Knowledge representation SETIA

SHUBHAM RE3004B33

Artificial Intelligence (AI), a branch of computer science, is the study of intelligent systems (i.e. software, computers, robots, etc.). Alternatively, it may be defined as "the study and design of intelligent agents", where an intelligent agent is a system that perceives its environment and takes actions that maximize its chances of success. ohn !c"arthy, who coined the term in #$%%, defines it as "the science and engineering of making intelligent machines". A& research is highly technical and specialised, deeply divided into subfields that often fail to communicate with each other. 'ome of the division is due to social and cultural factors( subfields have grown up around particular institutions and the work of individual researchers. A& research is also divided by several technical issues. )here are subfields which are focused on the solution of specific problems, on one of several possible approaches, on the use of widely differing tools and towards the accomplishment of particularapplications. )he central problems of A& include such traits as reasoning, knowledge, planning, learning, communication, perception and the ability to move and manipulate ob*ects. +eneral intelligence (or "strong A&") is still among the field,s long term goals. "urrently popular approaches include statistical

methods, computational intelligence and traditional symbolic A&. )here are an enormous number of tools used in A&, including versions of search and mathematical optimization, logic, methods based on probability and economics, and many others. )he field was founded on the claim that a central property of humans, intelligence - the sapience of .omo sapiens - can be so precisely described that it can be simulated by a machine. )his raises philosophical issues about the nature of the mind and the ethics of creating artificial beings, issues which have been addressed by myth, fiction andphilosophy since anti/uity. Artificial intelligence has been the sub*ect of tremendous optimism but has also suffered stunning setbacks, )oday it has become an essential part of the technology industry, providing the heavy lifting for many of the most difficult problems in computer science. History )hinking machines and artificial beings appear in +reek myths, such as )alos of "rete, the bronze robot of .ephaestus, and 0ygmalion,s +alatea. .uman likenesses believed to have intelligence were built in every ma*or civilization( animated cult images were worshiped in 1gypt and +reece and

Term paper Page 1

Knowledge representation SETIA

SHUBHAM RE3004B33

humanoid automatons were built by 2an 'hi, .ero of Alexandria and Al3 azari. &t was also widely believed that artificial beings had been created by 4bir ibn .ayy4n, udah 5oew and 0aracelsus. 6y the #$th and 78th centuries, artificial beings had become a common feature in fiction, as in !ary 'helley,s 9rankenstein or :arel ;apek,s <.=.<. (<ossum,s =niversal <obots). 0amela !c"orduck argues that all of these are examples of an ancient urge, as she describes it, "to forge the gods". 'tories of these creatures and their fates discuss many of the same hopes, fears andethical concerns that are presented by artificial intelligence. !echanical or "formal" reasoning has been developed by philosophers and mathematicians since anti/uity. )he study of logic led directly to the invention of the programmable digital electronic computer, based on the work of mathematician Alan )uring and others. )uring,s theory of computation suggested that a machine, by shuffling symbols as simple as "8" and "#", could simulate any conceivable (imaginable) act of mathematical deduction. )his, along with concurrent discoveries in neurology, information theory andcybernetics, inspired a small group of researchers to begin to

seriously consider the possibility of building an electronic brain. )he field of A& research was founded at a conference on the campus of >artmouth "ollege in the summer of #$%?. )he attendees, including ohn !c"arthy, !arvin !insky, Allen @ewell and .erbert 'imon, became the leaders of A& research for many decades. )hey and their students wrote programs that were, to most people, simply astonishing( "omputers were solving word problems in algebra, proving logical theorems and speaking 1nglish. 6y the middle of the #$?8s, research in the =.'. was heavily funded by the>epartment of >efense and laboratories had been established around the world. A&,s founders were profoundly optimistic about the future of the new field( .erbert 'imonpredicted that "machines will be capable, within twenty years, of doing any work a man can do" and !arvin !insky agreed, writing that "within a generation ... the problem of creating ,artificial intelligence, will substantially be solved". )hey had failed to recognize the difficulty of some of the problems they faced. &n #$AB, in response to the criticism of 'ir ames 5ighthill and ongoing pressure from the =' "ongress to fund more productive pro*ects, both the =.'. and 6ritish governments cut off all undirected exploratory research in A&. )he next

