CHAPTER 1

INTRODUCTION

Artificial Intelligence
1
 What is Artificial Intelligence?
 We call ourselves Homo sapiens was applied in 1758 by the father of modern
biological classification, Carolus Linnaeus, (species of modern human
beings/wise man) because our intelligence is so important to us.
 For thousands of years, we have tried to understand how we think, i.e., how a
matter can perceive, understand, predict, and manipulate a world far larger
and more complicated than itself.
 The field of Artificial Intelligence(Al) attempts not just to understand but also to
build intelligent entities.
 AI is one of the newest fields in science and engineering.
 Work started soon after World War II, and the name was coined in 1956.
 Along with molecular biology, AI is regularly cited as the field most liked by
scientists in other disciplines.
 A student in physics might feel that all the good ideas have already been taken
by Galileo, Newton, Einstein, and the rest.
Artificial Intelligence
2
 What is Artificial Intelligence?
 But, AI still has openings for several full-time Einsteins and Edisons.
 Al encompasses a huge variety of subfields ranging from the general (learning
and perception) to the specific such as playing chess, proving mathematical
theorems, writing poetry, driving a car on a crowded street, and diagnosing
diseases.
 AI is relevant to any intellectual task and is truly a universal field.
 In the Figure shown below, we can see eight definitions of AI, laid out along
two dimensions.
 The definitions on top are concerned with thought processes and reasoning,
and the ones on the bottom address behavior.
 The definitions on the left measure success in terms of fidelity(loyalty) to
human performance, on the right measure against an ideal performance
measure called rationality.
Artificial Intelligence
3
What is Artificial Intelligence?

Artificial Intelligence
4
 What is Artificial Intelligence?
 A system is rational if it does the right thing, given what it knows.
 Historically, all four approaches to AI have been followed, each by different
people with different methods.
 A human-centered approach must be in part an empirical science involving
observations and hypotheses about human behavior.
 A rationalist approach involves a combination of mathematics and engineering.

Artificial Intelligence
5
 Acting humanly: The Turing Test approach
 In 1950, Alan Turing introduced a test to check whether a machine can think
like a human or not, and this test is known as the Turing Test.
 In this test, Turing proposed that the computer can be said to be an intelligent
if it can mimic human response(fool the interrogator) under specific conditions.

Artificial Intelligence
6
 Acting humanly: The Turing Test approach
 The Turing test is based on a party game called Imitation game, with some
modifications.
 This game involves three players in which one player is computer, another
player is human responder, and the third player is a human Interrogator, who
is isolated from other two players and his job is to find that which player is
machine among two of them.
 The questions and answers can be like:

Interrogator: Are you a computer?

PlayerA (Computer): No
Interrogator: Multiply two large numbers such as (256896489*456725896)
Player A: Long pause and give the wrong answer.

Artificial Intelligence
7
 Acting humanly: The Turing Test approach
 A computer passes the test if a human interrogator, after posing some written
questions, cannot tell whether the written responses come from a person or from
a computer.
 Programming a computer to pass a test requires lot of work to be done.
 The computer need to possess the following capabilities:
 Natural language processing to enable it to communicate successfully in English.
 Knowledge representation to store 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.
 To pass the test, the computer need computer vision to perceive objects, and
robotics to manipulate objects and move about(most of AI composed of these six)
Artificial Intelligence
8
 Acting humanly: The Turing Test approach
Chatbots to attempt the Turing test:
 ELIZA was a Natural language processing computer program created by Joseph
Weizenbaum was one of the first chatterbots attempted the Turing Test.
 It was created to demonstrate the ability of communication between machine and
humans.
 It Parry was a chatterbot created by Kenneth Colby in 1972.
 Parry was designed to simulate a person with Paranoid schizophrenia(most
common chronic mental disorder).
 Parry was described as ELIZA with attitude was tested using a variation of the
Turing Test in the early 1970s.
 Eugene Goostman was developed in Saint Petersburg in 2001.
 In June 2012 at an event, Goostman won the competition promoted as largest-ever
Turing test content in which it has convinced 29% of judges that it was a human.
Artificial Intelligence
9
 Thinking humanly: The cognitive modeling approach
 If we say that a given program thinks like a human, we must have some way of
determining how humans think.
 So, we need to get inside the actual workings of human minds.
 There are three ways to do this:
 Through introspection: trying to catch our own thoughts as they go by.
 Through psychological experiments: observing a person in action.
 Through brain imaging: observing the brain in action.
 Once we have a sufficiently precise theory of the mind, it becomes possible to
express the theory as a computer program.
 If the input-output behavior of program matches corresponding human behavior,
then we can think it as an evidence that some of the program’s mechanisms could
also be operating in humans.
Artificial Intelligence
10
 Thinking humanly: The cognitive modeling approach
 For example, Allen Newell and Herbert Simon who developed GPS (General
Problem Solver) were not content merely to have their program solve problems
correctly.
 They were more concerned with comparing the trace of its reasoning steps to
traces of human subjects solving the same problems.
 The interdisciplinary field of cognitive science brings together computer models
from AI and experimental techniques from psychology to construct precise and
testable theories of the human mind.
 Cognitive science is another field, worthy of several textbooks and at least one
encyclopedia.
 We occasionally comment on similarities or differences between AI techniques and
human cognition.
 Real cognitive science is based on experimental investigation of actual humans or
animals.
Artificial Intelligence
11
 Thinking rationally: The “laws of thought” approach
 The Greek philosopher Aristotle was one of the first to attempt to codify right
thinking, that is irrefutable(impossible to deny or disprove) reasoning processes.
 His syllogisms provided patterns for argument structures that always yielded correct
conclusions when given correct premises.
 For example, “Socrates is a man; all men are mortal; therefore, Socrates is mortal.”
 These laws of thought were supposed to govern the operation of the mind.
 Logicians in the 19th century developed a precise notation for statements about all
kinds of objects in the world and the relations among them.
 By 1965, programs existed could solve any solvable problem described in logical
notation.
 The logicist tradition available with AI hopes to build on such programs to create
intelligent systems.

Artificial Intelligence
12
 Thinking rationally: The “laws of thought” approach
 There are two main obstacles to this approach.
 First, it is not easy to take informal knowledge and state it in the formal terms
required by logical notation, particularly when the knowledge is less than 100%
certain.
 Second, there is a big difference between solving a problem in principle and solving
it in practice.
 Even problems with just a few hundred facts can exhaust the computational
resources of any computer unless it has some guidance as to which reasoning
steps to try first.
 Though both of these obstacles apply to any attempt to build computational
reasoning systems, they appeared first in the logicist tradition.

