We already talked about the 1997 chess game between IBM’s Deep Blue and Kasparov. We also talked about AlphaGo and its milestone in front of millions of viewers and spectators in 2016. On December 7, 2017, something happened that could be categorized along the same lines, but at the same time, we shouldn’t. On that day, Google’s AlphaZero defeated Stockfish 8. Bear with me, I swear I’m not writing this to sound nerdy.
Stockfish 8 was the chess champion computer in 2016 and had information collected from different centuries of human play and decades of games run and simulated by computers. This computer was capable of calculating approximately 70 million moves[174] on a chessboard per second, yes, per second.
AlphaZero, at that time, was not only capable of executing only 80,000 moves per second but also up to about five hours before its competition against Stockfish 8, AlphaZero had never played chess, nor had it been loaded with prior information about the game. Not only did the famous Netflix series, The Queen’s Gambit, not yet exist, but Google’s creation didn’t even know about the existence of that strategy, which involves sacrificing a piece, usually a pawn, to gain a later advantage in the game.
AlphaZero was not loaded with prior instructions. It was simply set to play and compete against itself for hours. When facing Stockfish 8, which happened one hundred times consecutively, AlphaZero won 28 times, drew 72 times, and never lost[175]. This was possible due to the power of deep learning neural networks developed by the Google team. Checkmate world.
[174] Perier T. (2021) Analysis of Artificial Intelligence in Chess. Astrakhan. Viewed on February 15, 2023, at https://astrakhan.fr/en/blog/analysis-of-artificial-intelligence-in-chess.
[175] Klein, M. (2017). Google’s AlphaZero Destroys Stockfish In 100-Game Match. Chess.com. Viewed on February 6, 2022, at https://www.chess.com/news/view/google-s-alphazero-destroys-stockfish-in-100-game-match.