TURING's TEST
| Collection | International Encyclopedia of Systems and Cybernetics |
|---|---|
| Year | 2004 |
| Vol. (num.) | 2(2) |
| ID | ◀ 3658 ▶ |
| Object type | General information, Epistemology, ontology or semantics |
The English mathematician A. TURING proposed a test to find out if a given machine can be considered intelligent (1950).
As a synthesis, in the words of P. DENNING: the test is “… an imitation game in which an interrogator asks questions of a human being and a machine; if the interrogator is unable to distinguish between the two, the machine passes the test and is declared intelligent. TURING replaced the question ”Can a machine think?“ (Title however of his quite more exhaustive l950 essay — see 1956) with ”Can the interrogator distinguish the two in an imitiation game ?“ because he considered the former question so imprecise as to be meaningless. His own opinion was that by the year 2000 there would exist machines capable of fooling the interrogator for at least five minutes in 30 percent of the games played” (1990, p.100).
J. WEIZENBAUM's ironic book about his ELIZA program for psychological study. shows that the very acumen of the interrogator should also be tested in order to validate the test (1976).
At the time TURING devised his test, only sequential computers were in existence, and their potential performances where abysmally low, compared with the present ones. However, the algorithmic character of sequential computers programs still seems to rule out intelligence for this class of machines, even in chess playing or expert systems. However, the new connection machines and neural type networks, which are able to learn could easily reactualize TURING's test in a new way.