McCarthy, then employed at Dahrmounth College, his acquaintance from the studies and inventor of “neural nets” Marvin Minsky, Rochester, founder of the information theory Claude Shannon and unsigned Oliver Selfridge^17, wrote in August 1955. “A Proposal for the Dartmouth summer research project on artificial intelligence” and send it to a possible financier. Attractive and pretencious, previously almost unused fraze “artificial intelligence” which was coined by McCarthy^19 was soon accepted as a name for the entire field of computer technology.
3.1 LANGUAGE OF THE INTELLIGENT COMPUTERS
Proponents belived a computer can imitate every aspect of human intelligence and anounced attempts of solving many hard problems. They even anounced particular interests and plans. McCarthy belived it was necessary for the development of intelligent machines to apply standard methods of trial and error on “higher level of abstraction”. Just like humans use language for solving complex problems by making propositions and trying them, so would intelligent machines do as well. He intended to develop a language suitable for such use.^30 Already developed languages were easy to describe with informal mathematic and informal mathematic was easily translated into those languages, and it was also easy to test for correctness of the proof. The language of intelligent machines should also have some advantages of natural languages: be concisive, universality (in a natural language it is possible to define and adequately use any language), selfreferencing and propositional expression.
Probably from that period, there is a preserved short, undated McCarthy’s manuscript The programming problem which contains almost identical ideas, but also points out that language should be explicit: there should not be possibility for different interpretations of a procedure’s meaning.^21
Lisp can be viewed as an attempt to realize McCarthy’s ambition from 1955.
3.2 IN-BETWEEN IPL AND FORTRAN
The summer project was accepted and realized next, 1956. year. Despite participation of tenth of most famous researchers, expected breakthroughs were not achieved. Reasons of relative unsuccess were later explained by McCarthy with shortage of financial means, poor colaboration between researchers who hold to their own projects and difficultness of the problem which propponents underestimated. Minsky presented an idea for a geometry theorem solver. Ray Solomonoff started work on algorithmic complexity and Alex Bernstein presented a chess program. Instead of a work on the language, McCarthy presented “alfa-beta heuristic” for games like chess.^22
Despite not developing the announced language, McCarthy acquainted himself with work of Allen Newell, Cliff Shaw and Herbert Simon who prestened program LOGIC THEORIST (LT) written in INFORMATION PROCESSING LANGUAGE (IPL).^33 IPL supported single linked lists and recursions. Commands were calls to subprograms and could not be directly composited. McCarthy felt back then a need, even a passion, for “algebraic language” in which expressions would be written as in mathematic or Fortran, but which, like IPL, would make lists processing and recursion possible. Such language would significantly simplify expression analysis and subexpression refactoring compared to IPL.^34
17 McCarthy, The logic and philosophy of artificial intelligence, 1988. p3. 18 “Do not develop your artificial intelligence, but develop that intelligence which is from God. From the latter results virtue; from the former, cunning.” Giles, Cuhang Tzu - mystic, moralist and social reformer, 1889. p232. 19 Andresen, John McCarthy: father of AI, 2003, p84. 20 McCarthy et al.,/A proposal …/, 1955., p10. 21 Stoyan, Early LISP History (1956-59), 1984. p.300. 22 McCarthy, Dartmouth and beyond, 2006 23 Newell & Shaw, Programing the Logic Theory Machine, 1957 24 McCarthy, History of LISP, 1981., p. 174