12-08-2017, 01:25 AM
This doesn't have to do with Josh Waitzkin, but everything to do with chess. So I'm borrowing this thread based on that angle.
The top chess program for years has been Stockfish. But Google's new Alphazero just destroyed it in a hundred-game match.
What is fascinating about this is that Alphazero was not "taught" the game of chess; rather, it was told the rules of the game, then given four hours to play itself. With that little bit of "machine learning," it bested Stockfish in dramatic fashion, never losing a game.
I walked through the one game they show in this article. Some of Alphazero's moves defy logic -- at least the logic a human can muster.
Maybe I'm overreacting to this, but I sense this is one of those tipping points in AI, where we accelerate rapidly towards our new masters.
https://www.chess.com/news/view/google-s...game-match
The top chess program for years has been Stockfish. But Google's new Alphazero just destroyed it in a hundred-game match.
What is fascinating about this is that Alphazero was not "taught" the game of chess; rather, it was told the rules of the game, then given four hours to play itself. With that little bit of "machine learning," it bested Stockfish in dramatic fashion, never losing a game.
I walked through the one game they show in this article. Some of Alphazero's moves defy logic -- at least the logic a human can muster.
Maybe I'm overreacting to this, but I sense this is one of those tipping points in AI, where we accelerate rapidly towards our new masters.
https://www.chess.com/news/view/google-s...game-match