History of artificial intelligence

The history of artificial intelligence begins in antiquity with myths, stories and rumors of artificial beings endowed with intelligence and consciousness by master craftsmen. In the middle of the 20th century, a handful of scientists began to explore a new approach to this ancient idea based on their discoveries in neurology, a new mathematical theory of information, an understanding of control and stability called cybernetics and, above all, by the invention of the digital computer, a machine based on the abstract essence of mathematical reasoning.

The field of artificial intelligence research was born at a conference on the campus of Dartmouth College in the summer of 1956. Those who attended would become the leaders of AI research for many decades. Many of them predicted that a machine as intelligent as a human being would exist in no more than a generation and they were given millions of dollars to make this vision come true. Eventually it became obvious that they had grossly underestimated the difficulty of the project. In 1973, in response to the criticism of Sir James Lighthill and ongoing pressure from congress, the U.S. and British Governments stopped funding undirected research into artificial intelligence. Seven years later, the Japanese Government and American industry would provide AI with billions of dollars, but again the investors would be disappointed and by the late 80s the funding would dry up again. The cycle of boom and bust, of AI winters and summers, continues to the present day. Undaunted, there are those that make extraordinary predictions even now. [For example Harvtxt|Kurzweil|2005 argues that machines with human level intelligence will exist by 2029.]

Despite the rise and fall of AI in the perceptions of venture capitalists and government bureaucrats, AI has made continuous advances in all areas regardless of the climate, overcoming unexpected obstacles, reorienting itself to the light of new discoveries and riding the crest of the wave of increasing computer power. Progress has been slower than predicted but progress has continued nonetheless. Artificial intelligence problems that had begun to seem impossible in 1970 have been solved and the solutions are now used in successful commercial products.

It remains to be seen when or if an AI system will be built with a human level of intelligence. Alan Turing, in a famous 1950 paper, asked the question "can machines think?" and concluded: "We can only see a short distance ahead, but we can see plenty there that needs to be done."Harvnb|Turing|1950|p=460]

Precursors

Harvtxt|McCorduck|2004 writes "artificial intelligence in one form or another is an idea that has pervaded Western intellectual history, a dream in urgent need of being realized," expressed in humanity's myths, legends, stories, speculation and clockwork automatons. [Harvnb|McCorduck|2004|pp=5-35]

AI in myth, fiction and speculation

Mechanical men and artificial beings appear in Greek myths, such as the golden robots of Hephaestus and Pygmalion's Galatea. [Harvnb|McCorduck|2004|p=5,] In the Middle Ages, there were rumors of secret mystical or alchemical means of placing mind into matter, such as Geber's "Takwin", Paracelsus' homunculus and Rabbi Judah Loew's Golem. [Harvnb|McCorduck|2004|p=15-16,Harvnb|Buchanan|2005|p=50 (Judah Loew's Golem),Harvnb|McCorduck|2004|p=13-14 (Paracelsus),Harvnb|O'Connor|1994 (Geber's "Takwin")] By the 19th century, ideas about artificial men and thinking machines were developed in fiction, as in Mary Shelley's "Frankenstein" or Karel Čapek's "R.U.R. (Rossum's Universal Robots)", [] and speculation, such as Samuel Butler's "Darwin Among the Machines". []

Automatons

Realistic humanoid automatons were built by craftsman from every civilization, including Yan Shi, []
Hero of Alexandria, []
Al-Jazari [ [http://www.shef.ac.uk/marcoms/eview/articles58/robot.html A Thirteenth Century Programmable Robot] ] and Wolfgang von Kempelen. [] The oldest known automatons were the sacred statues of ancient Egypt and Greece. The faithful believed that craftsman had imbued these figures with very real minds, capable of wisdom and emotion—Hermes Trismegistus wrote that "by discovering the true nature of the gods, man has been able to reproduce it." [Quoted in Harvnb|McCorduck|2004|p=8. Harvnb|Crevier|1993|p=1 and Harvnb|McCorduck|2004|pp=6-9 discusses sacred statues.] [Other important automatons were built by Haroun al-Rashid Harv|McCorduck|2004|p=10, Jacques de Vaucanson Harv|McCorduck|2004|p=16 and Leonardo Torres y Quevedo Harv|McCorduck|2004|p=59-62]

Formal reasoning

In the 17th century, Thomas Hobbes, René Descartes and Gottfried Leibniz explored the possibility that all rational thought could be made as systematic as algebra or geometry. []
Hobbes famously wrote in "Leviathan": "reason is nothing but reckoning". []
Leibniz envisioned a universal language of reasoning (his "characteristica universalis") which would reduce argumentation to calculation, so that "there would be no more need of disputation between two philosophers than between two accountants. For it would suffice to take their pencils in hand, down to their slates, and to say each other (with a friend as witness, if they liked): "Let us calculate"." [] These philosophers had begun to articulate the physical symbol system hypothesis that would become the guiding faith of AI research.

Computer science

:Main|history of computer hardware|history of computer science

Calculating machines were built in antiquity and improved throughout history by many mathematicians, including (once again) philosopher Gottfried Leibniz. The first modern computers were the massive code breaking machines of the Second World War (such as Z3, ENIAC and Colossus). [Harvnb|McCorduck|2004|p=61-62, 64-66, Harvnb|Russell|Norvig|2003|p=14-15]

The birth of artificial intelligence 1943−1956

In the 1940s and 50s, a handful of scientists from a variety fields (mathematics, psychology, engineering, economics and political science) began to discuss the possibility of creating an artificial brain. The field of artificial intelligence research was founded as an academic discipline in 1956.

