History of artificial intelligence

From Wikipedia, the free encyclopedia

See also: timeline of artificial intelligence

The history of artificial intelligence begins in antiquity. Artificial human beings endowed with consciousness and intelligence have appeared in myths, stories and rumours for thousands of years.[1] 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.[2]

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, DARPA and the British Government 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.[3] Undaunted, there are those that make extraordinary predictions even now.[4]

But, 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 priorities in light of new discoveries and riding the crest of the wave of increasing computer power. Progress has been slower than predicted but has continued nonetheless. Artificial intelligence problems that had begun to seem impossible in 1970 have been solved and are now 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."[5]

artificial intelligence Portal
History of computing
Hardware before 1960
Hardware 1960s to present
Hardware in Soviet Bloc countries
Computer science
Operating systems
Personal computers
Laptops
Software engineering
Programming languages
Artificial intelligence
Graphical user interface
Internet
World Wide Web
Computer and video games
Timeline of computing
More...

Contents

[edit] Precursors

Pygmalion and Galatea (1890) by Jean-Léon Gérôme (1824–1904)
Pygmalion and Galatea (1890) by Jean-Léon Gérôme (1824–1904)

[edit] AI in myth and fiction

Main article: artificial intelligence in fiction

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.[6] Mechanical men and artificial beings appear in Greek myths, such as the golden robots of Hephaestus and Pygmalion's Galatea.[7] In the Middle Ages, it was believed that there were secret mystical or alchemical means of placing mind into matter, such as the Paracelsus' homunculus and Rabbi Judah Loew's Golem.[8] 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.[9]

Al-Jazari's programmable automata (1206 CE)
Al-Jazari's programmable automata (1206 CE)

[edit] Automatons

Main article: automaton

Realistic humanoid automatons were built by craftsman from every civilization, including Yan Shi,[10] Hero of Alexandria,[11] Al-Jazari[12] and Wolfgang von Kempelen.[13] 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."[14]

Gottfried Leibniz, who speculated that human reason could be reduced to calculation
Gottfried Leibniz, who speculated that human reason could be reduced to calculation

[edit] Formal reasoning

Main article: history of philosophy

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.[15] Hobbes famously wrote in Leviathan: "reason is nothing but reckoning".[16] 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."[17] These philosophers had begun to articulate the physical symbol system hypothesis that would become the guiding faith of AI research.

The ENIAC, at the Moore School of Electrical Engineering. (U.S. Army Photo)
The ENIAC, at the Moore School of Electrical Engineering. (U.S. Army Photo)

[edit] Computer science

Main articles: history of computer hardware and 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).[18]

The key insight was the Turing machine (Turing 1936), a simple theoretical construct that captured the essence of abstract symbol manipulation. The Church-Turing thesis implied that a mechanical device, shuffling symbols as simple as 0 and 1, could imitate any conceivable process of mathematical deduction. This would inspire a handful of scientists to begin discussing the possibility of thinking machines.[19]

[edit] The birth of artificial intelligence 1943−1956

The IBM 702: a computer used by the first generation of AI researchers.
The IBM 702: a computer used by the first generation of AI researchers.

A note on the sections in this article.[20]

The first computers cost millions of dollars, filled entire rooms and had less computing power than a modern clock or a thermostat. A number of researchers from many fields (mathematics, psychology, engineering and even political science) 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. In 1956, at a conference on the Dartmouth campus, the field of artificial intelligence was born.[21]

[edit] Cybernetics and early neural networks

In the early forties, two Princeton scientists attempted to create a mathematical description of the human brain. Walter Pitts and Warren McCulloch analyzed networks of idealized artificial neurons and showed how they might perform simple logical functions. They were the first to describe what later researchers would call a neural network.[22] Pitts and McCulloch were part of a research program called "cybernetics" that lasted from the 1940s (when Norbert Weiner defined the term) till the 1960s. Researchers developed robots, like W. Grey Walter's turtles and the Johns Hopkins Beast, that displayed rudimentary intelligence. These machines did not use computers or digital electronics; they were controlled entirely by analog circuitry.[23]

