Hubert L. Dreyfus (1929-2017) was an American philosopher, for most of his career at the University of California, Berkeley, who became the most persistent philosophical critic of artificial intelligence during its symbolic era. Trained in the phenomenology of Heidegger and Merleau-Ponty, he argued that human intelligence is grounded in embodied, situated skill and tacit know-how that cannot be captured by explicit rules and symbol manipulation - and that AI’s founders had simply assumed, without warrant, that it could.
His critique first appeared in a 1965 memo for the RAND Corporation titled “Alchemy and Artificial Intelligence,” hosted by the Computer History Museum. The provocative title compared AI researchers to alchemists: people reporting early, dramatic successes who mistake them for steps toward a goal they will never reach. He expanded the argument in the 1972 book “What Computers Can’t Do,” revised in 1979 and again in 1992 as “What Computers Still Can’t Do,” whose successive titles tracked his view that the field kept failing for the same reasons.
Dreyfus was treated with hostility by much of the AI community for years, and some of his specific predictions aged poorly. But his core point - that rule-based symbolic systems struggle with the open-ended, common-sense, bodily character of everyday competence - anticipated the problems that helped end the symbolic era, and it resonates with the later embodied-AI movement and with debates about what today’s statistical systems do and do not understand.
For business readers, Dreyfus is the canonical example of an outside critic whose objections were dismissed as obstructionist but turned out to identify real limits, a useful caution against treating early demos as proof that a hard problem is nearly solved.