The Dartmouth Conference 1956
The Dartmouth Conference of 1956 was a landmark summer workshop held at Dartmouth College in Hanover, New Hampshire, widely regarded as the founding moment of artificial intelligence as a formal scientific discipline. It matters because it gave the field its name, its founding ambitions, and its first community of researchers — shaping the direction of AI for decades to come. The remarkable and counterintuitive fact is that the conference was considered a partial failure by its own organizers at the time, yet it became one of the most consequential gatherings in the history of science.
Before 1956, the idea of making machines think was scattered across mathematics, psychology, philosophy, and engineering — there was no field, no name, and no shared vision. The Dartmouth Conference was the moment a small group of brilliant, ambitious researchers decided to treat machine intelligence as a single unified problem worth attacking together. The "aha moment" is this: a field of science can be invented — not just discovered — by the act of naming it, funding it, and gathering the right people in one room. What Dartmouth did was not solve AI; it created the space in which AI could be solved.
How It Works
The Dartmouth Conference was not a single dramatic event but a slow-burn gathering that unfolded over roughly six to eight weeks in the summer of 1956. Understanding what happened there requires understanding the world it emerged from.
The Proposal That Started Everything In 1955, John McCarthy, then a young mathematics professor at Dartmouth, co-wrote a funding proposal with Marvin Minsky, Claude Shannon, and Nathaniel Rochester. The proposal asked the Rockefeller Foundation for roughly 150,000 in today's money — to fund a summer research project. The central claim of that proposal was breathtaking in its confidence: > "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. The study is to proceed on the basis 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." That sentence did two things at once. First, it coined the term artificial intelligence — McCarthy preferred it over competing phrases like "complex information processing" or "thinking machines" because it was bold and unambiguous. Second, it stated a philosophical position: that intelligence is, at its core, a process that can be described, broken into steps, and therefore mechanized. This was not obvious at the time. Many scientists believed that human thought was fundamentally mysterious or biological in a way that resisted mechanical description.
Who Was in the Room The conference attracted a remarkable cross-section of mid-century intellectual life. The core attendees included:
- John McCarthy — the host and organizer, a mathematician obsessed with formal logic and computation
- Marvin Minsky — a polymath who had already built one of the earliest neural learning machines
- Claude Shannon — the father of information theory, who had essentially invented the mathematical language of communication
- Nathaniel Rochester — an IBM engineer who had designed one of the first commercial computers, the IBM 701
- Allen Newell and Herbert Simon — a psychologist-computer scientist duo from Carnegie Tech who arrived with a bombshell
- Arthur Samuel — an IBM researcher who had built a checkers-playing program that could improve through experience
- Ray Solomonoff — a brilliant and underappreciated theorist working on machine learning from first principles
- Oliver Selfridge — a pioneer in pattern recognition who had developed a model of how perception might work in machines
The attendance was fluid and informal. People came and went. Shannon attended only briefly. There was no single agenda, no required lectures, no formal proceedings that were published. Think of it less like a modern academic conference and more like an extended, unstructured intellectual salon.
The Main Themes Debated Despite the lack of formal structure, several big ideas dominated the conversations: 1. Symbolic reasoning — Can a machine manipulate abstract symbols according to logical rules and thereby "think"? Newell and Simon believed yes, and they had evidence. 2. Neural networks — Can a machine learn by mimicking the structure of the brain's neurons? Minsky leaned toward this, though he would later become a fierce critic. 3. Language — Can a machine understand and produce human language? McCarthy was drawn to this problem. 4. Game playing — Can a machine play chess or checkers well enough to learn general strategies? Shannon had written influential papers on chess-playing machines. 5. Creativity and self-improvement — Can a machine learn from its own mistakes and get better over time, the way humans do?
The Bombshell: Logic Theorist The single most dramatic moment of the conference came when Allen Newell and Herbert Simon demonstrated a program called the Logic Theorist. Written with programmer Cliff Shaw, the Logic Theorist was a computer program that could prove mathematical theorems in formal logic — on its own, using search and heuristic reasoning. It had already proven 38 of the 52 theorems in Bertrand Russell and Alfred North Whitehead's Principia Mathematica, and for one theorem it found a proof that was more elegant than the one published in the original book. This was extraordinary. No machine had ever done anything like it. Simon reportedly told his class at Carnegie Tech that winter: "Over the Christmas holiday, Al Newell and I invented a thinking machine." The Logic Theorist was not just a parlor trick — it was proof of concept that a digital computer could do something that looked, by any reasonable definition, like reasoning.
