Artificial Intelligence: History, Concepts, and the Future

Introduction to Artificial Intelligence

Not even a century ago — in fact, not even a half-century ago — few people could have imagined the present-day world with computers operating most government and business processes and the Internet running in millions of homes. Thus it would have been nearly impossible to comprehend artificial intelligence (AI) and the idea that scientists would try to create a machine capable of learning, adapting, reasoning, and correcting or improving itself. Whether or not this will become a reality is still unknown.

AI pioneer Chris Langton says that this "intelligent entity" will never be possible. He believes that "when scientists are faced with the choice of either admitting that the computer process is alive, or moving the goalposts to exclude the computer from the exclusive club of living organisms, they will choose the latter." Is this true? Will humans never admit that a computer can actually function as real life? Or will they instead decide there is nothing special about life, and that humanity can therefore be designed, built, and replicated? At least for the time being, there is no answer to this dilemma.

According to the American Association for Artificial Intelligence, AI is "the scientific understanding of the mechanisms underlying thought and intelligent behavior and their embodiment in machines." The evolution of this science became notable as early as 1821, when Charles Babbage stared at a table of logarithms and said, "I think that all these tables might be calculated by machinery." From that point on, he devoted his life to developing the first programmable computer.

Origins and Early Milestones

Much later, in 1943, Babbage's idea finally took hold when Warren McCulloch — a psychiatrist, cybernetician, philosopher, and poet — and Walter Pitts, a research student in mathematics, published an innovative paper combining early twentieth-century ideas on computation, logic, and the nervous system. The report promised to revolutionize both psychology and philosophy. The following year, those ideas were applied to develop the first American programmable computer, the Merck I.

It did not take long for British scientist Alan Turing to recognize the similarity between the computational process and human thinking. In his paper "Computing Machinery and Intelligence," he outlined the direction for the remainder of the century: developing computers for game playing, decision-making, natural language understanding, translation, theorem proving, and encryption code cracking.

In the 1950s, Newell, Shaw, and Simon created the program Logic Theorist (later called General Problem Solver), which used recursive search techniques — defining a solution in terms of itself. IBM developed the first program capable of playing a full game of chess in 1957. The following year, Newell, Shaw, and Simon observed: "There are now in the world machines that think, that learn and that create. Moreover, their ability to do these things is going to increase rapidly until — in a visible future — the range of problems they can handle will be co-extensive with the range to which the human mind has been applied" (Simon, p. 3).

The Turing Test and the Question of Intelligence

To help determine if and when a computer had actually become intelligent, Turing proposed the "imitation game," in which an interrogator would interview both a human being and a computer without knowing which was which, with all communication conducted through textual messages. Turing argued that if the interrogator could not distinguish between the two through questioning, it would be unreasonable to deny that the computer was intelligent. This concept is now commonly known as the Turing test for intelligence.

In 1967, a university computer won the first tournament match against a human chess player. World chess champion Gary Kasparov declared in 1988 that there was "no way" a grandmaster would be defeated by a computer in a tournament before 2000. Ten months later, he lost the bet. However, many people changed their position and argued that winning a championship chess game did not require "real" intelligence. For a number of observers, the connection between human and machine was becoming a little too close for comfort.

This was precisely why Turing had developed his test — any other attempt to define intelligence seemed to run into problems. The AI specialists themselves were not especially excited by the chess victory, because the computer — which relied on custom chips inside a machine — was seen as a kind of idiot savant, able to play a strong game of chess without any understanding of why it did what it did (Gershenfeld, 1999, p. 130).

As one critic noted, this represents a curious argument: it retroactively adds a clause to the Turing test, demanding that not only must a machine be able to match human performance at quintessentially intelligent tasks such as chess or conversation, but that the way it does so must also be deemed satisfactory (ibid).

Since then, the central question has shifted from whether an advanced computer can be built to what intelligence actually is. Fatmi and Young define intelligence as "that faculty of mind by which order is perceived in a situation previously considered disordered." Another definition holds that something intelligent should be able to improve its own processes (Martin, 2000, p. 46).

By the mid-1960s, AI was being pursued by researchers worldwide, though computer memory remained very limited. Perception and knowledge representation became the theme of many AI studies. For example, in the Blocks Micro World project, the computer program SHRDLU viewed a collection of pure geometric shapes through cameras, interpreted what it saw, and then manipulated blocks while expressing its perceptions, activities, and motivations. SHRDLU could respond to commands typed in natural English, such as "Will you please stack up both of the red blocks and either a green cube or a pyramid?" The program would plan a sequence of actions, and a robot arm would arrange the blocks accordingly. SHRDLU could correctly answer questions about its environment — for example, "Can a pyramid be supported by a pyramid?" (SHRDLU would attempt to stack two pyramids, fail, and report accordingly) and "Is there anything which is bigger than every pyramid but is not as wide as the thing that supports it?" (to which SHRDLU answered, "Yes, the blue block").

Although SHRDLU was initially regarded as a major breakthrough, its developer Winograd soon announced that the work was a dead end. The techniques pioneered in the program proved unsuitable for wider applications. Moreover, the appearance that SHRDLU gave of understanding the blocks and the English statements describing them was, in fact, false — SHRDLU had no idea what a red block actually was.

References

Copeland, B. J. (2000). What is AI? Retrieved May 21, 2004, from http://www.alanturing.net/turing_archive/pages/Reference%20Articles/what_is_AI/What%20is%20AI07.html

Freeman, D. (1990). Common sense and the computer. Discover, 11, 64–71.

Gershenfeld, N. (1999). When things start to think. New York: Henry Holt.

Kurzweil, R. (1999). Age of spiritual machines. New York: Penguin Press.

Leibniz, G. W. F. von (1969). On the general characteristic. In L. E. Loemmker (Ed.), Leibniz: Philosophical papers and letters. Boston: Reidel.

Martin, J. (2000). Alien intelligence. Washington, D.C.: Capital Press.

Simon, H., & Newell, A. (1958). Heuristic problem solving: The next advance in operations research. Operations Research, 6, 1–10.

Ward, M. (1999). Virtual organisms. New York: Thomas Dunne.