One of the students inspired by Pitts and McCulloch's work was a young Marvin Minsky, then a 24 year old graduate student. In 1951 (with Dean Edmonds) he built the first neural net machine, the SNARC. Minsky was to become one of the most important leaders and innovators in AI for the next 50 years.[24]

[edit] Turing's test

In 1950 Alan Turing published a landmark paper in which he speculated about the possibility of creating machines with true intelligence. 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 measure of intelligence has often been criticized but never improved upon: no universally accepted definition of intelligence has ever been put forward. The most electrifying aspect of the paper was that it clearly showed that intelligent machines were at least plausible: Turing anticipated and answered every major objection to the question "can machines think?" The Turing Test was the first serious proposal in the philosophy of artificial intelligence.[25] As such, he is often described as 'the father of AI'.

[edit] Newell and Simon's Logic Theorist

In 1955 Allen Newell and (future Nobel Laureate) Herbert Simon created the "Logic Theorist" (with help from J. C. Shaw). The program would eventually prove 38 of the first 52 theorems in Russell and Whitehead's Principia Mathematica, and find new and more elegant proofs for some.[26] Simon claimed that they had "solved the venerable mind/body problem, explaining how a system composed of matter can have the properties of mind."[27] This was an early statement of the philosophical position now called "Strong AI": that machines can contain minds just as human bodies do.[28]

[edit] Dartmouth Conference 1956: the birth of AI

The Dartmouth Conference of 1956 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" — an early statement of what would later be known as the "physical symbol system hypothesis".[29] 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.[30] At the conference, Newell and Simon debuted the "Logic Theorist" and McCarthy convinced the majority of attendees to accept "Artificial Intelligence" as the name of the field.[31] 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.[32][33]

[edit] The golden years 1956−1974

The years after the Dartmouth conference were an era of discovery, of sprinting across new ground. The programs that were developed during this time were, to most people, simply "astonishing"[34]: computers were solving algebra word problems, proving theorems in geometry and learning to speak English. Few at the time would have believed that such "intelligent" behavior by machines was possible at all.[35] Researchers expressed an intense optimism in private and in print, predicting that a fully intelligent machine would be built in less than 20 years.[36] Government agencies like ARPA poured money into the new field.[37]

[edit] The work

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

Reasoning as search

Many 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".[38]

The principle 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.[39]

Newell and Simon tried to capture a general version of this algorithm in a program called the "General Problem Solver". Other "searching" programs were able to accomplish impressive tasks like solving problems in geometry and algebra: Herbert Gelernter's Geometry Theorem Prover (1958) and SAINT written by Minsky's student James Slagle (1961).[40] Other programs searched through goals and subgoals to plan actions, like the STRIPS system developed at Stanford to control the behavior of their robot Shakey.[41]

An example of a semantic network
An example of a semantic network
Natural language and semantic nets

An important goal of AI research is to allow computers to communicate in natural languages like English. An early success was Daniel Bobrow's program STUDENT, which could solve high school algebra word problems by recognizing key phrases like "the sum of".[42]

In a semantic net, concepts (e.g. "house","door") are represented as nodes and relations between concepts (e.g. "has-a") are represented as links between the nodes. The first AI program to use a semantic net was written by Ross Quillian[43] and the most successful (and controversial) version was Roger Schank's Conceptual Dependency.[44]

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 repeated back what was said to her, rephrasing it using a few grammar rules and occasionally jumping back to earlier points in the conversation.[45]

Micro-worlds

In 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.[46]

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.[47]

[edit] The optimism

The first generation of AI researchers made these predictions about their work:

  • 1958, H. A. Simon and Allen Newell: "within ten years a digital computer will be the world's chess champion" and "within ten years a digital computer will discover and prove an important new mathematical theorem."[48]
  • 1965, H. A. Simon: "machines will be capable, within twenty years, of doing any work a man can do."[49]
  • 1967, Marvin Minsky: "Within a generation ... the problem of creating 'artificial intelligence' will substantially be solved."[50]
  • 1970, Marvin Minsky (in Life Magazine): "In from three to eight years we will have a machine with the general intelligence of an average human being."[51]

