next end up end toc toc mail

13.5.  The Turing Test

Some people do not consider the Turing Test as an indicative of whether a machine can think. In any case, no machine has passed the Turing Test, so far.

The test itself is quite straightforward. There are two subjects, one a human, the other a computer. They are hidden from a human interrogator, who does not know which is which. The interrogator is free to ask any question(s) to the subjects, and both respond in such a way as to not give away their identities (by printouts, for example). If after qustioning, the interrogator cannot determine which subject is human and which is the computer, then the computer will have passed the Turing Test, i.e., it will have demonstrated human intelligence. Implicitely, the interrogator will probably attempt to ``outsmart'' the computer subject by asking questions that require a human's common sense, creativity, or perhaps emotional response. These characteristics are, as far as we know, either impossible or extremely difficult to implement in a machine.

Computer scientists have found that machines are good in solving specialized problems -- or at least we know how to use machines to solve specialized problems -- but problems where ``common sense'' reasoning play an important role are still a challenge. Some of these problems include:

--   natural language understanding,
--   pattern recognition,
--   game playing,
--   medical diagnoses,
--   identifying objects in a picture,
--   advice in business,
--   text-to speech translation and vice-versa.

back end up end toc toc mail


Created by unroff & hp-tools. © by Hans-Peter Bischof. All Rights Reserved (1998).

Last modified: 27/July/98 (12:14)