Google and Deepmind’s Go-playing A.I., AlphaGo, represents a huge shift for artificial intelligence as a concept. The A.I. is based on machine learning and tree search, concepts that are in their infancy, and is designed to play the ancient game of Go, long thought of as nearly impossible for robots to play well due to the prohibitive amount of branching paths that’s possible in the game. This makes traditional A.I. tactics, such as brute-forcing, somewhat ineffective. AlphaGo, being based on newer technology, did not. After a string of wins against humans, even national champions and other A.I.s, they announced earlier in the year that the A.I. would be going up against the current Go world champion, Lee Sedol of Korea, in a series of planned matches. Earlier today, the A.I. won its first of five match against Sedol.
A bow at the beginning of the match may not have meant much to AlphaGo, but to Sedol, he was greeting a worthy opponent. Three and a half hours later, the historic match was over and AlphaGo had prevailed. It was a rough and tumble match. Both parties made some mistakes throughout, but where Sedol got nervous, needed to sip water and, at times, forgot key things, AlphaGo did not. All of the mistakes it made were due to its nature as an A.I., inherently less capable than a human of the kind of reasoning necessary to play the ancient and complex game of Go.
At a press conference after the round, Sedol seemed utterly crestfallen. Reporters described him as being “meek” and “on the verge of tears” as he talked about the match, saying, “I didn’t know AlphaGo would play such a perfect game.” The result rocked the Go world; many enthusiasts fully expected the A.I. to meet its match at the hands of the current world champion, the best human player that the game had to offer. The AlphaGo A.I., on the other hand, made use of “reinforcement learning” to learn from its mistakes in matches against other A.I.s and people, as well as against itself. The match with Sedol was no different, which means at this point that a win for the world champ is unlikely in the remaining four games.