Cybernetics and early neural networks

Turing's test

In 1950 Alan Turing published a landmark paper in which he speculated about the possibility of creating machines with true intelligence. [Harvnb|McCorduck|2004|pp=70−72,Harvnb|Crevier|1993|p=22−25,Harvnb|Russell|Norvig|2003|pp=2−3 and 948,Harvnb|Haugeland|1985|pp=6−9.See also ] He noted that "intelligence" is difficult to define and devised his famous Turing Test. If a machine could carry on a conversation (over a teletype) that was indistinguishable from a conversation with a human being, then the machine could be called "intelligent." This simplified version of the problem allowed Turing to argue convincingly that a "thinking machine" was at least "plausible" and the paper answered all the most common objections to the proposition. [Harvtxt|Norvig|Russell|2003|p=948 claim that Turing answered all the major objections to AI that have been offered in the years since the paper appeared.] The Turing Test was the first serious proposal in the philosophy of artificial intelligence.

ymbolic reasoning and the Logic Theorist

When access to digital computers became possible in the middle fifties, a few scientists instinctively recognized that a machine that could manipulate numbers could also manipulate symbols and that the manipulation of symbols could well be the essence of human thought. This was a new approach to creating thinking machines. [Harvnb|McCorduck|2004|pp=137-170, Harvnb|Crevier|pp=44-47]

Dartmouth Conference 1956: the birth of AI

The Dartmouth Conference of 1956 [Harvnb|McCorduck|2004|pp=111-136,Harvnb|Crevier|1993|pp=49-51 and] was organized by Marvin Minsky, John McCarthy and two senior scientists: Claude Shannon and Nathan Rochester of IBM. The proposal for the conference included this assertion: "every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it" — a clear statement of the philosophical position of AI research. [See Harvnb|McCarthy|Minsky|Rochester|Shannon|1955. Also see Harvnb|Crevier|1993|p=48 where Crevier states " [the proposal] later became known as the 'physical symbol systems hypothesis'". The physical symbol system hypothesis was articulated and named by Newell and Simon in their paper on GPS. Harv|Newell|Simon|1963 It includes a more specific definition of a "machine" as an agent that manipulates symbols. See the philosophy of artificial intelligence.] The participants included Ray Solomonoff, Oliver Selfridge, Trenchard More, Arthur Samuel, Allen Newell and Herbert Simon, all of whom would create important programs during the first decades of AI research. [Harvtxt|McCorduck|2004|p=129-130 discusses how the Dartmouth conference alumni dominated the first two decades of AI research, calling them the "invisible college".] At the conference Newell and Simon debuted the "Logic Theorist" and McCarthy persuaded the attendees to accept "Artificial Intelligence" as the name of the field. ["I won't swear and I hadn't seen it before," McCarthy told Pamela McCorduck in 1979. Harv|McCorduck|2004|p=114 However, McCarthy also stated unequivocally "I came up with the term" in a CNET interview. Harv|Skilling|2006] The 1956 Dartmouth conference was the moment that AI gained its name, its mission, its first success and its major players, and is widely considered the birth of AI. [Harvtxt|Crevier|1993|pp=49 writes "the conference is generally recognized as the official birthdate of the new science."]

The golden years 1956−1974

The work

There were many successful programs and new directions in the late 50s and 1960s. Among the most influential were these:

; Reasoning as searchMany AI programs used the same basic algorithm in the early years of AI research: to achieve some goal (like winning a game or proving a theorem) and they proceeded step by step towards it (by making a move or a deduction) as if searching through a maze, backtracking whenever they reached a dead end. This paradigm was called "reasoning as search". [Means-ends analysis, reasoning as search: Harvnb|McCorduck|2004|p=247-248. Harvnb|Russell|Norvig|2003|pp=59−61]

The principal difficulty was that, for many problems, the number of possible paths through the "maze" was simply astronomical (this is called a "combinatorial explosion"). Researchers would reduce the search space by using heuristics or "rules of thumb" that would eliminate those paths that were unlikely to lead to a solution. [Heuristic: Harvnb|McCorduck|2004|p=246, Harvnb|Russell|Norvig|2003|pp=21−22]

A semantic net represents concepts (e.g. "house","door") as nodes and relations among concepts (e.g. "has-a") as links between the nodes. The first AI program to use a semantic net was written by Ross Quillian [Harvnb|Crevier|1993|pp=79−83] and the most successful (and controversial) version was Roger Schank's Conceptual Dependency. [Harvnb|Crevier|1993|pp=164−172]

Perhaps the most interesting English speaking computer program was Joseph Weizenbaum's ELIZA, the first chatterbot. ELIZA could carry out conversations that were so realistic that users occasionally were fooled into thinking they were communicating with a human being and not a program. But in fact, ELIZA had no idea what she was talking about. She simply gave a canned response or repeated back what was said to her, rephrasing her response with a few grammar rules. [Harvnb|McCorduck|2004|pp=291-296, Harvnb|Crevier|1993|pp=134−139]

; Micro-worldsIn the late 60s, Marvin Minsky and Seymour Papert of the MIT AI Laboratory proposed that AI research should focus on artificially simple situations known as Micro-Worlds. They pointed out that in successful sciences like physics, basic principles were often best understood using simplified models like frictionless planes or perfectly rigid bodies. Much of the research focused on the so-called "blocks world," which consists of colored blocks of various shapes and sizes arrayed on a flat surface. [Harvnb|McCorduck|2004|pp=299-305, Harvnb|Crevier|1993|pp=83−102, Harvnb|Russell|Norvig|2003|p=19 and see also [http://www.alanturing.net/turing_archive/pages/Reference%20Articles/what_is_AI/What%20is%20AI06.html Micro-World AI] ]

