photo by chadmiller on flickr
For the first time, a computer has beaten a world champion in the game of Go. Google’s DeepMind program defeated grandmaster Lee Sedol after a three and a half hour battle. This accomplishment was once considered impossible, given the complexity of the game and the sheer number of permutations possible in the game.
What I find most interesting is the change in approach from IBM’s Deep Blue computer that beat chess champion Gary Kasparov in 1997. Deep Blue could analyze billions of potential future scenarios and pick the most logical and advantageous move at each step of the game. It essentially used brute force and statistics to pick the best path and stay ahead of Kasparov.
DeepMind takes a whole new approach (via: http://www.engadget.com/2016/03/09/google-deepmind-alphago-ai-go-champion/)
To prepare the system’s neural network, the AlphaGo team fed the computer 30 million moves from professional Go players so it could adopt its own strategies using a process of trial and error that is referred to in AI circles as reinforcement learning.
Google is using artificial intelligence and computational power in a very different way. Instead of muscling its way through the problem by analyzing as many future moves as possible (which would have likely been a futile exercise), it has learned how to play the game, to some degree like a human does. By watching millions of previously played moves, and using internal trial and error, it has learned what constitutes the best move in any given scenario. It can make an educated guess about what move to make even when the board layout is different than anything it has ever seen before.
This is extremely fascinating. We are getting tangibly close to computers that learn and think exactly the way our brains learn and think. This technique of watching past behaviors, churning internally through iterations of trial and error, and then using what it learns to face new, unseen challenges, can be used on countless difficult problems. You no longer have to completely solve the problem before you can start programming and computing solutions. As the old proverb teaches us, we have not given the computer a fish, we have taught it how to fish.