Term paper Page 2

Knowledge representation SETIA

SHUBHAM RE3004B33

few years would later be called the "A& winter" because funding for pro*ects was hard to find. 1ven though some A& systems are pure applications, most of them are created to be components of complex applications in which they include a form of intelligence. &n the early #$C8s, A& research was revived by the commercial success of expert systems, a form of A& program that simulated the knowledge and analytical skills of one or more human experts. 6y #$C% the market for A& had reached over a billion dollars. At the same time, apan,s fifth generation computer pro*ect inspired the =.' and 6ritish governments to restore funding for academic research in the field. .owever, beginning with the collapse of the 5isp !achine market in #$CA, A& once again fell into disrepute, and a second, longer lasting A& winter began. &n the #$$8s and early 7#st century, A& achieved its greatest successes, albeit somewhat behind the scenes. Artificial intelligence is used for logistics, data mining, medical diagnosis and many other areas throughout the technology industry. )he success was due to several factors( the increasing computational power of computers (see !oore,s law), a greater emphasis on solving specific subproblems, the creation of new ties between A& and other fields working on similar problems, and a new commitment by researchers to solid mathematical

methods and standards.

rigorous

scientific

Dn ## !ay #$$A, >eep 6lue became the first computer chess3playing system to beat a reigning world chess champion, +arry :asparov. &n 788%, a 'tanford robot won the>A<0A +rand "hallenge by driving autonomously for #E# miles along an unrehearsed desert trail. )wo years later, a team from "!= won the >A<0A =rban "hallenge when their vehicle autonomously navigated %% miles in an =rban environment while adhering to traffic hazards and all traffic laws. &n 9ebruary 78##, in a eopardyF /uiz show exhibition match, &6!,s /uestion answering system, Gatson, defeated the two greatest eopardy champions, 6rad <utter and :en ennings, by a significant margin. )he :inect, which provides a E> bodyHmotion interface for the Ibox E?8, uses algorithms that emerged from lengthy A& research as does the i0hones,s 'iri. Problems )he general problem of simulating (or creating) intelligence has been broken down into a number of specific sub3 problems. )hese consist of particular traits or capabilities that researchers would like an intelligent system to display. )he traits described below have received the most attention.

Term paper Page 3

Knowledge representation SETIA

SHUBHAM RE3004B33

Deduction, reasoning, problem solving 1arly A& researchers developed algorithms that imitated the step3by3 step reasoning that humans use when they solve puzzles or make logical deductions. 6y the late #$C8s and ,$8s, A& research had also developed highly successful methods for dealing with uncertain or incomplete information, employing concepts from probability and economics. 9or difficult problems, most of these algorithms can re/uire enormous computational resources H most experience a "combinatorial explosion"( the amount of memory or computer time re/uired becomes astronomical when the problem goes beyond a certain size. )he search for more efficient problem3solving algorithms is a high priority for A& research. .uman beings solve most of their problems using fast, intuitive *udgements rather than the conscious, step3by3step deduction that early A& research was able to model. A& has made some progress at imitating this kind of "sub3symbolic" problem solving( embodied agent approaches emphasize the importance of sensorimotor skills to higher reasoningJneural net research attempts to simulate the structures inside the brain that give rise to this skillJ statistical approaches to A& mimic

the probabilistic nature of the human ability to guess. Knowledge representation :nowledge representation and knowledge engineering are central to A& research. !any of the problems machines are expected to solve will re/uire extensive knowledge about the world. Among the things that A& needs to represent are( ob*ects, properties, categories and relations between ob*ectsJ situations, events, states and timeJ causes and effectsJ knowledge about knowledge (what we know about what other people know)J and many other, less well researched domains. A representation of "what exists" is an ontology(borrowing a word from traditional philosophy), of which the most general are called upper ontologies. Among the most difficult problems in knowledge representation are( Default reasoning and the qualification problem !any of the things people know take the form of "working assumptions." 9or example, if a bird comes up in conversation, people typically picture an animal that is fist sized, sings, and flies. @one of these things are true about all birds. ohn !c"arthy identified this problem in #$?$ as the /ualification problem( for any commonsense

