After a close final match, Google’s AlphaGo has defeated world champion Lee Se-dol in Google DeepMind’s Go tournament. This leaves the final tournament score at 4-1 in favor of AlphaGo. The overall outcome was decided during the third match, when AlphaGo won the three required matches consecutively. However, all five games were set to be played.
“It’s really difficult to see where [Se-dol] went wrong,” said Chris Garlock, managing editor of the American Go E-Journal, towards the end of the match.
AlphaGo and Se-dol both winning games muddies the waters for what the tournament really means. Of course, the artificial intelligence algorithm did win the first three and final games of the series, but Se-dol was able to adapt and prevent a total sweep. That fourth game, won be Se-dol, showed that while AlphaGo could simulate an extremely competent player, it still plays at a human level. Even the terse ending in this final match removes any idea of AlphaGo’s invulnerability.
However, it is sufficient to say that Google DeepMind has mastered the game of Go. This victory is nothing if not resounding, for the artificial intelligence startup acquired by Google in early 2014.
This victory cements the idea that artificial intelligence researchers can build highly advanced and adaptable software, able to defeat the most skilled humans that have dedicated their lives to a certain mental pursuit.
This event echoes the defeat of Garry Kasparov by IBM’s Deep Blue in 1997. However, Google has provided far more documentation about how they achieved their goal, publishing a paper in Nature earlier this year. Some on GitHub have even attempted to replicate the algorithm.
It’s hard to say what’s next for DeepMind. In recent interviews, CEO Demis Hassabis has hinted at games like StarCraft being the new frontier. The Google company likes to simulate artificial intelligence within video games, as it’s easy to replicate and control the environment.