Last year, DeepMind’s general purpose A.I. program, AlphaGo, annihilated world Go champion Lee Sedol 4-1 in a $1 million contest. It was considered by many to be a crowning achievement in A.I. research, and was lauded on most of 2016’s top tech charts, including our own. But according to a tweet by the company’s co-founder and de facto leader Demis Hassabis, the company isn’t done beating the proverbial pants off of players from around the world. In fact, AlphaGo’s been playing online under a pseudonym for a while now, and it’s essentially unbeatable.
According to Nature, two players who went by the names of Master(P) and Magister(P) were logging onto a pair of Go servers and destroying the competition at every level. Master(P) played 51 games (which included games against world champions like Gu Li), and won 50 of them. That last one was a draw because of what was probably a network glitch.In a recent tweet, which comes via New Scientist, Hassabis/DeepMind revealed that AlphaGo was indeed the player behind Master(P) and Magister(P). He also said that DeepMind is looking forward to playing more “official full-length” games in 2017 as well as exploring “the profound mysteries of the game further in this spirit of mutual enlightenment.”
— Demis Hassabis (@demishassabis) January 4, 2017
DeepMind’s enlightenment isn’t going to be limited to Go, however. The company is concurrently pursuing research in the video game world of StarCraft II, with the aim of using a general purpose A.I.—a program that learns how to play the game—to do to Starcraft II’s top players what it did to Go’s. Considering the fact that DeepMind is also owned by Google, their general A.I. will probably start helping in other areas of life as well. In fact, it’s already helped Google to save 40% on the energy usage of its data centers.
What do you think about AlphaGo’s continued domination? What game do you want to see DeepMind tackle next? Let us know in the comments below!
Images: Google DeepMind/YouTube