them more useful tools like expert
systems, speech recognition software etc.

43
 Intelligent Systems
Statistics
Artificial
Intelligence Theoretical
Physics
Signal Intelligent
Processing Systems Control
Theory

44
Robots
 AI Technologies and
Applications
AI TECHNOLOGIES

ROBOTICS EXPERT VISION SPEECH NLU CAI GAME LEARNING KE

SYSTEMS PLAYING
GPS

T. PROVING

Factory Fault Perception Understanding Comm. Learning Neural Acquisition

Automation Diagnosis and teaching Networks
Guidance Generation Grammer Text
Intelligent Parallel understanding
Autonomous Text to
Assistant Inspection Computing
Vehicals Speech Text
Genetic generation
Mechatronics Medical, Verification Speaker
Control, etc. Algorithms
Explanation
 Domain Specific Applications of
AI
APPLICATIONS

MEDICAL SCIENCE INDUSTRY MILITARY SERVICES SPACE FINANCIAL NATURAL

& ENGG. RESOURCES

Diagnosis Plant Factory Surveillance Intelligent Groupd Tax Prep. Prospecting

control M, P & S KB Access Ops. Aid Aids
Data Target
Treatment Intelligent ES Drilling
interpretation Trk. & Recog. P&S
Robots Ops.
Intelligent Traffic
Monitoring Autonomous Remote Intelligent Resource
design Inspection control
Vehicles Ops. Consultant Recovery
Knowledge Chem.
& Bio. Mechatronics Expert Sensor Resource
Automation Management
Synthesis Advisors Fusion
 AI
Technologies

49
 AI
Technologies

50
 AI
Technologies

51
 AI
Technologies

52
 AI
Technologies

53
AI Technologies
Robotics
Expert Systems
Computer
VisionRecognition
Speech
Automatic
Programming
Natural Language
Processing
Planning and Decision
Support
Intelligent Computer Aided
Instruction

54
Generic Applications of AI
Knowledge Human Interaction:
Management:
•Speech understanding
• Intelligent Database
•Speech generation
access
Learning and Teaching:
• Knowledge
acquisition •Computer aided
instruction
• Text understanding
•Intelligent computer
• Text generation
aided instruction
• Machine translation
•Learning from experience
• Explanation
•Concept generation
• Logical operations on
•Operation and
databases
maintenance instruction

55
 Generic Applications of AI

Fault Diagnosis and Communication:

Repair:
•Public access to large
•Humans databases via telephone
•Machines and speech
understanding
•Systems
•Natural Language
Computation: interfaces to computer
•Symbolic programs
Mathematics Operations of Machines
•“Fuzzy” operation and Complex Systems:

•Automatic •Factory Automation

programming •Mechatronics

56
 Generic Applications of AI

Autonomous Intelligent Design:

Systems:
•Systems
•Autonomous vehicles
•Equipment
Management:
•Intelligent Design
•Planning Aids
•Scheduling •Inventing
•Monitoring Visual Perception and
Guidance:
Sensor Interpretation and
Integration: •Inspection
•Developing meaning from •Identification
sensor data
•Verification
•Sensor fusion (integrating
multiple sensor inputs to
develop high level
interpretations) 57
 Generic Applications of AI

Visual Perception and Medical:

Guidance (Continued):
•Patient Monitoring
•Guidance
•Prosthetics
•Screening
•Artificial Sight
•Monitoring and Hearing
Intelligent Assistants: •Reading
•Medical Diagnosis, Machines for the
Maintenance Aids, and Blind
other Expert Systems •Medical Knowledge
•Expert System Automation
Building Tools Executive Assistance:
•Real Mail and Spot
Items on Importance
•Planning Aids
58
 Generic Applications of AI

Science And Industrial:

Engineering:
•Factory
•Discovering Management
•Physical and •Production
mathematical laws Planning and
Scheduling
•Determination of
regularities and •Intelligent Robots
aspects of interest
•Process Planning
•Chemical and
•Intelligent
Biological Synthesis
Machines
Planning
•Computer-Aided
•Test Management
Inspection
•Data Interpretation
•Mechatronics
•Intelligent Design
Aids 59
 Generic Applications of AI
Military: Military:
•Expert Advisers •Intelligent Robots
•Sensor Synthesis •Diagnosis and Maintenance
and Interpretation Aids
•Battle and Threat •Target Location and
Assessment Tracking
•Automatic Photo • Map Development Aids
Interpretation
•Intelligent Interactions with
•Tactical planning Knowledge Bases
•Military Financial:
Surveillance
•Tax Preparation
•Weapon-Target
•Financial Expert Systems
Assignment
•Intelligent Consultants
•Autonomous
Vehicles
60
 Generic Applications of AI
International: Natural Resources:
•Aids to Understanding •Prospecting Aids
and Interpretation
•Resource operations
•Goals, aspirations and
•Drilling Procedures
motives of different
countries and cultures •Resource Recovery
•Cultural models for Guidance
interpreting how others •Resource Management
perceive Using Remote Sensing
• Natural language Data
Translation
Services:
•Intelligent Knowledge
Base Access
•Airline Reservations
•Air Traffic Control and 61
 Generic Applications of AI

Space:
•Ground Operation Aids
•Planning and Scheduling Aids
•Diagnosis and Reconfiguration Aids
•Remote Operations of Spacecraft and
Space Vehicles
•Test Monitors
•Real-time Re-planning as required by
failures, changed conditions, or new
opportunities
•Automatic Subsystem Operations

62
 Playing Chess

A game of chess can be analyzed by a symbolic

program. Each chess piece is represented by a
symbol, and the rules governing legal moves are
represented by symbolic expressions.
 APPLYING AI
 Physics  Control  Computing
 VLSI Design  Broom  Speech
 Non Balancer Recognition
Destructive
 Prediction and  Intelligent
Testing Estimation Agent on the
 Element
 Backing of a Net
Classification Truck  Network
 Power Plant Management
 Load
Control and Routing
Measuring  Troubleshootin
 Mechanical
 Robotics
 Precision g
Design  Character
Manufacturing
Assembly Recognition
 Teaching
Physics
 Isotopes
Detection
 Circuit Analysis
 Learning
64