“Computing Machinery and Intelligence”

Alan Turing, Mind, vol. 59, no. 236, October 1950

Rather than asking whether machines think, let us ask whether they could behave, in writing, indistinguishably from a human.

I propose to consider the question, ‘Can machines think?’
Alan Turing, opening sentence of the paper

What the text says

Turing replaces a question he finds too vague with a protocol: the imitation game. An interrogator exchanges written messages with a human and a machine, without seeing them, and must guess which is which. If the machine deceives as often as a human would, on what would we still base our refusal to say that it thinks?

Most of the paper answers objections: the theological objection, the mathematical one, the argument from consciousness, and Lady Lovelace’s objection that a machine only does what it is told. Turing ends with a decisive intuition: rather than programming an adult mind, build a child-machine capable of learning.

Why read it now

Seventy-five years later, machines speak well enough for the question to be no longer theoretical. Turing’s paper shows that the real subject was never the machine, but our criteria: what we accept as a sign that there is a mind on the other side.

What remains open

If we can no longer tell a human from a machine, what have we lost: a certainty, or an illusion?