Alan Turing · Mathematics
Turing’s 1950 question and the ‘imitation game’ - how he proposed to replace the unanswerable question ‘can machines think?’ with a concrete behavioural test, founding the philosophy of artificial intelligence.
In 1950, Alan Turing published a paper in the philosophy journal Mind with a disarmingly direct opening: ‘I propose to consider the question, “Can machines think?”’ It is among the most influential papers in the history of philosophy and the founding document of the field of artificial intelligence. But Turing immediately did something characteristically shrewd. He noticed that the question ‘Can machines think?’ is hopelessly vague - it depends entirely on what we mean by ‘machine’ and, far worse, by ‘think,’ words whose ordinary meanings are slippery, contested, and bound up with our human self-image. To debate the question as posed, he saw, would be to argue endlessly about definitions and never reach a conclusion. So instead of trying to define ‘thinking’ - a task he regarded as futile - Turing proposed to replace the question with a different one: a concrete, operational, decidable question about behaviour. Could a machine, in a specific test, behave so much like a thinking human that we could not tell the difference? This reframing - from the metaphysical question of what thinking is to the empirical question of what a machine can do - is Turing’s great methodological contribution to the problem of machine intelligence, and it set the terms of the debate for the next seventy years and counting.
Why did Turing make conversation - open-ended natural language - the arena of his test, rather than, say, chess or calculation? The choice is deep and deliberate. Language is perhaps the most distinctively human capacity, and to converse well across the full range of human topics requires, or seems to require, almost everything we associate with intelligence: understanding of meaning, knowledge of the world, reasoning, memory, the ability to follow an argument, wit, the grasp of context and nuance, even something like emotional and social understanding. A narrow skill like arithmetic or chess can be mastered by a machine through brute calculation without anything we would call general intelligence; but to hold up a human side of a wide-ranging conversation - to answer unexpected questions sensibly, to get jokes, to discuss a poem, to reason about a novel situation, to maintain a coherent personality across an hour of dialogue - seems to demand the whole apparatus of a thinking mind. Turing recognised that conversation is a kind of universal probe: through it, the interrogator can test any aspect of intelligence they choose, because anything a mind can do can be talked about. The Turing test is therefore not a test of one ability but a test that can examine all abilities, indirectly, through their expression in language. By making indistinguishable conversation the criterion, Turing set a bar that he believed no machine could clear without genuine, general intelligence - and made the open-endedness of human language the guardian against cheap tricks.
Much of Turing’s 1950 paper is devoted to anticipating and answering objections to the idea that machines might think - and his treatment is so thorough and prescient that it still frames the debate. He considered, and rebutted, a series of contrary views he named with characteristic wit. The ‘Theological Objection’: that thinking requires an immortal soul God gives only to humans (Turing finds this unpersuasive and even impious in limiting God’s power). The ‘Heads in the Sand’ Objection: that the consequences of thinking machines would be too dreadful, so we hope they are impossible (not an argument, Turing notes, merely a wish). The ‘Mathematical Objection’: that Gödel’s and Turing’s own limitative theorems show machines have limits humans lack (Turing replies that humans, too, make errors and have limits, and that the argument is far from conclusive). ‘Lady Lovelace’s Objection’: that a machine can only do what it is programmed to do and can never originate anything or surprise us (Turing replies that machines surprise him constantly, and that ‘originating’ may be less mysterious than it seems). The ‘Argument from Consciousness’: that a machine cannot truly feel, only simulate (Turing replies that we cannot directly verify consciousness in other humans either, yet do not doubt them). Turing did not claim to have proved machines can think; he claimed that the objections were weaker than they appeared, and famously predicted that by the year 2000 machines would play the imitation game well enough that an average interrogator would have no more than a 70 per cent chance of identifying them after five minutes. The prediction was roughly borne out in spirit, and his framing of the objections remains the indispensable map of the entire field.
This is the opening of the lesson. The rest — the dialogue, the primary source, and the recall — is in the app.
You learned about Turing’s 1950 paper, the ‘imitation game’ (the Turing test), and how he reframed the question of machine intelligence behaviourally. Explain the test, why Turing proposed it, and the main objections he anticipated.
Leads to Rene Descartes.
Begin this lesson →epoché — a humanities education that remembers you.