“There is an entity that cannot be defeated”, pronounced the former Go world champion, Lee Sedol, in his recent announcement of retirement from the sport, a board game that is popular in South Korea and China. The entity he refers to is AlphaGo, an artificial intelligence (AI) that in 2016 defeated, for the first time ever, a top-ranked Go player, winning 4 games to 1.
For AI, this was a momentous achievement due to the complexity of Go compared to other board games such as Chess, where computers have been able to defeat grandmasters since 1997. Go had evaded researchers due to two problems: the number of possible moves a player can make (an average of 250 moves compared to Chess’s 35) and in evaluating the strength of a board position in a computer program. Continue reading