After conquering the game of Go, the AI-driven AlphaGo computer program is showing that it can easily dominate the chess world, too.
The latest iteration of AlphaGo bested a leading computer chess program known as Stockfish. Over the course of 100 matches, it won 28, and came to a draw on the remaining 72, never losing once.
Researchers with DeepMind, the Alphabet-owned company behind the computer program, detailed the accomplishment in a paper out this week. This new version of AlphaGo, called AlphaZero, taught itself chess by using a machine-learning technique called "reinforcement learning," where the program played the game against itself. In a mere four hours, AlphaZero optimized its play to the point the program began outperforming Stockfish, according to the paper.
Chess enthusiasts have been impressed with the program's immediate dominance.
"It's a fantastic achievement with dazzling implications," tweeted Garry Kasparov, a former champion who famously competed and lost against IBM's chess program Deep Blue back in 1997.
"On the 6th of December, 2017, AlphaZero took over the chess world," tweeted Simon Williams, another chess grandmaster.
It's also a notable feat for AI research and underscores the potential for the technology to translate from one domain to another. The AlphaGo program, for instance, was originally designed to play the Chinese board game of Go and optimized its play by competing with human players. However, AlphaGo researchers experimented with the program's ability to teach itself to play, without any need for human feedback, which ultimately translated to chess, too.
When playing the game, AlphaZero will search the board for 80,000 positions per second. That may sound like a lot, but it's actually far fewer than the 70 million positions the Stockfish program will seek. However, AlphaZero is focused more on quality than quantity. It may be studying fewer positions, but it's looking for the most promising ones, an approach that worked against Stockfish.
In addition to chess, AlphaZero also learned to master the Japanese board game known as Shogi, which is similar to chess. Perhaps to no one's surprise, AlphaZero beat the top computer program for Shogi by winning 90 out of the 100 matches.