Its name is AlphaZero; it defeated the world’s best chess program in four hours. Before the game, he knew nothing, but how the chess pieces are able to move.
Now AlphaZero is the best chess player on the Earth. This system created by Deepmind did not lose a single game of 100 games played in the tournament with Stockfish 8, the best chess program.
AlphaZero is a modified version of AlphaGo Zero, an artificial intelligence that recently won the tournament of Go with the famous AlphaGo. The latter won the world’s best human players.
The system works almost identically with AlphaGo Zero, but instead of it, the machine is programmed to play chess and shogi, another Japanese strategic game. AlphaZero acquired the needed skills without any assistance. Unlike all previous chess programs, it had no empirical data, no database with the archives of the chess games, no knowledge of chess strategies. He just knew the purpose of the game.
AlphaZero has learned not only how to play chess in four hours, but how to become an absolute chess champion. AlphaZero just began to play on its own repeatedly improving its skills. If it is possible to count in human terms, it spent about 1400 years playing chess during these four hours. The system can handle 800,000 positions per second, while Stockfish 8 can handle 70 million moves per second.
After the first success, it surpassed the AlphaGo playing Go in eight hours. At last, he won the program Elmo that used to be a champion in the game of shogi in two hours.