Google’s French Go-champion-beating AlphaGo artificial intelligence will take on the Go world No 1 in a live broadcast from Seoul, South Korea. The contest will begin on 9 March and offers a $1m prize.
Founder of Google’s DeepMind AI development house Demis Hassabis announced the news on Twitter, which will see the company’s AlphaGo system take on 32-year-old South Korean Lee Se-dol, the current highest ranked Go player in the world, in a five-match game with $1m (£687,000) at stake.
Thrilled to officially announce the 5-game challenge match between #AlphaGo and Lee Sedol in Seoul from March 9th-15th for a $1M prize!
— Demis Hassabis (@demishassabis) February 4, 2016
The matches will be broadcast live on YouTube on 9, 10, 12, 13 and 15 March.
AlphaGo successfully beat the three-time European Go champion Fan Hui 5–0 in a series of games at DeepMind’s King’s Cross headquarters last October. But it is possible that ninth-dan Lee, who turned pro in 1996 at the age of 13 and has a 71.8% winning percentage, could beat the AI.
Lee said: “I have heard that Google DeepMind’s AI is surprisingly strong and getting stronger, but I am confident that I can win at least this time.”
Unlike previous systems that have attempted to compute the next move from all available moves and games in chess, for instance, AlphaGo beat all expectations by using “Deep learning” to build its own understanding of Go. It picks moves based on what it thinks is most likely to win. It used a similar technique to learn how to play retro video games.
Should AlphaGo beat Lee, it could arguably be ranked as the best Go player in the world. If Lee wins he will net the $1m prize in the year that he intended to retire and move to the US to promote Go.
