Humanity is the biggest winner of the upcoming Go match between South Korean champion Lee Se-dol and DeepMind's artificial intelligence (AI) AlphaGo, Alphabet executive chairman Eric Schmidt has said during a surprise appearance at a Seoul press conference.
The 33-year-old Lee will on Wednesday play his first match of five against AlphaGo -- which defeated the European Go champion 5-0 last October. The match will broadcast live worldwide via YouTube, with match results recorded on Google's Blog. Four more matches are due to be played on March 10, 12, 13, and 15, with the winner taking $1 million. If AlphaGo wins, the money will be donated to Unicef and STEM for research and education.
The game has been dubbed "the match of the century" by South Korean press and will be the first AI versus human match to be live-streamed online.
"Whatever the outcome may be, the winner will be humanity," said Schmidt. "As artificial intelligence and machine learning develop further, each and every person will become smarter and more talented."
Schmidt said that living as a computer engineer his whole life, he initially believed something like the upcoming Go match would happen in the 1960s, but the past 30 years have been an ice age for AI. However, due to new algorithms and faster computing power coupled with more investments and human resources spent on AI in the last 10 years, the area has seen great developments, citing services such as Google Translate and Google Photo as examples.
He also singled out DeepMind's Reinforcement Learning Framework technologies for praise and for making the match possible.
DeepMind CEO Demis Hassabis, who arrived in Seoul on Monday, briefed Lee and the media on the preparation his team has made ahead of the matches. "Our ultimate goal is to use the technology beyond games," he said.
Lee remained confident of his win a day ahead of the match but looked markedly more nervous than his previous appearance in front of reporters.
"I don't think I can win 5:0. After hearing the explanation about the algorithm [of AlphaGo by DeepMind CEO] I feel I should be on the edge [during the match]," he said.
"I have been in countless matches, but this unfamiliar feeling that I am getting is a first," he said. "I like it because it's new, but since my opponent is not human, I am preparing in a different way than my previous matches."
Lee said since he cannot read the "feelings or aura" of AlphaGo, he has played by himself for one to two hours a day in preparation.
Go, a Chinese board game, has 10761 possible games, compared with 10120 for chess, and had been considered impossible for computers to learn due to having too many variables and being based on instinct.
AlphaGo uses a Monte-Carlo tree search to predict possible moves, with the neural networks suggesting moves and judging board positions.
DeepMind, a London-based AI start-up, was bought by Google back in 2014.