[edit] The money

In June of 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.[52] ARPA made similar grants to Newell and Simon's program at CMU and to the Stanford AI Project (founded by John McCarthy in 1963).[53] Another important AI laboratory was established at Edinburgh University by Donald Michie in 1965.[54] These four institutions would continue to be the main centers of AI research (and funding) in academia for many years.[55]

The money was proffered with few strings attached: J. C. R. Licklider, then the director of ARPA, felt that his organization should "fund people, not projects!" and allowed researchers to pursue whatever directions might interest them.[56] The freewheeling atmosphere at MIT gave birth to the hacker culture.[57] However, this "hands off" approach would soon come to an end.

[edit] The first AI winter 1974−1980

In the 70s, AI was subject to critiques and financial setbacks. AI researchers had failed to appreciate the difficulty of the problems they face. Their tremendous optimism had raised expectations impossibly high, and when the results they had promised failed to materialize, funding for AI disappeared.[58] At the same time, the field of connectionism (or neural nets) was shut down almost completely for 10 years by Marvin Minsky's devastating criticism of perceptrons.[59]

Despite the difficulties with public perception of AI in the late 70s, new ideas were explored in logic programming, commonsense reasoning and many other areas.[60]

[edit] 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 of the programs were, in some sense, "toys".[61] 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.[62]

  1. 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.[63] Hans Moravec argued in 1976 that computers were still millions of times too weak to exhibit intelligence, and 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.[64]
  2. 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.[65]
  3. 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 information. No one in 1970 could build a database so large and no one knew how a program might learn so much information.[66]
  4. Moravec's paradox: It would eventually dawn on many AI researchers working with vision and robotics that normal intuitions about which problems were "easy" or "hard" did not apply to AI. 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.[67]
  5. 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, but they were forced to give up some of their precious certainty.[68]

[edit] The end of funding

See also: AI Winter

The agencies that funded AI research (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.[69] 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.[70] (The report specifically mentioned the combinatorial explosion problem as a reason for AI's failings.)[71] DARPA was deeply disappointed with researchers working on the Speech Understanding Research program at CMU and canceled an annual grant of three million dollars. By 1974, funding for AI projects was hard to find.[72]

Hans Moravec blamed the crisis on the unrealistic predictions of his colleagues: "Many researchers were caught up in a web of increasing exaggeration. 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."[73]

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.[74]


[edit] Critiques from across campus

See also: Philosophy of artificial intelligence

A number of philosophers had strong objections to the claims being made by AI researchers. One of the earliest was John Lucas, who argued that Gödel's incompleteness theorem showed that a physical symbol system (such as a computer program) could never see the truth of certain statements, while a human being could.[75] Hubert Dreyfus ridiculed the broken promises of the 60s and critiqued the assumptions of AI, arguing that human reasoning actually involved very little "symbol processing" and a great of deal embodied, instinctive, unconscious "know how".[76] John Searle's Chinese Room argument, presented in 1980, attempted to show that a program (or any physical symbol system) could not be said to "understand" the symbols that it uses; that the symbols have no meaning for the machine, and so the machine can never be truly intelligent.[77]

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."[78] Dreyfus, who taught at MIT, was given a cold shoulder: he later said that AI researchers "dared not be seen having lunch with me."[79]

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."[80] He began to have serious ethical doubts about AI when Kenneth Colby wrote DOCTOR, a chatterbot therapist. Weizenbaum was disturbed that Colby saw his mindless program as a serious therapeutic tool. A feud began, and the situation was not helped when Colby did not credit Weizenbaum for his contribution to the program. Eventually Weizenbaum would publish a thoughtful moral critique of AI.[81]

[edit] 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 made optimistic claims 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 Minsky and Papert's 1969 book Perceptrons. They showed that there were severe limitations to what perceptrons could do and that Frank Rosenblatt's claims had been grossly exaggerated. The effect of the book was devastating: virtually no research at all was done in connectionism for 10 years. Eventually, the work of Hopfield and others 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.[82]