Term paper Page 4

Knowledge representation SETIA

SHUBHAM RE3004B33

rule that A& researchers care to represent, there tend to be a huge number of exceptions. Almost nothing is simply true or false in the way that abstract logic re/uires. A& research has explored a number of solutions to this problem. The breadth of commonsense nowledge )he number of atomic facts that the average person knows is astronomical. <esearch pro*ects that attempt to build a complete knowledge base of commonsense knowledge(e.g., "yc) re/uire enormous amounts of laborious ontological engineering - they must be built, by hand, one complicated concept at a time. A ma*or goal is to have the computer understand enough concepts to be able to learn by reading from sources like the internet, and thus be able to add to its own ontology. The subsymbolic form of some commonsense nowledge !uch of what people know is not represented as "facts" or "statements" that they could express verbally. 9or example, a chess master will avoid a

particular chess position because it "feels too exposed" or an art critic can take one look at a statue and instantly realize that it is a fake. )hese are intuitions or tendencies that are represented in the brain non3consciously and sub3symbolically. :nowledge like this informs, supports and provides a context for symbolic, conscious knowledge. As with the related problem of sub3 symbolic reasoning, it is hoped that situated A&, computational intelligence, or statistical A& will provide ways to represent this kind of knowledge Planning &ntelligent agents must be able to set goals and achieve them. )hey need a way to visualize the future (they must have a representation of the state of the world and be able to make predictions about how their actions will change it) and be able to make choices that maximize the utility (or "value") of the available choices. &n classical planning problems, the agent can assume that it is the only thing acting on the world and it can be certain what the conse/uences of its action may be. .owever, if the agent is not the only actor, it must periodically ascertain whether the world matches its predictions and it must change its plan as this becomes necessary,

Term paper Page 5

Knowledge representation SETIA

SHUBHAM RE3004B33

re/uiring the agent to reason under uncertainty. !ulti3agent planning uses the cooperation and competition of many agents to achieve a given goal. 1mergent behavior such as this is used by evolutionary algorithms and swarm intelligence. !earning !achine learning is the study of computer algorithms that improve automatically through experience and has been central to A& research since the field,s inception. =nsupervised learning is the ability to find patterns in a stream of input. 'upervised learning includes both classification and numerical regression. "lassification is used to determine what category something belongs in, after seeing a number of examples of things from several categories. <egression is the attempt to produce a function that describes the relationship between inputs and outputs and predicts how the outputs should change as the inputs change. &n reinforcement learning the agent is rewarded for good responses and punished for bad ones. )hese can be analyzed in terms of decision theory, using concepts like utility. )he mathematical analysis of machine learning algorithms and their performance is a branch of theoretical computer science known as computational learning theory

"atural language processing @atural language processing gives machines the ability to read and understand the languages that humans speak. A sufficiently powerful natural language processing system would enable natural language user interfaces and the ac/uisition of knowledge directly from human3written sources, such as &nternet texts. 'ome straightforward applications of natural language processing include information retrieval (or text mining) and machine translation. A common method of processing and extracting meaning from natural language is through semantic indexing. &ncreases in processing speeds and the drop in the cost of data storage makes indexing large volumes of abstractions of the users input much more efficient. #otion and manipulation )he field of robotics is closely related to A&. &ntelligence is re/uired for robots to be able to handle such tasks as ob*ect manipulation and navigation, with sub3problems of localization (knowing where you are, or finding out where other things are), mapping(learning what is around you, building a map of the environment), and motion planning (figuring out how to get there) or path planning (going from one point

Term paper Page 6

Knowledge representation SETIA

SHUBHAM RE3004B33

in space to another point, which may involve compliant motion 3 where the robot moves while maintaining physical contact with an ob*ect). $ocial intelligence Affective computing is the study and development of systems and devices that can recognize, interpret, process, and simulate humanaffects. &t is an interdisciplinary field spanning computer sciences, psychology, and cognitive science. Ghile the origins of the field may be traced as far back as to early philosophical in/uiries into emotion, the more modern branch of computer science originated with <osalind 0icard,s #$$% paper on affective computing. A motivation for the research is the ability to simulateempathy. )he machine should interpret the emotional state of humans and adapt its behaviour to them, giving an appropriate response for those emotions. 1motion and social skills play two roles for an intelligent agent. 9irst, it must be able to predict the actions of others, by understanding their motives and emotional states. ()his involves elements of game theory, decision theory, as well as the ability to model human emotions and the perceptual skills to detect emotions.) Also, in an effort to facilitate human3computer interaction, an intelligent machine

might want to be able to display emotions-even if it does not actually experience them itself-in order to appear sensitive to the emotional dynamics of human interaction. %reativity A sub3field of A& addresses creativity both theoretically (from a philosophical and psychological perspective) and practically (via specific implementations of systems that generate outputs that can be considered creative, or systems that identify and assess creativity). <elated areas of computational research are Artificial intuition and Artificial imagination. &eneral intelligence !ost researchers think that their work will eventually be incorporated into a machine with general intelligence (known as strong A&), combining all the skills above and exceeding human abilities at most or all of them. A few believe that anthropomorphic features like artificial consciousness or an artificial brain may be re/uired for such a pro*ect. !any of the problems above may re/uire general intelligence to be considered solved. 9or example, even a straightforward, specific task

