The term is coined: the Dartmouth conference of 1956
It began in 1950, when British mathematician Alan Turing asked "Can machines think?" and proposed a test for it. But the field's actual founding moment was a summer workshop at Dartmouth College in 1956: a small group of researchers around John McCarthy and Marvin Minsky deliberately coined the term "Artificial Intelligence" there - partly to attract funding for a new, standalone field. Naming and marketing were part of the field from day one.
Early approaches: rules instead of learning
The first decades of AI research followed a completely different idea from today's models: instead of learning from data (see "How does a language model actually 'think'?"), people hard-coded rules and expert knowledge into the system - "symbolic AI". A well-known example is ELIZA (1966), a simple program that mimicked conversational responses through basic pattern matching. Many users at the time attributed real understanding to ELIZA, even though only text patterns were behind it - an early version of the hype that "Spotting AI hype vs. real value" describes today.
Boom and bust: the AI winters
In the 1980s, an offshoot of symbolic AI saw a commercial boom: "expert systems", which encoded individual specialists' knowledge as rule sets and were deployed in businesses. The problem: these systems scaled poorly and were helpless in unexpected cases. When the high expectations weren't met, funding and interest collapsed sharply more than once - first as early as the early 1970s (the UK's Lighthill report sharply criticized the field's meagre progress), then again in the late 1980s. These phases became known as "AI winters".
Why this matters for you as a decision-maker
This history isn't a side note: it shows that enthusiasm for AI has surged and collapsed before, whenever promises outpaced actual capability. Today's language-model boom (more on that in the next module) differs from earlier cycles mainly through the sheer volume of data and compute - not because "AI" was suddenly reinvented. That helps you read vendor claims like "we reinvented AI" (see "Spotting AI hype vs. real value" and "See through vendor pitches") with the right amount of distance.