[edit] 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.[83] In 1963, J. Alan Robinson had discovered a simple method to implement deduction on computers, the resolution and unification algorithm. However, straight forward 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.[84] 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.[85] 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.[86]

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.[87] 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.[88]

[edit] 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. As Gerald Sussman said, "using precise language to describe essentially imprecise concepts doesn't make them any more precise."[89] Schank described their "anti-logic" approaches as "scruffy", as opposed to the "neat" paradigms used by McCarthy, Kowalski, Feigenbaum, Newell and Simon.[90]

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.[91] Many years later object-oriented programming would adopt the essential of idea "inheritance" from AI research on frames.

[edit] Boom 1980–1987

In the 1980s a form of AI program called "expert systems" was adopted by corporations around the world, and, 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.

A Hopfield net with four nodes.
A Hopfield net with four nodes.

[edit] 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.[92][32]

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.[93][32]

[edit] 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.[94]

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.[95]

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.[96] 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.[97]

[edit] The money returns: the fifth generation project

In 1981, the Japanese Ministry of International Trade and Industry set aside $850 million dollars for the Fifth generation computer project. Their objectives were to write programs and build machines that could carry on conversations, translate languages, interpret pictures, and reason like human beings.[98] Much to the chagrin of scruffies, they chose Prolog as the primary computer language for the project.[99]

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.[100][32] DARPA responded as well, tripling its investment in AI between 1984 and 1988.[101]

[edit] 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. Douglas Lenat began a multi-decade project to encode commonsense knowledge. Rodney Brooks and Hans Moravec, researchers from the related field of robotics, argued for an entirely new approach to artificial intelligence.

[edit] AI winter

Researchers who had survived the funding cuts of 1974, such as Roger Schank and Marvin Minsky, became concerned that enthusiasm for expert systems had spiraled out of control, and that disappointment would certainly follow. In 1984, the term AI winter was coined to describe an era of pessimism and funding cuts after an era of high expectations and hype.[102]

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.[103]

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.[104]

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 in 2001, and may not be met by 2011. As with other AI projects, expectations had run much higher than what was actually possible.[105]

[edit] Cyc: an encyclopedia of common sense

In 1988, Doug Lenat made this announcement: "I would like to present a surprisingly compact, powerful, elegant set of reasoning methods that form a set of first principles which explain creativity, humor and common sense reasoning ... but, sadly, I don't believe they exist. So, instead, this paper will tell you about Cyc, the massive knowledge base project that we've been working on at MCC for the last four years."[106] Lenat wanted to attack the commonsense knowledge problem directly, by creating a massive database that would contain all the mundane facts that the average person knows. He believed, as many others do, 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 originally expected to take only two person-centuries, but all indications are that it will take much longer.[107]

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

Main articles: Nouvelle AI, behavior-based AI, situated, and embodied cognitive science

In the late 80s, a number of researchers advocated a completely new approach to artificial intelligence, based on robotics. 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."[108]

The approach revived ideas from cybernetics and control theory that had been unpopular since the sixties. David Marr came 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 was cut short by leukemia in 1980.[109]

In a 1990 paper 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."[110] 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.[111]

[edit] 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. AI was more cautious and more successful than it had ever been.

[edit] 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. In 1995 the VaMP car of Ernst Dickmanns drove up to 158 kilometers in fast traffic without human intervention, reaching speeds up to 180 kilometers per hour (112 mph). Ten years later, a Stanford robot won the DARPA Grand Challenge by driving autonomously for 131 miles along an unrehearsed desert trail. At the turn of the millennium the first sophisticated walking humanoid robots were built. 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.[112] 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.[113] Thanks to Moore's law, the fundamental problem of "raw computer power" was slowly being overcome.

