Textbook of AI

1929–2017, Philosopher of mind

Hubert Lederer Dreyfus was an American philosopher whose 1972 book What Computers Can't Do: The Limits of Artificial Intelligence (revised 1979 and as What Computers Still Can't Do, 1992) was the most sustained and influential philosophical critique of the symbolic AI programme. Drawing on Heidegger, Merleau-Ponty and Wittgenstein, Dreyfus argued that human intelligence depends on embodied skill and background know-how that resist explicit formalisation; the symbolic-AI assumption that intelligence consists in symbol manipulation could not, he claimed, capture the embodied character of human understanding.

Dreyfus's predictions about specific shortcomings of symbolic AI, failure of generality, brittleness, inability to handle commonsense, proved largely correct. The mainstream AI community reacted to his critique with hostility through the 1970s and 1980s, and Dreyfus accumulated a personal collection of letters from AI researchers calling him incompetent and worse. With the decline of pure symbolic AI and the rise of statistical and embodied approaches in the 1990s, Dreyfus's standing improved.

Dreyfus spent the bulk of his career at UC Berkeley, where he taught Heidegger and Merleau-Ponty. His brother Stuart Dreyfus is a co-author of his AI critiques and an applied mathematician who contributed to dynamic programming.