This paradigm led to innovative work in machine vision by Gerald Sussman (who led the team), Adolfo Guzman, David Waltz (who invented "constraint propagation"), and especially Patrick Winston. At the same time, Minsky and Papert built a robot arm that could stack blocks, bringing the blocks world to life. The crowning achievement of the micro-world program was Terry Winograd's SHRDLU. It could communicate in ordinary English sentences, plan operations and execute them. [Harvnb|McCorduck|2004|pp=300-305, Harvnb|Crevier|1993|pp=84−102, Harvnb|Russell|Norvig|2003|p=19]

The optimism

The money

In June 1963 MIT received a $2.2 million grant from the newly created Advanced Research Projects Agency (later known as DARPA). The money was used to fund project MAC which subsumed the "AI Group" founded by Minsky and McCarthy five years earlier. ARPA continued to provide three million dollars a year until the 70s. [Harvnb|Crevier|1993|pp=64−65]
ARPA made similar grants to Newell and Simon's program at CMU and to the Stanford AI Project (founded by John McCarthy in 1963). [Harvnb|Crevier|1993|p=94] Another important AI laboratory was established at Edinburgh University by Donald Michie in 1965. [Harvnb|Howe|1994] These four institutions would continue to be the main centers of AI research (and funding) in academia for many years. [Harvnb|McCorduck|2004|p=131, Harvnb|Crevier|1993|p=51. McCorduck also notes that funding was mostly under the direction of alumni of the Dartmouth conference of 1956.]

The money was proffered with few strings attached: J. C. R. Licklider, then the director of ARPA, believed that his organization should "fund people, not projects!" and allowed researchers to pursue whatever directions might interest them. [Harvnb|Crevier|1993|p=65] This created a freewheeling atmosphere at MIT that gave birth to the hacker culture, [Harvnb|Crevier|1993|pp=68−71 and Harvnb|Turkle|1984] but this "hands off" approach would not last.

The first AI winter 1974−1980

The problems

In the early seventies, the capabilities of AI programs were disturbingly limited. Even the most impressive could only handle trivial versions of the problems they were supposed to solve; all the programs were, in some sense, "toys". [Harvnb|Crevier|1993|p=146] AI researchers had begun to run into several fundamental limits that could not be overcome in the 1970s. Although some of these limits would be conquered in later decades, others still stymie the field to this day. [Harvnb|Russell|Norvig|2003|pp=20−21]
# Limited computer power: There was not enough memory or processing speed to accomplish anything truly useful. For example, Ross Quillian's successful work on natural language was demonstrated with a vocabulary of only "twenty" words, because that was all that would fit in memory. [Harvnb|Crevier|1993|pp=146−148, see also Harvnb|Buchanan|2005|p=56: "Early programs were necessarily limited in scope by the size and speed of memory"] Hans Moravec argued in 1976 that computers were still millions of times too weak to exhibit intelligence. He suggested an analogy: artificial intelligence requires computer power in the same way that aircraft require horsepower. Below a certain threshold, it's impossible, but, as power increases, eventually it could become easy. [Harvnb|Moravec|1976. McCarthy has always disagreed with Moravec, back to their early days together at SAIL. He states "I would say that 50 years ago, the machine capability was much too small, but by 30 years ago, machine capability wasn't the real problem." in a CNET interview. Harv|Skillings|2006]
# Intractability and the combinatorial explosion. In 1972 Richard Karp (building on Stephen Cook's 1971 theorem) showed there are many problems that can probably only be solved in exponential time (in the size of the inputs). To find optimal solutions to these problems required unimaginable amounts of computer time except when the problems were trivial. This almost certainly meant that many of the "toy" solutions used by AI would probably never scale up into useful systems. [Harvnb|Russell|Norvig|2003|pp=9,21−22 and Harvnb|Lighthill|1973]
# Commonsense knowledge and reasoning. Many important artificial intelligence applications like vision or natural language required simply enormous amounts of information about the world: the program needed to have some idea of what it might be looking at or what it was talking about. This required that the program know most of the same things about the world that a child does. Researchers soon discovered that this was a truly "vast" amount of information. No one in 1970 could build a database so large and no one knew how a program might learn so much information. [Harvnb|McCorduck|2004|pp=300 & 421, Harvnb|Crevier|1993|pp=113−114, Harvnb|Moravec|1988|p=13, Harvnb|Lenat|1989 (Introduction) and Harvnb|Russell|Norvig|2003|p=21]
# Moravec's paradox: It would eventually dawn on many AI researchers working with vision and robotics that tasks like proving theorems or solving geometry problems were easy for computers to carry out, but supposedly "simple" tasks like recognizing a face or crossing a room without bumping into anything were extremely difficult. This helped explain why research in these areas had made so little progress by the middle 1970s. [Harvnb|McCorduck|2004|p=456, Harvnb|Moravec|1988|pp=15−16]
# The frame and qualification problems. AI researchers (like John McCarthy) who used logic discovered that they could not represent ordinary deductions that involved planning or default reasoning without making changes to the structure of logic itself. They developed new logics (like non-monotonic logics and modal logics) to try to solve the problems. [Harvnb|McCarthy|Hayes|1969, Harvnb|Crevier|1993|pp=117−119]

