AlphaGo AI wins contest against Go world champion Lee Sedol

Wired : Google DeepMind’s AlphaGo artificial intelligence program has defeated Korean Go grandmaster and top-ranked human player Lee Sedol in a best-of-five competition. AlphaGo won the first three rounds before Sedol finally claimed victory in the fourth game Sunday. The victories by AlphaGo are the first by an AI over an elite human player, and they came in dominating fashion. However, Lee Sedol’s win showed that the AI had not completely outstripped human competitors. The game Sedol won was similar to a previous game in which the grandmaster had made a key mistake; this time Sedol managed to avoid major mistakes and capitalize on AlphaGo’s poor moves. AlphaGo’s AI uses a deep neural network . The program initially learned how to play Go—an ancient board game in which players take turns placing black and white tokens on a grid in an attempt to control the most space on the board—by analyzing a massive library of moves played by humans. After the initial training, AlphaGo switched to reinforcement learning in which it played games against itself to develop its own unique strategies that provided the highest likelihood of success. The final round of AlphaGo vs. Sedol is planned for Tuesday.