I’ve been writing a fair bit lately about IBM’s Watson project, so it seems appropriate to mark this anniversary of a significant milestone in one of IBM’s earlier machine intelligence projects. As an article at Wired reminds us, it was fifteen years ago today that IBM’s Deep Blue system won its first game of chess against the grand master Garry Kasparov, on February 10, 1996. Kasparov went on to win that match, winning four games to Deep Blue’s two. He did not fare as well in a re-match in 1997.
Kasparov lost a rematch to Deep Blue the following year — his first match loss ever to any kind of opponent.
Deep Blue’s victory generated a fair amount of publicity at the time. Chess, especially to those who don’t play it, seems like a complex game; for a computer, though, it is a simpler problem than Jeopardy!, because its rules are very well defined. Essentially, DeepBlue was able to evaluate more potential moves, and their consequences for future moves, than a person could.
Superficially, Jeopardy! seems easier; most educated people can get a couple of responses right in a typical half-hour episode. But the essentially unlimited problem domain, and the trickiness of natural language, make it a very hard problem for a machine.