4.

Modern definition: AI is defined as the branch of computer science

dealing with symbolic and
non-algorithmic method of problem solving."
NOTE: Alworks with 'pattern-matching" methods which attempttodescribe object event
|andprocess in terms of their qualitative nature and logical and combinational relationship.
Please understand that intelligence" is acatchy word herc. Intelligence mcans an integrated sum
of all knowledge and feats, acquired through study and cxpericnce. Also note that the food for this
intelligence is knowledge.
AI task domain
Basically, Al covers three types of tasks
(a) Mundane tasks
(b) Fomal tasks
(c) Expert tasks
Let usunderstand these tasks now.
I. Mundane tasks
Alhas some mundane tasks also like,
(a) Perception of vision and speech.
(b) Natural Language understanding.
(c) Common sense Reasoning
(d) Robotics
II. Formal tasks
AIformal tasks include:
(a) Game playing, e.g.,chess, 8-queens problems, water-jug problem etc.
(b) Mathematics -geometry, calculus etc.
III. Expert tasks
It includes:
(a) Engineering field.
(b) Scientific analysis.
(c) Medical and financial analysis.
What AI is NOT?
1. Al is not the study and creation of conventional computer systems.
All programs exhibitsome degree of intelligence and an Alprogram must go beyond this degree
of intelligence also. This is not the case. Remember that human beings are more intelligent than our
intelligent systems also as we only have created them.
2. Al is not the study of mind, nor of body, nor of languages, as found in fields of psychology
physiology, cognitive science or linguistics.
Although there is some overlap between these fields and Al yet the goal of Al is to develop a
computer system that is capable of perfornming intelligent tasks effectively and etticiently.
So, what Al will include? It includes areas like:
(a) Robotics
(b) Memory Organization,
(c) Knowledge representation,
(d) Storage and recall,
(e) Learning models,
(f) Inference techniques,
(g) Common sense reasoning
 (h) Decision making.
(i) Patterm recognition,
(i) Searching,
(k) Speech recognition,
() Speech Synthesis.
History of AI
Let us see the history of Al in 1950s, 1960s, 1970, 1980s and 1990s in a tabular form now.
Year Developments made
1950s o Birth of AI
o First Neural Net Simulator (Minsky).
O GPS General Purpose Problem-Solver (Simon)
o GTP GeometryTheorem Prover- (Gelertber),
input diagrams, backward reasoning.
o SAINT Symbolic integration
1960s OANALOGY-Could solve IQ test puzzle
o STUDENT could solve algebraic word problems
o SHRDLU couldmanipulate blocks using robotic arm.
o STRIPS aproblem-solver planner.
oMinsky and Papert -demonstratedthe limitations ofneural nets.
1970s oConceptual dependency theory (Shank)
oFrames (Minsky)
o Machine Learning: ID3(Quinlan),
AM (Lenat)
o Expert System got success.
o DENDRAL by Feigenbaum--to identify molecular structure.
oMYCIN by Shortliffe and Buchanan - to diagonose infectious
blood diseases.
1980s oCheaper computing, so AI Software a feasible success, neural
nets revisited.
o XCON (by Mc Sermott)-Saved $40m per year.
o Neural computing-back propogation (Werbos), associative
memory (Hopfield), logic programming developed and used.
1990s o Embedded intelligent systems, agents.
oHybrid approaches: logic + neural nets + genetic algorithms
+ fuzzy logic
o CYC(By Lenat)-a far-reaching project tocapture common
sense reasoning.
oSociety of Mind (Minsky)-Intelligence is aproduct of complex
interactions of simple agents.
o Deep Blue (formerly Deep Thought)-defeated Kasparov in
speed chess in 1997.
4 Artificial Intelligence -A Practical Approach
Let us see the entire history now, from 1941to 1991.
The birth of AI AI-based hardware
Start of DOD's sells $425 million
advanced research to companies
projects
First Al system beats
Electronic Dartmouth human chess
conference
|computer First expert system master

1941 1949 1956 1958 1963 1968 1970 1972 1986 1991

Lisp language AImilitary systems

Microworld
First commercial developed programn SHRDLU
used effectively in
stored program DESERT STORM
created
computer PROLOG language
revealed
Logic theorist
developed

Fig. 1.1 History ofAI.

