A. M. Turing · 1950
The paper that asked whether machines can think, and answered with a party game about telling a man from a woman.
What the lessons say
The Turing Test Was Never About Machines vs Humans — It Was About Men Pretending to Be Women
Here's why this matters: everyone assumes Turing pitted machines against humans, but he actually built the test on a man-versus-woman guessing game first.
Here's how it works: three players—man A, woman B, interrogator C. C asks written questions, like "will X tell me the length of their hair?" A, the man, has to answer as if he's the woman, maybe lying about "shingled hair, nine inches." B can even shout "I'm the real woman, don't listen to him!"—but that's useless since A can say the same thing. Then Turing swaps A for a machine. The real question becomes: does the machine fool C as often as the man fooled C pretending to be a woman?
So the "thinking machine" question quietly becomes a question about successful imitation, not consciousness.
Turing Refused to Define 'Thinking' — Because He Thought the Question Was a Trap
Here's why this matters: Turing dodges a trap most people walk right into when they argue about AI—defining "thinking."
Instead of debating definitions, he swaps the question entirely. He sets up the Imitation Game: a man, a woman, and an interrogator asking typed questions to guess who's who. Then he asks—what if a machine replaces the man? If the interrogator gets fooled just as often, that's your new benchmark.
Notice the move: he's not asking "does the machine think?" He's asking "can it perform indistinguishably?" That's testable. That's measurable. He sidesteps the Gallup-poll problem of relying on vague common usage.
Next, you'll see how this test actually plays out—and why "imitation" became such a controversial stand-in for "intelligence."
Turing's Original Test Wasn't About Machines At All—It Was About Gender
Here's why this matters: everyone cites Turing as asking "can machines think," but he actually replaced that question with a game about gender deception—and that detail changes everything about how you should read the paper.
Here's the setup: a man, a woman, and an interrogator who questions both through typewritten notes—no voice, no visuals. The interrogator tries to identify who's who; the man tries to fool them into guessing wrong. Turing then swaps the man for a machine and asks: does the interrogator get fooled just as often? Notice what this does—it sidesteps defining "thinking" entirely by turning intelligence into a behavioral, text-only imitation contest. That's the real innovation—not the machine part, but reframing an unanswerable philosophical question into a testable game of deception.
Next, we'll unpack why Turing restricted the machines to digital computers specifically.
Turing Casually Mentions Cloning a Human From Skin Cells... In 1950
Here's why this matters: Turing's paper isn't just history, it's the blueprint for how we still think about testing AI today.
So here's how he sets it up. First, he ditches the vague question "can machines think" because it's unanswerable, and replaces it with something testable: the imitation game. Three players, an interrogator asking typed questions to a hidden man and woman, trying to guess who's who. Then he swaps the man for a machine. Notice his reasoning for excluding humans from qualifying as "machines" - he casually mentions you could theoretically clone a person from a single skin cell, but that still wouldn't count as building a thinking machine. That's 1950, mentioning cloning.
Next up, he narrows "machine" specifically to digital computers, setting up his defense of that choice.
Turing Predicted Human Cloning... in 1950
Here's why this matters: Turing anticipated cloning debates decades early, and he shows you exactly how to define "thinking machine" without getting tangled in biology.
Watch his method. First, he rules out a trick answer—growing a human from a skin cell wouldn't count as "building a thinking machine," even though it's biologically stunning. That's his way of narrowing the definition to digital computers specifically. Second, he breaks down what a digital computer actually is: it mimics a human "computer" who follows fixed rules from a rulebook. Three parts matter—the store (like paper and memory), the executive unit (which does operations, simple or complex), and the control (which enforces the rulebook's order). Third, he reframes the real question: not whether today's machines pass the test, but whether any imaginable computer could.
Next, he'll build his famous test around that exact question.
