In the late 1980s, the expert-systems and Lisp-workstation boom deflated: specialized hardware lost to general workstations, and many rule bases failed to deliver ROI.
Companies failed or pivoted; conferences and magazines spoke of an “AI winter.” Meanwhile, in vision, speech, and data mining, statistical methods and networks (with backprop) advanced with less fanfare.
That winter prepared ML’s ideological ground: fewer AGI promises, more benchmarks, datasets, and probability. The “AI” brand would rehabilitate only after public milestones like Deep Blue and later ImageNet.
