Google engineers unveil their latest feat: their program, AlphaGo, which defeated South Korean Go champion Lee Sedol. Note that this is the first of five Go games to occur before a winner is crowned. The talking point is how remarkable it is to configure a program to play this game, let alone defeat a professional. This Chinese board game, as old as age itself, is known for its layers of complexity.
READ MORE // Segway Unveils the Personal Robot of the Future
Commentators noted how close the game was.
Lee could have won, had he changed one or two early decisions in the game. AlphaGo was by no means miles ahead of Lee at any point. Lee has won 18 world championships, and has played professionally since the age of 12. He remarked his surprise of losing, and his early mistake which cost him the game.
Lee described Go’s early strategy as “excellent”: Go executed a move Lee found “unconventional”, a move no human would think of – or dare do – in that situation. Even so, Lee does not seem perturbed by the loss. He’s already mentioned his ache for the four remaining games.
The Google team behind AlphaGo refer to themselves as DeepMind. What makes AlphaGo special, in their words, is its “reinforcement learning” aspect.
This term implies how the machine plays against itself to “adjust its neutral networks based on trial and error”. In addition to this, Go narrows down its next move from an infinite amount of choices, to less than a dozen. Then cross references what decision will damage its long-term and short-term game plan. Ultimately this results in making a move.
The engineers behind Go stated that the mechanics behind this program can help a lot of other problems, both third-world and health-care related. However, they’ve not been clear on which problems they mean, or how.