Sir: William Hartston has a very interesting analysis of why Kasparov was able to beat Deep Blue (19 February). He argues that Deep Blue was unable to recognise the concept of a defensive wall. He goes on to argue that a human could learn such a concept by means of examples but that machines could not generalise from the examples.
Machine-learning programs have been developed since the Fifties, and there are many cases of programs learning concepts from examples. Grandmasters have spent many years learning about chess, and it is almost certainly the case that any machine that can adequately beat a grandmaster will have to use sophisticated learning programs and a lot of experience.
Department of Computer Studies
University of Glamorgan
Pontypridd, Mid GlamorganReuse content