I, Pool Shark (The Globe and Mail, Anne McIlroy, Sat. July 24, 2004, p. F9) is a science story about Michael Greenspan‘s work on a pool playing AI/robot called “Deep Green”!
What is the connection between games and AI? Chinook, the checkers playing AI developed at the U of Alberta was the first “Man-Machine Champion” – the Guinness Book of Records says so. (See Chinook (ACJ Extra) by James Propp.) Deep Blue, got more attention as it mastered a more popular game – chess. Deep Blue is an IBM Research project that repeatedly beat Garry Kasparov in chess, demonstrating that computers could win at games we think of as complex and indicative of intelligence. Now Deep Green takes on a game with motor skils – pool.
Playing games is a paradigmatic human activity, for which reason, developing computers that can play them with us makes an interesting history. Games, in so far as they work within the constraints of simple rules and worlds, are easier to simulate on a computer than “real world” situations like conversation so they can stand in for Turing Tests. Programming a super-computer to win at some popular game is a way to get attention and support for an AI research project. I am tempted to say that just as we teach children through structured play, we are teaching computers by designing them to master one game after another. The difference is that the same child learns one game after another, and we don’t expect them to master games. Computer systems, in contrast, are being optimized for one game at a time and being developed until they beat the best.
What if the same system were taught to master one game after another? Could we see emergent properties independent of any particular game? What might be the games we would choose to teach it?
For an interesting essay on Deep Blue and popular culture see, The End of an Era, The Beginning of Another? Hal, Deep Blue and Kasparov by David G Stork.
Comments are closed.