The end of funding

The agencies that funded AI research (such as the British government, DARPA and NRC) became frustrated with the lack of progress and eventually cut off almost all funding for undirected research into AI. The pattern began as early as 1966 when the ALPAC report appeared criticizing machine translation efforts. After spending 20 million dollars, the NRC ended all support. [Harvnb|McCorduck|2004|p=280-281, Harvnb|Crevier|1993|p=110, Harvnb|Russell|Norvig|2003|p=21 and Harvnb|NRC|1999 under "Success in Speech Recognition".] In 1973, the Lighthill report on the state of AI research in England criticized the utter failure of AI to achieve its "grandiose objectives" and led to the dismantling of AI research in that country. [Harvnb|Crevier|1993|p=117, Harvnb|Russell|Norvig|2003|p=22, Harvnb|Howe|1994 and see also Harvnb|Lighthill|1973.] (The report specifically mentioned the combinatorial explosion problem as a reason for AI's failings.) [Harvnb|Russell|Norvig|2003|p=22, Harvnb|Lighthill|1973, John McCarthy wrote in response that "the combinatorial explosion problem has been recognized in AI from the beginning" in [http://www-formal.stanford.edu/jmc/reviews/lighthill/lighthill.html Review of Lighthill report] ]
DARPA was deeply disappointed with researchers working on the Speech Understanding Research program at CMU and canceled an annual grant of three million dollars. [Harvnb|Crevier|1993|pp=115−116 (on whom this account is based). Other views include Harvnb|McCorduck|2004|p=306-313 and Harvnb|NRC|1999 under "Success in Speech Recognition".] By 1974, funding for AI projects was hard to find.

Hans Moravec blamed the crisis on the unrealistic predictions of his colleagues. "Many researchers were caught up in a web of increasing exaggeration." [Harvnb|Crevier|1993|p=115. Moravec explains, "Their initial promises to DARPA had been much too optimistic. Of course, what they delivered stopped considerably short of that. But they felt they couldn't in their next proposal promise less than in the first one, so they promised more."] However, there was another issue: since the passage of Mansfield Amendment in 1969, DARPA had been under increasing pressure to fund "mission-oriented direct research, rather than basic undirected research." The creative, freewheeling exploration that had gone on in the 60s would not be funded by DARPA. The money was directed to specific projects with clear objectives, like autonomous tanks and battle management systems. [Harvnb|NRC|1999 under "Shift to Applied Research Increases Investment." While the autonomous tank was a failure, the battle management system proved to be enormously successful, saving billions in the first Gulf War, repaying the investment and justifying the DARPA's pragmatic policy, at least as far as DARPA was concerned.]

Critiques from across campus

These critiques were not taken seriously by AI researchers, often because they seemed so far off the point. Problems like intractability and commonsense knowledge seemed much more immediate and serious. It wasn't clear what difference "know how" or "intensionality" made to an actual program. Minsky said of Dreyfus and Searle "they misunderstand, and should be ignored." [Quoted in Harvnb|Crevier|1993|p=143] Dreyfus, who taught at MIT, was given a cold shoulder: he later said that AI researchers "dared not be seen having lunch with me." [Quoted in Harvnb|Crevier|1993|p=122] Joseph Weizenbaum, the author of ELIZA, felt his colleagues' treatment of Dreyfus was unprofessional and childish. Although he was an outspoken critic of Dreyfus' positions, he "deliberately made it plain that theirs was not the way to treat a human being." ["I became the only member of the AI community to be seen eating lunch with Dreyfus. And I deliberately made it plain that theirs was not the way to treat a human being." Joseph Weizenbaum, quoted in Harvnb|Crevier|1993|p=123.]

Perceptrons and the dark age of connectionism

A perceptron was a form of neural network introduced in 1958 by Frank Rosenblatt, who had been a schoolmate of Marvin Minsky at the Bronx High School of Science. Like most AI researchers, he was optimistic about their power, predicting that "perceptron may eventually be able to learn, make decisions, and translate languages." An active research program into the paradigm was carried out throughout the 60s but came to a sudden halt with the publication of Minsky and Papert's 1969 book "Perceptrons". They showed that there were severe limitations to what perceptrons could do and that Frank Rosenblatt's predictions had been grossly exaggerated. The effect of the book was devastating: virtually no research at all was done in connectionism for 10 years. Eventually, a new generation of researchers would revive the field and thereafter it would become a vital and useful part of artificial intelligence. Rosenblatt would not live to see this, as he died in a boating accident shortly after the book was published.

The neats: logic, Prolog and expert systems

Logic was introduced into AI research as early as 1958, by John McCarthy in his Advice Taker proposal. [] In 1963, J. Alan Robinson had discovered a simple method to implement deduction on computers, the resolution and unification algorithm. However, straightforward implementations, like those attempted by McCarthy and his students in the late 60s, were especially intractable: the programs required astronomical numbers of steps to prove simple theorems. [Harvnb|McCorduck|2004|p=51, Harvnb|Crevier|1993|pp=190−192] A more fruitful approach to logic was developed in the 70s by Robert Kowalski at the University of Edinburgh, and soon this led to the collaboration with French researchers Alain Colmerauer and Phillipe Roussel who created the successful logic programming language Prolog. [Harvnb|Crevier|1993|pp=193−196] Prolog uses a subset of logic (Horn clauses, closely related to "rules" and "production rules") that permit tractable computation. Rules would continue to be influential, providing a foundation for Edward Feigenbaum's expert systems and the continuing work by Alan Newell and Herbert Simon that would lead to Soar and their unified theories of cognition. [Harvnb|Crevier|1993|pp=145−149,258−63]

Critics of the logical approach noted, as Dreyfus had, that human beings rarely used logic when they solved problems. Experiments by psychologists like Peter Wason, Eleanor Rosch, Amos Tversky, Daniel Kahneman and others provided proof. [] McCarthy responded that what people do is irrelevant and pointed out that we don't need machines that think as people do, we need machines that can solve problems that people normally solve by thinking. [An early example of McCathy's position was in the journal Science where he said "This is AI, so we don't care if it's psychologically real" (see [http://books.google.com/books?id=PEkqAAAAMAAJ&q=%22we+don't+care+if+it's+psychologically+real%22&dq=%22we+don't+care+if+it's+psychologically+real%22&output=html&pgis=1 see Science at Google Books] ), and he recently reiterated his position at the AI@50 conference where he said "Artificial intelligence is not, by definition, simulation of human intelligence" (see [http://www.engagingexperience.com/2006/07/ai50_ai_past_pr.html McCarthy's presentation at AI@50] )]

