The «Dark Ages of AI» panel and the «AI Winter»
At AAAI-84 in Austin (6-10 August 1984) a panel discussed the worry that expectations of AI were too high. In his opening Drew McDermott described a worst case — the collapse of Strategic Computing and the Fifth Generation, failed startups, Texas Instruments and Schlumberger losing interest, «you can't get money for anything connected with AI» — and called it the «AI Winter» «by some». The transcript was printed in AI Magazine in the fall of 1985.
Why it matters
The winter here is a forecast, not an event: no cut had yet happened, and the DARPA representative (Ohlander) expects government funding to continue. The document's weight is that it fixes the mechanism in advance: a backlash against inflated promises, not the result of one decision. This is an editorial assessment.
Participants: Drew McDermott (Yale; opening and chair), M. Mitchell Waldrop (Science; the press), Ron Ohlander (DARPA), Roger Schank (Yale), B. Chandrasekaran (Ohio State), John McDermott (Carnegie-Mellon). McDermott: «This condition, called the "AI Winter" by some»; he rates the scenario as unlikely and urges «disciplining ourselves» and educating the public. Ohlander: DARPA's basic programme is funded at the same level with a «sliding piece» over the next few years; he expects a «shakeout» among machine makers and expert-system companies but that government will sustain or increase funding; «We're pretty much riding the whirlwind». A later speaker: «Will there be an AI Winter? I think there may be an AI dusk». What the record does not claim. The words «AI Winter» are given in the opening as someone else's phrase («by some»); the text read does not say who coined it or whether it was voiced publicly earlier, and no earlier document with it was found. The day of the panel within the conference is not printed. The transcript is a record of a talk edited for print in 1985. The collapse of the Lisp machine market (1987) is a separate event.