Perhaps those same experts should have given Google’s AI company, DeepMind, more credit when it bolstered AlphaGo enough for it to score a 5-0 victory over European Go champ Fan Hui last October. The champion may now have found a chink in its AI’s armour.
Advertisement
Google’s artificial intelligence program secured its win for the fourth time out of five matches played opposite human “Go” champion Sedol.
“The victor here, no matter what happens, is humanity”, Alphabet chairman Eric Schmidt said prior to the first match, and that certainly seems to ring true now the challenge – and $1 million purse riding on it – has been decided. That work was featured in a front cover article in the science journal Nature in January 2016.
Lee played as black for the first time in the tournament – possibly to try and confuse the AlphaGo AI – but he still lost.
“I can’t accept that it has better playing skills, but it has a concentration level which humans can’t beat easily”, Lee told the Korea’s Yonhap News Agency on Tuesday following his loss in game five. AlphaGo attacked the right center by seeking to build a large territory there, while Lee built a territory in the lower-right side of the board according to his strategy which reportedly occupies more areas in an early phase. Players place stones onto the board aiming to capture opponent’s stones by completely surrounding them and eventually win by controlling more territory than the opponent.
The SWA, which has over 1,000 members, even held post-game sessions to analyse the four-hour matches.
But Lee surprised the computer midway through game 4, on move 78, with a placement choice that impressed the human commentators and threw off the A.I.
I don’t know how to start or what to say today, but I think I would have to express my apologies first.
The final match was too close to call until the very end. In an interview with UK’s The Guardian, Hassabis explained: “It’s the combination of deep learning, neural network stuff, with reinforcement learning: so learning by trial and error, and incrementally improving and learning from your mistakes and your error”.
AlphaGo, a computer program developed by Google has beaten a South Korean master of the ancient board game Go.
Advertisement
It’s a major milestone for artificial-intelligence research: Go is a simple game but has been notoriously hard for computers to master because of the sheer number of potential moves. It involves two players who take turns putting white and black stones on a grid of 19 lines by 19 lines.