The scruffies: frames and scripts

Among the critics of McCarthy's approach were his colleagues across the country at MIT. Marvin Minsky, Seymour Papert and Roger Schank were trying to solve problems like "story understanding" and "object recognition" that "required" a machine to think like a person. In order to use ordinary concepts like "chair" or "restaurant" they had to make all the same illogical assumptions that people normally made. Unfortunately, imprecise concepts like these are hard to represent in logic. Gerald Sussman observed that "using precise language to describe essentially imprecise concepts doesn't make them any more precise." [Harvnb|Crevier|1993|pp=175] Schank described their "anti-logic" approaches as "scruffy", as opposed to the "neat" paradigms used by McCarthy, Kowalski, Feigenbaum, Newell and Simon. [Neat vs. scruffy: Harvnb|McCorduck|2004|pp=421-424 (who picks up the state of the debate in 1984). Harvnb|Crevier|1993|pp=168 (who documents Schank's original use of the term). Another aspect of the conflict was called "the procedural/declarative distinction" but did not prove to be influential in later AI research.]

In 1975, in a seminal paper, Minsky noted that many of his fellow "scruffy" researchers were using the same kind of tool: a framework that captures all our common sense assumptions about something. For example, if we use the concept of a bird, there is a constellation of facts that immediately come to mind: we might assume that it flies, eats worms and so on. We know these facts are not always true and that deductions using these facts will not be "logical," but these structured sets of assumptions are part of the "context" of everything we say and think. He called these structures "frames". Schank used a version of frames he called "scripts" to successfully answer questions about short stories in English. [Harvnb|McCorduck|2004|pp=305-306, Harvnb|Crevier|1993|pp=170−173, 246 and Harvnb|Russell|Norvig|2003|p=24. Minsky's frame paper: Harvnb|Minsky|1974.] Many years later object-oriented programming would adopt the essential idea of "inheritance" from AI research on frames.

Boom 1980–1987

In the 1980s a form of AI program called "expert systems" was adopted by corporations around the world and knowledge became the focus of mainstream AI research. In those same years, the Japanese government aggressively funded AI with its fifth generation computer project. Another encouraging event in the early 1980s was the revival of connectionism in the work of John Hopfield and David Rumelhart. Once again, AI had achieved success.

The rise of expert systems

An expert system is a program that answers questions or solves problems about a specific domain of knowledge, using logical rules that are derived from the knowledge of experts. The earliest examples were developed by Edward Feigenbaum and his students. Dendral, begun in 1965, identified compounds from spectrometer readings. MYCIN, developed in 1972, diagnosed infectious blood diseases. They demonstrated the feasibility of the approach. [Harvnb|McCorduck|2004|pp=327-335 (Dendral), Harvnb|Crevier|1993|pp=148−159, Harvnb|Russell|Norvig|2003|pp=22−23]

Expert systems restricted themselves to a small domain of specific knowledge (thus avoiding the commonsense knowledge problem) and their simple design made it relatively easy for programs to be built and then modified once they were in place. All in all, the programs proved to be "useful": something that AI had not been able to achieve up to this point. [harvnb|Crevier|1993|pp=158−159 and Harvnb|Russell|Norvig|2003|p=23−24]

In 1980, an expert system called XCON was completed at CMU for the Digital Equipment Corporation. It was an enormous success: it was saving the company 40 million dollars annually by 1986. [Harvnb|Crevier|1993|p=198] Corporations around the world began to develop and deploy expert systems and by 1985 they were spending over a billion dollars on AI, most of it to in-house AI departments. An industry grew up to support them, including hardware companies like Symbolics and Lisp Machines and software companies such as IntelliCorp and Aion. [Harvnb|McCorduck|2004|pp=434-435, Harvnb|Crevier|1993|pp=161−162,197−203 and Harvnb|Russell|Norvig|2003|p=24]

The knowledge revolution

The power of expert systems came from the expert knowledge they contained. They were part of a new direction in AI research that had been gaining ground throughout the 70s. "AI researchers were beginning to suspect—reluctantly, for it violated the scientific canon of parsimony—that intelligence might very well be based on the ability to use large amounts of diverse knowledge in different ways," [Harvnb|McCorduck|2004|p=299] writes Pamela McCorduck. " [T] he great lesson from the 1970s was that intelligent behavior depended very much on dealing with knowledge, sometimes quite detailed knowledge, of a domain where a given task lay". [{Harvnb|McCorduck|2004|pp=421] Knowledge based systems and knowledge engineering became a major focus of AI research in the 1980s. [Knowledge revolution: Harvnb|McCorduck|2004|pp=266-276, 298-300, 314, 421, Harvnb|Russell|Norvig|pp=22-23]

The 1980s also saw the birth of Cyc, the first attempt to attack the commonsense knowledge problem directly, by creating a massive database that would contain all the mundane facts that the average person knows. Douglas Lenat, who started and led the project, argued that there is no shortcut ― the only way for machines to know the meaning of human concepts is to teach them, one concept at a time, by hand. The project was not expected to be completed for many decades. [Cyc: Harvnb|McCorduck|2004|p=489, Harvnb|Crevier|1993|pp=239−243, Harvnb|Russell|Norvig|2003|p=363−365 and Harvnb|Lenat|Guha|1989]

The money returns: the fifth generation project

Other countries responded with new programs of their own: England began the ₤350 million Alvey project and a consortium of American companies formed the Microelectronics and Computer Technology Corporation (or "MCC") to fund large scale projects in AI and information technology. [harvnb|Crevier|1993|pp=240.] DARPA responded as well, founding the Strategic Computing Initiative and tripling its investment in AI between 1984 and 1988. [Harvnb|McCorduck|2004|pp=426-432, Harvnb|NRC|1999 under "Shift to Applied Research Increases Investment"]