Artificial Intelligence
13
 Acting rationally: The rational agent approach
 An agent is just something that acts.
 All computer programs do something, but computer agents are expected to do
more such as operate autonomously, perceive their environment, persist over a
prolonged time period, adapt to change, and create and pursue goals.
 A rational agent is one that acts so as to achieve the best outcome or, when there is
uncertainty, the best expected outcome.
 In the laws of thought approach to AI, the emphasis was on correct inferences.
 Making correct inferences is 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 and in some situations,
there is no provably correct thing to do, but something must still be done.
 There are also ways of acting rationally that cannot be said to involve inference.
Artificial Intelligence
14
 Acting rationally: The rational agent approach
 For example, recoiling from a hot stove is a reflex action that is usually more
successful than a slower action taken after careful deliberation.
 All the skills needed for the Turing Test also allow an agent to act rationally.
 Knowledge representation and reasoning enable agents to reach good decisions.
 We need to be able to generate comprehensible sentences in natural language to
get by in a complex society.
 We need learning not only for erudition, but also because it improves our ability to
generate effective behaviour.
 The rational-agent approach has two advantages over the other approaches.
 First, it is more general than the laws of thought approach because correct
inference is just one of several possible mechanisms for achieving rationality.
 Second, it is more amenable to scientific development than are approaches based
on human behavior or human thought.
Artificial Intelligence
15
 Acting rationally: The rational agent approach
 The standard of rationality is mathematically well defined and completely
general and can be unpacked to generate agent designs that provably
achieve it.
 Human behavior, on the other hand is well adapted for one specific
environment and is defined by the sum total of all the things that humans do.
 One important point to keep in mind that achieving perfect rationality.
 That is always doing the right thing, which is not feasible in complicated
environments.
 The computational demands are just too high.
 We will adopt the working hypothesis that perfect rationality is a good
starting point for analysis.

Artificial Intelligence
16
 The Foundations of Artificial Intelligence
 We will study the history of the disciplines that contributed ideas, viewpoints,
and techniques to the development of AI.
 Like any history, it is forced to concentrate on a small number of people,
events, and ideas and to ignore others that also were important.
Philosophy:
1. Can formal rules be used to draw valid conclusions?
2. How does the mind arise from a physical brain?
3. Where does knowledge come from?
4. How does knowledge lead to action?
 Aristotle (384-322 B.C.) was the first to formulate a precise set of laws
governing the rational part of the mind.

Artificial Intelligence
17
 The Foundations of Artificial Intelligence
 He developed an informal system of syllogisms for proper reasoning, which
in principle allowed one to generate conclusions mechanically, given initial
premises.
 Ramon Lull had the idea that useful reasoning could actually be carried out
by a mechanical artifact.
 Thomas Hobbes proposed that reasoning was like numerical computation,
that we add and subtract in our silent thoughts.
 Around 1500, Leonardo da Vinci designed but did not build a mechanical
calculator, but the recent reconstructions have shown the design to be
functional.
 The first known calculating machine was constructed around 1623 by the
German scientist Wilhelm Schickard, although the Pascaline, built in 1642
by Blaise Pascal is more famous.

Artificial Intelligence
18
 The Foundations of Artificial Intelligence
 Leibniz built a calculator that could add, subtract, multiply, and take roots,
whereas the Pascaline could only add and subtract.
 Some speculated that machines might not just do calculations, but they are
able to think and act on their own.
 Thomas Hobbes suggested the idea of an artificial animal arguing “For what is
the heart but a spring and the nerves, but so many strings and the joints, but
so many wheels.”
 It’s one thing to say that the mind operates at least in part according to logical
rules, and to build physical systems that emulate some of those rules.
 Ren’e Descartes gave the first clear discussion of the distinction between
mind and matter and of the problems that arise.
 One problem with a purely physical conception of the mind is to leave little
room for free will, and if the mind is governed entirely by physical laws, then it
has no more free will.
Artificial Intelligence
19
 The Foundations of Artificial Intelligence
 The final element in the philosophical picture of the mind is the connection
between knowledge and action.
 This question is vital to AI because intelligence requires action and reasoning.
 Moreover, only by understanding how actions are justified, can we understand
how to build an agent whose actions are justifiable? (or rational).
 Aristotle argued that actions are justified by a logical connection between
goals and knowledge of the action’s outcome.
 But how does it happen that thinking is sometimes accompanied by action
and sometimes not, sometimes by motion, and sometimes not?
 It looks as if almost the same thing happens as in the case of reasoning and
making inferences about unchanging objects.
 Aristotle’s algorithm was implemented 2300 years later by Newell and Simon
in their GPS program, now it is called as a regression planning system.
Artificial Intelligence
20
 The Foundations of Artificial Intelligence
Mathematics:
1. What are the formal rules to draw valid conclusions?
2. What can be computed?
3. How do we reason with uncertain information?
 Philosophers staked out some of the fundamental ideas of AI, but the leap to a
formal science required a level of mathematical formalization in the area logic,
computation, and probability.
 The idea of formal logic can be traced back to the philosophers of ancient
Greece, but its mathematical development began with the work of George
Boole, who worked out the details of propositional, or Boolean logic.
 In 1879, Gottlob Frege extended Boole’s logic to include objects and
relations, creating the first order logic that is used today.

Artificial Intelligence
21
 The Foundations of Artificial Intelligence
 Alfred Tarski introduced a theory of reference that shows how to relate the
objects in a logic to objects in the real world.
 The first nontrivial algorithm is thought to be Euclid’s algorithm for computing
greatest common divisors.
 The word algorithm comes from al-Khowarazmi, a Persian mathematician of
the 9th century, who also introduced Arabic numerals and algebra to Europe.
 Boole and others discussed algorithms for logical deduction, and later efforts
were made to formalize general mathematical reasoning as logical
deduction.
 In 1931, Godel showed that limits on deduction do exist.
 His incompleteness theorem showed that in any formal theory, there are true
statements that are undecidable in the sense that they have no proof within
the theory.
Artificial Intelligence
22
 The Foundations of Artificial Intelligence
 Besides logic and computation, the third great contribution of mathematics to AI
is the theory of probability.
 The Italian Gerolamo Cardano first framed the idea of probability, describing it in
terms of the possible outcomes of gambling events.
 In 1654 Blaise Pascal showed how to predict the future of an unfinished
gambling game and assign average payoffs to the gamblers.
 Probability quickly became an invaluable part of all the quantitative sciences,
helping to deal with uncertain measurements and incomplete theories.
 James Bernoulli, Pierre Laplace, and others advanced the theory of probability,
and introduced new statistical methods.
 Thomas Bayes proposed a rule for updating probabilities in the light of new
evidence.
 Bayes’ rule underlies most modern approaches to uncertain reasoning in AI
systems.
Artificial Intelligence
23
 The Foundations of Artificial Intelligence
Economics:
1. How should we make decisions so as to maximize payoff?
2. How should we do this when others may not go along?
3. How should we do this when the payoff may be far in the future?
 The science of economics got its start in 1776, when Scottish philosopher Adam
Smith published An Inquiry into the Nature and Causes of the Wealth of Nations.
 While the ancient Greeks and others made contributions to economic thought,
Smith was the first to treat it as a science, and using the idea that economies
can be thought of as consisting of individual agents maximizing their own
economic well-being.
 Most people think of economics as being about money, but economists say that
they are really studying how people make choices that lead to preferred
outcomes.
Artificial Intelligence
24
 The Foundations of Artificial Intelligence
 When McDonald’s offers a hamburger for a dollar, they are asserting that they
would prefer the dollar and hoping that customers will prefer the hamburger.
 The mathematical treatment of preferred outcomes or utility was first formalized
by Leon Walras and was improved by Frank Ramsey and later by John von
Neumann and Oskar Morgenstern.
 Decision theory which combines probability theory with utility theory, provides a
formal and complete framework for decisions made under uncertainty.
 That is, in cases where probabilistic descriptions appropriately capture the
decision maker’s environment.
 This is suitable for large economies where each agent need pay no attention to
the actions of other agents as individuals.
 For small economies, the situation is much more like a game, the actions of one
player can significantly affect the utility of another(either positively or negatively).

