The pattern is a boom, a disappointment, then a freeze. In 1973 the mathematician Sir James Lighthill wrote a report for the UK’s Science Research Council that judged most AI research to have fallen short of its promises, and British funding for the field was cut back. That June the BBC broadcast a debate between Lighthill and AI researchers including Donald Michie, which Edinburgh’s AI institute still archives. A second slump came in the late 1980s, when companies that had bought expensive rule-based expert systems found them hard to maintain, and the specialized hardware market built around them collapsed.
The phrase itself came from researchers worried about a repeat. At the 1984 AAAI conference, a panel later published as “The Dark Ages of AI” debated whether the commercial excitement of the time was setting the field up for another fall. It was.
A business version is easy to picture. A company funds an ambitious AI program on the promise of near-human results, the pilots underwhelm, and the budget for the whole category disappears for years, including the projects that were working. That is why the history matters to builders today: winters punish the honest projects along with the overhyped ones.
On Chain of Thought the term comes up as a live question rather than history. In episode 22, at 27:11, Databricks’ Denny Lee says people are talking about the idea of an AI winter, and argues the real limit over the next couple of years is energy, not inference cost. That is his view of the bottleneck, not a forecast that a winter is coming. The episode was sponsored by Galileo.
Today’s equivalent worry is usually framed as an AI bubble: heavy investment ahead of proven returns. For how guests separate hype from what ships, see Is the AI Agent Bubble Real?. Related history: symbolic AI and the Turing test.
Hear it from the guest
“And people are talking about the idea of AI winter, and I think the really limiting factor is gonna be energy for the next couple years”
Quotes lightly edited to remove filler words.
Sources
- AI Magazine: The Dark Ages of AI, a panel discussion at AAAI-84 — Record of the 1984 AAAI panel, chaired by Drew McDermott, where researchers debated whether inflated expectations would bring on a downturn in AI funding and interest; checked October 24, 2026.
- University of Edinburgh AIAI: the 1973 Lighthill controversy debate — Archive of the June 1973 BBC debate at the Royal Institution between Sir James Lighthill, Donald Michie and others, held after Lighthill's report criticized AI research in the UK; checked October 24, 2026.
