WUZHEN, CHINA --- A computer wasn't supposed to be able to beat a grandmaster at the ancient game of Go for at least another decade. But AlphaGo, an artificially intelligent system designed by Google-owned DeepMind, did just that. In its public debut last year at a tournament in Seoul, AlphaGo thrashed Lee Sedol, the best player of the past decade. Now AlphaGo is back, facing off in China against the world's top player to show just how much further machine-approximated intuition has advanced in the past year, and WIRED is there.
Tomorrow morning, AlphaGo is set to play 19-year-old Ke Jie in Wuzhen, a town crisscrossed by canals 80 miles west of Shanghai. China is the birthplace of Go, a 2,500-year-old game that remains exceedingly popular across Asia. Like last year’s match in Seoul, this three-game contest will generate enormous attention. But this year, it won’t likely be much of a contest. Though he is now ranked higher than Lee Sedol, Ke Jie doesn’t stand much of a chance against AlphaGo. This week is more an opportunity to see how far AlphaGo has come---and how far it can go.
AlphaGo learns to play Go by analyzing professional moves and, in essence, playing against itself, an artificial intelligence technique known as reinforcement learning. AlphaGo gets better as time goes on and now has several more months of training under its belt---not to mention any improvements made by the DeepMind team since the match in Seoul. In January, via the internet, DeepMind secretly matched AlphaGo against several of the world’s top players, including Ke Jie. The machine won its first 50 games and drew the 51st (but only because of a lost internet connection).
In Wuzhen, the suspense will come not so much from wondering who (or what) will win but in seeing how much AlphaGo has evolved over the past year. The machine may reveal some of its new self while playing the 19-year-old Chinese grandmaster. But the greater insight will likely come from Demis Hassabis, the CEO of DeepMind, who will be in China along with many other members of his team. Hassabis and David Silver, the lead researcher on the AlphaGo project, are set to speak between games one and two.
Between games two and three, Google will host additional matches in which the machine will play alongside various grandmasters. In Seoul, AlphaGo’s inhuman style of play paradoxically suggested ways the system could help push human players to new levels—a proxy for the way AI more generally could help humans excel. The same AI techniques underpinning AlphaGo are also reinventing everything from internet services to health care to robotics. This week, the world will get a glimpse of how much more change AI has the potential to unleash.