The revival of connectionism

In 1982, physicist John Hopfield was able to prove that a form of neural network (now called a "Hopfield net") could learn and process information in a completely new way. Around the same time, David Rumelhart popularized a new method for training neural networks called "backpropagation" (discovered years earlier by Paul Werbos). These two discoveries revived the field of connectionism which had been largely abandoned since 1970. [harvnb|Crevier|1993|pp=214−215.] Harvnb|Russell|Norvig|2003|p=25]

The new field was unified and inspired by the appearance of "Parallel Distributed Processing" in 1986—a two volume collection of papers edited by Rumelhart and psychologist James McClelland. Neural networks would become commercially successful in the 1990s, when they began to be used as the engines driving programs like optical character recognition and speech recognition. [harvnb|Crevier|1993|pp=215−216.]

Bust: the second AI winter 1987−1993

The business community's fascination with AI rose and fell in the 80s in the classic pattern of an economic bubble. The collapse was in the "perception" of AI by government agencies and investors — the field continued to make advances despite the criticism. Rodney Brooks and Hans Moravec, researchers from the related field of robotics, argued for an entirely new approach to artificial intelligence.

AI winter

The term "AI winter" was coined by researchers who had survived the funding cuts of 1974 when they became concerned that enthusiasm for expert systems had spiraled out of control and that disappointment would certainly follow. [Harvnb|Crevier|1993|pp=203. AI winter was first used as the title of a seminar on the subject for the Association for the Advancement of Artificial Intelligence.]

The first indication of a change in weather was the sudden collapse of the market for specialized AI hardware in 1987. Desktop computers from Apple and IBM had been steadily gaining speed and power and in 1987 they became more powerful than the more expensive Lisp machines made by Symbolics and others. There was no longer a good reason to buy them. An entire industry worth half a billion dollars was demolished overnight. [Harvnb|McCorduck|2004|p=435, Harvnb|Crevier|1993|pp=209−210]

Eventually the earliest successful expert systems, such as XCON, proved too expensive to maintain. They were difficult to update, they could not learn, they were "brittle" (i.e., they could make grotesque mistakes when given unusual inputs), and they fell prey to problems (such as the qualification problem) that had been identified years earlier. Expert systems proved useful, but only in a few special contexts. [Harvnb|McCorduck|2004|p=435 (who cites institutional reasons for their ultimate failure), Harvnb|Crevier|1993|pp=204−208 (who cites the difficulty of truth maintenance, i.e., learning and updating), Harvnb|Lenat|Guha|1989|loc=Introduction (who emphasizes the brittleness and the inability to handle excessive qualification.)]

In the late 80s, the new management of the Strategic Computing Initiative cut funding to AI "deeply and brutally" [Harvnb|McCorduck|2004|pp=430-431] in favor of other projects that seemed more likely to produce immediate results.

By 1991, the impressive list of goals penned in 1981 for Japan's Fifth Generation Project had not been met. Indeed, some of them, like "carry on a casual conversation" had not been met by 2008.Harvnb|McCorduck|2004|p=441, Harvnb|Crevier|1993|p=212. McCorduck writes "Two and a half decades later, we can see that the Japanese didn't quite meet all of those ambitious goals."] As with other AI projects, expectations had run much higher than what was actually possible.

The importance of having a body: Nouvelle AI and embodied reason

In the late 80s, several researchers advocated a completely new approach to artificial intelligence, based on robotics. [Harvnb|McCorduck|2004|pp=454-462] They believed that, to show real intelligence, a machine needs to have a "body" — it needs to perceive, move, survive and deal with the world. They argued that these sensorimotor skills are essential to higher level skills like commonsense reasoning and that abstract reasoning was actually the "least" interesting or important human skill (see Moravec's paradox). They advocated building intelligence "from the bottom up." [Harvtxt|Moravec|1988|p=20 writes: "I am confident that this bottom-up route to artificial intelligence will one date meet the traditional top-down route more than half way, ready to provide the real world competence and the commonsense knowledge that has been so frustratingly elusive in reasoning programs. Fully intelligent machines will result when the metaphorical golden spike is driven uniting the two efforts."]

The approach revived ideas from cybernetics and control theory that had been unpopular since the sixties. Another precursor was David Marr, who had come to MIT in the late 70s from a successful background in neurology to lead the group studying vision. He rejected all symbolic approaches ("both" McCarthy's logic and Minsky's frames), arguing that AI needed to understand the physical machinery of vision from the bottom up before any symbolic processing took place. Marr's work would be cut short by leukemia in 1980. [Harvnb|Crevier|1993|pp=183−190.]

In a 1990 paper [http://people.csail.mit.edu/brooks/papers/elephants.pdf Elephants Don't Play Chess] , robotics researcher Rodney Brooks took direct aim at the physical symbol system hypothesis, arguing that symbols are not always necessary since "the world is its own best model. It is always exactly up to date. It always has every detail there is to be known. The trick is to sense it appropriately and often enough." [Harvnb|Brooks 1990|p=3] In the 80s and 90s, many cognitive scientists also rejected the symbol processing model of the mind and argued that the body was essential for reasoning, a theory called the embodied mind thesis. [See, for example, Harvnb|Lakoff|Turner|1999]

AI 1993−present

The field of AI, now more than a half a century old, finally achieved some of its oldest goals. It began to be used successfully throughout the technology industry, although somewhat behind the scenes. Some of the success was due to increasing computer power and some was achieved by focusing on specific isolated problems and pursuing them with the highest standards of scientific accountability. Still, the reputation of AI, in the business world at least, was less than pristine. Inside the field there was little agreement on the reasons for AI's failure to fulfill the dream of human level intelligence that had captured the imagination of the world in the 1960s. Together, all these factors helped to fragment AI into competing subfields focussed on particular problems or approaches, sometimes even under new names that disguised the tarnished pedigree of "artificial intelligence." [Harvtxt|McCorduck|2004|p=424 discusses the fragmentation and the abandonment of AI's original goals.] AI was both more cautious and more successful than it had ever been.

