AIster · Prologue · Breakthroughs

Turing asks whether machines can think

The paper that gave AI its most famous yardstick: can a machine pass for a person?

Key facts

Date

Oct 1950

Era

Prologue

Interest

Breakthroughs

Threads

Model wars, Safety & governance

Who

Alan M. Turing

Where

Mind, vol. LIX, no. 236, pp. 433–460

Opening line

“I propose to consider the question, ‘Can machines think?’”

Original name

The imitation game

A question swapped for a game

Alan Turing's paper in the October 1950 issue of Mind opens with the line: “I propose to consider the question, ‘Can machines think?’” Turing set aside the argument over what the words “machine” and “think” mean and replaced the question with something that could actually be tried. He called it the imitation game.

The set-up has three participants kept apart from one another: a machine, a human, and a human judge. The judge talks to the other two through typed messages only, and both try to convince the judge that they are the human. If the judge cannot reliably tell which is which, the machine has passed.

A dated prediction

Turing did not stop at the thought experiment. He put a forecast on the record: “I believe that in about fifty years' time it will be possible to programme computers, with a storage capacity of about 10^9, to make them play the imitation game so well that an average interrogator will not have more than 70 percent chance of making the right identification after five minutes of questioning.”

That sentence is unusually concrete for a philosophy paper. It names a timescale, a memory size, a success rate and a length of conversation, which is part of why it is still quoted.

The name came later

The paper calls the set-up the imitation game. The label “Turing test” was attached to the proposal later, and it stuck.

Why it still matters

The paper's lasting value is the way it framed the debate. It suggested judging a machine by what it does in conversation and setting aside what goes on inside it. Critics have long argued that imitating a person in text is not the same as understanding anything, and that argument is still unresolved.

For anyone catching up on today's chatbots, this is the starting point. Many of the questions people now ask about large language models, such as whether fluent text proves intelligence and how you would know, were first put in a testable form here, more than seventy years ago.

Questions people ask

What is the Turing test?

It is the name later given to Alan Turing's 1950 “imitation game”, in which a human judge exchanges typed messages with a hidden person and a hidden machine. If the judge cannot reliably tell which is the machine, the machine is said to have passed.

What did Turing predict about computers by the year 2000?

He wrote that in about fifty years computers with a storage capacity of about 10^9 could play the imitation game so well that an average interrogator would have no more than a 70 per cent chance of identifying them correctly after five minutes of questioning.

Sources

1. Mind (Oxford Academic), I.—Computing Machinery and Intelligence, 1950-10-01
2. Stanford Encyclopedia of Philosophy, The Turing Test, 2003-04-09