Google's AI Continues to Wallop Expert Human Go Player
Google DeepMind's Go-playing artificial intelligence program AlphaGo has once again defeated champion Go player Lee Se-dol. After losing two games in a row to the computer, it seems unlikely that Lee can come back in the five-game match and win the prize of $1 million.
Go is an incredibly difficult game for a computer program to master because of the vast number of possible moves, and the success of AlphaGo is a major landmark in the development of AI. The system uses two neural networks simultaneously to calculate the probability of various moves leading to a victory and consider board position to narrow down the move options to consider at all.
Lee played a slower game today, considering each of his moves for longer than he did in the game yesterday. But it may have come back to hurt him, as Lee ran into trouble with the clock toward the end of the game, and commentators suggested that he didn't have time to read all the moves on the part of the board where he was laying his pieces. Both players used up the full two-hours of allowed game time, and then entered a kind of overtime known as "byo-yomi" periods.
Though DeepMind CEO Demis Hassabis said that the game was "unbelievably exciting and incredibly tense," Lee himself indicated that he never felt like he was in control.
"Yesterday I was surprised, but today I am more than that; I am quite speechless," Lee told reporters after the loss. "Today I really feel that AlphaGo played a near-perfect game. There was not a moment where I thought AlphaGo's moves were unreasonable."
The next match is scheduled for March 12 at 11 p.m. EST. You can watch it live on the DeepMind YouTube channel.
