Alan Turing

Picture a single machine, imaginary, absurdly simple. It is a strip of paper tape, infinitely long, divided into squares. A little head sits over one square. It can read the symbol there, erase it, write a new one, and shuffle the tape one square left or right. It follows a tiny table of rules — if you see this and you are in that state, do the following — and nothing else. No screen, no keyboard, no cleverness. A machine a child could operate.

Now here is the shock. Anything that can be computed at all — every sum, every chess move, every weather forecast, every video streaming to your phone right now — can be computed by that ribbon of tape and its shuffling head. Not "in principle, roughly." Exactly. The device in your pocket, with its billions of transistors, does nothing the paper tape could not do, given enough time and patience. It is only faster.

A twenty-three-year-old English mathematician dreamed up that machine in 1936, not to build it, but to answer a question in pure logic. In the process he did three things almost no one manages once: he settled a two-thousand-year-old dream about the limits of reasoning, he invented the computer as an idea decades before anyone welded one together, and he handed philosophy its sharpest modern question — can a machine have a mind? His name was Alan Turing, and the country he saved would, within twenty years, hound him to his death.

The Life

Alan Mathison Turing was born in London, in Paddington, in June 1912, into the world of the British Empire's servant class — the respectable, over-stretched, perpetually-abroad middle. His father worked in the Indian Civil Service, the vast bureaucracy that ran British India, and for much of Alan's childhood both parents were on the far side of the world. He and his brother were boarded out to a retired army couple, the Wards, in an English seaside town. It was ordinary for families of that kind and quietly bleak for the children: a boy raised by kind strangers, his parents a face on the dock every few years. A certain institutional loneliness settled into him early, and it never entirely left.

He was sent to Sherborne, an old English boarding school that prized Latin, Greek, and the smooth production of gentlemen. Turing wanted only mathematics and science, and the school regarded this the way it might regard a boy who wanted only to eat with his hands. His headmaster wrote a line that ought to be carved somewhere as a warning about every institution that mistakes conformity for education: "If he is to be solely a Scientific Specialist, he is wasting his time at a Public School." The system that produced the twentieth century's most consequential mind looked at that mind and saw a boy in the wrong classroom.

Then came Christopher Morcom. At Sherborne, Turing fell in — intellectually, and with the whole undefended force of a lonely teenager's heart — with a brilliant older boy, gifted in science, quick where Turing was awkward. It was Turing's first love, though the word would not have been sayable. In 1930 Morcom died, suddenly, of bovine tuberculosis contracted from infected milk years earlier. He was eighteen. Turing was flattened. He wrote to Morcom's mother for years, and the letters circle one question with heartbreaking persistence: where has Christopher's mind gone? Was it only the brain, now stopped — or was there a spirit, and if so, what was it, and what was it made of? A grieving, rigorously materialist boy trying to reason his way to whether the person he loved still, in any sense, existed. Hold onto that question. It is, quietly, the seed of everything he later thought about machines and minds.

He went up to King's College, Cambridge, and there he could breathe. King's in the 1930s was a pocket of tolerance in an intolerant country — the world of the economist John Maynard Keynes and the novelist E. M. Forster, where a young man's homosexuality could be lived, among friends, with a freedom unthinkable outside those walls. Turing thrived, and at twenty-two was elected a Fellow of the college. At twenty-three, in 1936, he wrote the paper described in a moment — On Computable Numbers — and the history of thought forked.

He crossed the Atlantic to Princeton for a PhD under the logician Alonzo Church, whose own work had reached the same summit by a different road; their names are now welded together in the "Church–Turing thesis." His Princeton thesis pushed into the wilderness that Kurt Gödel (chapter 233) had opened — the strange country of what mathematics cannot settle. From Gödel he had inherited a haunting: the discovery that truth can outrun proof. Turing turned it into the sharper discovery that some questions no machine can ever decide.

Then the war. From 1939 Turing worked at Bletchley Park, the country house north of London where Britain's codebreakers fought a silent, decisive battle against German ciphers. The great prize was Enigma, the German machine that scrambled military orders into apparent gibberish, resettable daily to astronomically many configurations. Polish mathematicians had cracked early Enigma first — Marian Rejewski's bomba deserves its honour here — and handed their work to Britain on the eve of invasion. Turing, with Gordon Welchman, designed the Bombe, an electromechanical machine that tore through Enigma's settings by exploiting tiny logical leaks in how the Germans used it. It was mathematics and engineering fused into a single humming cabinet.

