The field was named to win a turf war. Everything that followed was shaped by that choice.
The Problem With the Name That Already Existed
Before John McCarthy wrote a word of that Rockefeller Foundation proposal, there was already a perfectly good name for the field he wanted to build. It was called cybernetics. Norbert Wiener, an MIT mathematician of formidable reputation, had coined the term in 1948 and built an influential framework around it — the study of how systems, biological or mechanical, regulate themselves through feedback and control. By 1955, cybernetics was the established vocabulary for anyone thinking seriously about intelligent machines.
McCarthy didn't want it. In his own words, preserved in his archives at Stanford, he wrote that one of the reasons for inventing the term "artificial intelligence" was to escape association with cybernetics — and to avoid having either to accept Norbert Wiener as a guru or having to argue with him. That sentence is worth sitting with. The field of AI was not named because the name was the most accurate description of what the researchers were attempting. It was named, in part, to keep a difficult personality out of the room.
The Co-Signatories Didn't Even Agree With It
The politics didn't end with Wiener. Shannon preferred the term "automata studies." Newell and Simon, who would arrive at Dartmouth with the only working program anyone brought that summer, continued to use "complex information processing" for years afterward. Three of the four most important figures in the room had reservations about the name they had co-signed onto. The term "artificial intelligence" was McCarthy's invention, McCarthy's positioning, and — crucially — McCarthy's grant application strategy. It won not because everyone agreed it was right, but because McCarthy was the one who wrote the proposal.
This is not a trivial footnote. The name shaped the expectations. "Artificial intelligence" implies a destination — machine behaviour indistinguishable from human thought. "Automata studies" or "complex information processing" implies a method — systematic, incremental, modest in its claims. The destination framing is what made the field legible to funders and journalists. It is also what made every shortfall feel like a failure rather than progress.
Who Actually Showed Up, and For How Long
The conference ran for eight weeks across the summer of 1956 on Dartmouth's campus in Hanover, New Hampshire. McCarthy reported to the Rockefeller Foundation that Shannon and Rochester would attend for two weeks at the beginning and two weeks at the end — not the full duration. Newell and Simon joined for just two weeks in the middle. The gathering that is remembered as the founding moment of a field was, in practice, a loosely attended seminar where most of the famous names passed through briefly.
Newell and Simon, who only came for a few days, were the stars of the show. They presented the Logic Theorist — the only working program anyone brought — and compared its outputs with results from human subjects. It was met with a lukewarm reception. The program that history now treats as the first proof of concept for AI, the one that proved theorems more elegantly than Bertrand Russell, did not impress the room. The attendees were more interested in debating the shape of the field than in examining the evidence that something was already working inside it. It is one of the stranger ironies in the history of science.
What the Summer Actually Produced
No landmark paper came out of Dartmouth. No formal proceedings were published. The researchers went home with something less tangible and more durable: a shared vocabulary, a shared conviction, and the beginning of separate research programs that would scatter across American universities and define the next decade. McCarthy's phrase had floated through two months of loose talk and hard disagreement without breaking, and by the time the early papers began to cite the Dartmouth meeting, the words were already doing administrative work. The name had stuck. The field had a centre of gravity it hadn't had before.
What it also had — and this is the part the history books tend to skip — was a set of ambitions that were calibrated to the best result anyone had seen, not to the average difficulty of the problems ahead. One program had proved theorems. The conclusion drawn was not "theorem-proving is tractable." The conclusion drawn was "intelligence is tractable." That is a much larger leap than it looks.
Why It Still Matters
The pattern that came out of that summer — confident naming, high ambitions, spectacular early results, and conclusions that outran the evidence — did not stay in 1956. It became the template. Every AI hype cycle since has followed the same structure: a tractable problem gets solved, the solution gets treated as a window onto the general case, and the timeline gets compressed accordingly.
Understanding where that pattern came from, and why it was almost inevitable given how the field was named and founded, is the difference between reading AI headlines and actually understanding them.
The Video Goes Deeper
The post gives you the room where it happened. The video gives you the logic — step by step — of why the early results looked so convincing, and precisely where the reasoning went wrong. It also includes a practical two-question framework you can apply to any AI capability claim you encounter at work.
If you've been following the series, this is where the history starts to feel uncomfortably familiar.
