We have noticed you are using an ad blocker
To continue providing news and award winning journalism, we rely on advertising revenue.
To continue reading, please turn off your ad blocker or whitelist us.
Google's artificial intelligence (AI) system that went head to head with the world champion of the board game Go has won for the first time, marking a milestone for computing power. AlphaGo – the program developed at Google's UK subsidiary DeepMind – beat three-time world champion Fan Hui five times in a row in front of Nature magazine's editors.
Go is a popular East Asian board game believed to be notoriously difficult to crack, even more so than chess. The game involves placing black and white tiles on board, while attempting to remove as many of one's opponents tiles as possible. So far, no AI computer system had been able to beat a human at the game.
Google announced the unprecedented victory in a research paper in Nature on 27 January. Dr Tanguy Chouard, a senior editor of the science journal who was present during the game as a reviewer, described the atmosphere of the game as: "one of the most exciting moments of my career. But with the usual mixed feelings ... in the quiet room downstairs, one couldn't help but root for the poor human being beaten."
AlphaGo's victory comes a decade before the most optimistic experts' estimates. Remi Coulom, AI developer of Crazy Stone, previously believed to be the leading contender in cracking the Go game, had said it would take at least 10 more years for any intelligent machine to beat a human champion of the game without a handicap, the Guardian reported.
DeepMind developers said that AlphaGo won because it was programmed to approach the challenge in a completely different way, in comparison to other AI systems. The AI program apparently used "Deep Learning" mechanisms to create its own understanding of the game, rather than rely on pre-existing game analysis. "AlphaGo looks ahead by playing out the rest of the game in its imagination, many times over," said David Silver, co-lead researcher with DeepMind.
AlphaGo's next challenge in the Go game, according to Google's AI team, would be to defeat reigning world champion Lee Se-dol in an upcoming game. Although Lee is aware of how strong AlphaGo is likely to be, he said he feels confident about winning "at least this time". He also added, "regardless of the result, it will be a meaningful event in Baduk (Korean translation for Go game) history."