In Hut 8 he led the assault on the hardest problem of all: naval Enigma, the cipher guarding the U-boats that were strangling Britain's Atlantic lifeline, sinking the ships that carried the nation's food and fuel. Breaking it meant reading the wolf-packs' orders and steering convoys around them. The arithmetic here is tonnage and lives: ships not sunk, sailors not drowned, a starving island kept fed. Historians — the official intelligence historian Harry Hinsley foremost — have estimated that the codebreaking shortened the war by two to four years and saved many millions of lives. These are estimates, and should be flagged as such; counterfactual history is guesswork. But no serious account calls Bletchley's contribution small, and Turing was near its centre.

The man himself, at Bletchley, was gloriously odd. He chained his tea mug to a radiator so no one would take it; in hay-fever season he cycled to work in a gas mask to keep out the pollen. When called to meetings in London he sometimes ran the forty miles rather than trust the trains — no eccentric jog, this: Turing was a serious distance runner who once clocked a marathon in two hours and forty-six minutes, only about eleven minutes off the Olympic standard of his day. He proposed marriage to a fellow codebreaker, the mathematician Joan Clarke, and then, with characteristic honesty, told her he was homosexual and withdrew — the whole episode handled between them with more tenderness than the age deserved.

For all this he received an OBE and thirty years of silence. Bletchley's work was so secret that the greatest war effort of his life could not be spoken of; the fame he had earned was locked in a government vault while he lived. Remember that, too, for the end.

After the war he designed, at the National Physical Laboratory, the ACE — a detailed 1945 blueprint for a stored-program computer, one that keeps its instructions in the same memory as its data. (Priority here is tangled: the American mathematician John von Neumann described a similar architecture at nearly the same moment, and the fair account gives both men their due.) At Manchester he helped bring one of the first real stored-program computers to life and wrote its programming manual — arguably the first working software philosopher. In 1950 he published, in the philosophy journal Mind, the paper on machine intelligence that every artificial-intelligence course still opens with. And in 1952 he turned to biology — the mind kept prising open new sciences, one after another.

Then, in January 1952, his house was burgled. Reporting it to the police, Turing mentioned, without much apparent alarm, his relationship with a young man named Arnold Murray. In the Britain of 1952 that was a confession to a crime. He was charged with "gross indecency" — the same statute under which Oscar Wilde had been ruined half a century before. Offered a choice between prison and "organo-therapy," Turing accepted the second: a course of oestrogen injections, a chemical castration, meant to suppress desire. The physical effects — impotence, the growth of breast tissue — he recorded and endured with a scientist's flat clarity. His security clearance was revoked; in the paranoid Cold War, homosexual men were treated as blackmail risks and purged from secret work. The man who had done more than almost anyone to save the state was barred from its secrets.

On 8 June 1954 he was found dead at home. Beside the bed lay a half-eaten apple; the cause of death was cyanide poisoning. He was forty-one. The inquest ruled suicide. His mother insisted to the end that it was an accident — he tinkered with chemicals, was careless, might have inhaled cyanide by mishap. The truth is genuinely unsettled: the scholar Jack Copeland has shown that the case for a clean suicide verdict is weaker than the legend assumes — no note was left, the tests were incomplete. The standard reading remains suicide, and it carries real weight given what the previous two years had done to him. But honesty requires saying the door is not fully closed.

The afterlife is a long national act of repentance, told here in two lines. In 2009 the British Prime Minister issued a formal apology; in 2013 Turing received a royal pardon; in 2017 "Turing's law" pardoned thousands of other men convicted under the same dead statute; in 2021 his face went onto the Bank of England's fifty-pound note. And computing's highest honour, its Nobel, is called the Turing Award. The state that broke him now prints him as money.

The Big Ideas

On Computable Numbers: defining the computer before it existed. In the 1920s the great mathematician David Hilbert had posed a dream called the Entscheidungsproblem — German for "decision problem." The dream was this: find a single mechanical method, a fixed recipe, that could take any mathematical statement and grind out, yes or no, whether it is provable. A universal truth-machine for mathematics. Gödel had already wounded a related hope in 1931 by showing that no consistent system rich enough for arithmetic can prove all the truths it contains. Turing came to finish the Entscheidungsproblem off — and to do it, he first had to answer a question nobody had properly asked: what does it even mean to follow a "mechanical method"?

