Go champion battles Google DeepMind AlphaGo computer

One of the world's top players of the game Go will take home $1 million US if he beats a Google computer in a five-match tournament that started Wednesday.

Lee Sedol, 33, a professional South Korean Go player who has the secondmost international Go titles in the world, will face Google's AlphaGo in Seoul, South Korea, for the first time. The game was live streamed on YouTube and lasted about five hours.

Subsequent matches will take place over the next week. Updates will be posted on theGoogle Asia Pacific blog.

Go is a two-player strategy board game that originated in ancient China. Players place black and white pieces on a square grid, and each person tries to take more territory than the other. It is considered far more complex than chess.

While computers have been beating human chess champions for nearly two decades, theyhave only played Go at the level of human amateurs until very recently.

AlphaGo, created by Google's DeepMind artificial intelligence team, made headlines when the company announced in the journal Nature in January that it had beat three-time European Go champion and Chinese professional player Fan Hui in a five-game match.

Sedol is ranked much higher than Fan internationally, but he expressed some trepidation about playing AlphaGo ahead of the match.

Human mistakes

"Because humans are human, they make mistakes," he said at a news conference. "If there are human mistakes, I could lose."

Lee added that in playing against a machine, the absence of visualcues that human players use to read the reactions and psychology oftheir opponents puts him in unfamiliar territory.

"In a human versus human game, it is important to read the otherperson's energy and force. But in this match, it is impossible toread such things. It could feel like I'm playing alone."

Demis Hassabis, founder and CEO of Google DeepMind, saidanother advantage AlphaGo has is that it will never get tired or intimidated.

Go is very challenging for computers because of the huge number of possible moves. AlphaGo works by not considering all moves, but narrowing down its possible moves, based on what it has "learned" from its past mistakes, as well as from data from about 100,000 Go games available online.

If AlphaGo wins, it will also take home $1 million US, but the money will be donated to UNICEF, STEM and Go societies, wroteHassabisin a blog post.

• Computer shows human-like learning for writing characters