In this artitical, Doctor Turning explored the question "Can machines think?" through a game called "imitation game". The game is played by three people, a man(A), a woman(B) ,and an interrogator(C).C is separated from A and B, its task is to tell who is A(the man) and who is B through asking questions and analyze the answers. A is supposed to trick C to make the wrong judgement(e.g, "I‘m wearing ear studs")and B is supposed to help C make the right judgement and avoid A's traps("Don't listen to him! he's lying!").But of course, since C does not who is A and who is B, so C can't really tell who's trying to trick it and who's helping it.(Turning provided more details and experimental conditions, I didn't list them all here).
The purpose of this game is to see when facing such a condition(which means pure logic won't help very much here), how would a robot "think" and analyze the situation in order to get the right answer? Turning did not give any specific result on this imitation game, instead , he's just trying to materialize the way of "thinking" of a robot into a controllable experiment situation so people can have more ground to base when think over this kind of question.
He also mentioned another situation, that is a robot who participates in the game is supposed to trick its interrogator into believing it's a human; while a human who also participates in this game, tries to trick the interrogator into believing that he/she is a robot.The result, however, is not very favourable to human. For example, when an interrogating question like"please multiply 34957 to 23457" is given, a robot can still pretend to be a human by pausing for a while, pretending that it is calculating on a piece of paper, or simply by answering:"I don't have a pencil and paper with me, I can't do this calculation" while a human would immediately get revealed since a robot is supposed to give out an answer in a split second,but a human just can't.
Turning agrees that a robot might not do very well on test like"what do you think of Shakespeare?". They may give out some very limited and rigid answers. But Turning believes that "the criticism that a machine cannot have much diversity of behavior is just a way of saying that it cannot have much storage capacity.Until fairly recently a storage capacity of even a thousand digits was very rare." He agrees that the way a robot functions is certainly not the same as the nervous system; but "by observing the results fo its own behavior it can modify its own programs so as to achieve some purpose more effectively." So the ideas that one day people might not be able to differentiate a robot from a human in the imitation game or in other fields "are possibilities of the near future, rather than Utopian dreams."
(note: Turning has written a complete and detailed essay on this topic, the points I mentioned here is only part of his research.)
Main source:
Turning, Alan Mathison. Ed.Luger, George F. Computation and Intelligence: Collected Readings.
Menlo Park/Cambridge/London/AAAI Press/ The MIT Press, 1995.
No comments:
Post a Comment