On your marks, get set, play Go. But not against a computer. They’re too clever for us now.
Google’s artificial intelligence program AlphaGo on Thursday beat a human opponent for the second time in a row at the ancient, highly complex board game of Go. And not just any human. Lee Sedol, AlphaGo’s human opponent, is the number two Go player in the world.
Thursday’s game was the second of five that AlphaGo is scheduled to play against Lee as part of the Google DeepMind Challenge. Tucked away in the Four Seasons Hotel in Seoul, South Korea, the two will continue to go head to head into next week as millions of onlookers around the world keep vigil over a live stream.
Like all good science experiments, Google’s victory basically didn’t count if it couldn’t be replicated. The fact that AlphaGo has won two games in a row proves that its first victory wasn’t just a fluke. “Yesterday I was surprised but today it’s more than that, I am quite speechless,” Lee told the BBC.
The next match is scheduled for Saturday, and if AlphaGo wins, Google will be declared the overall champion.
The drama surrounding AlphaGo’s victory is all down to the fact that Go is thought to present one of the stiffest possible challenges to computers. To put it in context, it’s a game for people who think chess is too easy. The victory has also come as a surprise to everyone, as it wasn’t thought that artificial intelligence, the science of computers that more closely mimic human smarts, was ready to take on humans at Go just yet. It’s a sign that AlphaGo is smarter than we thought.
“Hard for us to believe,” tweeted DeepMind Founder and CEO Demis Hassabis. “AlphaGo played some beautiful creative moves in this game. Mega-tense.” Lee agreed, saying he felt as though “AlphaGo played a nearly perfect game.”
SpaceX and Tesla founder Elon Musk tweeted his congratulations to the DeepMind team following the first win, pointing out that many AI experts thought computers were still a decade away from being able to beat humans at the game.
It might seem worrying that AlphaGo is much more advanced than even its creators gave it credit for. What else might it be capable of? But AlphaGo isn’t an all-powerful, autonomous thinking machine that can outmaneuver the human mind at every turn. It has been programmed very specifically to master the game of Go.
Even in light of this major breakthrough, it’s still early days for AI. Research is pressing ahead, even though experts can’t agree whether we should be scared of its future capabilities or not.
Musk might have congratulated Google, but he’s one of a number of high-profile voices within the fields of science and tech, including the likes of Bill Gates and Stephen Hawking, who have cautioned against the potential danger that AI might pose.
Facebook Chief Executive Mark Zuckerberg, on the other hand, said last month that he believes we have nothing to fear. He pointed out that we’re “nowhere near understanding how intelligence actually works,” never mind replicating and beating it. Facebook has a number of AI enterprises it’s working on itself, ranging from facial recognition technology to Zuckerberg’s own pet project of creating a personal AI butler.