1.1 SCOPE OFAI
AIhas a very wide scope. Someof its applications are discussed below:
1.1.1 Games
According to Newell and Simon (1976), the essential basis for human problem
solving is to
systematically explore a space of problem states i.e., successive and alternative stages
problem solving. For example, the different board configurations in a in the
steps in a reasoning process. This space is called as state space. chess game or intermediate
then searched to find a final answer. This space of alternative solutions is
Please understand that games can generate extremely large
techniques to search for our solution in this space. search spaces. So, we need powerul
These
note that intelligence resides in heuristics only. So, AI techniques are called as heuristics. Also
helps here.
1.1.2 Theorem Proving
It is a formal system wherein logic
lends
solved by using logical axioms and treatingitself to automation. A large variety of problems can be
problem
proving has devised powerful solution heuristics and instances as theorems to be proved. Theorem
For example, design and reduced the complexity of the search space.
to such an approach. verification of logic circuits,controlof complex systems etc., will
respornd
Many modern theorem provers function as
humans todo the task of decomposition of large 'intelligent assistants'. By this we mean that it allows
for searching the space of
possible
problems into subproblems and divising heuristicS
proofs. On the
simpler but stilldemanding task of proving other hand, the theorem prover then performs th
lemmas, verifying smaller guesses and
formal aspects of a proof outlined by the humarns. completing the
1.1.3 Natural Language Processing (NLP)
It is a subfield of AI which deals with the
methods of communicating with a computer in ones
oWn natural language. This will fill the gap between the
need not be computer literate to humans and the machines. So, now one
communicate with it.
1.14 Vision and Speech Processing
Computer vision is acomputation intensive process. It involves multiple transformations. This relies
moreon classical A(methods of symbolic
processing.
Speech understanding requires recognition of basic speech patterns. These patterns are
against lexicon patterns for recognition. Developing systems that understand speech hasmatched
continuing goal ofAI researchers. been a
1.1.5 Robotics
According to the Robot Institute of America (RIA) "A robot is a
multifunctional, manipular that is designed to move materials, parts, tools orreprogrammable,
specialized devices
through various programmned motions for the performance of a variety of tasks."
Today, we need intelligent Robots to make them smaller devices. So, intelligence needs to be
embedded into it. So, now vision is very essential. These intelligent robots as well as AI promise us to
solveour complex problems easily.
1.1.6 Expert Systems(ES)
They are knowledge intensive programs that solve problems in a domain that needs a good technical
expertise. As per the definition given by British Computer Society's Committee of the specialist
group on Expert Systems
Theembodiment within acomputer of a knowledge-based component from an expert skill
in such a form that the machine can offer intelligent advice or take an intelligent decision about
a processing function. A desirable additional characteristic which many would regard
fundamental, is thecapability of the system on demand to justify its own line of reasoning in a
manner directly intelligible to the inquirer. The style adopted to attain these characteristics is
rule-based programming."
So, we find from this definition that an Al specialist or a knowledge engineer or expert system
designers are responsible for implementing knowledge in aprogram, both effectively and intelligently.
For examples,
1. DENDRAL developed at Stanford University in late 1960s. It was designed to infer the
structure of organic molecules from their chemical formulas and mass spectrographic information
about the chemical bonds present in the molecules. As these organic molecules tend to be very
large, so the number of possible structures for these molecules tend to be huge. DENDRAL
addresses this problem of a large search space by applying the heuristics knowledge of expert
chemists to the structure elucidation problem.
DENDRAL used domain specific knowledge for this.
2. MYCIN It uses the expert medical knowledge to diagnose and prescribe treatment for
spinal meningitis and bacterial infections of the blood. It was developed at Stanford in mid
1970s. It provides clear and logical explanations of itsreasoning. It usesacontrol structure
appropriate to the specific problem domain.
3. PROSPECTOR-Aprogram for detemining the probable location and the type of oredeposits
based on geological information about a site.
Please note that most expert systems have been written for relatively specialized expert
level domains.
1.2 AITECHNIQUES
I. Search Knowledge
body of facts and principles accumulated by humankind or
Knowledge can be defined as the
state of knowing. For example, in biological organisms. knowledge isstored as
the act, fact or representations
structures of interconnected neurons. The structures correspond to symbolic
complex
possessed by the organism, the facts, rules and so on. Please note that an average
of the knowledge and contains an estimated number of 10'" neurons.
Also
human brain weighs about 3.3 pounds
neurons and their interconnection capabilities provide about 10'" bits of potential
note that these
knowledge is stored as symbolic structures but in
storage capacity. On the other hand, in computers
form of collections of magnetic spotsand voltage states.
Knowledge is of three types as shown below:
Knowledge

