Google's DeepMind artificial intelligence unit scored another major win, as its AlphaGo program just defeated world champion Go Player Lee Sedol (9-dan pro).
In 2015, DeepMind's AI successfully pitted the AlphaGo against European champion Fan Hui (2-dan pro) and won all five Go matches. We already knew back then that Lee Sedol would be AlphaGo's next formidable opponent and suspense was building up ahead of the big competition.
The AI was said to have a less than 5 percent chance of winning, but AlphaGo proved its worth once again in an epic AI vs. Man match.
AlphaGo has now defeated the legendary Lee Sedol in the first of five scheduled matches, and the stakes are higher than ever.
Go is an ancient Chinese strategy game that's considered the world's toughest game to master, so it's an immense challenge for AI. This complex game has far more possible combinations and moves than chess and an average Go game includes roughly 150 turns, whereas an average game of chess usually has about 80 turns.
The latest Go match between Google's AlphaGo and Go champion Lee Sedol marks a tremendous step forward for artificial intelligence in a hugely complex game, echoing and amplifying the resounding success that IBM's Deep Blue achieved back in 1997, when it defeated chess master Garry Kasparov. Deep Blue scored another win back in 2011 when Watson won Jeopardy, but AlphaGo heralds even more impressive advancements.
Facing Google's DeepMind AlphaGo AI player, Lee Sedol resigned after roughly three and a half hours of fierce gameplay. Sedol is the first 9-dan pro Go player to go up against an AI opponent, and the starting match tips the scales in favor of AI. Nevertheless, Sedol still has four more matches to turn things around and try to beat AlphaGo.
While AlphaGo had a rather defensive strategy in its confrontations with European champion Fan Hui in October 2015, it has now adopted a more aggressive approach. In fact, both AlphaGo and Lee Sedol played aggressively, according to 9-dan pro Go player Michael Redmond, who was one of the commenters.
The most challenging part for the Go-playing AI is the fact that Go is not about brute calculations and clear computations. A computer would excel at that, but Go requires intuition and reaction, which is more human-like. To get better in this area, AI programs such as DeepMind's AlphaGo or Facebook's AI Go player rely on deep neural networks to guide their strategy and "imagination."
It remains to be seen how AlphaGo will fare in the next four matches against Lee Sedol, but it's definitely off to a strong start. Check out the video below to see the historic match from Tuesday.
