What Computer Chess Still Has to Teach Us: The Game That Will Not Go

Abstract

Computer chess started as a promising domain of research in Artificial Intelligence (AI) more than five decades ago. The basic idea was that as a ‘thinking sport’ it would be a good challenge toward better understanding and potentially simulating (human) cognition in machines. Unfortunately, it was soon discovered that computers could be made to play chess (and certain other games like it) using computational methods quite different from how humans are thought to think. This spawned a competitive computer chess gaming industry and in 1997, the world chess champion was defeated by an IBM supercomputer. Since then, computer chess has seen further improvement with programs playing at the strong grandmaster level even on desktop machines. In the field of AI, attention has therefore shifted to even more-complex games like Go in the hope that computational approaches toward them will succeed where chess had apparently failed. In this paper I challenge that contention. I have reasons to believe that chess still has some things to teach us not only about computation and its limits but also about the human mind and how it probably works. As such, this is not a technical paper but rather one on ‘computational philosophy’.