His move was pure genius and looks, at first, like a step sideways. Instead of building a device, he analysed a human being computing. Picture a clerk working through a long calculation with pencil and paper. At any moment he looks at a few symbols, he is in some particular state of mind, he writes or erases a symbol, he shifts his attention, and he follows fixed rules. Strip that down to its bones — the tape of paper, the reading head, the finite list of internal states, the table of rules — and you have what we now call a Turing machine. Turing had, in his own careful words, taken "a man in the process of computing" and replaced him with a machine. The achievement was not engineering. It was definition: he had captured, precisely, what "an effective procedure" — a step-by-step mechanical method — actually is. Gödel himself, who did not hand out compliments freely, later said that with Turing's work "the correct definition of mechanical computability was established beyond any doubt." The word definition is the whole point; Turing didn't describe a machine so much as pin down a concept that had floated, undefined, since antiquity.

Then came the lemma that built the modern world. Turing showed you could design one special machine — a universal machine — that reads, on its tape, a description of any other Turing machine, and then behaves exactly like it. One machine to simulate them all. The trick is that the program is written on the same tape as the data; instructions and information are made of the same stuff. That is the stored-program concept, and it was born not as a gadget but as a step in a proof. Every general-purpose computer ever built is a physical instance of Turing's universal machine. The line from that 1936 abstraction to the phone in your hand is not a metaphor. It is direct.

And the payoff for Hilbert's dream was a funeral. Turing asked whether a machine could look at any other machine and decide, in general, whether it will eventually halt or run forever. He proved, by a beautiful self-swallowing argument in the family of Gödel's diagonal trick, that no such machine can exist — the halting problem is undecidable. Assume a halt-detector exists, feed it a cleverly built machine that does the opposite of whatever the detector predicts about itself, and you get a contradiction. So there is no universal decision method. The Entscheidungsproblem's answer is no. Notice the rhyme with Gödel: truth outruns proof (Gödel), and halting outruns deciding (Turing). Two limit-theorems of the century, cousins, each marking a fence beyond which no mechanical procedure can go. It is a strange and permanent fact that the man who defined the computer's power did so in the very act of proving its limits.

One footnote earns its keep. Turing's machines, Church's lambda-calculus, and a third definition called recursive functions all turned out to describe exactly the same set of computable things. That convergence is the Church–Turing thesis: the claim that this shared notion really is what "computable" means. Its status is worth a careful line — it is not a theorem you can prove but not merely a definition either; it sits in between, a claim about the world that every attempt to compute has so far confirmed.

Computing Machinery and Intelligence: how to ask whether a machine can think. In 1950 Turing published the paper that founded the philosophy of artificial intelligence, and its first move is the philosophically deep one. He judged the plain question — "Can machines think?" — to be, in his words, too meaningless to deserve discussion, because "think" and "machine" are too loose to argue about. So he replaced it with a test. In the imitation game, an interrogator sits at a text terminal and exchanges typed messages with two hidden respondents, one human and one machine, trying to tell which is which. If, over time, the machine's answers cannot be reliably distinguished from a person's, then — Turing proposed — quibbling about whether it "really" thinks has lost its grip.

This deserves auditing both ways, because it is doing something subtle. Its elegance is that it is behaviourist: it refuses to peer inside and judges only by performance, sidestepping the swamp of defining thought. Its cost is that it brackets the very things some people care about most — whether the machine is conscious, whether it understands rather than merely produces the right marks. Turing knew this. The test was never a claim that passing it proves an inner life; it was a proposal to stop demanding a proof no one could give for other humans either.

The paper's real genius is its structure: Turing anticipates nine objections and answers them in advance. The theological one (thinking needs a God-given soul) he brushes aside. The mathematical one is his own inheritance turned against him: doesn't Gödel show there are truths a machine can never reach, so machines must be inferior minds? Turing's reply is dry — humans hit their own walls and make errors constantly; a fallible machine is not thereby disqualified. (This objection has a long afterlife: the philosophers Lucas and later Roger Penrose would revive it, and it hinges on exactly the disjunction Gödel himself drew in chapter 233 — one line is enough to mark it.) To the Lady Lovelace objection — that machines can only do what we explicitly order, and so can never originate anything — Turing replies that machines constantly surprise him, and that a machine could be built to learn. He names Ada Lovelace with warmth; her 1843 note on Babbage's engine is honoured, not mocked, as the first serious statement of the doubt he is answering. To the argument from consciousness — Professor Jefferson's line that a machine cannot be said to think until it writes a sonnet out of felt emotion — Turing turns the knife gently: you cannot verify the inner feelings of other people either, yet you grant them minds; the other-minds problem cuts both ways.

