The First AI Winter

First, with winters, it's important to note that they're more about funding than sentiment per se. Certainly capital and sentiment can go hand in hand. But just because AI capital mostly went away, no doubt damaging spirit, it in no way whatsoever permanently destroyed morale.

Just to recap, from 1969 onwards, AI funding went nearly exclusively to its symbolic, expert system branch. The only issue is that after a few years, that camp's limitations became quite apparent. With apparently nowhere at all for the funds to go as symbolic enthusiasm waned, we enter the first AI winter proper. Now, AI does not completely disappear the rest of the decade as later moments (MYCIN, frames and scripts) will show, but things do get quieter for a bit.

Hard to believe but even the US military with its unlimited budget and keen (KEEN) interest in smarter weapons stopped its funding. The Mansfield Amendment of November 1969 had limited Defense Department support to research of “direct and apparent” military use.

But the real death knell was that pesky Lighthill Report. Its influence extended well beyond the UK as it prompted a Pentagon panel to review DARPA’s own AI program. The result? DARPA moved money out of robotics, hurting Stanford Research Institute. DARPA also ended its five-year Speech Understanding Research program without the planned follow-on.

But it's important to note that AI winters are best thought of as funding dry spells, slower periods, rather than full-fledged cataclysms of nothingness. (again, see the many future moments in this decade about AI). Plenty of evidence suggests that despite wintery conditions, grassroots efforts expanded during this time. The historian Thomas Haigh disputes talk of AI winters and notes that SIGART, the ACM’s AI interest group, grew from 1,241 members in 1973 to 3,500 by mid-1978. People might have cared even if dollars didn't at this time.