London 1912 – Wilmslow 1954

Alan Turing

"Can machines think?" is too vague to answer. Replace it with a game: if a machine can hold a conversation that no judge can tell apart from a person's, what further reason could we have for denying that it thinks?

The view

Turing is best known for two achievements: the abstract model of computation now called the Turing machine (1936), and his wartime work breaking German codes at Bletchley Park. In 1950 he published a paper in the philosophy journal Mind, "Computing Machinery and Intelligence", that has shaped the philosophy of mind ever since. It opens with a single sentence:

I propose to consider the question, "Can machines think?" Turing, "Computing Machinery and Intelligence" (1950)

Turing thought the words "machine" and "think" were too slippery for a direct answer. Instead he proposed the imitation game. A judge exchanges typed messages with two hidden players, one human and one machine, and tries to work out which is which. If the machine can regularly fool the judge, Turing argued, we should credit it with thinking. He predicted that by about the year 2000, machines would play well enough that an average judge would have no better than a seventy per cent chance of picking correctly after five minutes of questions.

The argument

Most of the paper is a reply to objections, and many are still raised today. The theological objection says thinking is a function of the soul. The "heads in the sand" objection says the consequences would be too dreadful. Lady Lovelace's objection says a machine can only do what we tell it. The argument from consciousness says a machine might act as if it thinks but would feel nothing. Turing's reply to the last is the heart of his position. We have no direct access to other people's consciousness either. We credit them with minds because of how they behave, especially how they talk. To demand more of a machine than we demand of each other is to adopt a double standard. Play a round below.

The questions in the figure echo the ones Turing imagined, including his arithmetic question. In his sample dialogue the machine pauses before answering and then gets the sum slightly wrong, because a machine imitating a person must imitate human weaknesses too.

Where he meets Dennett

Dennett was one of Turing's strongest defenders. In "Can Machines Think?" (1985) he argued that the test is much harder than critics assume. Passing it in an open-ended conversation, on any subject the judge likes, requires a grasp of the world so wide and flexible that it would be absurd to deny the machine any understanding. Critics who say a machine could pass by trickery, he argued, underestimate what the trickery would have to do. Turing's refusal to treat consciousness as a hidden extra beyond what a system can do is Dennett's position too.

Where they part

Hardly at all on principle, but the world has changed. Large language models now produce fluent conversation of a kind Turing predicted, and Dennett, late in life, worried less about whether they think than about what they do to us. He warned against "counterfeit people", systems designed to pass as human, arguing that they could undermine the trust that conversation depends on. The imitation game, he suggested, had become a real-world hazard as well as a thought experiment. Turing asked whether a machine could pass for a person. Dennett's last question was what happens to a society when machines routinely do.

See every view compared →

Further reading

• A. M. Turing, "Computing Machinery and Intelligence", Mind 59 (1950): short, witty and readable.
• Andrew Hodges, Alan Turing: The Enigma (1983; Vintage, 2014): the standard biography.
• Daniel C. Dennett, "Can Machines Think?" (1985), reprinted in Brainchildren (MIT Press, 1998).
• The Turing test, Stanford Encyclopedia of Philosophy.