AlphaGo A.I. Beats World Champion Go Player

In October of 2015, Google DeepMind’s artificial intelligence played Fan Hui, Europe’s Go champion. Go is an ancient Chinese board game originating over 2500 years ago with more board configurations than atoms in the universe. This artificial intelligence, known as AlphaGo, beat the European champion 5-0, a feat never before done by an artificial intelligence.

Normally, artificial intelligences can use their brute force through the different possibilities of moves in simpler games, such as chess, and determine which move would eventually end in favor of the AI. This is not the case in Go, as in each turn there can be over 250 different possible moves, so calculating multiple moves in the future is nearly impossible. Because of this, AlphaGo utilizes two different neural networks, one called the “policy” network and one called the “value” network. The policy network chooses a few moves to analyze further from a large pool of possible moves whilst the value network reduces the depth of the search and only analyzes the few selected moves from the Policy network and calculates how that move will affect the game 20-30 moves ahead instead of 300 moves ahead like normal artificial intelligences would. This makes the AI work more like normal humans but still maintains the benefits of a supercomputer.

More recently, however, AlphaGo got the chance to face the best Go player alive, Lee Sedol. The system was able to learn by playing not only other players but also itself, similar to the WOPR from the 1983 film War Games. AlphaGo and Lee Sedol faced off in the Four Seasons Hotel in Seoul, South Korea on March 9th, 10th, 12th, 13th, and 15th on Google DeepMind’s Youtube channel and on live television throughout Asia. In the games, which have been watched almost seven million times, AlphaGo proved its superiority beating the world champion 4-1. This proved to be a big accomplishment in the artificial intelligence world. No AI has ever beaten a professional Go player without a handicap, making this the first time ever a computer has been able to best a human at one of the most complex games ever created.