[edit] Intelligent agents

A new paradigm called "intelligent agents" became widely accepted during the 90s.[114] Although earlier researchers had proposed modular "divide and conquer" approaches to AI,[115] the intelligent agent did not reach its modern form until Judea Pearl, Alan Newell and others brought concepts from decision theory and economics into the study of AI.[116] When the economist's definition of a rational agent was married to computer science's definition of an object or module, the intelligent agent paradigm was complete.

An intelligent agent is a system that perceives its environment and takes actions which maximizes 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.[117]

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.[116]

[edit] Victory of the neats

AI researchers began to develop and use sophisticated mathematical tools more than they ever had in the past. 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. Russell & Norvig (2003) describe this as nothing less than a "revolution" and "the victory of the neats."[118] For some, it was a rejection of "good old fashioned" symbolic AI.

Among the new tools were Bayesian networks, hidden Markov models, information theory, stochastic modeling and classical optimization. Judea Pearl's highly influential 1988 book[119] brought probability and decision theory into AI. Precise mathematical descriptions were made of "computational intelligence" paradigms like neural networks and evolutionary algorithms.[118]

[edit] 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,[120] such as data mining, industrial robotics, logistics,[121] speech recognition,[122] banking software,[123] medical diagnosis[123] and Google's search engine[124] to name a few.

The field of AI receives little or no credit for these successes. Now no longer considered a part of AI, each has been reduced to the status of just another item in the tool chest of computer science.[125] 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."[126] (This is called the "AI effect" and is expressed most succinctly by Tesler's Theorem: "AI is whatever hasn't been done yet.")[127]

In fact, many researchers in AI today deliberately call their work by other names, such as informatics, knowledge-based 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: 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."[128][129][130]

[edit] Where is HAL 9000?

In 1968, Arthur C. Clark and Stanley Kubrick had imagined that by the year 2001, 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.[131]

By the 21st century, a combination of forces had fragmented the field into a number of related disciplines, all pursuing goals that were once a part of AI.[132] Marvin Minsky complained in 2006 that the central problems, like commonsense reasoning, were being neglected, while the majority of researchers pursued things like commercial applications of neural nets or genetic algorithms. "So the question is why we didn't get HAL in 2001? I think the answer is I believe we could have."[133]

