sourced story
1987-1993Primary source · 3 sourcesWell documented

The Lisp machine crash brings a second AI winter

Cheap workstations catch up to $100,000 specialized AI hardware almost overnight

On the timeline · around 1987-1993 · Symbolic AI, Booms and WintersSymbolic AI, Booms and WintersThe Statistical and Machine Learning TurnThe Lisp machine crash brings a second AI winter19751980198519901995

Quick facts

Key companies affected
Symbolics, Lisp Machines Inc.
Government program cut
DARPA Strategic Computing Initiative
Symptom
Knowledge acquisition bottleneck in expert systems

What happened

Through the mid-1980s, companies including Symbolics and Lisp Machines Inc. sold specialized workstations, known as Lisp machines, built to run the AI programming language Lisp efficiently, priced from tens of thousands to over $100,000. By 1987, general-purpose workstations from vendors like Sun Microsystems had closed the performance gap at a fraction of the cost, and the dedicated Lisp-hardware market collapsed within about a year; Lisp Machines Inc. filed for bankruptcy in 1987 and Symbolics followed in 1993. At the same time, expert systems such as DEC's XCON, despite genuine early success managing tens of thousands of rules, proved brittle and expensive to maintain as their rule sets grew, a problem researchers called the knowledge acquisition bottleneck. DARPA, which had funded AI heavily through its Strategic Computing Initiative since 1983, cut funding sharply after concluding the initiative would not deliver on its original goals.

Why it matters

The second AI winter pushed most surviving researchers away from ambitious 'strong AI' goals toward narrower, more achievable 'weak AI' problems, a retreat that indirectly set the stage for the statistical, data-driven methods that would eventually succeed where symbolic rule systems had stalled.

How we know

Nils Nilsson's history of the field, written by a Stanford AI researcher who lived through the period, documents the Lisp-machine collapse and XCON's rule count directly, and an MIT case study of Symbolics' finances shows its declining revenue against rising commodity-hardware competition.

Sources

See something wrong? . Corrections with a source get fixed fastest.

Part of a timelineArtificial Intelligence30 events · From a wartime theory of neurons to machines that write, paint, and fold proteinsView all →