Computer Beats Master at Go
“The path for machine victory over the humans who play the ancient Chinese game of Go was not achieved through mathematical superiority, because Go is a very different game from chess.
Rather than the 8 × 8 grid for chess, Go is played on a 19 × 19 board, with each player having dozens of black or white stones. Each stone has the same value—unlike chess, in which the pieces are not all equal. The rules of Go are fairly straightforward—the two players try to surround each other’s stones and take territory from each other. However, because of the size of the grid, the number of potential positions in Go is staggering—considerably larger than the number of atoms in the Universe.
This sheer complexity is why intuition is so often cited as a key factor in winning the game, and why a computer program beating one of the best Go players that ever lived was considered so significant. As players add more stones to the board, the number of possible countermoves and counter-countermoves grows exponentially. As a result, brute-force “look-ahead” computing approaches to solving Go just can’t look far enough ahead: computers aren’t big enough. The Universe isn’t big enough.
AlphaGo® is the AI program that beat South Korean Go master Lee Sedol (b. 1983) in March 2016, in four out of five games, by adopting the same sort of strategic search strategies a human would. The program was created by the Google DeepMind team that evolved from Google’s acquisition of British company DeepMind Technologies, a British AI company that built a neural network to play video games like a human.
Lee Sedol did win once, however, so the computer did not dominate the match. In game four, white move 78, Lee Sedol found AlphaGo’s Achilles’ heel and made a move that so thoroughly confused the system that it started to make rookie mistakes, not recovering in time to save the game. The irony is that Sedol placed the stone where he did because AlphaGo had put him in a position where he saw no alternative move to make.”
SEE ALSO Computer Is World Chess Champion (1997)
“Go is played on a 19 × 19 board, with one player using black stones and the other using white stones, all possessing the same value.”
Byford, Sam. “Why Is Google’s Go Win Such a Big Deal?” The Verge, March 9, 2016. https://www.theverge.com/2016/3/9/11185030/google-deepmind-alphago-go-artificial-intelligence-impact.
House, Patrick. “The Electronic Holy War.” New Yorker online, May 25, 2014. https://www.newyorker.com/tech/elements/the-electronic-holy-war.
Koch, Christof. “How the Computer Beat the Go Master.” Scientific American online, March 19, 2016. https://www.scientificamerican.com/article/how-the-computer-beat-the-go-master.
Moyer, Christopher. “How Google’s AlphaGo Beat a World Chess Champion.” Atlantic online, March 28, 2016. https://www.theatlantic.com/technology/archive/2016/03/the-invisible-opponent/475611.