Artificial Intelligence
25
 The Foundations of Artificial Intelligence
 Von Neumann and Morgenstern’s development of game theory included the
surprising result that, for some games a rational agent should adopt policies that
are randomized.
 Unlike decision theory, the game theory does not offer an unambiguous
prescription for selecting actions.
 For the most part, economists did not address the third question listed earlier,
namely, how to make rational decisions when payoffs from actions are not
immediate but instead result from several actions taken in sequence.
 This topic was pursued in the field of operations research, which emerged in
World War II from efforts in Britain to optimize radar installations, and later found
civilian applications in complex management decisions.
 The work of Richard Bellman formalized a class of sequential decision problems
called Markov decision processes.

Artificial Intelligence
26
 The Foundations of Artificial Intelligence
 Work in economics and operations research has contributed much to the
notion of rational agents, but still for many years AI research developed
entirely on separate paths.
 One reason was the complexity of making rational decisions.
 The pioneering AI researcher Herbert Simon won the Nobel Prize in
economics in 1978 for his work showing that models based on strategies in
making decisions that are good enough rather than laboriously calculating
an optimal decision gave a better description of human behavior.
 Since the 1990s there has been a increase of interest in decision-theoretic
techniques for agent systems.

Artificial Intelligence
27
 The Foundations of Artificial Intelligence
Neuroscience:
1. How do brains process information?
 Neuroscience is the study of the nervous system, particularly the brain.
 Although the exact way in which the brain enables thought is one of the
great mysteries of science, the fact that it does enable thought has been
appreciated for thousands of years because of the evidence that strong
blows to the head can lead to mental incapacitation.
 Also, it is known that human brains are somehow different.
 In about 335 B.C. Aristotle wrote that, “Of all the animals, man has the
largest brain in proportion to his size.”
 Still, it was not until the middle of the 18th century that the brain was widely
recognized as the seat of consciousness.
 Before that, candidate locations included the heart and the spleen.
Artificial Intelligence
28
 The Foundations of Artificial Intelligence
 Paul Broca’s study of aphasia (speech deficit) in brain damaged patients in
1861 demonstrated the existence of localized areas of the brain responsible
for specific cognitive functions.
 In particular, he showed that speech production was localized to the portion
of the left hemisphere now called Broca’s area.
 By that time, it was known that the brain consisted of nerve cells, or neurons,
but it was not until 1873 that Camillo Golgi developed a staining technique
allowing the observation of individual neurons in the brain.
 This technique was used by Santiago Ramon Y Cajal in his pioneering
studies of the brain’s neuronal structures.
 Nicolas Rashevsky was the first to apply mathematical models to the study of
the nervous system.

Artificial Intelligence
29
The Foundations of Artificial Intelligence

Artificial Intelligence
30
 The Foundations of Artificial Intelligence
 We now have some data on the mapping between areas of the brain and the
parts of the body that they control or from which they receive sensory input.
 Such mappings are able to change over the course of a few weeks, and
some animals seem to have multiple maps.
 Moreover, we do not fully understand how other areas can take over
functions when one area is damaged.
 There is almost no theory on how an individual memory is stored.
 The measurement of intact brain activity began in 1929 with the invention of
the electroencephalograph (EEG) by Hans Berger.
 The recent development of functional Magnetic Resonance Imaging (fMRI)
gives the detailed images of brain activity to the neuroscientists enabling
measurements that correspond in interesting ways to ongoing cognitive
processes.

Artificial Intelligence
31
 The Foundations of Artificial Intelligence
 These are augmented by advances in single-cell recording of neuron activity.
 Individual neurons can be stimulated electrically, chemically, or even optically
allowing neuronal input-output relationships to be mapped.
 With all these advances, we are still a long way from understanding how
cognitive processes actually work.
 The conclusion is that a collection of simple cells can lead to thought, action,
and consciousness, in otherwords, brains cause minds.
 The only real alternative theory is mysticism, that minds operate in some
mystical realm (unexplainable) that is beyond physical science.
 Brains and digital computers have somewhat different properties.
 Following figure shows that computers have a cycle time that is a million
times faster than a brain.

Artificial Intelligence
32
 The Foundations of Artificial Intelligence
 The brain makes up for that with far more storage and interconnection than
even a high-end personal computer although the largest supercomputers
have a capacity that is similar to the brain’s.
 Even with a computer of unlimited capacity, we still could not know how to
achieve the brain’s level of intelligence.

Artificial Intelligence
33
 The Foundations of Artificial Intelligence
Psychology:
1. How do humans and animals think and act?
 The origins of scientific psychology are usually traced to the work of the
German physicist Hermann von Helmholtz and his student Wilhelm
Wundt.
 Helmholtz applied the scientific method to the study of human vision, and
his Handbook of Physiological Optics is even now described as the most
important on the physics and physiology of human vision.
 In 1879, Wilhelm Wundt opened the first laboratory of experimental
psychology, at the University of Leipzig.
 Wilhelm Wundt insisted on carefully controlled experiments in which his
workers would perform a perceptual or associative task while
introspecting on their thought processes.

