Go world champion Lee still wins AlphaGo in fourth round

South Korean world champion Lee Sedol still managed to beat Google’s DeepMind program AlphaGo in the fourth round in the Chinese board game go. Lee can no longer win the competition, but he has shown that AlphaGo is not yet unbeatable.

The fourth match in the go competition of human versus artificial intelligence ended in a surrender by AlphaGo after almost five hours of play and 180 moves in total. At move 78, Lee managed to mount an offensive that AlphaGo seemed unprepared for and the subsequent move by the artificial intelligence, number 79, was a mistake, according to the commentators. The moment AlphaGo threw in the towel, which it does when it estimates the chance of a victory lower than 20 percent, the Korean audience in the Four Seasons Hotel in Seoul burst into loud applause.

In the afterthought, it was said that Lee no longer felt the tension that he suffered, especially in the third round. After all, it was no longer about the entire competition because AlphaGo has already won the majority of the best-of-five series. Nevertheless, the South Korean has shown that artificial intelligence in the field of playing go against human opponents is not yet perfect. AlphaGo, which runs on an array of 170 gpus and 1200 cpus, is one step closer to that goal thanks to this defeat. AlphaGo’s deep neural networks can analyze the course of this competition and learn from it.

The last of the five go-pots will be played next Tuesday. The $1 million prize money won by Google will go to various charities. The news item about the third round of the go competition is also accompanied by extensive background information about Google’s artificial intelligence AlphaGo.