The funding letter that named a field

On 31 August 1955, four researchers signed a proposal for a summer research project. John McCarthy was a young mathematician at Dartmouth College. Marvin Minsky was at Harvard, Nathaniel Rochester at IBM and Claude Shannon at Bell Labs. They asked for money for a two-month study by about ten people at Dartmouth in the summer of 1956.

The proposal is only a few pages long, and it is where the phrase "artificial intelligence" appears to have been used for the first time. Its starting idea was bold: that every aspect of learning and intelligence could, in principle, be described so precisely that a machine could be made to imitate it.

It listed the problems to work on: getting computers to use language, building networks loosely modelled on brain cells, forming abstract ideas, having machines improve themselves, and the role of randomness in creativity. The authors wrote that they thought a carefully chosen group could make a significant advance on one or more of these in a single summer.

The workshop went ahead in 1956. It did not solve these problems. It did give the field its name and brought together several people who would lead AI research for decades.

Almost every topic in that list is still being worked on. Chatbots that use language, neural networks, and systems that improve from experience are all on it. What the authors got wrong was the timescale. A summer turned into roughly seventy years, and the field went through long quiet periods when promises ran ahead of results.

For someone using a chatbot today, the lesson is about the word itself. "Artificial intelligence" was a label chosen for a research proposal, not a description of something that already existed. Today it covers everything from spam filters to chatbots. When a product says it uses AI, that tells you very little. It is better to ask what the system actually does, what it learned from, and how well it works on your kind of task.

The idea behind this caseLesson: What is AI, really? →

1. 1When a product says "powered by AI", ask what exactly it does and what it was trained on.
2. 2Be wary of firm timelines for big breakthroughs. Experts have been too optimistic since 1955.
3. 3Judge AI tools on your own real tasks, not on the label or the promise.
4. 4Remember that "AI" covers many different systems. A chatbot, a face filter and a fraud detector work in very different ways.