In the summer of 1956, a small group of mathematicians and engineers gathered on the campus of Dartmouth College in Hanover, New Hampshire, with an audacious agenda. They intended to figure out how to make machines think. They gave themselves about two months. The field they accidentally founded is still working on the problem nearly seventy years later, which tells you something about both their ambition and their optimism.
Four Names and a Proposal
The Dartmouth Summer Research Project on Artificial Intelligence was organized by four men whose careers would shape computing for decades. John McCarthy, then a young assistant professor of mathematics at Dartmouth, was the prime mover. He was joined by Marvin Minsky, a Harvard junior fellow who would go on to co-found the MIT AI Laboratory; Nathaniel Rochester, the chief architect of IBM’s first commercial scientific computer, the IBM 701; and Claude Shannon, the Bell Labs polymath whose 1948 work had already founded information theory.
In 1955 the four submitted a funding request to the Rockefeller Foundation. The document opened with a sentence that has since become one of the most quoted in the history of computing:
“We propose that a 2 month, 10 man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire.”
It was in that proposal that McCarthy coined the term “artificial intelligence.” The choice was partly strategic. McCarthy later admitted he wanted a name that staked out new territory and avoided association with existing work he found limiting, particularly Norbert Wiener’s cybernetics. The label was deliberately broad, and it stuck. The Rockefeller Foundation, for its part, funded the project at a modest level, well short of the budget the organizers had hoped for.
The Conjecture at the Heart of It
The proposal’s intellectual core was a single, breathtakingly confident claim:
“The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it.”
This was the founding bet of the field. Intelligence, the organizers wagered, was not a mystical property of biology but a process that could be formalized and mechanized. They expected to attack a cluster of problems together: how computers might use language, how to build neural networks, how machines could improve themselves, abstraction, and randomness in creative thought. The famous line that “a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer” reads today as a masterpiece of underestimation.
What Actually Happened
The reality was looser than the plan. Rather than a tightly run workshop, Dartmouth functioned more like an extended series of visits, with participants drifting in and out across roughly six to eight weeks. The “10 man study” never assembled as a fixed cohort. Attendees over the summer included future luminaries such as:
- Arthur Samuel, the IBM researcher whose checkers-playing program was an early demonstration of machine learning.
- Ray Solomonoff, who pursued ideas that later matured into algorithmic probability.
- Oliver Selfridge, a pioneer of pattern recognition.
- Allen Newell and Herbert Simon of the RAND Corporation and Carnegie Tech.
Newell and Simon arrived with something the others mostly had on paper: a working program. Their Logic Theorist, developed with the programmer Cliff Shaw, could prove theorems from Whitehead and Russell’s Principia Mathematica, and in at least one case found a proof more elegant than the original. It is widely regarded as the first artificial intelligence program ever written. Curiously, Newell and Simon were somewhat outsiders to the Dartmouth in-crowd, and their concrete achievement did not dominate the gathering the way hindsight suggests it should have.
The Optimism Problem
It is easy, and a little unfair, to mock the timeline. The Dartmouth organizers were not naive; they simply could not see how hard the problems were, because no one had yet tried to solve them rigorously. Tasks that seemed trivial to a human, recognizing a face, understanding an offhand sentence, reasoning with common sense, turned out to require enormous machinery that the field would spend decades only beginning to build.
That gap between confident prediction and stubborn reality became a recurring pattern. The overpromising at Dartmouth set a template for cycles of hype and disappointment, including the “AI winters” of funding collapse in the 1970s and again in the late 1980s. The conjecture was sound; the schedule was science fiction.
Why 1956 Still Counts
For all that, Dartmouth earned its place as the founding moment of artificial intelligence, and not merely because it provided the name. It gave a scattered set of inquiries a shared identity, a research agenda, and a community. The people in that room and the ones who passed through it went on to build the institutions, MIT, Stanford, Carnegie Mellon, where the discipline would actually grow. McCarthy would invent the Lisp programming language; Minsky would shape the field’s theory for a generation; Newell and Simon would win a Turing Award.
The meeting matters less for what it accomplished in a summer than for what it declared: that thinking was a fit subject for engineering. That declaration, more than any single program, is the inheritance.
The Dartmouth founders bet that intelligence could be described precisely enough for a machine to reproduce it. Today, systems that write, reason, and code are evidence that the bet was placed on the right table, even if it took far longer to pay out than a New Hampshire summer. Their conjecture remains the quiet premise behind every line of this sentence and every machine that helped shape it.