Term paper Page 7

Knowledge representation SETIA

SHUBHAM RE3004B33

like machine translation re/uires that the machine read and write in both languages (@50), follow the au thor,s argument (reason), know what is being talked about (knowledge), and faithfully reproduce the author,s intention (social intelligence). A problem like machine translation is considered "A&3complete". &n order to solve this particular problem, you must solve all the problems. Approaches )here is no established unifying theory or paradigm that guides A& research. <esearchers disagree about many issues. A few of the most long standing /uestions that have remained unanswered are these( should artificial intelligence simulate natural intelligence by studying psychology or neurologyK Dr is human biology as irrelevant to A& research as bird biology is L to aeronautical engineeringK "an intelligent behavior be described using simple, elegant principles (such as logic or optimization)K Dr does it necessarily re/uire solving a large number of completely unrelated problemsK "an intelligence be reproduced using high3level symbols, similar to words and ideasK Dr does it re/uire "sub3symbolic" processingK ohn .augeland, who

coined the term +D9A& (+ood Dld3 9ashioned Artificial &ntelligence), also proposed that A& should more properly be referred to as synthetic intelligence, a term which has since been adopted by some non3+D9A& researchers. Probabilistic methods for uncertain reasoning !any problems in A& (in reasoning, planning, learning, perception and robotics) re/uire the agent to operate with incomplete or uncertain information. A& researchers have devised a number of powerful tools to solve these problems using methods from probability theory and economics. 6ayesian networks are a very general tool that can be used for a large number of problems( reasoning (using the 6ayesian inference algorithm), learning (using theexpectation3maximization algorithm), planning (using decision networks) and perception (using dyna mic 6ayesian networks). 0robabilistic algorithms can also be used for filtering, prediction, smoothing and finding explanations for streams of data, helping perception systems to analyze processes that occur over time (e.g., hidden !arkov modelsor :alman filters). A key concept from the science of economics is "utility"( a measure of how valuable something is to an intelligent agent. 0recise mathematical

Term paper Page 8

Knowledge representation SETIA

SHUBHAM RE3004B33

tools have been developed that analyze how an agent can make choices and plan, using decision theory, decision analysis, information value theory. )hese tools include models such as !arkov decision processes, dynamic decision networks, game theory and mechanism design.

0oole, !ackworth +oebel #$$C, pp. AH7#

5uger M 'tubblefield 788B, pp. 7E%H7B8

'eferences 1. ( >efinition of A& as the study

of intelligent agents( 0oole, !ackworth M +oebel #$$C, p. #, which provides the version that is used in this article. @ote that they use the term "computational intelligence" as a synonym for artificial intelligence.

)he definition used in this article, in terms of goals, actions, perception and environment, is due to <ussell M @orvig (788E). Dther definitions also include knowledge and learning as additional criteria.

3. ( Although

<ussell M @orvig (788E) (who prefer the term "rational agent") and write ")he whole3 agent view is now widely accepted in the field" (<ussell M @orvig 788E, p. %%). @ilsson #$$C
a b

controversy (see "revier p. %8)), !c"arthystates une/uivocally "& came up with the term" in a cOnet interview. ('killings 788?) !c"arthy first used the term in the proposal for the >artmouth conference, which appeared in #$%%. (!c"arthy et al. #$%%) 4. ( !c"arthy,s definition of A&(

there is on this

some point (#$$E,

!c"arthy 788A

2. N

)he intelligent agent paradigm(

5. ( 0amela !c"orduck

<ussell M @orvig 788E, pp. 7A, E7H%C, $?CH$A7

(788B, pp. B7B) writes of "the rough shattering of A& in subfieldsvision, natural language, decision theory, genetic algorithms, robotics ... and these with own sub3subfield-that

Term paper Page 9

Knowledge representation SETIA

SHUBHAM RE3004B33

would hardly have anything to say to each other."

6. N a b )his list of intelligent traits is

based on the topics covered by the ma*or A& textbooks, including(

<ussell M @orvig 788E 5uger M 'tubblefield 788B 0oole, !ackworth +oebel #$$C @ilsson #$$C M

Term paper Page 10