When Google DeepMind’s AlphaGo AI started beating world champions at one of the most ancient board games in history, there were some worries that players and fans of the game would begrudge the victory. Curiously, those players and fans were actually more receptive than outsiders, seeing AlphaGo’s wins as an opportunity to change the way Go has been played and thought of for centuries. Now the Google subsidiary is formalizing and gathering those lessons into a new AlphaGo Teach tool that tries to help humans play like the AI pro.
Some will argue that games like chess and Go are really less about strategy and more about memorizing and replaying moves, many of which have been known to “win”. Some even have names. Of course, there’s still a good amount of strategic thinking involved in trying to out-think your opponent and think dozens of steps ahead.
Human brains, however, have nothing on the speed and near infinite memo, y of a computer. Especially one that’s been “trained” to learn and master the art of Go. It has forced many masters to rethink their strategies, which, in turn could give the next generation of players an edge, especially with this new DeepMind tool.
AlphaGo Teach is an interactive website that offers the lessons that the AI has learned over the course of its training and its competition with human masters of the game. It boasts of analysis of “6,000 of the most popular opening moves”, and just in recent history, mind. It also has the data the AI has gathered from analyzing 231,000 human versus human games and, of course, the 75 games it played with humans.
The tool lets Go players test out moves and see what AlphaGo thinks of those moves, including the probability of winning, depending on whether you’re playing black or white. All in all, Deepmind hopes that the tool will help human players open their minds to new ways of playing the game, ways that an AI has already found for them.