Computers have been capable of beating the best human chess players for almost two decades, but a week-long stretch earlier this month marked the first instance of an artificial intelligence system toppling a champion Go player. Google’s AlphaGo defeated Go grandmaster Lee Sedol 4-1 in a historic five-game match that drew comparisons to Deep Blue’s famous triumph over chess world champion Garry Kasparov in 1997.
Go is an ancient game in which players attempt to establish “territories” by placing black or white pieces on a 19-by-19 grid. The simplistic appearance of a Go board, however, is merely superficial. While chess players may have 10-50 different legal moves on most turns, Go players can grapple with over 300. Also, because Go typically lasts much longer than chess, these numbers really start to add up; the number of possible positions on a Go board greatly outstrips the total number of atoms in the visible universe.
This staggering level of complexity posed a significant challenge for programmers. Computers excel at “brute force” calculations, but even the most powerful supercomputers would take hours to enumerate and analyze all of the possible contingencies when looking only a few moves ahead. Humans, on the other hand, tend to rely on fuzzier, less algorithmic modes of thought when dealing with such large numbers. This strategy makes us better than computers at many different tasks, such as being a lawyer or navigating a crowded grocery store. Unfortunately, it appears that AlphaGo has beaten us at our own game.
Acknowledgments: Many thanks to David Jacobson, a data scientist at HG Data, for providing his expertise and commentary on the topic.
Managing Correspondent: Christopher Gerry
Media Coverage: Google’s AI Wins Fifth and Final Game Against Go Genius Lee Sedol – WIRED; AlphaGo beats human Go champ in milestone for artificial intelligence – The Los Angeles Times