Google's artificial intelligence (AI) computer system won its third straight game against the world's top Go grandmaster to take the series. DeepMind's AlphaGo used game-changing machine learning to teach itself the ancient game and win the million dollar prize.
In the past computers including IBM's Watson have defeated humans in various games. They include checkers, chess, Scrabble, and Jeopardy, but this was the first time an AI system defeated one of the world's top Go grandmasters in a game that is more complex than chess and requires players to guess their opponent's next moves.
The historic match took place between AlphaGo and South Korean Lee Sedol. Many of the machine learning techs in the system are already used by big Internet companies so the supercomputer's victory shows the potential of future AI.
A few years ago industry experts thought a machine would defeat a top Go player about a decade later. However, London-based DeepMind used a pair of powerful types of machine learning that allowed computers to perform self-learning to teach themselves how to play the old game, according to Wired.
AlphaGo won the first two matches. However, Lee got a big advantage when he played Game 3 with the black stones since he could make the contest's first move.
Lee had another advantage. Google could not tweak AlphaGo's code during the match, but Lee was able to adjust his strategy after the first two games.
Lee and other Go grandmasters reportedly stayed up almost all night on the eve of Game 3 evaluating the supercomputer's moves and looking for weaknesses. However, English commentator Michael Redmond suggested it would be better for Lee to keep using his regular game plan.
Many Google VIPs were present at the series. They include Google chairman Eric Schmidt at Game 1 and co-founder Sergey Brin at Game 3.
This was not the first time a supercomputer has defeated a top-notch human opponent. In 2011 IBM's Watson beat two former Jeopardy champions. The machine's three-day total of $77,147 was over three times higher than the second place total.
IBM's supercomputer stored 200 million pages of content on 4 terabytes (TB) of disk space. It also was powered with around 2,800 processor cores and 6 million logic rules to pick the best answers, according to PC World.