Can Machines Think?

Alan Turing

Alan Turing in his famous “Computing Machinery and Intelligence” (published in 1950) initially addresses the question “can machine think?” but later he forms the problem in terms of a game, which is called the “Imitation Game”. The reason for that approach is to avoid the ambiguity between the concepts “machine” and “think”.

The “Imitation Game” defines that three players are allowed to play: a man (A), a woman (B) and an interrogator (C) who may be of either sex. The object of the game is the determination of the sex between A and B by C. The interrogator is based in a closed room and he/she is allowed to put questions to A and B and give the conclusive answer in the form of determination of their gender identity. The point is for C to think and then deduce the answer from the way his questions about the physique of the persons in question are answered. A and B voices since they reveal their gender identity must be “transferred” into a typewritten form. Therefore, if the game requires a considerable thinking process and also if the A and B are replaced with a machine, the question that must be asked is if C is able to distinguish between the machine and the persons. Hence, Turing initial question “can machines think” after the replacement of the “Imitation Game” concludes to the determination criterion which states that a machine can think arises from the interrogator’s inability to distinguish the difference between the machine and the person.

The argument “machines can never make mistakes”.

Turing avoids to answer the simple question whether or not machine can make mistakes. He adopts “a more sympathetic attitude” and he introduces the distinction between errors of functioning and errors of conclusion. He mentions that the question whether or not machines can make mistakes depends on a confusion between those two kinds of mistakes. Firstly, he states that errors of functioning are based on an electrical or mechanical fault, which causes the machine to operate not the way that it was designed to. Considering a philosophical perspective and assuming that those errors could not be arise because of the mathematical fiction of those “abstract machines” then “machines can never make mistakes”. However, he emphasizes that in case they are physical objects (as practical machines are) and therefore not theoretical discussions then those errors are inevitable in case of mechanical or electrical faults. Secondly, Turing talks about errors of conclusion that “can only arise when some meaning is attached to the output signals of the machine”. An example of this kind of errors is having an output “30” in response to the input “15+16”. Humans make many errors of this second class and Turing points out that machines could also make errors of conclusion. If a machine is programmed to do for example mathematical operations in the way that people actually solve mathematical operations then the machine might as humans, answer to some mathematical operations with an incorrect result.