Term Artificial Intelligence
The term artificial intelligence is the name of a research field and its goal, introduced by John McCarthy in the August 1955 proposal for a summer research project at Dartmouth College, signed by McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon, and then publicized by the workshop held there in the summer of 1956. It names an aim, making a machine behave in ways that would be called intelligent if a person behaved that way, rather than any particular technology, which explains both why the label has lasted and why it has been argued over ever since.
McCarthy's own account was that the name was a deliberate choice of a neutral banner. In the mid-1950s the territory was divided among several labels and communities. Norbert Wiener's cybernetics, from 1948, was oriented toward feedback, control and analog systems; automata theory concerned abstract machines and was the subject of the volume Automata Studies, edited by Shannon and McCarthy (Princeton, 1956); and a growing group working on programs that played games or proved theorems had no common name. McCarthy later said he wanted to avoid being tied to Wiener's cybernetics and was dissatisfied that the submissions to Automata Studies were not much about intelligence, which pushed him to a name that foregrounded the goal. The proposal used the new phrase in its title and stated the working conjecture that any feature of learning or intelligence could in principle be described precisely enough for a machine to simulate it.
Alternatives were available and were used by serious people. Allen Newell and Herbert Simon, whose Logic Theorist was the most visible working result at Dartmouth, preferred complex information processing and kept using it for years because it described what the programs did, handling symbols, without asserting that the result was intelligent. Machine intelligence became the British and Edinburgh label, notably in the Machine Intelligence series that Donald Michie began in 1967, and heuristic programming and, later, machine learning, knowledge engineering and cognitive science each claimed part of the area. Alan Turing's 1950 paper had spoken of machines that think, and Minsky's 1961 Steps toward artificial intelligence helped consolidate the phrase in the literature.
The name stuck for institutional rather than intellectual reasons. It was short and ambitious, it arrived with a funded workshop, and the people who attended, McCarthy at Stanford and Minsky at MIT, soon led laboratories that used it in their names, which made it the default term for grant-makers and journalists. It was also accommodating: because it defined the field by a goal, it could absorb new methods, whether logic, search, neural networks or statistical learning, without changing its name.
The critiques were part of its history from early on. Hubert Dreyfus's 1965 RAND paper Alchemy and artificial intelligence and his 1972 book What computers can't do attacked the assumptions of the program; James Lighthill's 1973 report to the British Science Research Council, Artificial Intelligence: A General Survey, led to sharp cuts in British funding; and Drew McDermott's 1976 essay Artificial intelligence meets natural stupidity criticized wishful labels, naming a program's component after a human faculty it only vaguely resembles. A different complaint, often attributed to Larry Tesler, is the AI effect, in which any capability that works, such as chess play or optical character recognition, stops being called intelligence. Some researchers, including Michael Jordan, have argued that the label obscures what the systems are, and prefer terms such as machine learning or intelligence augmentation.
Two misconceptions are common. One is that the term was coined to describe a settled definition or a single technology; it was a funding-proposal phrase for a research bet. The other is that it was coined in 1956 at the workshop itself; the phrase is in the 1955 proposal, while the 1956 event is when it gained a community. The neighbors are distinct entries: the Dartmouth proposal is the document, the Dartmouth conference is the event, and this entry concerns the word.
Read the term artificial intelligence as a flag rather than a definition: a name picked in 1955 to rally researchers around a goal, which has stayed useful precisely because it never specified a method.