After AlphaGo also won the third round against Chinese top player Ke Jie, the DeepMind team of Google has announced that this was the last go match that the artificial intelligence will play.
AlphaGo and Ke Jie played their last match against each other on Saturday. AlphaGo already had a 2-0 lead and was therefore already certain of his victory, but in the context of pushing the AI algorithms to the limit, the last game was equally important. According to a report from Google, Ke Jie went for a ‘3:3 point’ strategy this time. That was an unusual decision that may have been motivated by the Chinese player’s already disadvantaged position. The strategy, however, is one that AlphaGo itself masters well, so that it offered no solace for Ke Jie.
Thanks to the 3-0 victory over the top player, the question of whether the artificial intelligence is powerful enough to beat human players in the complicated board game has now been definitively answered in the eyes of Google. This is the highest point that AlphaGo can reach at this point and therefore this was the last game. AlphaGo has played a total of 50 matches.
The DeepMind team behind AlphaGo will now work on general algorithms to try to rid the world of major issues, such as diseases, energy conservation and the invention of new materials. In a 2016 interview, DeepMind founder Demis Hassabis said StarCraft would be a good next challenge for the team and its AI. No further mention has been made of this, so it remains to be seen whether this will eventually continue.
The team also promises to publish another research report in which it outlines the latest improvements to the AlphaGo algorithm and also highlights the potential applications outside the board game. In addition, a teaching tool will also be put online, in which the analyzes that AlphaGo makes of the go situations are explained and made clear. Ke Jie, the top player who was last defeated by AlphaGo, has also promised to provide his analysis per move, so that interested parties can fully experience the entire match from the perspective of both parties.
AlphaGo has used strategies that have not been seen before in the go world. Some opponents also described the game of the AI as the ‘Go from a distant future’. These strategies have been developed, among other things, by having AlphaGo play against itself as training. Google DeepMind is going to make a collection of 50 match registrations in which the AI plays against itself, also freely available. It is expected that the go community will be able to pick up new ideas and strategies from the images. The first ten jars are already available.