MIT Artificial Intelligence Laboratory

Contribution to AI

Unmetered access to a machine, handed to students who picked their own problems, was itself a method here. The Incompatible Timesharing System, running from 1967, had no passwords and no file protection, and the working habits that grew on it — code as something you read, alter and hand on — descend directly to the free software movement Richard Stallman announced from the lab in 1983. The signature technical move was the microworld: shrink a domain until a programme can hold all of it, then demand full linguistic and physical competence inside those bounds. Blocks-world work fused parsing, reference and planning into one connected problem, and the failure of those results to survive contact with the open world later became the standard case against symbolic methods. Greenblatt's chess programme won a game against a rated human in tournament play in 1967, the first time that happened. Minsky and Papert's Perceptrons, in 1969, established what loop-free networks provably could not compute and was read for a decade as a verdict on connectionism generally. The lab then built hardware to run its own software, and the personal Lisp workstations it designed left MIT in 1979 and 1980 as Symbolics and LMI, supplying the machines the expert-systems industry ran on and, when demand vanished, the clearest marker of its collapse.

Drafted with AI and edited by hand (claude-opus-5, reviewed 2026-08).