[edit] Notes

  1. ^ McCorduck 2004, pp. 3-35
  2. ^ Among the researchers who laid the foundations of the theory of computation, cybernetics, information theory and neural networks were Claude Shannon, Norbert Weiner, Warren McCullough, Walter Pitts, Donald Hebb, Donald McKay, Alan Turing and John Von Neumann. McCorduck 2004, pp. 51-107, Crevier 1993, pp. 27-32, Russell & Norvig 2003, pp. 15, 940, Moravec 1988, p. 3.
  3. ^ Crevier 1993, pp. 4-7 (the founders of AI), Crevier 1993, pp. 48-50 (the conference, of which he writes "the conference is generally recognized as the official birthdate of the new science."), Crevier 1993, pp. 108-109, (their predictions), Crevier 1993, pp. 64-65 (the money), Crevier 1993, pp. 115-117 (the first AI winter), Crevier 1993, pp. 197-208 (the second AI winter)
  4. ^ See Kurzweil 2005, p. 262, where he predicts that computers will achieve "strong AI" in 2029-2040
  5. ^ Turing 1950
  6. ^ McCorduck 2004, pp. 5-35
  7. ^ McCorduck 2004, p. 5, Russell & Norvig 2003, p. 939
  8. ^ McCorduck 2004, p. 15-16, Buchanan 2005, p. 50 (Judah Loew's Golem). McCorduck 2004, p. 13-14 (Paracelsus)
  9. ^ McCorduck 2004, pp. 17-25, Project Gutenberg eBook Erewhon by Samuel Butler.
  10. ^ Needham 1986, p. 53
  11. ^ McCorduck 2004, p. 6
  12. ^ A Thirteenth Century Programmable Robot
  13. ^ McCorduck 2004, p. 17 and see also Levitt 2000
  14. ^ Quoted in McCorduck 2004, p. 8. Crevier (1993), p. 1) and McCorduck (2004), pp. 6-9) discusses sacred statues.
  15. ^ McCorduck 2004, pp. 37-46, Haugaland 1986, chpt. 2, Russell & Norvig 2003, p. 6 and Buchanan 2005, p. 53
  16. ^ McCorduck 2004, p. 42, Hobbes 1651, chapter 5
  17. ^ McCorduck 2004, p. 41, Berlinski 2000, p. 12, Buchanan 2005, p. 53, Russell & Norvig 2003, p. 6
  18. ^ Russell & Norvig 2003, p. 14-15
  19. ^ McCorduck 2004, pp. 63-64, Crevier 1993, pp. 22-24, Russell & Norvig 2003, p. 8
  20. ^ The starting and ending dates of the sections in this article are adopted from Crevier 1993 and Russell & Norvig 2003, p. 16−27. Themes, trends and projects are treated in the period that the most important work was done.
  21. ^ Crevier 1993, pp. 3−4,48−50 and Russell & Norvig 2003, p. 16−17
  22. ^ Crevier 1993, p. 30, Russell & Norvig 2003, p. 15−16 and see also Pitts & McCullough 1943
  23. ^ Moravec 1988, p. 3, Crevier 1993, pp. 27−28 and Russell & Norvig 2003, pp. 15,940
  24. ^ Crevier 1993, pp. 34−35 and Russell & Norvig 2003, p. 17
  25. ^ Turing 1950, McCorduck 2004, pp. 70−72, Haugeland 1985, pp. 6−9, Crevier 1993, p. 22−25, Russell & Norvig 2003, pp. 2−3 and 948
  26. ^ Crevier 1993, pp. 44−46 and Russell & Norvig 2003, p. 17
  27. ^ Crevier 1993, p. 46 and Russell & Norvig 2003, p. 17
  28. ^ Russell & Norvig 2003, p. 947,952
  29. ^ See A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence McCarthy et al. 1955. Also see 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. (Newell & Simon 1963) It includes a more specific definition of a "machine" as an agent that manipulates symbols. See the philosophy of artificial intelligence.
  30. ^ Crevier 1993, pp. 39−41 and Russell & Norvig 2003, p. 17
  31. ^ Crevier writes of McCarthy "He lays no claim to having coined the phrase and admits that it may have been used casually before." (Crevier 1993, p. 50) However, McCarthy states unequivocally "I came up with the term" in a c|net interview. (See Getting Machines to Think Like Us.)
  32. ^ a b c d Russell & Norvig 2003, p. 25.
  33. ^ (Crevier 1993, pp. 49−50), who writes "the conference is generally recognized as the official birthdate of the new science."
  34. ^ Russell and Norvig write "it was astonishing whenever a computer did anything remotely clever." Russell & Norvig 2003, p. 18
