The Improbability of Genuine Thinking Machines

Tim Crane at the TLS:

Brian Cantwell Smith’s new book is a provocative expression of scepticism about these recent claims on behalf of AI, from a distinguished practitioner in the field. His overall argument is based on a distinction between what he calls “reckoning” and “judgment”. Reckoning is understood here in its original etymological sense: as calculation, like addition and subtraction. Judgment, by contrast, is something more. It is described by Smith as “an overarching, systemic capacity or commitment, involving the whole commitment of the whole system to the whole world”. Our thinking involves not just some kind of simple on-off representation of things around us, but an entire emotional and value-laden involvement with the world itself. Computers have none of this. As the philosopher John Haugeland (a major influence on Smith) used to say, “computers don’t give a damn”. Giving a damn is a precondition of “judgment” in Smith’s sense, and anything that amounted to a real AGI would need to exercise judgment, and not simply calculate.

more here.