Korean Go master, Lee Se-dol, has announced his retirement from the ancient strategy game. Lee cites unbeatable artificial intelligence as his reason for calling it quits from professional competition.
In the 2,500-old game, players take alternating turns to place black and white stones on a 19 x 19 grid. Throughout the match, players battle for territory by surrounding their opponent’s pieces to eliminate them from the board. There are virtually unlimited combinations of moves in Go, which is why it requires so much strategy.
Consequently, the game has long been considered a benchmarking opportunity for AI. Go is too complex for simple calculations, making it much harder for machines to master than other board game classics like chess and checkers. Now, it seems that AlphaGo needs a new challenge.
The Match that Changed Everything
Lee Se-dol has faced off with AI in the past. In 2016, he lost to Google’s AlphaGo bot in a highly publicized match that showed how far the technology has come. Lee won a single game out of five in the series. He remains the only human player to have beaten AlphaGo. However, Lee has also speculated that his victory was because of a “bug” in the code that prevented AlphaGo from being able to counter one specific, unusual move appropriately.
Lee explained his difficult decision to retire. “With the debut of AI in Go games, I’ve realized that I’m not at the top, even if I become the number one through frantic efforts. Even if I become the number one, there is an entity that cannot be defeated.”
His announcement ends an illustrious 24-year career. The 36-year old won 32 domestic and 18 international tournaments before resigning this month. Now, it seems that Lee has joined the other side. He is helping develop a new South Korean AI program, HanDol, which has already defeated the top five Go players in the country.
AI Still Evolving
Since its victory over Lee, AlphaGo has only gotten more sophisticated. Google’s creation recently defeated China’s Ke Jie in a best-of-three series that featured some of the best overall play ever, according to Google engineers. AlphaGo’s analysis concluded that the first 50 moves of the game were “virtually perfect.” Overall, the first 100 moves were the best that the software has ever faced.
Experts didn’t expect AI to beat human players at Go for many years. AlphaGo opened the floodgates in 2016 with its victory and has set new expectations for the public about artificial intelligence. Since then, AI has won matches against professionals in “Starcraft II,” written pop music, and solved a Rubik’s Cube in less than half a second.
Today, we are able to feed vast quantities of data into computers and create ample opportunities for deep learning algorithms to study human behavior. The challenge now is understanding why AI programs make the decisions that they do.
Overcoming the explainability, or “black box,” issue is the next big hurdle to overcome. When that happens, developers will be able to create much more effective algorithms and take AI to new levels.