  35. ^ Crevier 1993, pp. 52−107, Moravec 1988, p. 9 and Russell & Norvig 2003, p. 18−21.
  36. ^ Crevier 1993, pp. 108−109 and Russell & Norvig 2003, p. 21.
  37. ^ Crevier 1993, pp. 52−107, Moravec 1988, p. 9
  38. ^ Russell & Norvig 2003, pp. 59−61
  39. ^ Russell & Norvig 2003, pp. 21−22
  40. ^ Crevier 1993, pp. 51−58,65−66 and Russell & Norvig 2003, pp. 18−19
  41. ^ Crevier 1993, pp. 95−96, Moravec 1988, pp. 14−15
  42. ^ Crevier 1993, pp. 76−79 and Russell & Norvig 2003, p. 19.
  43. ^ Crevier 1993, pp. 79−83
  44. ^ Crevier 1993, pp. 164−172
  45. ^ Crevier 1993, pp. 134−139
  46. ^ Crevier 1993, pp. 83−102, Russell & Norvig 2003, p. 19 and see also Micro-World AI
  47. ^ Crevier 1993, pp. 84−102 and Russell & Norvig 2003, p. 19
  48. ^ Simon & Newell 1958, p. 7−8 quoted in Crevier 1993, p. 108. See also Russell & Norvig 2003, p. 21
  49. ^ Simon 1965, p. 96 quoted in Crevier 1993, p. 109
  50. ^ Minsky 1967, p. 2 quoted in Crevier 1993, p. 109
  51. ^ Minsky strongly believes he was misquoted. See Crevier 1993, p. 96 and Darrach 1970.
  52. ^ Crevier 1993, pp. 64−65
  53. ^ Crevier 1993, p. 94
  54. ^ Howe 1994
  55. ^ Crevier 1993, p. 51
  56. ^ Crevier 1993, p. 65
  57. ^ Crevier 1993, pp. 68−71 and Turkle 1984
  58. ^ Crevier 1993, pp. 100−144 and Russell & Norvig 2003, pp. 21−22
  59. ^ Crevier 1993, pp. 102−105 and Russell & Norvig 2003, p. 22
  60. ^ Crevier 1993, pp. 163−196
  61. ^ Crevier 1993, p. 146
  62. ^ Russell & Norvig 2003, pp. 20−21
  63. ^ Crevier 1993, pp. 146−148, see also Buchanan 2005, p. 56: "Early programs were necessarily limited in scope by the size and speed of memory"
  64. ^ 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 Getting machines to think like us
  65. ^ Russell & Norvig 2003, pp. 9,21−22 and Lighthill 1973
  66. ^ Crevier 1993, pp. 113−114, Moravec 1988, p. 13, Lenat 1989 (Introduction) and Russell & Norvig 2003, p. 21
  67. ^ Moravec 1988, pp. 15−16
  68. ^ McCarthy & Hayes 1969, Crevier 1993, pp. 117−119
  69. ^ Crevier 1993, p. 110, NRC 1999 under "Success in Speech Recognition" and Russell & Norvig 2003, p. 21.
  70. ^ Lighthill 1973, Crevier 1993, p. 117, Russell & Norvig 2003, p. 22 and Howe 1994
  71. ^ Lighthill 1973, Russell & Norvig 2003, p. 22. John McCarthy wrote in response that "the combinatorial explosion problem has been recognized in AI from the beginning" in Review of Lighthill report
  72. ^ Crevier 1993, pp. 115−116
  73. ^ Crevier 1993, p. 115
  74. ^ NRC 1999 under "Shift to Applied Research Increases Investment."
  75. ^ Lucas 1961, Hofstadter 1980, pp. 471−477, Crevier 1993, p. 22
  76. ^ Dreyfus 1972, Crevier 1993, pp. 120−132
  77. ^ Searle 1980, Crevier 1993, pp. 269−271
  78. ^ Crevier 1993, p. 143
  79. ^ Crevier 1993, p. 122
  80. ^ "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 Crevier 1993, p. 123.
  81. ^ Crevier 1993, pp. 132−144, McCorduck 2004, pp. 356-373, Russell & Norvig 2003, p. 961 and Weizenbaum 1976
  82. ^ Crevier 1993, pp. 102−105, McCorduck 2004, pp. 104−107, Russell & Norvig 2003, p. 22
  83. ^ McCorduck 2004, p. 51, Russell & Norvig 2003, pp. 19, 23
  84. ^ Crevier 1993, pp. 190−192
  85. ^ Crevier 1993, pp. 193−196
  86. ^ Crevier 1993, pp. 145−149,258−63
  87. ^ Wason (1966) showed that people do poorly on completely abstract problems, but if the problem is restated to allowed the use of intuitive social intelligence, performance dramatically improves. (See Wason selection task) Tversky, Slovic & Kahnemann (1982) have shown that people are terrible at elementary problems that involve uncertain reasoning. (See list of cognitive biases for several examples). Eleanor Rosch's work is described in Lakoff 1987
