He came to believe that if you were going to figure out how intelligence worked you had always to remember the particular tasks for which mind had evolved in the first place: running away from predators and toward mates and food. A mind’s first task, in other words, was to control a body. The idea of pure thought was biologically incoherent: cognition was always embodied. In the early days of A.I., intelligence had for the most part been talked about as the ability to do things that A.I. researchers found hard, like proving theorems and playing chess. Things that small children found easy, such as walking around without bumping into walls, or telling the difference between a stuffed animal and a table, were not thought of as requiring any interesting sort of intelligence at all. But then some researchers started to build robots, and they discovered that programming childlike skills like walking was actually extremely difficult—harder than chess. – The mind expanding ideas of Andy Clark, Larissa MacFarquhar (The New Yorker, 2018)