Dartmouth Summer Workshop

In the summer of 1956, about a dozen scientists met at Dartmouth College in the US state of New Hampshire. They intended to work together for several weeks on one question: Can a machine be built that thinks? The meeting was organized by the mathematician John McCarthy, together with three colleagues. In their funding proposal, they wrote a term that had not existed before: “artificial intelligence.” This gave an entire field of research its name. The meeting itself did not produce finished results, but it turned scattered individual interests into a discipline.

The birth certificate of a field

Before 1956, the idea of thinking machines certainly already existed. But it was spread across many disciplines: mathematics, electrical engineering, psychology, linguistics. Everyone worked under a different label, such as “cybernetics” or “automata theory.” McCarthy deliberately chose the new name because he wanted to distance himself from these older schools of thought.

A shared name has practical consequences. One can apply for funding for it, name professorships after it, and found journals about it. That is exactly what happened in the years that followed. The participants of Dartmouth built the first AI labs at universities such as MIT, Stanford, and Carnegie Mellon. These three institutions shaped the field for decades.

Today the workshop also stands for a recurring pattern: grand promises and long waiting periods. The proposal stated that a carefully selected team could make significant progress in a single summer. In reality, it took nearly seventy years to reach systems like today’s chatbots. Anyone who reads in the news that AI is “on the verge of a breakthrough” should be aware of this history.

What the participants had set out to do

The 1956 proposal named seven topics. These included language, artificial neural networks, learning ability, creativity, and the question of how a machine can improve itself. The underlying assumption was clearly stated: every aspect of intelligence could be described so precisely that a machine could replicate it. To this day, this assumption has been neither proven nor disproven.

In practice, the meeting turned out differently than planned. There was no fixed program and no joint work spanning weeks. Participants came and went, gave talks, and held discussions. The proceedings resembled an open summer-guest stay more than a conference with a schedule.

Nevertheless, one result was demonstrated: the “Logic Theorist” by Allen Newell and Herbert Simon. This program proved mathematical theorems independently by systematically trying out possible proof steps. It is regarded as the first working AI program ever. The other participants, however, reacted with more reserve than enthusiasm.

Why the date appears in every AI retrospective

When the media report on the history of artificial intelligence, the timeline almost always begins in 1956. Accordingly, the workshop will be mentioned frequently at its seventieth anniversary in 2026. Companies also like to invoke the date to place their own products within a long tradition.

In school lessons and introductory lectures, Dartmouth serves as the starting point of a narrative. The spirit of optimism was followed by two phases known as “AI winters”: periods in the 1970s and late 1980s during which funding was cut because the promises had not been fulfilled. It was only from around 2012 onward that the field gained significant momentum again through learning networks.

A common misconception holds that artificial intelligence was invented at Dartmouth. What was mainly invented there was the name. Important groundwork came from Alan Turing, who had already asked in 1950 whether machines can think. The workshop was thus less a beginning than a gathering point where scattered ideas were given a common title.