Source
Turing (1950), Computing Machinery and Intelligence, LIT-394 (Mind 59:433–460), read in full in NOTE-355.
What was actually shown
The replacement (§1). “Can machines think?” is “too meaningless to deserve discussion”. In its place: a man (A), a woman (B) and an interrogator (C) communicate by teleprinter; “What will happen when a machine takes the part of A in this game?” Turing never calls this a test of thinking, and never says “Turing test” (NOTE-355, Corrections).
The reduction (§5). A digital computer with adequate storage and speed mimics any discrete-state machine, “considerations of speed apart”. So the question becomes whether one computer, suitably programmed, can “play satisfactorily the part of A in the imitation game, the part of B being taken by a man”.
The prediction (§6). In about fifty years, at storage about 10^9, “an average interrogator will not have more than 70 per cent. chance of making the right identification after five minutes of questioning”.
The objections (§6). The ones that bear on the account:
- Consciousness (Jefferson). Pressed fully, it is solipsism, and most holders would accept the game, as a viva voce, rather than that. Turing adds that he does not think there is “no mystery about consciousness”; the mystery need not be solved first.
- Mathematical. Any one machine has questions it cannot answer, but “it has only been stated, without any sort of proof, that no such limitations apply to the human intellect”.
- Informality of behaviour. It confuses rules of conduct with laws of behaviour, and a 1000-unit Manchester programme already defeats prediction from its replies.
- Lady Lovelace. “Machines take me by surprise with great frequency.”
What Turing concedes (§7). “I have no very convincing arguments of a positive nature to support my views.” The onion and atomic-pile passages are “recitations tending to produce belief”.
What could have come out otherwise. The prediction could, and it is the paper’s only empirical claim. The account itself is an argument about what we would be entitled to say.
What this does not say
- That passing the game is a sufficient condition for thinking. Turing replaces the question; he does not define thought. Block (n. 13) reads him as settling for passing as sufficient, and that is the reading THEORY-092 denies. The parity argument is what would make it sufficient, and only for an entitlement to say “thinks”.
- That the machine is conscious. The reply to Jefferson is about what objectors would accept, not about what is there. Nagel’s claim that a functional analysis is compatible with absent experience (LIT-096) is not engaged.
- That parity holds for an imitator. §2 assumes “the best strategy is to try to provide answers that would naturally be given by a man”, and calls the alternative unlikely to matter. The parity argument assumes the machine’s answers come about the way a man’s do. A machine trained on human output, whose objective is imitation, is the case the assumption does not cover (NOTE-355, Open questions).
- That the game has been passed. A claim that it has should be measured against §6: an average interrogator, five minutes, at most 70 per cent right, with a human competitor in B’s place.
Connections
- Rival: psychologism (THEORY-092). Block’s lookup machine matches
every actual and counterfactual response and “has the intelligence of a
toaster”, so there is a reason to withhold intelligence that does not
apply to people: how the responses are produced. The two cannot both be
right. THEORY-092 does not declare
corrects, since both are Proposed. - Searle (THEORY-638) answers the other-minds reply in one line: the question is what is attributed, not how it is known. He also grants that we would rationally ascribe intentionality to a humanlike robot until we learned it ran a formal program.
- Dennett (THEORY-538) gives a criterion of the same family for belief: anything whose behaviour the intentional strategy predicts reliably and voluminously is a believer.
- Mimicry. THEORY-023 holds that mimicry undercuts behavioural evidence for AI consciousness. It is aimed at the assumption of §2, which this account states and does not defend.