The Turing TestThe Turing Test

In 1950, Alan Turing opened his paper 'Computing Machinery and Intelligence' with a provocation: the question 'can machines think?' was too vague to answer. Rather than argue about definitions of thinking, he replaced the question with a game.

The setup was borrowed from a Victorian party game. An interrogator exchanges typed messages with two hidden participants, one human and one machine, and must decide which is which. If the machine can fool the judge as often as a person could, Turing argued, quibbling about whether it 'really' thinks becomes pointless.

Turing published this in Mind, a philosophy journal, not an engineering one. That was deliberate. He spent most of the paper demolishing objections he anticipated, from theological arguments to Lady Lovelace's claim that machines can only do what they are told. Several of those rebuttals still get recycled in AI debates today.

He also made a concrete prediction: by the year 2000, machines with enough storage would play the imitation game well enough that an average interrogator would have no better than a 70 percent chance of correct identification after five minutes of questioning. He was early, but not absurdly so.

Turing never saw any of it tested. He died in 1954, two years after being prosecuted by the British government for homosexuality. The test he proposed almost as a thought experiment became the most famous benchmark in computing, one that modern chatbots have arguably passed and philosophers still argue tells us nothing at all.

Either way, the framing won. When people today judge an AI system by conversing with it, they are playing Turing's game, whether they know it or not.