Test of Turing

The test of Turing is a proposal for a test of Artificial intelligence having faculty to imitate the human conversation. Described by Alan Turing in 1950 in its publication “Computing machinery and intelligence”, this test consists in putting in verbal confrontation human with a computer and another human at the blind man. If the man who engages the conversations is not able to say which is the computer and which is the other man, one can consider that the software of the computer passed successfully the test. That implies that the computer and the man will try to have an appearance human Sémantique. To preserve the simplicity and the universality of the test, the conversation is limited to a textual exchange between the protagonists.

History

The test was inspired by a play of imitation in which a man and a woman go in separate parts and the guests try to discuss with the two protagonists by writing questions and by reading the answers which are returned to them. In this play the man and the woman try to convince the guests who they are both of the women.

In the Turing beginning imagined this test to answer its existential question: “can a machine think? ”, by giving a more concrete interpretation of its question.

An interesting idea of its proposal for a test is that the answers must be given in definite time intervals. He imagines that is necessary so that the observer cannot establish a conclusion which is based on the fact that a computer can answer more quickly than a man, especially on questions of mathematics.

Objections and answers

Turing itself suggested many objections which can be made with the test and an answer in its initial publication gave some:

theological Objection : The thought would be the innate fact of the heart with which the man would only be equipped, and thus the machine could not think. Turing answers that he does not see any reason for which God could not give to a computer a heart if it wished it.
Argument of the conscience : This argument suggested by the professor Jefferson Lister says that “no machine can write a Sonnet or compose a Concerto because of the absence of emotion, and even by aligning notes randomly, one cannot say that a machine can equalize an human brain”. The answer of Turing is that we them men do not have any means of knowing truly the experiment of the emotions of any other individual but oneself, and thus which should accept to us the test.
Originality : one of the principal objections is that the computers are unable to have originality. Turing answers that the computers can surprise the human ones, in particular when the consequences of various facts are not immediately recognizable.
Formalism : This argument says that each system controlled by Loi S can thus be foreseeable and not really intelligent. Turing answers that this amounts confusing laws of the behavior with general rules of control.
Extrasensory perception : Turing seems to suggest that there is evidence of extrasensory perceptions. However it estimates that ideal conditions can be created, in which these perceptions would not affect the test and thus would be negligible.

Predictions and tests

Turing predicted that the computers would be one day able to pass the test. In fact it estimated that in year 2000 of the machines with 128 Mo of memory would be able to mislead approximately 30% of the human judges during a 5 minutes test. He predicted that the human ones at this time there, would not see the expression “machine intelligent” like contradictory. He as predicted as acquisition by training of the computers would be also important to build powerful computers, an assertion which is regarded today as plausible by the contemporary researchers in Artificial intelligence.

In 2004 no computer passed the test of Turing as tel. Of the programmes of simple conversation such as ELIZA misled in human believing to speak with the other human ones, with abstract expressions, like the club-footed AOLiza . In spite of that, such “successes” are not comparable with the test of Turing. Most of the time the man does not have reason of suspecter which he does not speak with human, whereas for the test of Turing the judge tries to determine in an active way nature of the entity with which he converses. The referred cases relate to in particular IRC, where the fact of engaging a futile conversation and without significance, is current. Moreover, much of participants in IRC converse in English without it being their native tongue, which makes even easier the fact of being misled by club-footed, by thinking not all of including/understanding or while not being sufficiently informed of the existence even of the club-footed ones.

The Prix Loebner is an annual competition which makes it possible to determine best the program able to pass the test of Turing. Although the price is allotted each year to the judged software nearest to the human conversation, of which ALICE was a winner on several occasions, the price allotted to the system which would pass the test of forever allotted Turing.

Origin of the name

In the publication of Turing, the term “Play of imitation” is used for its proposal for a test. The name of “Test of Turing” seems to be invented in 1968 by Arthur C. Clarke in its news of science fiction from which was drawn the film “2001, the odyssey of space”.

See too

Captcha, public test of entirely automatic Turing having for goal to discriminate human computers.
Price Loebner
Mark V Shaney (computer program)
Test of Turing reversed, also called Test of Gnirut
Computationnalisme

External bonds

Converse with Eliza (English)
Encyclopédie of Philosophy of Stanford (English) the Test of Turing (English), by G. Oppy and D. Dowe.
Page on the Test of Turing the last articles, bonds, and other information on the Test of Turing (English).
Turing Test: 50 years after return over one half-century of work on the Test of Turing (English).
Bet between Kapor and Kurzweil, including/understanding their respective point of view in detail (English).
In what the Test of Turing is the greatest dead end for the Artificial intelligence by Blay Witby (English).
a humorous vision by proving the not-intelligence of a Twinkie (brioche) (English).
Article of NewYork Times taking again the broad outlines of AD INTERIM share 1 and share 2

References

“Computing machinery and intelligence” by Alan Turing, Mind , vol. LIX, No 236, October 1950, pp. 433-460. Online version: , * There exists a translation of this article: Turing (Alan Mr.), " Computers and the intelligence" , in Thought and machine , ED of the Field Small valley, 1983: '' Minds and machine '', 1964, p. 39-67.
Computers and the Intelligence , in, pp. 133-175
The Emperor' S New Mind by Roger Penrose, a book treating on these subjects (in English)

Simple: Turing test

Random links: