Alphabet’s DeepMind subsidiary announced that its AI-powered computer program that beat the world’s best players of the ancient board game Go a few years ago has extended its dominance to chess and shogi, the Japanese version of chess. This shows why Google is considered the top destination for AI researchers and underscores why Facebook was also interested in acquiring DeepMind, as we reported in 2014.
The announcement also highlights the speed of DeepMind’s progress in AI. DeepMind’s AlphaGo, which beat the world’s best Go player in May 2017, was trained using data about how humans played the game. But the new version, called AlphaZero, has no prior knowledge and learns by playing games millions of times against itself.
Alphabet has already seen a return on its DeepMind acquisition by using its AI algorithms to reduce power consumption in its data centers. But the technology and research underpinning AlphZero has more interesting potential, even if commercial applications remain unclear.