Computer program challenges human champions in game of Go

Science : One of the “great unsolved problems in AI” may have just been solved: the creation of a computer program that can beat the best human players at the ancient game of Go. London-based company DeepMind has developed a program called AlphaGo , which recently bested European Go champion Fan Hui. Go has long defied computer analysis because the board on which it is played is so large and the number of possible moves is astronomical. So the program’s creators introduced machine learning tools that mimic the connections of neurons in the human brain. Those “deep neural networks” allow the computer to move beyond the random exploration of all possible moves by learning to evaluate board positions and discern better moves from poorer ones. AlphaGo’s next milestone will be its playoff in March against Lee Sedol, the top Go player in the world.