Artificial Intelligence
34
 The Foundations of Artificial Intelligence
 The careful controls went a long way towards making psychology a science,
but the subjective nature of the data made it unlikely that an experimenter
would ever disconfirm his or her own theories.
 Biologists studying animal behavior lacked introspective data and developed
an objective methodology, as described by H. S. Jennings in his work
Behavior of the Lower Organisms (exhibits simple body structure and simple
organization).
 Applying this viewpoint to humans, the behaviorism movement led by John
Watson rejected any theory involving mental processes on the grounds that
introspection (observation) could not provide reliable evidence.
 Behaviorists insisted on studying only objective measures of the percepts (or
stimulus) given to an animal and its resulting actions (or response).
 Behaviorism discovered a lot about rats and pigeons but had less success at
understanding humans.
Artificial Intelligence
35
 The Foundations of Artificial Intelligence
 Cognitive psychology which views the brain as an information
processing device, can be traced back to the works of William
James.
 Helmholtz also mentioned that perception involved a form of
unconscious logical inference.
 The cognitive viewpoint was largely eclipsed by behaviorism in the
United States, but at Cambridge’s Applied Psychology Unit directed
by Frederic Bartlett, cognitive modeling was able to flourish.
 The Nature of Explanation by Bartlett’s student and successor
Kenneth Craik reestablished the legitimacy of such mental terms as
beliefs and goals, arguing that they are just as scientific as using
pressure and temperature to talk about gases.
 Craik specified the three key steps of a knowledge-based agent.

Artificial Intelligence
36
 The Foundations of Artificial Intelligence
1. The stimulus must be translated into an internal representation.
2. The representation is manipulated by cognitive processes to derive
new internal representations.
3. These are in turn retranslated back into action.
 If the organism carries a small-scale model of external reality and its own
actions within its head, it is able to try out various alternatives, conclude
which is the best of them, react to future situations before they arise, use
the knowledge of past events to deal with the present and future, and in
every way to react in a much fuller, safer, and more competent manner to
the emergencies faced by it.
 After Craik’s death, his work was continued by Donald Broadbent, whose
book on Perception and Communication was one of the first works to
model psychological phenomena as information processing.

Artificial Intelligence
37
 The Foundations of Artificial Intelligence
 In the United States, the development of computer modeling led to the
creation of the field of cognitive science.
 This field was started at a workshop in September 1956 at MIT, just two
months after the conference where AI itself was born.
 At the workshop, George Miller presented The Magic Number Seven, Noam
Chomsky presented Three Models of Language, and Allen Newell and
Herbert Simon presented The Logic Theory Machine.
 These three papers showed how computer models could be used to address
the psychology of memory, language, and logical thinking.
 Now, it is a common view among psychologists that “a cognitive theory
should be like a computer program”.
 That is, it should describe a detailed information processing mechanism so
that some cognitive function can be implemented.

Artificial Intelligence
38
 The Foundations of Artificial Intelligence
Computer engineering:
1. How can we build an efficient computer?
 For AI to succeed, we need two things: intelligence and an artifact.
 The computer has been the artifact of choice.
 The modern digital electronic computer was invented independently and almost
simultaneously by scientists in three countries embattled in World War II.
 The first operational computer was the electromechanical built in 1940 by Alan
Turing’s team for a purpose of deciphering German messages.
 In 1943, the same group developed the Colossus, a powerful general-purpose
machine based on vacuum tubes.
 The first operational programmable computer was the Z-3 invented by Konrad
Zuse in Germany in 1941.
Artificial Intelligence
39
 The Foundations of Artificial Intelligence
 Zuse also invented floating-point numbers and the first high-level programming
language called Plankalkul.
 The first electronic computer, the ABC was assembled by John Atanasoff and
his student Clifford Berry between 1940 and 1942 at Iowa State University.
 The ENIAC, developed as part of a secret military project at the University of
Pennsylvania by a team including John Mauchly and John Eckert that proved
to be the most influential forerunner of modern computers.
 Since then, each generation of computer hardware has brought an increase in
speed and capacity and a decrease in price.
 Performance doubled every 18 months or so until around 2005, when power
dissipation problems led manufacturers to start multiplying the number of CPU
cores rather than the clock speed.
 Current expectations are that future increases in power will come from massive
parallelism, which is a curious convergence with the properties of the brain.
Artificial Intelligence
40
 The Foundations of Artificial Intelligence
 The first programmable machine was a loom, devised in 1805 by
Joseph Marie Jacquard, that used punched cards to store instructions
for the pattern to be woven.
 In the mid 19th century, Charles Babbage designed two machines,
neither of which he completed.
 The Difference Engine was intended to compute mathematical tables
for engineering and scientific projects.
 It was finally built and shown to work in 1991 at the Science Museum in
London.
 Babbage’s Analytical Engine was more ambitious which included
addressable memory, stored programs, and conditional jumps and was
the first artifact capable of universal computation.
 Babbage’s colleague Ada Lovelace, daughter of the poet Lord Byron,
was perhaps the world’s first programmer.
Artificial Intelligence
41
 The Foundations of Artificial Intelligence
 The programming language “Ada” is named after her.
 She wrote programs for the unfinished Analytical Engine and even
speculated that the machine could play chess or compose music.
 AI also owes a debt to the software side of computer science, which has
supplied the operating systems, programming languages, and tools needed
to write modern programs.
 But this is one area where the debt has been repaid.
 Work in AI has pioneered many ideas that have made their way back to
mainstream computer science including time sharing, interactive
interpreters, personal computers with windows and mice, rapid
development environments, the linked list data type, automatic storage
management, and key concepts of symbolic, functional, declarative, and
object-oriented programming.

Artificial Intelligence
42
 The Foundations of Artificial Intelligence
Control theory and cybernetics:
1. How can artifacts operate under their own control?
 Ktesibios of Alexandria built the first self-controlling machine, a water clock
with a regulator that maintained a constant flow rate.
 This invention changed the definition of what an artifact could do.
 Previously, only living things could modify their behavior in response to
changes in the environment.
 Other examples of self-regulating feedback control systems include the
steam engine governor, created by James Watt and the thermostat,
invented by Cornelis Drebbel who also invented the submarine.
 The mathematical theory of stable feedback systems was developed in the
19th century.

Artificial Intelligence
43
 The Foundations of Artificial Intelligence
 Norbert Wiener was the important person to create the control theory.
 Wiener was a brilliant mathematician who worked with Bertrand Russell among
others, before developing an interest in biological and mechanical control
systems and their connection to cognition.
 Like Craik, Wiener and his colleagues Arturo Rosenblueth and Julian Bigelow
challenged the behaviorist orthodoxy.
 They viewed purposive behavior as arising from a regulatory mechanism trying
to minimize error which is the difference between current state and goal state.
 In the late 1940s, Wiener along with Warren McCulloch, Walter Pitts, and John
von Neumann, organized a series of influential conferences that explored the
new mathematical and computational models of cognition.
 Wiener’s book Cybernetics became a bestseller and awoke the public to the
possibility of artificially intelligent machines.

Artificial Intelligence
44
 The Foundations of Artificial Intelligence
Linguistics:
1. How does language relate to thought?
 In 1957, B. F. Skinner published Verbal Behavior.
 This was a comprehensive detailed account of the behaviorist approach to
language learning, written by the foremost experts in the field.
 But a review of the book became as well known as the book itself and served
to almost kill off interest in behaviorism.
 The author of the review was the linguist Noam Chomsky, who had published a
book on Syntactic Structures.
 Chomsky pointed out that the behaviorist theory did not address the notion of
creativity in language, it did not explain how a child could understand and make
up sentences that he or she had never heard before.