Then comes the paper's under-read second half, and it is prophetic. Don't try to program an adult mind, Turing says. Educate a child-machine. Build something simple and teachable, then bring it up — with rewards and punishments, with teaching, letting it modify itself through experience. He is sketching, in 1950, what we now call machine learning: intelligence grown rather than engineered. Read against today's large language models — systems that are, in essence, trained rather than programmed, exactly as he urged — the prophecy is uncanny. It is worth stating soberly what he actually predicted, because it is precise: he guessed that in about fifty years' time, machines with a storage capacity of about a billion units would play the imitation game well enough that an average interrogator would have no more than a 70 per cent chance of a correct identification after five minutes of questioning. Not "machines will think by 2000." A specific, hedged, five-minute, 70-per-cent bet. Modern systems arguably meet and exceed that narrow target — while leaving wide open the very things Turing was careful never to claim the test measured: genuine understanding, genuine mind. That modesty was his, and much of the culture that inherited his test has lost it. (The sharpest reply was still to come: the philosopher John Searle's "Chinese Room," chapter 275, would argue that manipulating symbols correctly is not the same as understanding them — the debate Turing set the terms for, decades early.)

Machine minds and the mind–body question. Underneath the test lies the boy writing to Mrs. Morcom. The question that broke his heart at eighteen — where did Christopher's mind go; is it the brain, or something else? — is the same question, grown up, that the 1950 paper answers. The tender materialism of those grief-letters became, by the end, a picture of mind as pattern and process rather than special stuff: if what a mind does can be captured as the right organisation of states and transitions, then the substrate — neurons, valves, silicon — need not matter. This is functionalism before functionalism had a name: mind defined by what it does, not what it is made of. The philosopher Hilary Putnam (chapter 262) would later build "machine-state functionalism" on precisely this foundation and acknowledge the ancestor — one line is enough to plant it.

Turing's thinking here was richer than the neat test suggests. In a 1948 report he sketched what he called "unorganised machines" — networks of simple connected units that could be trained into structure, a startling early anticipation of the neural networks that drive AI today. In the cybernetics circle of his day — the Ratio Club, where mathematicians and biologists met to talk about brains and machines — he argued about whether intelligence needed a body and a world to move in, or could live in pure abstraction. He floated the idea that intelligence is partly a cultural search, something a community of machines might grow together rather than a lone brain computing in the dark. And he let a random element into the child-machine, a flicker of chance in the learning — a suggestive gesture toward the old problem of free will.

Morphogenesis: how the leopard gets its spots. In 1952, with characteristic refusal to stay in one field, Turing published The Chemical Basis of Morphogenesis and opened a third science. The puzzle: how does a smooth, uniform ball of identical cells in an embryo turn into a striped, spotted, fingered, structured animal? Where does the pattern come from, if everything starts the same? Turing's answer was mathematics of exquisite economy. Imagine two chemicals diffusing through tissue, one activating, one inhibiting, spreading at different rates and feeding back on each other. He showed that such a system, starting from near-perfect uniformity, will spontaneously break its own symmetry and settle into stable patterns — spots, stripes, whorls. Pattern from sameness, order from a differential equation. This is the founding paper of mathematical developmental biology, and it was long treated as beautiful speculation — until, in the 2010s, biologists confirmed genuine "Turing patterns" governing the spacing of everything from fish stripes to the ridges on the roof of a mouse's mouth to the arrangement of digits on a limb. (He was also, at his death, working on phyllotaxis, the Fibonacci spirals of leaves and seeds — unfinished.) The same mind that defined computation and interrogated thought also worked out, on the side, why the tiger is striped.

Before & After

Turing stands at the meeting-point of several old rivers. The dream he killed was ancient: Leibniz (chapter 137) had imagined a calculus ratiocinator, a logic so perfect that disputes could be settled by saying "let us calculate" — Turing proved no such universal calculation exists. Behind that stands Hobbes (chapter 126), who had declared, blunt and materialist, that all reasoning is just reckoning, adding and subtracting in the head — the ancestor of the idea that thought is computation. And behind the imitation game stands, of all people, Descartes (chapter 127), the great dualist: in the Discourse he argued you could always tell a machine from a human because a machine could never use language responsively, as a person does. That is the Turing Test run in reverse — a language test offered to rule out machine minds three centuries before Turing offered one to admit them. La Mettrie, in the materialist circle around Diderot (chapter 153), had pushed Descartes' animal-machine all the way to l'homme machine, man the machine — the frank materialism Turing quietly inherited. Babbage and Ada Lovelace, in the nineteenth century, built and imagined the mechanical engines he honoured by name. And the logic itself came down from Frege (chapter 184) and Russell (chapter 195), through Gödel (chapter 233), whose incompleteness is the fence beside which Turing built his own.