Milestones and Moore's Law

On 11 May 1997, Deep Blue became the first computer Chess-playing system to beat a reigning world Chess champion, Gary Kasparov. [Harvnb|McCorduck|2004|pp=480-483] In 2005, a Stanford robot won the DARPA Grand Challenge by driving autonomously for 131 miles along an unrehearsed desert trail. After many years of effort, such milestones were finally achieved. These successes were not due to some revolutionary new paradigm, but mostly on the tedious application of engineering skill and on the tremendous power of computers today. [Harvnb|Kurzweil|2005|p=274 writes that the improvement in computer chess, "according to common wisdom, is governed only by the brute force expansion of computer hardware."] In fact, Deep Blue's computer was 10 million times faster than the Ferranti Mark I that Christopher Strachey taught to play chess in 1951. [Cycle time of Ferranti Mark I was 1.2 milliseconds, which is arguably equivalent to about 833 flops. Deep Blue ran at 11.38 gigaflops (and this does not even take into account Deep Blue's special-purpose hardware for chess). "Very" approximately, these differ by a factor of 10^7.] Thanks to Moore's law, the fundamental problem of "raw computer power" was slowly being overcome.

Intelligent agents

An intelligent agent is a system that perceives its environment and takes actions which maximize its chances of success. The simplest intelligent agents are programs that solve specific problems. The most complicated intelligent agents would be rational, thinking human beings. The intelligent agent paradigm defines AI research "the study of intelligent agents". This is a generalization of some earlier definitions of AI: it goes beyond studying human intelligence; it studies all kinds of intelligence. [This is how the most widely accepted textbooks of the 21st century define artificial intelligence. See Harvnb|Russell|Norvig|2003|p=32 and Harvnb|Poole|Mackworth|Goebel|1998|p=1]

The paradigm gave researchers license to study isolated problems and find solutions that were both verifiable and useful. It provided a common language to describe problems and share their solutions with each other, and with other fields that also used concepts of abstract agents, like economics and control theory. It was hoped that a complete agent architecture (like Newell's SOAR) would one day allow researchers to build more versatile and intelligent systems out of interacting intelligent agents. [Harvnb|McCorduck|2004|p=478]

Victory of the neats

AI researchers began to develop and use sophisticated mathematical tools more than they ever had in the past. [Harvnb|McCorduck|2004|p=486-487, Harvnb|Russell|Norvig|2003|p=25-26] There was a widespread realization that many of the problems that AI needed to solve were already being worked on by researchers in fields like mathematics, economics or operations research. The shared mathematical language allowed both a higher level of collaboration with more established and successful fields and the achievement of results which were measurable and provable; AI had become a more rigorous "scientific" discipline. Harvtxt|Russell|Norvig|2003 describe this as nothing less than a "revolution" and "the victory of the neats."Harvnb|Russell|Norvig|2003|p=25−26] [Harvtxt|McCorduck|2004|p=487: "As I write, AI enjoys a Neat hegemony."]

Judea Pearl's highly influential 1988 book [Harvnb|Pearl|1988] brought probability and decision theory into AI. Among the many new tools in use were Bayesian networks, hidden Markov models, information theory, stochastic modeling and classical optimization. Precise mathematical descriptions were also developed for "computational intelligence" paradigms like neural networks and evolutionary algorithms.

AI behind the scenes

Algorithms originally developed by AI researchers began to appear as parts of larger systems. AI had solved a lot of very difficult problems and their solutions proved to be useful throughout the technology industry, [Harvnb|NRC|1999 under "Artificial Intelligence in the 90s", and Harvnb|Kurzweil|2005|p=264] such as
data mining,
industrial robotics,
logistics, [Harvnb|Russell|Norvig|2003|p=28]
speech recognition, [For the new state of the art in AI based speech recognition, see [http://www.economist.com/science/tq/displaystory.cfm?story_id=9249338 Are You Talking to Me?] ] banking software,"AI-inspired systems were already integral to many everyday technologies such as internet search engines, bank software for processing transactions and in medical diagnosis." Nick Bostrom, [http://www.cnn.com/2006/TECH/science/07/24/ai.bostrom/ AI set to exceed human brain power] CNN.com (July 26, 2006)] medical diagnosisand Google's search engine. [For the use of AI at Google, see [http://news.com.com/Googles+man+behind+the+curtain/2008-1024_3-5208228.html Google's man behind the curtain] , [http://news.com.com/Google+backs+character-recognition+research/2100-1032_3-6175136.html Google backs character recognition] and [http://news.com.com/Spying+an+intelligent+search+engine/2100-1032_3-6107048.html Spying an intelligent search engine] .]