Artificial Intelligence
45
 The Foundations of Artificial Intelligence
 Chomsky’s theory based on syntactic models going back to the Indian linguist
Panini could explain this and unlike previous theories, it was formal enough
that it could be programmed in principle .
 Modern linguistics and AI were born at about the same time, and grew up
together, intersecting in a hybrid field called computational linguistics or
Natural Language Processing(NLP).
 The problem of understanding language soon turned out to be more complex
than it seemed in 1957.
 Understanding language requires an understanding of the subject matter and
context, not just an understanding of the structure of sentences.
 This might seem obvious, but it was not widely appreciated until the 1960s.
 Much of the early work in knowledge representation (the study of how to put
knowledge into a form that a computer can reason with) was tied to language
and informed by research in linguistics.
Artificial Intelligence
46
 The History of Artificial Intelligence
The gestation of artificial intelligence (1943-1955):
 The first work that is recognized as AI was done by Warren McCulloch and
Walter Pitts (1943).
 They got the information from three sources such as knowledge of the basic
physiology and function of neurons in the brain, a formal analysis of
propositional logic, and Turing’s theory of computation.
 They proposed a model of artificial neurons in which each neuron is
characterized as being “on” or “off,” with a switch to “on” occurring in
response to stimulation by several neighboring neurons.
 The state of a neuron was conceived of as “factually equivalent to a
proposition which proposed its adequate stimulus.”
 They showed that any computable function could be computed by some
network of connected neurons, and that all the logical connectives (and, or,
not, etc.) could be implemented by simple net structures.
Artificial Intelligence
47
 The History of Artificial Intelligence
 McCulloch and Pitts also suggested that suitably defined networks could learn.
 Donald Hebb demonstrated a simple updating rule for modifying the connection
strengths between neurons.
 His rule now called as Hebbian learning, remains an influential model till today.
 Two undergraduate students at Harvard, Marvin Minsky and Dean Edmonds,
built the first neural network computer (SNARC - Stochastic Neural Analog
Reinforcement Calculator) in 1950.
 This computer used 3000 vacuum tubes and a surplus automatic pilot
mechanism from a B-24 bomber to simulate a network of 40 neurons.
 Later at Princeton, Minsky studied universal computation in neural networks.
 His Ph.D. committee was skeptical(doubtful) about whether this kind of
work should be considered mathematics, but von Neumann reportedly
said, “If it isn’t now, it will be someday.”
Artificial Intelligence
48
 The History of Artificial Intelligence
 Minsky was later to prove influential theorems showing the limitations
of neural network research.
 There were a number of early examples of work that can be
characterized as AI, but Alan Turing’s vision was the most influential.
 He gave lectures on the topic earier to 1947 at the London
Mathematical Society and articulated a persuasive agenda in his 1950
article “Computing Machinery and Intelligence.”
 There, he introduced the Turing Test, machine learning, genetic
algorithms, and reinforcement learning.
 He proposed the Child Programme idea, explaining “Instead of trying
to produce a programme to simulate the adult mind, why not try to
produce one which simulated the child’s?”

Artificial Intelligence
49
 The History of Artificial Intelligence
The birth of Artificial Intelligence (1956):
 John McCarthy after receiving his Ph D there in 1951, and working for two
years as an instructor, moved to Stanford and then to Dartmouth College,
which was to become the official birthplace of the field Artificial Intelligence.
 McCarthy convinced Minsky, Claude Shannon, and Nathaniel Rochester to
help him bring together U.S. researchers interested in automata theory,
neural nets, and the study of intelligence.
 They organized a two-month workshop at Dartmouth in 1956.
 They proposed that a 2 month, 10 man study of artificial intelligence be
carried out during 1956 at Dartmouth College in Hanover, New Hampshire.
 The study is to proceed on the basis of 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.

Artificial Intelligence
50
 The History of Artificial Intelligence
 An attempt was made to find how to make machines use language, form
abstractions and concepts, solve kinds of problems now reserved for
humans, and improve themselves.
 There were 10 attendees in all, including Trenchard More from Princeton,
Arthur Samuel from IBM, and Ray Solomonoff and Oliver Selfridge from MIT.
 Two researchers from Carnegie Tech, Allen Newell and Herbert Simon stole
the show.
 Although the others had ideas and in some cases programs for particular
applications such as checkers, Newell and Simon already had a reasoning
program, the Logic Theorist (LT), about which Simon claimed, “We have
invented a computer program capable of thinking non-numerically, and
thereby solved the venerable mind-body problem.”
 Soon after the workshop, the program was able to prove most of the
theorems appear Russel’s book in Principia Mathematica.
Artificial Intelligence
51
 The History of Artificial Intelligence
 Russell was delighted when Simon showed him that the program had come
up with a proof for one theorem that was shorter than the one in Principia.
 The editors of the Journal of Symbolic Logic were less impressed, and they
rejected a paper coauthored by Newell, Simon, and Logic Theorist.
 The Dartmouth workshop did not lead to any new breakthroughs, but it did
introduce all the major people to each other.
 For the next 20 years, the field would be dominated by these people and
their students and colleagues at MIT, CMU, Stanford, and IBM.
 Looking at the proposal for the Dartmouth workshop, we can see why it
was necessary for AI to become a separate field.
 Why couldn’t all the work done in AI have taken place under the name of
control theory or operations research or decision theory, which have
objectives similar to those of AI?
Artificial Intelligence
52
 The History of Artificial Intelligence
 Why isn’t AI a branch of mathematics?
 The first answer is that AI from the start embraced the idea of
duplicating human faculties such as creativity, self-improvement, and
language use.
 None of the other fields were addressing these issues.
 The second answer is methodology.
 AI is the only one of these fields that is clearly a branch of computer
science (although operations research does share an emphasis on
computer simulations), and AI is the only field to attempt to build
machines that will function autonomously in complex, changing
environments.

Artificial Intelligence
53
 The History of Artificial Intelligence
Early enthusiasm, great expectations (1952-1969):
 The early years of AI were full of successes in a limited way.
 Given the primitive computers and programming tools of the time and
the fact that only a few years earlier computers were seen as things
that could do arithmetic and no more, it was astonishing whenever a
computer did anything remotely clever.
 The intellectual establishment preferred to believe that “a machine can
never do X.”
 Newell and Simon’s early success was followed up with the General
Problem Solver (GPS).
 Unlike Logic Theorist, this program was designed from the start to
imitate human problem-solving protocols.

