Skip to main content
Intelligences Plurielles

Applied philosophy of AIThinking human-AI relations beyond dichotomies

Français

règle : 2 branches · ℓ × 0,72 · 9 générations · graine : alan-turing

Thinker profile · Mathematics, logic, theoretical computer science

Alan Turing

British mathematician and logician, father of theoretical computer science and of artificial intelligence. By replacing “can machines think?” with a test of indistinguishability, he opened a debate that conversational AI makes more alive than ever.

1912–1954 · United Kingdom · Mathematics, logic, analytic philosophy, AI

Alan Turing occupies a singular place: he is at once the theoretical founder of computer science and the first to have posed the question of machine intelligence in terms that still structure our debates — seventy-five years later.

From the paper machine to the computer

In 1936, at the age of twenty-four, Turing published “On Computable Numbers”. To solve a problem in mathematical logic, he invented a paper machine: an abstract automaton that reads and writes symbols on a tape according to simple rules. He demonstrated that a single machine of this kind — the universal machine — can carry out any computation, provided it is given the description of it. The modern computer is the material realisation of this idea: our machines are not sophisticated calculators, they are universal machines.

During the Second World War, Turing put this power of abstraction to concrete use: at Bletchley Park, his work on breaking Enigma helped to shorten the conflict. After the war, he took part in the design of the first electronic computers, before turning to mathematical biology and the forms of living things.

The imitation game

In 1950, in the philosophy journal Mind, Turing published “Computing Machinery and Intelligence”. The question “can machines think?” struck him as badly posed — too laden with metaphysics. In its place he put a game: an interrogator converses in writing with two hidden interlocutors, a human and a machine. If the interrogator cannot tell them apart, what remains of the idea that the machine does not think?

In the same article, Turing anticipates and discusses nine objections, among them Ada Lovelace's: a machine only does what it is ordered to do, it creates nothing. Turing replies that this argument holds for fixed-program machines — not for learning machines, capable of surprising their own designer. The reply, formulated half a century before machine learning, has since taken on a singular resonance.

A question still open

Turing defended a continuity between the human and the machine: intelligence would be a function, not a substance; the difference, a matter of degree and not of nature. One may contest this position — a whole philosophical tradition sets about doing so. But it was he who fixed the terms of the discussion: since conversational systems began sustaining dialogues hard to distinguish from those of a human, his test has left the laboratories to become an everyday experience — and the question of what, exactly, it proves has never been so alive.

Persecuted by the British justice system for his homosexuality, sentenced to chemical castration in 1952, Turing died two years later, at the age of forty-one. His posthumous rehabilitation does not efface what his death cost to thought.

The core thought

Intelligence is a function, not a substance. If a machine is behaviourally indistinguishable from a human, then thought must be granted to it: between the human and the machine, the difference is one of degree — of complexity — and not of nature.

  • The “imitation game” (1950), which became the Turing Test, replaces the metaphysical question “can machines think?” with an operational criterion: the indistinguishability of behaviours in dialogue.
  • The universal machine (1936) establishes that any computation is a mechanical process executable by one and the same abstract automaton — which opens up the theoretical possibility of thinking machines.
  • Against Ada Lovelace's objection (“a machine only does what it is ordered to do”), Turing replies that learning machines, unlike fixed-program machines, can surprise their own programmer.

Method A behavioural and operational criterion: judging intelligence by its observable manifestations rather than by its supposed nature.

Major works

  • On Computable Numbers, with an Application to the Entscheidungsproblem (1936) — The founding article of theoretical computer science, published in the Proceedings of the London Mathematical Society: it introduces the universal machine, an abstract automaton capable of executing any computable algorithm — the conceptual foundation of the modern computer.

  • Computing Machinery and Intelligence (1950) — Published in the journal Mind, the philosophical birth certificate of artificial intelligence: it proposes the imitation game and anticipates then refutes nine major objections, among them Ada Lovelace's.

Profile co-created — Matthieu Ferry ⇄ AI