The field of AI receives little or no credit for these successes. Many of AI's greatest innovations have been reduced to the status of just another item in the tool chest of computer science. [Harvnb|McCorduck|2004|p=423, Harvnb|Kurzweil|2005|p=265, Harvnb|Hofstadter|1979|p=601]
Nick Bostrom explains "A lot of cutting edge AI has filtered into general applications, often without being called AI because once something becomes useful enough and common enough it's not labeled AI anymore." [ [http://www.cnn.com/2006/TECH/science/07/24/ai.bostrom/ AI set to exceed human brain power] CNN.com (July 26, 2006)]

Many researchers in AI today deliberately call their work by other names, such as informatics, knowledge-based systems, cognitive systems or computational intelligence. In part, this may be because they considered their field to be fundamentally different from AI, but also the new names help to procure funding. In the commercial world at least, the failed promises of the AI Winter continue to haunt AI research, as the New York Times reported in 2005: "Computer scientists and software engineers avoided the term artificial intelligence for fear of being viewed as wild-eyed dreamers." [ cite news |first=John |last=Markoff |title=Behind Artificial Intelligence, a Squadron of Bright Real People |url=http://www.nytimes.com/2005/10/14/technology/14artificial.html?ei=5070&en=11ab55edb7cead5e&ex=1185940800&adxnnl=1&adxnnlx=1185805173-o7WsfW7qaP0x5/NUs1cQCQ |work= |publisher=The New York Times |date=2005-10-14 |accessdate=2007-07-30 ] [Alex Castro (2007) [http://www.economist.com/science/tq/displaystory.cfm?story_id=9249338 Are you talking to me?] The Economist Technology Quarterly (June 7, 2007)] [Patty Tascarella, [http://www.bizjournals.com/pittsburgh/stories/2006/08/14/focus3.html?b=1155528000%5E1329573 Robotics firms find fundraising struggle, with venture capital shy] . Pittsburgh Business Times (August 11, 2006)]

Where is HAL 9000?

In 1968, Arthur C. Clark and Stanley Kubrick had imagined that by the year , a machine would exist with an intelligence that matched or exceeded the capability of human beings. The character they created, HAL-9000, was based on hard science: many leading AI researchers also believed that such a machine would exist by the year 2001. [Harvnb|Crevier|1993|pp=108−109]

Marvin Minsky asks "So the question is why didn't we get HAL in 2001?" [He goes on to say: "The answer is, I believe we could have ... I once went to an international conference on neural net [s] . There were 40 thousand registrants ... but ... if you had an international conference, for example, on using multiple representations for common sense reasoning, I've only been able to find 6 or 7 people in the whole world." Marvin Minsky, in [http://technetcast.ddj.com/tnc_play_stream.html?stream_id=526 It's 2001] ] Minsky believes that the answer is that the central problems, like commonsense reasoning, were being neglected, while most researchers pursued things like commercial applications of neural nets or genetic algorithms. John McCarthy, on the other hand, still blames the qualification problem. [See [http://www.engagingexperience.com/2006/07/ai50_ai_past_pr.html McCarthy's presentation at AI@50] ] For Ray Kurzweil, the issue is computer power and, using Moore's Law, he predicts that machines with human-level intelligence will appear by 2029. [Harvnb|Kurzweil|2005] Jeff Hawkins argues that neural net research ignores the essential properties of the human cortex, preferring simple models that have been successful at solving simple problems. [Harvnb|Hawkins|Blakeslee|2004] There are many other explanations and for each there is a corresponding research program underway.

Alan Turing's quote from 1950 still applies in the 21st century: "We can only see a short distance ahead, but we can see that there is much to be done.

Notes

References

Wikimedia Foundation. 2010.

Игры ⚽ Поможем написать реферат

Look at other dictionaries:

Artificial intelligence — AI redirects here. For other uses, see Ai. For other uses, see Artificial intelligence (disambiguation). TOPIO, a humanoid robot, played table tennis at Tokyo International Robot Exhibition (IREX) 2009.[1] Artificial intelligence ( … Wikipedia
Outline of artificial intelligence — The following outline is provided as an overview of and topical guide to artificial intelligence: Artificial intelligence (AI) – branch of computer science that deals with intelligent behavior, learning, and adaptation in machines. Research in AI … Wikipedia
Timeline of artificial intelligence — See also: History of artificial intelligence To 1900 1960s 1990s 2000 and Beyond Notes References* * * * * * * * * * * * * * * * * * * *. * * * * * * * * * * * * * * * * * * *External links* [http://www.idsia.ch/ juergen/ai.html AI history… … Wikipedia
artificial intelligence — the capacity of a computer to perform operations analogous to learning and decision making in humans, as by an expert system, a program for CAD or CAM, or a program for the perception and recognition of shapes in computer vision systems. Abbr.:… … Universalium
Artificial intelligence (video games) — In strategy games like Freeciv, the game AI must deal with large amounts of information Game artificial intelligence refers to techniques used in computer and video games to produce the illusion of intelligence in the behavior of non player… … Wikipedia
Topic outline of artificial intelligence — Artificial intelligence (AI) is a branch of computer science that deals with intelligent behavior, learning, and adaptation in machines . Research in AI is concerned with producing machines to automate tasks requiring intelligent behavior. The… … Wikipedia
Artificial intelligence in fiction — This is a sub article of Artificial intelligence (AI), describing the different futuristic portrayals of fictional artificial intelligence in books and film.Since Artificial Intelligence has a potential equal to or greater than our own, it can… … Wikipedia
Artificial Intelligence Applications Institute — The Artificial Intelligence Applications Institute (AIAI) at the School of Informatics at the University of Edinburgh is a non profit technology transfer organisation that promotes the benefits of the application of Artificial Intelligence… … Wikipedia
History of artificial life — The idea of human artifacts being given life has fascinated mankind for as long as men have been recording their myths and stories. Whether Pygmalion or Frankenstein, mankind has been fascinated with the idea of artificial life. Pre computer… … Wikipedia
MIT Computer Science and Artificial Intelligence Laboratory — Established July 1, 2003 July 1, 1963 (as Project MAC) Field of Research Computer science … Wikipedia

Academic Dictionaries and Encyclopedias

History of artificial intelligence

Look at other dictionaries:

Share the article and excerpts

Academic Dictionaries and Encyclopedias

Wikipedia

History of artificial intelligence

Look at other dictionaries:

Share the article and excerpts

Direct link