Artificial Intelligence
54
 The History of Artificial Intelligence
 Within the limited class of puzzles it could handle, it turned out that the
order in which the program considered subgoals and possible actions
was similar to that in which humans approached the same problems.
 Hence, GPS was probably the first program to embody the thinking
humanly approach.
 The success of GPS and subsequent programs as models of cognition
led Newell and Simon to formulate the famous physical symbol system
hypothesis, which states that “a physical symbol system has the
necessary and sufficient means for general intelligent action.”
 What they learnt is that any system (human or machine) exhibiting
intelligence must operate by manipulating data structures composed of
symbols.
 At IBM, Nathaniel Rochester and his colleagues produced some of the
first AI programs.
Artificial Intelligence
55
 The History of Artificial Intelligence
 Starting in 1952, Arthur Samuel wrote a series of programs for checkers
(draughts) that finally learned to play at a strong amateur level.
 Also, he disproved the idea that computers can do only what they are told to,
but his program quickly learned to play a better game than its creator.
 The program was demonstrated on television in February 1956, creating a
strong impression.
 Like Turing, Samuel had trouble finding computer time.
 Working at night, he used machines that were still on the testing floor at
IBM’s manufacturing plant.
 John McCarthy moved from Dartmouth to MIT and there he made three
crucial contributions in one historic year(1958).
 In MIT AI Lab, McCarthy defined the high-level language “Lisp” which
became the dominant AI programming language for the next 30 years.
Artificial Intelligence
56
 The History of Artificial Intelligence
 With Lisp, McCarthy had the tool he needed, but access to scarce and
expensive computing resources was also a serious problem.
 In response, he and others at MIT invented time sharing.
 Also in 1958, McCarthy published a paper entitled Programs with Common
Sense, in which he described the Advice Taker, a hypothetical program that
can be seen as the first complete AI system.
 The Advice Taker embodied the central principles of knowledge
representation and reasoning that it is useful to have a formal, explicit
representation of the world and its workings and to be able to manipulate that
representation with deductive processes.
 James Slagle’s SAINT program was able to solve closed-form calculus
integration problems.
 Tom Evans’s ANALOGY program solved geometric analogy problems that
appear in IQ tests.
Artificial Intelligence
57
 The History of Artificial Intelligence
 Daniel Bobrow’s STUDENT program solved algebra story problems such as
the following:
 If
the number of customers Tom gets is twice the square of 20 percent of the
number of advertisements he runs, and the number of advertisements he runs is
45, then what is the number of customers Tom gets?

 The most famous microworld was the blocks world, which consists of a set of
solid blocks placed on a tabletop, as shown in Figure 1.4.
 A typical task in this blocks world is to rearrange the blocks in a certain way,
using a robot hand that can pick up one block at a time.
 SHRDLU, An early natural-language understanding computer program
developed by Terry Winograd at MIT in 1968-1970 is used to do this task.
 The blocks world was home to the vision project of David Huffman, the vision
and constraint-propagation work of David Waltz, the learning theory of Patrick
Winston, the natural-language-understanding program of Terry Winograd, and
the planner of Scott Fahlman.
Artificial Intelligence
58
The History of Artificial Intelligence

Artificial Intelligence
59
 The History of Artificial Intelligence

 The name SHRDLU comes from ETAOIN SHRDLU, which is the order in
which the letter keys on a Linotype machine are set up, based on how
often they are in English.
Artificial Intelligence
60
 The History of Artificial Intelligence
A dose of reality (1966–1973):
 From the beginning, AI researchers were not shy about making predictions
of their coming successes.
 Herbert Simon in 1957 is quotes:
 It is not my aim to surprise or shock you, but the simplest way I can summarize
is to say that, there are machines in the world that can think, learn and create.
 Their ability to do these things is going to increase rapidly until in a visible future,
the range of problems they can handle will be coextensive with the range to
which the human mind has been applied.

 Simon also made more concrete predictions that, within 10 years a

computer would be chess champion, and a significant mathematical
theorem would be proved by machine.
 These predictions came true (or approximately true) within 40 years rather
than 10 years.
Artificial Intelligence
61
 The History of Artificial Intelligence
 Simon’s overconfidence was due to the promising performance of early
AI systems on simple examples.
 But, in most of the cases these early systems turned out to fail
miserably when they are tried out on the problems that are large and
more difficult problems.
 The first kind of difficulty arose because most early programs knew
nothing of their subject matter and they were succeeded by means of
simple syntactic manipulations.
 Initially, it was thought that simple syntactic transformations based on
the grammars of Russian and English, and word replacement from an
electronic dictionary would be sufficient to preserve the exact meaning
of sentences.
 The fact is that accurate translation requires background knowledge in
order to resolve ambiguity and establish the content of the sentence.
Artificial Intelligence
62
 The History of Artificial Intelligence
 The second kind of difficulty was the intractability of many of the problems that
AI was attempting to solve.
 Most of the early AI programs solved the problems by trying with different
combinations of steps until the solution was found.
 This strategy worked initially because microworlds contained very few objects
and hence very few possible actions and very short solution sequences.
 Before the theory of computational complexity was developed, it was thought
that scaling up to the larger problems was simply a matter of faster hardware
and larger memories.
 The development of resolution theorem proving was dampened when the
researchers failed to prove theorems involving more than a few dozen facts.
 The fact that a program can find a solution in principle does not mean that the
program contains any of the mechanisms needed to find it.

Artificial Intelligence
63
 The History of Artificial Intelligence
 The illusion of unlimited computational power was not confined to problem-
solving programs.
• Early experiments in machine evolution (now called genetic algorithms)
were based on the undoubtedly correct belief that by making an
appropriate series of small mutations (transformations) to a machine-code
program, one can generate a program with good performance for any task.
 Then the idea was to try random mutations with a selection process to
preserve the mutations that are useful.
 Even after spending thousands of hours of CPU time, almost no progress
was demonstrated.
 Modern genetic algorithms use better representations and have shown
more success.

Artificial Intelligence
64
 The History of Artificial Intelligence
 A third difficulty arose because of some limitations on the basic structures
used to generate the intelligent behavior.
 For example, Minsky and Papert’s book Perceptrons, proved that
although perceptrons (a simple form of neural network) could be shown
to learn anything they were capable of representing, they could represent
very little.
 In particular, a two-input perceptron could not be trained to recognize
when its two inputs were different.
 Although their results did not apply to more complex, multilayer networks,
research funding for neural-net research soon decreased to almost nil.
 Also, the new back-propagation learning algorithms for multilayer
networks that were to cause a resurgence in neural-net research were
discovered first in 1969.