Two Cambridge encounters rhyme. In 1939 Turing sat in Wittgenstein's (chapter 208) lectures on the foundations of mathematics, and the two argued — gloriously, past each other — about what happens when a formal system contains a hidden contradiction. Turing, the working mathematician, worried a contradiction could make a bridge fall down; Wittgenstein, chasing something else entirely, thought that was to misunderstand what mathematics is. Neither convinced the other, and the transcript is a treasure precisely because two great minds are speaking different languages at full volume. And there is the sadder rhyme of Ramsey (chapter 227), the other Cambridge comet, brilliant and dead young — a shape Turing's own life would trace.

Forward, the lines fan out enormously. Putnam (chapter 262) would turn Turing's mind-as-pattern into full-blown functionalism. Searle (chapter 275) would build the great counter-argument. The Dartmouth conference of 1956 would name "artificial intelligence" and set off its long cycle of winters and springs. Complexity theory — the study of not just what can be computed but how expensively, with its famous unsolved question of P versus NP — is Turing's direct grandchild. So is mathematical biology. And the persecution he suffered became, decades on, a landmark in the long argument over gay rights, the state's apology a marker of how far a society can be dragged toward its own better sense.

Why It Still Matters

You are living inside Turing's 1936 paper. The universality he proved as a lemma is why one device — a laptop — can be a typewriter, a cinema, a darkroom, a telephone, and a games console, merely by loading different programs into the same memory. That is the universal machine, and you own several.

His limit-theorem is not a museum piece either. Every time a piece of software cannot, even in principle, be checked for all bugs in advance — every time a computer scientist says "that problem is undecidable" — they are invoking the halting problem. The fences Turing built still hold.

But the reason his name is everywhere in 2026 is the 1950 paper, and here the moment demands sobriety, not hype. We now have machines trained rather than programmed — his child-machine, grown up — that hold fluent conversations, pass professional exams, and write in a manner most people, most of the time, cannot distinguish from a human's. In the narrow terms of his five-minute wager, the imitation game is arguably won. And this has, if anything, sharpened rather than settled the philosophical questions, because Turing's test was always a test of behaviour, and the hard question was always whether behaviour is enough.

Here is where a distinction from the chapter just concluded earns its keep. Wilfrid Sellars (chapter 241) drew a line between the space of reasons — the domain of justification, of giving and asking for grounds — and mere causation, one event making another happen. When a modern system produces a correct, well-argued answer, is it reasoning — moving in the space of reasons, answerable for its claims — or is it an extraordinarily sophisticated causal process that outputs reason-shaped text without ever occupying that space? Turing's test cannot tell them apart, and was honest that it couldn't. That is not a flaw anyone has fixed; it is the live question under every debate about whether these machines "understand." The person who framed the question most cleanly, seventy-five years ago, also marked its limit most honestly — and the modesty that made him say the test measured performance and not mind is exactly the modesty missing from most of the noise around us now.

And then there is the man. The century's most consequential intellect died unknown for his greatest work, criminalised for his love, poisoned by his own state's chemistry or his own hand or an accident nobody can now be sure of. The society that owed him its survival made him a criminal. It is possible to admire the mathematics and still let that verdict stand without softening: a country that could not tell the difference between a threat and a hero destroyed the man who had best understood the difference between a mind and a machine. The apple by the bed in 1954 and the face on the fifty-pound note in 2021 are fifty years and one long apology apart, and the distance between them is the measure of how wrong it is possible to be.

Over to You

If a machine can converse so well you cannot tell it from a person, on what grounds — if any — would you still insist it doesn't "really" understand? And could you meet that same standard for the person sitting next to you?

Turing thought the honest question wasn't "can machines think?" but "can you tell the difference?" Is that a brilliant reframing or a sleight of hand that changes the subject?

His mind-as-pattern idea says what a mind is made of doesn't matter, only how it's organised. Do you believe a mind could run on silicon — or is there something about being made of living stuff that you're not ready to give up?

The state that owed Turing its survival criminalised him. When a society honours someone fifty years too late, is that repentance worth something — or just a cheaper way of feeling better about an unforgivable thing?

Where to Go Next

Start with the source: Turing's 1950 paper, Computing Machinery and Intelligence (in the journal Mind) — remarkably readable, funny, and generous with objections; you need no mathematics to follow it.

A modern way in: Andrew Hodges, Alan Turing: The Enigma — the definitive biography, humane and rigorous, and the basis for the film The Imitation Game.

The thread forward: From the question of whether a machine can have a mind we turn to a man who asked whether a human life, in a silent universe, can have meaning at all — Albert Camus (chapter 243) and the philosophy of the absurd. The pendulum swings from machine minds to the meaning of life, and the chapter almost writes itself.