AI History
The Room-Sized Brain That Taught Us to Fear Thinking Machines
The moment a computer beat a human at chess for the first time, in 1997, was not the beginning of the AI story — it was closer to the middle of a tale that starts in a New Hampshire farmhouse in the summer of 1956.
The Idea
Most people trace artificial intelligence to recent breakthroughs — neural networks, large language models, the sudden arrival of systems that can write and reason. But the field has a precise birthday: the Dartmouth Conference of 1956, where a small group of mathematicians and computer scientists gathered with the audacious conviction that every aspect of human intelligence could, in principle, be simulated by a machine. John McCarthy, who coined the term 'artificial intelligence,' believed they might crack the problem in a single summer. They did not. What they launched instead was a decades-long cycle of wild optimism, collapsed funding, and quiet revival that researchers now call the 'AI winters' and 'AI springs.' The first winter came in the 1970s, when early programs that seemed magical — solving algebra problems, translating simple sentences — hit walls no one had anticipated. Language, it turned out, was not logic. Common sense was not a database you could fill in. The second winter followed the expert systems boom of the 1980s, when rule-based programs that could diagnose rare diseases or configure computer hardware became expensive, brittle, and impossible to maintain. What makes this history genuinely surprising is how much was understood, philosophically and mathematically, decades before hardware caught up. The ideas were right. The world just wasn't fast enough yet.
In the World
In 1950, six years before the Dartmouth gathering, Alan Turing published a paper in the journal Mind that opened with a question so precise it became the century's most famous thought experiment: 'Can machines think?' He immediately sidestepped it, replacing it with what he called the Imitation Game — a test of whether a machine could conduct a written conversation indistinguishable from a human's. Turing was not speculating idly. He had spent the war years at Bletchley Park breaking Nazi cipher codes with electromechanical machines, and he understood, viscerally, that computation could do things that felt like intelligence without anyone inside doing the thinking. What is less remembered is Turing's prediction: he believed that by the year 2000, machines would pass his test well enough to fool an ordinary questioner about thirty percent of the time. He was roughly right about the capability, wrong about the timeline. More striking still, Turing anticipated almost every major objection to machine intelligence in that 1950 paper — the theological objection, the 'machines can only do what we program them to' objection, the consciousness objection — and answered each with care. The field spent the next seventy years essentially catching up to arguments he had already made, and refuted, before most of its pioneers had finished school.
Why It Matters
Understanding AI as a history — not a sudden eruption, but a stuttering, century-long argument — changes how you relate to the technology around you now. The hype cycle you are living through is not new. The fear is not new. The breathless claims and the inevitable disappointments follow a pattern that has repeated at least three times before this moment. That does not mean the current wave is just another false dawn; the underlying compute and data shifts are genuinely different in scale. But it does mean that the people making the boldest promises now are not the first to make them, and that measured scepticism has historically been the more accurate stance. It also opens a more interesting question than 'will AI take my job' — which is: what does it mean that the hardest things for machines to learn turned out not to be chess or calculus, but conversation, humour, and knowing what another person probably meant? The shape of machine difficulty is, in a way, a mirror held up to the shape of human intelligence.
A Question to Ponder
If the things that turned out to be hardest for AI — ambiguity, context, common sense — are precisely the things we rarely think of as intelligence, what does that suggest about what intelligence actually is?
Get a new one of these every morning.
Start learning with Thinkable