Short answer
The term "artificial intelligence" was coined by John McCarthy, a young mathematician at Dartmouth College, in a funding proposal dated August 31, 1955 that he wrote with Marvin Minsky, Nathaniel Rochester and Claude Shannon. The proposal asked for a summer workshop at Dartmouth in 1956, an event many people regard as the founding moment of AI as a field.
Follow-up question What did the 1956 Dartmouth workshop set out to do?
The name "artificial intelligence" comes from a short document typed in the summer of 1955. John McCarthy, who was teaching mathematics at Dartmouth College in New Hampshire as a young assistant professor, wanted to bring a small group of researchers together to work on thinking machines. With three co-authors he wrote a proposal to the Rockefeller Foundation, dated August 31, 1955, titled "A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence." That document is credited with introducing the term.
The co-authors were well chosen. Marvin Minsky was at Harvard, Nathaniel Rochester worked at IBM, and Claude Shannon, one of the founders of information theory, was at Bell Telephone Laboratories. Their names gave a bold idea a serious hearing.
Why McCarthy picked that name
In the early 1950s, research on thinking machines went by several labels: cybernetics, automata theory, complex information processing. Each pointed to a different picture of what such machines would be.
McCarthy chose "artificial intelligence" partly because it was neutral. "Automata theory" felt too narrow, and "cybernetics" was strongly tied to analog feedback systems and to the mathematician Norbert Wiener, whose influence McCarthy did not want to inherit. A fresh name let the new group define its own questions.
That choice has had a long life. The word "intelligence" invites comparisons with people that no program can yet live up to, but the name stuck, and today it covers everything from chess programs to the language models in this year's news.
What the proposal actually says
The opening paragraph is the part most often quoted, and it is ambitious. The authors asked for "a 2 month, 10 man study of artificial intelligence" during the summer of 1956. Their working assumption was that every part of learning, and every other feature of intelligence, could in principle be described precisely enough for a machine to imitate it.
They listed seven areas to explore. Here they are, with a plain-language note on each:
| Topic in the proposal | What it means in everyday words |
|---|---|
| Automatic computers | The real limit is our programs, not machine speed |
| How a computer can be programmed to use a language | Getting machines to work with words and rules of reasoning |
| Neuron nets | Arranging simple artificial "neurons" so they form concepts |
| Theory of the size of a calculation | Measuring how efficient a method is, instead of trying every answer |
| Self-improvement | Machines that get better at their own tasks |
| Abstractions | Forming general ideas from sensory and other data |
| Randomness and creativity | Whether a dash of guided randomness explains creative thinking |
Read that list again with this year in mind. Language, neural networks, learning and creativity are still the headline topics. The budget, by contrast, was modest: the authors asked for $13,500.
Try it: Take the seven topics in the table and, for each one, write down one product or research result you have heard about recently. Which topics have clear examples, and which still feel wide open?
Who came, and what happened
The workshop did not run as a tidy ten-person seminar. According to notes kept by the participant Ray Solomonoff, the project ran for about eight weeks, starting around June 18 and ending near August 17, 1956, and people came and went. Besides the four organizers, those who took part or visited included Ray Solomonoff himself, Herbert Simon, Allen Newell, Arthur Samuel, Oliver Selfridge and Trenchard More, among others. McCarthy wrote beforehand that the group would focus on how to program a calculator to form concepts and generalizations, while noting that plans could change once everyone arrived.
It was less a structured conference than an extended brainstorming session. No single breakthrough came out of that summer. What came out was a community, a shared set of problems and a name.
What did the 1956 Dartmouth workshop set out to do?
It set out to test a bold guess: that intelligence could be described so exactly that a machine could simulate it. The proposal names concrete goals: machines that handle language, build concepts and abstractions, tackle problems only people could then solve, and get better on their own. The authors believed a carefully chosen group working together for one summer could make a significant advance on at least one of those problems.
That was very optimistic. The problems on the 1955 list are hard, and in 2019 most of them are still open research areas.
For a beginner, the useful lesson is the method rather than the timetable. The Dartmouth group broke a huge question, "can machines think?", into smaller ones that could be studied: language, learning, efficiency, abstraction, creativity. When you read about a new AI system, try asking which of those smaller questions it actually answers. That habit keeps the bold name in perspective.
Sources
- A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence — J. McCarthy, M. L. Minsky, N. Rochester, C. E. Shannon (copy on John McCarthy's Stanford site), August 31, 1955
- Dartmouth workshop — Wikipedia (archived April 2019), January 13, 2019