Why the Conference "Failed" and Why It Didn't Matter The original ambition — to crack the major problems of AI in a single summer — was never remotely achieved. The participants did not agree on a unified approach. There was no manifesto, no roadmap, no grand synthesis. Shannon and some others were skeptical that the problems were as tractable as McCarthy and Simon believed. But in the years that followed, Dartmouth became mythologized as the founding moment precisely because of what its attendees went on to accomplish. McCarthy went to MIT and then Stanford, founding AI labs that shaped generations of researchers. Minsky co-founded the MIT AI Lab. Newell and Simon built the General Problem Solver and won the Turing Award and Nobel Prize. Shannon's information theory became the bedrock of modern computing. Arthur Samuel's checkers program became a landmark in machine learning. The network of people, ideas, and institutions that grew out of those six weeks in New Hampshire defined the shape of AI research for the next fifty years.
McCarthy was a mathematician at Dartmouth who organized the 1956 conference and deliberately chose the phrase "artificial intelligence" over alternatives to give the field a sharp, ambitious identity. He later invented the programming language LISP, which became the dominant language of AI research for decades, and developed the concept of time-sharing computing systems that allowed multiple users to share a single computer. A surprising personal fact: McCarthy was also a committed environmentalist and futurist who wrote extensively about the long-term prospects of human civilization and space colonization — he believed AI would eventually solve the problem of human mortality.
Minsky was one of the most brilliant and provocative figures at the conference, having already built the SNARC — the first neural network learning machine — as a Harvard graduate student in 1951. He co-founded the MIT Artificial Intelligence Laboratory with McCarthy in 1959, which became the most influential AI research center in the world, and his 1969 book Perceptrons (co-authored with Seymour Papert) was so critical of neural networks that it is often credited with setting back connectionist AI research by a decade. A remarkable personal fact: Minsky was a skilled pianist and composer who saw deep connections between musical creativity and machine intelligence, and he counted author Arthur C. Clarke among his close friends — Clarke based the HAL 9000 computer in 2001: A Space Odyssey partly on conversations with Minsky.
Shannon was already one of the most celebrated scientists in America by the time of the Dartmouth Conference, having published his landmark paper on information theory in 1948 that defined how information could be measured, stored, and transmitted mathematically. His 1950 paper "Programming a Computer for Playing Chess" was one of the first serious technical treatments of machine game-playing and directly inspired many of the conversations at Dartmouth. A delightful personal fact: Shannon was famous for riding a unicycle through the halls of Bell Labs while juggling, and he built elaborate mechanical toys including a machine whose sole function was to reach out and turn itself off — he called it the "Ultimate Machine."
Intuition
Imagine a group of explorers who each believe there is a great continent somewhere to the west, but they have each been sailing from different ports, in different boats, with different maps. Some are navigating by stars, some by ocean currents, some by the flight patterns of birds. Each is making real progress — but they don't know about each other, and no one has given the continent a name yet. Now imagine someone calls a meeting. They rent a small inn for the summer, bring the best navigators together, and says: "Let's call this continent Intelligence. Let's agree it exists. Let's share our maps." No one finds the continent that summer. The meeting breaks up with a lot of arguments and unfinished sentences. But now everyone knows each other's names. They share their maps. They hire each other's students. They fight at future conferences in that productive, energizing way that only happens when smart people disagree about something they all care about. That is Dartmouth. The continent — machine intelligence — was not discovered in 1956. But for the first time, it had a name, a community, and a flag planted on the shore. Everything that followed — the AI winters, the neural network revolutions, the large language models — traces its lineage back to those six weeks in New Hampshire. The key mental shift is this: great scientific movements are not just found, they are organized. Dartmouth was an act of collective imagination as much as science.
Historical Development
Origins · 1 minhow this was discovered
Why It Matters
Prerequisites
What This Leads To
See Also
Alan Turing and the Turing Test · Claude Shannon and Information Theory · The History of Neural Networks · The First AI Winter · Symbolic AI and Expert Systems