|
To all those who say Kasparov is playing the programmers; that's not quite right. The cool things in computing these days are done using adaptive systems, which are systems that develop themselves over time. The computer would have learned how to play the game; it's not going through set algorithms the programmers have created (Glytch suggested that Kasparov was up against 'damn near every strategy ever though up', which is far from the truth. He's only up against the ones Fritz has learned and developed himself). Yeah, the system will only be as good as the programmers who wrote it, but the programmers don't have to be good at chess for the program to be great. They just have to be good at writing adaptive systems. I'd wager that Fritz could absolutely thump every one of the programmers who wrote it, every time.
dnd, I don't know much about the architecture of Fritz, but it's unlikely that it is a neural network. At the moment, they're not very good at chess (next year, I'll possibly be working on a neural net that plays checkers, and even that is pretty challenging). More likely, it'd be some kind of genetic algorithm.
(They've also done some nifty things recently with game-playing cellular automata, but they're still a long way off world-class)
As for Go, that's the holy grail of computer gaming agents. There are way too many possibilities in the game for the computer to search even a fraction of them. The game seems to be played more by intuition than strategy; even great Go players can't really explain how they play well. So what you're trying to teach the computer isn't intelligence, which they're fairly good at, as much at it is intuition and judgement, which they have quite a hard time with.
__________________
Strewth
|