Skip to main content

Google’s AlphaGo AI is about to face off against the world’s best Go player

Google’s AlphaGo AI is about to face off against the world’s best Go player

Share this story

Last week’s I/O conference saw Google reiterate its positioning as an AI-first company, a move it first highlighted a year previously. But perhaps the first event that made the public take note of the rise of AI, or artificial intelligence, came a few months earlier than that. An AI called AlphaGo scored a momentous victory over Lee Se-dol, one of the all-time best players of Go, the staggeringly complex ancient Chinese board game.

In defeating Lee 4-1, DeepMind, the British startup acquired by Google in 2014 and developer of AlphaGo, achieved something that many computer scientists believed would be decades away. But the matter wasn’t quite settled among the Go community. The world’s current number one, a 19-year-old Chinese player named Ke Jie, came out after AlphaGo’s first win declaring that he believed he could defeat the AI; Ke toned down his rhetoric as AlphaGo further demonstrated his dominance over the series, but the point remains that AlphaGo hadn’t quite beaten the best humanity had to offer.

This week, the matter will be settled once and for all. Ke Jie and AlphaGo will face off in a three-game match in Wuzhen, China, as part of the Future of Go Summit being held by Google. As well as the Go matches, the event will host a series of panels from many of Google’s artificial intelligence and machine learning luminaries, who we can expect to expand on many of the initiatives Google laid out last week at I/O.

AlphaGo’s capabilities will further be tested in two exhibition matches with intriguing stipulations. The first, Pair Go, will see two Chinese pros face off against each other alongside an AlphaGo teammate: since Go strategy requires immense foresight and strategy, the helping hand could be more of a hindrance in practice for both human and computer, removing their abilities to plan ahead. Much of the drama last year came from AlphaGo’s unorthodox, baffling moves that later turned out to be inspired and even beautiful, so it will be fascinating to see how the AI performs under these conditions.

The second exhibition match, Team Go, is more straightforward: five Chinese pro players will basically gang up on AlphaGo, working together in an attempt to take the AI down. Will the collaborative approach help the human players by providing an avenue for new ideas, or will it be a case of too many people cluttering the strategy? Either way, it’ll make for an opponent unlike any AlphaGo has faced to date. And both stipulation matches will demonstrate not only a wider range of AlphaGo’s capabilities than before, but how the arrival of AI has changed the way that humans play Go.

Most eyes will be on the Ke Jie showdown, however, and with good reason. If Ke wins, it won’t exactly diminish DeepMind’s achievement, but it may prompt a slowing of the AI hype train that has occasionally threatened to derail over the past year. If AlphaGo wins, on the other hand, it’ll confirm that the game of Go has been essentially beaten, and no doubt prompt further handwringing about the machines taking over.

“Instead of diminishing the game, as some feared, artificial intelligence has actually made human players stronger and more creative,” DeepMind CEO and co-founder Demis Hassabis said upon the announcement of the Future of Go Summit. “It’s humbling to see how pros and amateurs alike, who have pored over every detail of AlphaGo’s innovative game play, have actually learned new knowledge and strategies about perhaps the most studied and contemplated game in history. Clearly, there remains much more to learn from this partnership between Go’s best human players and its most creative AI competitor.“

What’s going to happen? Last time, no-one really knew what to predict, but AlphaGo’s strong showing against Lee Se-dol makes it the clear favorite to take the three-game match — it’s likely to be even stronger with another year or so of machine learning under its artificial belt. There’s also the small matter that the AI has already beaten Ke Jie multiple times before in online games conducted under a pseudonym, though that’s a very different setting than the formality of an official match.

Ke Jie (far right) explains the beauty of AlphaGo's moves to Google CEO Sundar Pichai
Ke Jie (far right) explains the beauty of AlphaGo's moves to Google CEO Sundar Pichai

“My thoughts are that both Ke Jie and the team will have a low chance of winning because when AlphaGo played 60 games online, no one managed to expose its weakness,” says Lee Ha-jin, secretary general of the International Go Federation until last year and a pro player herself. “If I had to choose, I would say the team has a better chance.”

Ultimately, of course, the promise of AI is about more than solving board games. This week you can expect to hear DeepMind speak about its health initiatives, and Google should go into further detail about how machine learning is already improving many of its products. But the summit will also serve as a fascinating focal point for the impact of AI; an abstract board game created thousands of years ago in China is now right at the center of what could be one of the most profound technological developments of our time. Go may not be the endgame, but it’s a beguiling, engrossing context for these conversations to be held in.

“I am still excited to see the games!” says Lee Ha-jin, who doesn’t particularly seem to mind if AlphaGo prevails. “Go is like a journey and what happens in the journey is more interesting than the result.”

The Verge is in Wuzhen covering the Future of Go Summit. Various events and games will be streamed at this link.