Alpha Zero, an AI created by Google’s subsidiary DeepMind beat Stockfish 8, the world’s best chess-playing computer program, after having taught itself how to play the game in under four hours.
In a 100-game contest, Alpha Zero won 28 games – 25 with white pieces and three with black – and drew 78.
Last Tuesday, DeepMind published a paper describing their program that achieved a superhuman level of play in chess, shogi (a similar Japanese board game) and Go.
“Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi as well as Go, and convincingly defeated a world-champion program in each case,” said the paper’s authors, which include DeepMind founder Demis Hassabis, who was a child chess prodigy.
“We have been playing this game for over 600 years, and all it takes for it is four hours,” joked former Indian world champion Viswanathan Anand about Alpha Zero’s win. The “four hours”, he said, was just “a nice punchline”.
“The more relevant thing is it figured everything from scratch,” Anand added. “That is more scary and promising if you look at it.”
Here’s the full video:
For over a thousand years, the game’s creators – us, humans – were its greatest masters. Then, when we created computers, they took over from us. With the increase in power of their engines resulted a surge in their intelligence. IBM’s computer program Deep Blue supercomputer defeated Garry Kasparov in May 1997. Ever since, it has become almost impossible for humans to defeat computers in chess.
“It’s a remarkable achievement, even if we should have expected it after AlphaGo,” former world champion Kasparov told Chess.com. “We have always assumed that chess required too much empirical knowledge for a machine to play so well from scratch, with no human knowledge added at all.”
