AlphaGo, the Go-playing AI by DeepMind, which itself is a division of Google parent company Alphabet, has defeated the top-ranked Go player in the world, 19-year-old Ke Jie. There are still two matches to go, but humanity’s already on the back foot.
Go is an ancient game that boils down to trying to capture enemy stones in a circle of allied stones that’s played on a gridded board. The game was considered to be a decade away or more from being mastered by artificial intelligence, but AlphaGo turned out to be an incredibly fast learner. It defeated Lee Sedol, one of the top-ranked Go players in the world, in a surprise upset last year.
Ke Jie’s loss comes as part of a planned summit in China focusing on both Go and AI. There are a variety of Go formats that will be played at the event, but the AlphaGo vs. Ke Jie matches are the pinnacle. The two will face off again on Thursday, and once more on Friday. Due to the time difference, that means staying up late the night before for anyone that wants to watch live in the United States. While the expectation was that Jie would ultimately lose all three games, especially if the AI had only improved since facing Sedol, the first match was technically rather close: Jie lost by just half a point.
A win's a win, however. But it seems like why Jie ended up as close as he did to victory is partly due to the fact that he studied AlphaGo’s previous moves very carefully — and even used some of them against it. “At the very beginning, I made fierce, targeted moves, two 3-3 moves,” Jie said at a post-match press conference. “I copied some moves that AlphaGo liked to use in past games.”
Ultimately, the idea is that the same kind of machine learning at the heart of AlphaGo and the improvements made as part of its quest to conquer the game of Go could then be applied to other real-life problems with higher stakes like climate science or disease analysis that are just as if not more complex. For now, AlphaGo will continue to dominate the one game it is very, very good at, and there’s no telling what it might conquer next.
