Elephants Don't Play Chess
Brooks argued that intelligence grew out of movement and survival rather than symbolic reasoning, so it must be built from the bottom up.
Why it matters
The most direct statement that the field had spent more than thirty years on the wrong problem, and it came from inside, from MIT.
The title points at the remark that it would be unfair to deny an elephant an intelligence worth studying just because it does not play chess. Brooks recalls the chronology of evolution: life arose about 3.5 billion years ago, fish about 550 million, man 2.5 million, agriculture 19,000 and writing less than 5,000 years ago. The paper prints agriculture as 19000 years ago; the same passage in Brooks's article "Intelligence without representation" (Artificial Intelligence 47, 1991, received September 1987) says 10,000 years ago. The record gives the figure as printed here. From this he concludes that problem solving, language, expert knowledge and reasoning are rather simple once movement and perception are in place, and that this foundation is the hard part. For mobility, vision and survival he cites Moravec; the thesis is close to Moravec's, but stated as an engineering programme, physically grounded systems built from the bottom up, rather than as an observation.