Procedural Declarat1ve
or Heuristic
Operational Relational knowledge
knowledge knowiedge

Fig. 1.2 Types of Knowledge

Let us define these types now.
1. Procedural or Operational Knowledge
It is defined as a compiled knowledge related to the performanceof some task. For example,
steps to solve a quadratic equation are expressed as a procedural knowledge.
2. Declarative or Relational Knowledge
It is apassive knowledge expressed as statements of facts about the world. For example, Personnel
data in adatabase. Such data are explicit pieces of independent knowledge.
3. Heuristic Knowledge
Heuristics means using some rules of thumb or tricks or some strategies to simplify the solution
to problems. We acquire this after much of the
experience.
Some terminologies related to
knowledge
We define certain terms that will be used again
and again here. They are as tollOWS:
1. Knowledge and data: Adoctor has both
knowled ge as
record whereas knowledge is what he has leaned in hiswell as data. Here, data is thepatient s
by Feigenbaum and medical college. This was explained
Mc.Corduck.
2. Belief: It is defined as essentially any
represented. It may be true or false. meaningful and coherent expression that can be
3. Hypothesis: Itis defined as a justified belief that is not known to be
with some supporting evidences but it true. It is backed up
may still be false.
1.2 AI TECHNIQUES

I. Search Knowledge accumulated by humankind or

body of facts and principles
Knowledge can be defined as the biological organisms, knowledge is stored as
For example, in
the act, fact or state of knowing. ncurons. Thcstructures correspond to symbolic representations
complex structures of interconnected facts, rules and so on. Please note that an average
organism, the
of the knowledge possessed by the contains an estimated number of 10" neurons. Also
pounds and
human brain weighs about 3.3 capabilities provide about 10'" bits of potential
these neurons and their interconnection
note that
On the other hand, in computers knowledge is stored as symbolic structures but in
storage capacity.
form of collections of magnetic spots and voltage states.
Knowlcdge is of three types as shown below:
Knowledge

