The field acquired its name at a summer workshop, and the name has been a problem since

A proposal, a summer, and a label that stuck

In the mid-1950s a small group of researchers proposed a summer study at Dartmouth College, on the premise that every aspect of learning and intelligence could in principle be described precisely enough for a machine to simulate it. The gathering took place in 1956, ran through the summer, and is conventionally treated as the moment the field began.

The phrase artificial intelligence appears in that proposal, and the workshop is where it entered circulation. It was not obviously the natural choice at the time; work on similar problems was proceeding under headings such as cybernetics, automata studies and complex information processing, each carrying its own community and assumptions.

Accounts from participants suggest the new phrase served partly to mark out territory distinct from those existing traditions, and partly because it was arresting. Both motives are ordinary academic behaviour. The consequence was not ordinary at all.

The workshop did not do what the name implies

The gathering itself was loosely organised. People came and went, there was no single agenda, and by most retrospective accounts it produced no breakthrough and little consensus. Its importance is retrospective: it assembled people who went on to found laboratories, train students and define research programmes for the next thirty years.

The proposal’s ambitions were extraordinary by any standard, sketching progress on language, abstraction, self-improvement and creativity within a period measured in months. This was not cynicism. It reflected a genuine and widely shared belief that the essential problems were within reach.

That belief was wrong about the timescale by an amount that is difficult to overstate, and the pattern it established — sincere confidence, ambitious framing, slower reality — recurs so regularly in the field’s history that it deserves to be treated as a structural feature rather than a series of individual misjudgements.

A name that promises the destination

Most disciplines are named after their subject matter or their method. This one is named after an outcome it has not achieved, which creates a permanent gap between what the label announces and what the work consists of.

The practical effect is that every system built in the field is measured, by the public and often by funders, against the word rather than against its actual specification. A program that schedules deliveries extremely well is described as artificial intelligence, invites comparison with human intelligence, and is found wanting on a criterion nobody designing it ever adopted.

It also produces a persistent goalpost problem. Once a capability is achieved it stops seeming like intelligence and becomes ordinary software, so the field is left permanently defined by whatever remains unsolved. Chess, translation and handwriting recognition were each considered hallmarks of intelligence until they worked.

Alternatives were proposed and none of them took

Various participants and later researchers preferred other terms. Suggestions along the lines of complex information processing, machine intelligence or computational rationality were offered at different times, and one occasionally sees a deliberate preference for machine learning or statistical modelling as a way of avoiding the loaded word.

None displaced it. The phrase is memorable, it communicates ambition to funders and to the public, and it had a thirty-year head start by the time serious objections accumulated. Terminology rarely gets fixed by argument once it has entered general use.

The current situation is a mild absurdity: practitioners frequently describe their work as machine learning or statistics precisely to avoid the connotations of the field’s own name, while the same name is used enthusiastically in every announcement about their results.

What the naming episode is worth as history

It is a useful corrective to the sense that scientific fields emerge from a considered account of their subject. This one was named at a workshop by a handful of people making a rhetorical choice, and the choice has shaped expectations, funding cycles and public argument for the better part of a century.

It is also a reminder to read claims about the field with attention to which sense of the word is in play. Nearly every dispute about whether something counts as real artificial intelligence is a dispute about a term that was never defined precisely enough to adjudicate.

The most defensible use is the deflationary one: the field is the study of getting machines to do things that, for the moment, they are not obviously good at. That is unglamorous, historically accurate, and considerably less likely to mislead.