Artificial Intelligence
65
 The History of Artificial Intelligence
Knowledge-based systems: The key to power? (1969-1979):
 The picture of problem solving that had arisen during the first decade of
AI research was of a general-purpose search mechanism trying to use
elementary reasoning steps to find complete solutions.
 These approaches were called as weak methods because, they do not
scale up to large or difficult problem instances.
 The alternative to weak methods is to use more powerful, domain-
specific knowledge that allows larger reasoning steps and can easily
handle typically occurring cases in narrow areas of expertise.
 One can say that to solve a hard problem, you must know the answer
already.
 The DENDRAL program was an early example of this approach.
 DENDRAL helps chemists identify unknown organic molecules.
Artificial Intelligence
66
 The History of Artificial Intelligence
 It was developed at Stanford, where Ed Feigenbaum, Bruce Buchanan,
and Joshua Lederberg to solve the problem of inferring molecular structure
from the information provided by a mass spectrometer.
 The input to the program consists of the elementary formula of the
molecule (e.g., C6H13NO2) and the mass spectrum giving the masses of the
various fragments of the molecule generated when it is bombarded by an
electron beam.
 For example, the mass spectrum might contain a peak at m = 15,
corresponding to the mass of a methyl (CH3) fragment.
 The naive version of the program generated all possible structures
consistent with the formula, and then predicted what mass spectrum would
be observed for each, comparing this with the actual spectrum.
 This was intractable for even moderate-sized molecules.

Artificial Intelligence
67
 The History of Artificial Intelligence
 The DENDRAL researchers consulted analytical chemists and found that they
worked by looking for well-known patterns of peaks in the spectrum that
suggested common substructures in the molecule.
 For example, the following rule is used to recognize a ketone (C=O) subgroup
(which weighs 28):
if there are two peaks at x1 and x2 such that
a) x1 + x2 = M + 28 (M is the mass of the whole molecule);
b) x1 − 28 is a high peak;
c) x2 − 28 is a high peak;
d) At least one of x1 and x2 is high.
then there is a ketone subgroup
 Recognizing that the molecule contains a particular substructure reduces the
number of possible candidates enormously.

Artificial Intelligence
68
 The History of Artificial Intelligence
 DENDRAL was powerful because,
• All the relevant theoretical knowledge to solve these problems has been
mapped from its general form to efficient special forms.

 The significance of DENDRAL was that it was the first successful

knowledge-intensive system and its expertise derived from large
numbers of special-purpose rules.
 Later systems also incorporated the main theme of McCarthy’s Advice
Taker approach, the clean separation of the knowledge (in the form of
rules) from the reasoning component.
 With this knowledge, Feigenbaum and Co at Stanford began the
Heuristic Programming Project (HPP) to investigate the extent to which
the new methodology of expert systems could be applied to other areas
of human expertise.

Artificial Intelligence
69
 The History of Artificial Intelligence
 The next major effort was in the area of medical diagnosis.
 Feigenbaum, Buchanan, and Dr. Edward Shortliffe developed MYCIN to diagnose
blood infections.
 With about 450 rules, MYCIN was able to perform well like some experts, and
considerably better than junior doctors.
 It has two major differences from DENDRAL.
 First, unlike the DENDRAL rules, no general theoretical model is existed from
which the MYCIN rules could be deduced.
 They are to be acquired from interviewing the experts, who in turn acquired them
from textbooks, other experts, and direct experience of cases.
 Second, the rules had to reflect the uncertainty associated with medical knowledge.
 MYCIN incorporated a calculus of uncertainty called certainty factors, which fit well
with how doctors assessed the impact of evidence on the diagnosis.
Artificial Intelligence
70
 The History of Artificial Intelligence
 The growth of applications to real-world problems caused increase in
the demands for workable knowledge representation schemes.
 A large number of different representation and reasoning languages
were developed.
 Some of these were based on logic, for example, the Prolog language
became popular in Europe, and the PLANNER in the United States.
 By following Minsky’s idea of frames, others have adopted a more
structured approach, assembling facts about particular object and
event types and arranging the types into a large taxonomic hierarchy
analogous to a biological taxonomy.

Artificial Intelligence
71
 The History of Artificial Intelligence
AI becomes an industry (1980-present):
 The first successful commercial expert system, called R1/XCON started its
operation at the Digital Equipment Corporation in 1982.
 An expert system is a computer program that uses artificial intelligence (AI)
technologies to simulate the judgment and behavior of a human or an
organization that has expertise and experience in a particular field.
 Expert systems are intended to complement, not replace the human experts.
 R1/XCON automatically selects and orders for computer components and / or
new computer systems based on customer specifications.
 By using R1/XCON the company was saving an estimated $40 million a year.
 DEC’s AI group had deployed 40 expert systems with more on the way by 1988.
 DuPont had 100 in use and 500 in development, saving about $10 million a year.
Artificial Intelligence
72
 The History of Artificial Intelligence
 Every major U.S. corporation had its own AI group and was either using or
investigating expert systems.
 In 1981, the Japanese announced the fifth-generation project, a 10-year plan to
build intelligent computers running Prolog.
 In response, the United States formed the Microelectronics and Computer
Technology Corporation (MCC) as a research consortium to assure national
competitiveness.
 In both cases, AI was part of a broad effort, including chip design and human-
interface research.
 In Britain, the Alvey report reinstated the funding that was cut earlier.
 But, in all these three countries, the projects never met their ambitious goals.
 The AI industry boomed from a few million dollars in 1980 to billions of dollars in
1988, including hundreds of companies building expert systems, vision systems,
robots, and software and hardware specialized for these purposes.
Artificial Intelligence
73
 The History of Artificial Intelligence
The return of Neural Networks (1986-present):
 In the mid-1980s at least four different groups reinvented the back-
propagation learning algorithm that was first found in 1969 by Bryson
and Ho.
 The algorithm was applied to many learning problems in computer
science and psychology, and the dissemination of the results in the
collection Parallel Distributed Processing caused great excitement.
 These connectionist models of intelligent systems were seen by some
as direct competitors both to the symbolic models promoted by Newell
and Simon and to the logicist approach of McCarthy and Co.
 It might seem obvious that at some level, humans manipulate the
symbols.

Artificial Intelligence
74
 The History of Artificial Intelligence
 Terrence Deacon’s book The Symbolic Species suggests that this is the
defining characteristic of humans, but the most ardent (enthusiastic)
connectionists questioned whether symbol manipulation had any real
explanatory role in the detailed models of cognition.
 This question remains unanswered, but the current view is that
connectionist and symbolic approaches are complementary to each other,
but not competing.
 As seen with the separation of AI and cognitive science, modern neural
network research has bifurcated into two fields, one concerned with
creating effective network architectures and algorithms and understanding
their mathematical properties, the other concerned with modeling of the
empirical properties of actual neurons and ensembles (set) of neurons.