Procedural Declarative
Or Heuristic
Operational Relational knowledge
knowledge knowledge
Fig. 1.2 Types of Knowledge
Let us define these types now.
1. Proceural or Operational Knowledge
It isdefined as a compiled knowledge related to the performance of sometask. For example,
steps to solve a quadratic equation are expressed as a procedural knowledge.
2. Declarative or Relational Knowledge
It is a passive knowledge expressed as statements of facts about the world.
data in a database. Such data are explicit pieces of For example, Personnel
independent knowledge.
3. Heuristic Knowledge
Heuristicsmeans using some rules of thumb or tricks or some
to problems. We acquire this after much of the strategies to simplifythe solution
experience.
Some terminologies related to knowledge
We define certain terms that will be Used
again and again here, They are as follows:
1. Knowledge and data: Adoctor has
record whereas knowledge is what hebothhasknowledgein as well as data. Here, data is the patient's
learned his medical college. This was
by Feigenbaum and Mc.Corduck. explained
2. Belief: It is defined as essentially any
represented. It may be true or false. meaningful and coherent expression that can be
3. Hypothesis: It isdefined as a justified belief
that is not known to be true. It is
with some supporting evidences but it may stillbe false. backed up
 4.
Knowledge: It is a true justified belief.
5. Meta Knowledge: It is the knowledge about the knowledge.
6. Epistemology: It is the study of the nature of knowledge.
Al technique is a method that exploits knowledge that should be represented in such a way that:
1. It captures generalizations.
2. It can be understood by people who must provide it.
3. It can be easily modified to correct errors and to incorporate changes.
4. Itcan be used in a great many situations even if it is not totally accurate or complete.
5. It can be used to help overcome its own sheer bulk by helping to narrow the range of possibilities
that must usually be considered.
Please note that it is possible to solve AI problems without using AI techniques but the solutions
would be inefficient. Also it is possible to apply AI techniques to the solution of non Al problems
and this willbe agood thing for those problems that possess same characteristics as Al problems
have.
Those systems that depend on a rich base of knowledge to perform difficult tasks are known
as knowledge-based systems.
Aknowledge based system (KBS) consists of three main components as shown in Fig. 1.3.
Query
Input-Output Inference Knowledge
Unit Control Unit base

Fig. 1.3 Components of a KBS

Explanation: KBSs get their power from the expert knowledge that has been coded into facts,
rules, heuristics and procedures. The knowledge is stored in a knowledge base only. Since it is stored
separately (from I/O unit and Inference CU)so we can casily add new knowledge to this knowledge
base or refine existing knowledge without recompiling the control and inferencing programs. So, now
the construction as well as the maintenance of KBS becomes very simple.
Knowledge -base is different from a database. Let us tabulate the differences between them
now.

Data-base (DB) Knowledge Base (KB)

1. It is defined as a collection of data 1. It has information at a higher level of
representing facts abstraction.

2. It is larger than a KB. 2. It is smaller than a DB.

3. Changes are fast 3. Changes are gradual.

4. All information needs to be stated explicitly. 4. It has the power of inferencing.
5. It is mainttained for operational purposes. 5. It is used for data analysis and planning.
6. Knowledge is represented by relational, 6. Knowledge is represented by logic or rules.
network or hierarchical model.

Knowledge may be represented at various levels as shown in Fig. 1.4.

 Artificial Intelligence-A Practical Approach

Mental image

Written text

Character string

Binary numbers

Magnetic spots
Fig. 1.4 Levels of Knowledge Representation
Please note, however, that any choice of representation wll depend on the type of problem to
be solved and the inference methods available.
Notonly this, knowledge organization in memory is also very vital. The KBS discussed in Fig. 1.3
may require several (tens of thousands) of facts and rules to perform their intended tasks. So, now it
is mandatory to casily search, locate and retrieve appropriate facts andrules from this KB. We use
indexingsto speed up searches wecan form groups of knowledge of similar type and then makea
pointer to point to this group. So, now instead of searching for an entire KB, only a small portion of
it may be accessed.
We need tomanipulate this knowledge also. Usually, the user gives some input and thus, initiates
asearch for a goal or decision. Please understand that thisrequires that known facts in the KB
belocated, compared (matched) and altered somehow. Also note that this process may set up
some sub-goals which requires some more inputs and so on until a final solution is found. This
reauires aform of inference or deduction using the knowledge and inferringg rules, Please note here
that all forms of reasoning require certaln amount of searching and matching, Also note that
searching and metaching both consume the maximum time of computation in AI systems. So,
ue need to have the best searching technique today so that we can avoid this
problem during searching. combinatorial explosion
Now. the question is how to acquire knowledge? Knowledge may be
textbooks, references, reports, technical research papers and so on and acquired from sources like
to be useful, it should be
accurate, completle, inconsIstent and So on. Our KBS depends on ahigh
SucceSs.
guality knowledge tor ther