PROUDfOOT | 289
just a machine’. Deep Blue showed ‘a very human sense of danger’.^6 Disconcerted, Kasparov
went on to lose the match. A computer, it was declared, had ‘passed a Turing chess test.
A grandmaster cannot know whether his hidden opponent is another grandmaster or a com-
puter program’.^7
In 1950 Turing presented his test as it is generally known today—an imitation game extend-
ing his 1948 experiment to include questions on ‘almost any one of the fields of human endeav-
our that we wish to include’. (For an expanded introduction to the test, see Chapter 25.) In his
paper ‘Computing machinery and intelligence’ he first introduced an experiment in the form
of two simultaneous interviews, again with three players in total: an interrogator questions
two contestants, a man and a woman. Both contestants are again hidden, and the interrogator
conducts the interviews by teleprinter. The male contestant’s goal is to fool the interrogator into
misidentifying him as the woman. Having described this setup, Turing said:
We now ask the question, ‘What will happen when a machine takes the part of [the man] in this
game?’ Will the interrogator decide wrongly as often when the game is played like this as he does
when the game is played between a man and a woman? These questions replace our original,
‘Can machines think?’.
Satisfactory performance by a machine in such a game is a ‘criterion for “thinking” ’, Turing
said.^8 The machines that he had in mind were digital computers. If some computer does well in
this computer-imitates-human game, machines can think.
Rules of the game
Much of the vast literature on the Turing test consists of conflicting accounts of Turing’s design
for the 1950 imitation game. Several commentators claim that the computer’s task is to imitate
a man (since the computer is introduced as taking the man’s place in a man-imitates-woman
game), or to imitate a woman (since that is the man’s goal), or even to imitate a man who is
imitating a woman. The arguments offered on behalf of the claim that the machine’s task is to
imitate a woman include: this makes the test harder for the machine (which must convince the
interrogator both that it is human and that it is female); it makes the test easier for the machine
(since an interrogator looking out for ‘female’ behaviour is less likely to spot ‘machine’ behav-
iour); and a ‘gender-twisting’ test would have ‘resonated deeply’ with Turing.^9 However, Turing,
when he said that the point of the test is to see if a computer can ‘imitate a brain’, made it clear
that the machine’s task is in fact to imitate a human being (male or female).^10
Turing provided guidelines for all three players. The interrogator is to be ‘average’—just as
the interrogator in the chess-playing imitation game is to be a ‘rather poor’ chess player. This
excludes experts who might easily spot typical computer strategies. The interrogator is also
allowed to make comments such as ‘I put it to you that you are only pretending to be a man’.
This, along with the freedom to ask questions on almost any subject, prevents a simple pro-
gram’s doing well in the game merely by generating pre-programmed replies to trigger words
on a few topics. (The ‘Eugene Goostman’ program, alleged to pass the test on the sixtieth anni-
versary of Turing’s death, simulated a 13-year-old Ukrainian boy speaking English as a second
language; this restricted questions to those that such a contestant would understand, violating
Turing’s guidelines.) In turn the computer is ‘permitted all sorts of tricks’—it would, Turing
said, ‘have to do quite a bit of acting’. Turing suggested, for example, that if the computer were