Artificial Intelligence
75
 The History of Artificial Intelligence
AI adopts the scientific method (1987-present):
 Recent years have seen a revolution in both the content and the methodology
of work in artificial intelligence.
 It is more common to build on existing theories than to propose brand new
ones, to base claims on rigorous theorems or hard experimental evidence
rather than on intuition, and to show relevance to real-world applications rather
than toy examples.
 AI was founded in part as a rebellion against the limitations of existing fields
like control theory and statistics, but now it is embracing those fields.
 In terms of methodology, AI has come firmly under the scientific method.
 To be accepted, hypotheses must be subjected to rigorous empirical
experiments, and the results must be analyzed statistically for their importance.
 It is now possible to replicate experiments by using shared repositories of test
data and code.
Artificial Intelligence
76
 The History of Artificial Intelligence
 The field of speech recognition illustrates the pattern.
 In 1970s, variety of different architectures and approaches were tried.
 Many of these were ad hoc and fragile, and they were demonstrated on
only a few selected examples.
 In recent years, approaches based on Hidden Markov Models (HMM)
have come to dominate the area.
 Two aspects of HMMs are relevant.
 First, they are based on a rigorous mathematical theory.
 This has allowed speech researchers to build on several decades of
mathematical results developed in other fields.
 Second, they are generated by a process of training on a large collection
of real speech data.
Artificial Intelligence
77
 The History of Artificial Intelligence
 This ensures that the performance is robust, and in rigorous blind tests the
HMMs have been improving their scores steadily.
 Speech technology and the related field of handwritten character recognition
are already making the transition to industrial and consumer applications.
 Note that there is no scientific claim that humans use HMMs to recognize
speech but, HMMs provide a mathematical framework for understanding the
problem and support the engineering claim that they work well in practice.
 Machine translation follows the same course as speech recognition.
 In the 1950s there was initial enthusiasm for an approach based on
sequences of words, with models learned according to the principles of
information theory.
 That approach fell out of favor in the 1960s, but returned in the late 1990s
and now dominates the field.

Artificial Intelligence
78
 The History of Artificial Intelligence
 Much of the work on neural nets in the 1980s was done in an attempt to scope
out what could be done and to learn how neural nets differ from traditional
techniques.
 Using improved methodology and theoretical frameworks, the field arrived at
an understanding in which neural nets can now be compared with the
techniques from statistics, pattern recognition, and machine learning, and the
most promising technique can be applied to each application.
 As a result, data mining technology has spawned a vigorous new industry.
 The Bayesian network formalism was invented to allow efficient representation
and rigorous reasoning with uncertain knowledge.
 This approach largely overcomes many problems of the probabilistic reasoning
systems of the 1960s and 1970s, now it dominates AI research on uncertain
reasoning and expert systems.

Artificial Intelligence
79
 The History of Artificial Intelligence
The emergence of intelligent agents (1995-present):
 Encouraged by the progress in solving the subproblems of AI, researchers have
also started to look at the whole agent problem again.
 The work of Allen Newell, John Laird, and Paul Rosenbloom on SOAR (Security
Orchestration, Automation, and Response) is the best known example of a
complete agent architecture.
 One of the most important environments for intelligent agents is the Internet.
 AI systems have become so common in Web-based applications that the “-bot”
suffix has entered everyday language.
 Moreover, AI technologies underlie many Internet tools such as search engines,
recommender systems, and Web site aggregators(Curator, Google News).
 One consequence of trying to build complete agents is the realization that the
previously isolated subfields of AI might need to be reorganized somewhat when
their results are to be tied together.
Artificial Intelligence
80
 The History of Artificial Intelligence
 It is now widely appreciated that sensory systems (vision, sonar,
speech recognition, etc.) cannot deliver perfectly reliable information
about the environment.
 Hence, reasoning and planning systems must be able to handle
uncertainty.
 A second consequence of the agent perspective is that AI has been
drawn into much closer contact with other fields, such as control
theory and economics, that also deal with agents.
 Recent progress in the control of robotic cars has derived from a
mixture of approaches ranging from better sensors, control-theoretic
integration of sensing, localization and mapping, as well as a degree
of high-level planning.
 Despite these successes, some influential founders of AI, have
expressed discontent with the progress of AI.
Artificial Intelligence
81
 The History of Artificial Intelligence
 They think that AI should put less emphasis on creating ever-improved
versions of applications that are good at a specific task such as driving
a car, playing chess, or recognizing speech.
 Instead, they believe AI should return to its roots of striving for, in
Simon’s words “machines that think, that learn and that create.”
 A related idea is the subfield of Artificial General Intelligence or AGI
which held its first conference and organized the Journal of Artificial
General Intelligence in 2008.
 AGI looks for a universal algorithm for learning and acting in any
environment and has its roots in the work of Ray Solomonoff, who was
one of the member of Dartmouth conference held in1956.
 Guaranteeing that what we create is really Friendly AI is also a concern.

Artificial Intelligence
82
 The History of Artificial Intelligence
The availability of very large data sets (2001–present):
 Throughout the 60-year history of computer science, the emphasis has been
on the algorithm as the main subject of study.
 But some recent work in AI suggests that for many problems, it makes more
sense to worry about the data and be less picky about what algorithm to apply.
 This is true because of the increasing availability of very large data source, for
example, trillions of words of English and billions of images from the Web or
billions of base pairs of genomic sequences.(Used to decipher the genetic
material found in an organism or virus)
 One influential paper in this line was Yarowsky’s work on word-sense
disambiguation.
 For example, given the use of the word plant in a sentence, does that refer to
flora or factory?

Artificial Intelligence
83
 The History of Artificial Intelligence
 Previous approaches to the problem had relied on human-labeled examples
combined with machine learning algorithms.
 Yarowsky showed that the task can be done with accuracy above 96% with
no labeled examples at all.
 Instead, given a very large corpus of unannotated text and just the
dictionary definitions of the two senses, “works, industrial plant” and “flora,
plant life”, one can label examples in the corpus, and from there bootstrap
to learn new patterns that help label new examples.
 Banko and Brill show that techniques like this perform even better as the
amount of available text goes from a million words to a billion and that the
increase in performance from using more data exceeds any difference in
algorithm choice, a mediocre algorithm with 100 million words of unlabeled
training data outperforms the best-known algorithm with 1 million words.

Artificial Intelligence
84
 The History of Artificial Intelligence
 As another example, Hays and Efros discuss the problem of filling in holes
in a photograph.
 Suppose you use Photoshop to mask out an ex-friend from a group photo,
but now you need to fill in the masked area with something that matches
the background.
 Hays and Efros defined an algorithm that searches through a collection of
photos to find something that will match.
 They found the performance of their algorithm was poor when they used a
collection of only ten thousand photos, but crossed a threshold into
excellent performance when they grew the collection to two million photos.
 Work like this suggests that the knowledge bottleneck in AI, the problem of
how to express all the knowledge that a system needs.

Artificial Intelligence
85
 The History of Artificial Intelligence
 It may be solved in many applications by learning methods rather than
hand-coded knowledge engineering, provided the learning algorithms
have enough data to go on.
 Reporters have noticed the surge of new applications and have written
that AI Winter may be yielding to a new Spring.
 As Kurzweil writes, “today, many thousands of AI applications are
deeply embedded in the infrastructure of every industry.”

Artificial Intelligence
86