  88. ^ 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 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 McCarthy's presentation at AI@50)
  89. ^ Crevier 1993, pp. 175
  90. ^ Crevier 1993, pp. 168
  91. ^ Minsky 1974, Crevier 1993, pp. 170−173, 246 and Russell & Norvig 2003, p. 24
  92. ^ Crevier 1993, pp. 214−215.
  93. ^ Crevier 1993, pp. 215−216.
  94. ^ Crevier 1993, pp. 148−159 and Russell & Norvig 2003, pp. 22−23
  95. ^ Crevier 1993, pp. 158−159 and Russell & Norvig 2003, p. 23−24.
  96. ^ Crevier 1993, p. 198
  97. ^ Crevier 1993, pp. 161−162,197−203 and Russell & Norvig 2003, p. 24
  98. ^ Crevier 1993, pp. 211, Russell & Norvig 2003, p. 24 and see also Feigenbaum & McCorduck 1983
  99. ^ Crevier 1993, pp. 195
  100. ^ Crevier 1993, pp. 240.
  101. ^ NRC 1999 under "Shift to Applied Research Increases Investment"
  102. ^ Crevier 1993, pp. 203. AI winter was first used as the title of a seminar
  103. ^ Crevier 1993, pp. 209−210
  104. ^ Crevier 1993, pp. 204−208, Lenat & Guha 1989, Introduction
  105. ^ Crevier 1993, pp. 212
  106. ^ Lenat & Guha 1989, pp. xvii
  107. ^ Crevier 1993, pp. 239−243, Russell & Norvig 2003, p. 363−365 and Lenat & Guha 1989
  108. ^ 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."
  109. ^ Crevier 1993, pp. 183−190
  110. ^ Brooks 1990, p. 3
  111. ^ For example, Lakoff & Turner 1999
  112. ^ 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."
  113. ^ 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.
  114. ^ "The whole-agent view is now widely accepted in the field" Russell & Norvig 2003, p. 55. The intelligent agent paradigm is discussed in major AI textbooks, such as: Russell & Norvig 2003, pp. 32−58, 968−972, Poole, Mackworth & Goebel 1998, pp. 7−21, Luger & Stubblefield 2004, pp. 235−240
  115. ^ For example, both John Doyle (Doyle 1983) and Marvin Minsky's popular classic The Society of Mind (Minsky 1986) used the word "agent". Other "modular" proposals included Rodney Brook's subsumption architecture, object-oriented programming and others.
  116. ^ a b Russell & Norvig 2003, pp. 27, 55
  117. ^ This is how the most widely accepted textbooks of the 21st century define artificial intelligence. See Russell & Norvig 2003, p. 32 and Poole, Mackworth & Goebel 1998, p. 1
  118. ^ a b Russell Norvig, p. 25−26
  119. ^ Pearl 1988
  120. ^ NRC 1999 under "Artificial Intelligence in the 90s", and Kurzweil 2005, p. 264
  121. ^ Russell & Norvig 2003, p. 28
  122. ^ For the new state of the art in AI based speech recognition, see Are You Talking to Me?
  123. ^ a b "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, AI set to exceed human brain power CNN.com (July 26, 2006)
  124. ^ For the use of AI at Google, see Google's man behind the curtain, Google backs character recognition and Spying an intelligent search engine.
  125. ^ Kurzweil 2005, p. 265
  126. ^ AI set to exceed human brain power CNN.com (July 26, 2006)
  127. ^ As quoted in Hofstadter 1979:601. Larry Tesler actually feels he was misquoted: see his note at the bottom of Larry Tesler's Resume
  128. ^ Markoff, John. "Behind Artificial Intelligence, a Squadron of Bright Real People", The New York Times, 2005-10-14. Retrieved on 2007-07-30. 
  129. ^ Alex Castro (2007) Are you talking to me? The Economist Technology Quarterly (June 7, 2007)
  130. ^ Patty Tascarella, Robotics firms find fundraising struggle, with venture capital shy. Pittsburgh Business Times (August 11, 2006)
  131. ^ Crevier & 1993 108−109
  132. ^ "In many cases, formalization and specialization have led to fragmentation." Russell & Norvig 2003, p. 26
  133. ^ He goes on to say: "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 It's 2001

[edit] References