In March 2016, DeepMind’s AlphaGo program played a five-game match against Lee Sedol, one of the world’s strongest Go players, in Seoul, South Korea. According to DeepMind’s own account, AlphaGo won 4-1 in a match watched by over 200 million people worldwide.
Go had long been considered a grand challenge for artificial intelligence because the number of possible positions is astronomically large, far too many to search by brute force the way chess engines do. AlphaGo combined deep neural networks, which learned to evaluate board positions and suggest moves, with a tree-search method, and it improved further by playing games against itself. The detailed research was published in the journal Nature as “Mastering the game of Go with deep neural networks and tree search.”
The win was a landmark comparable to Deep Blue beating Kasparov at chess in 1997, but harder, and it signaled that learning-based AI could conquer problems once thought to require human intuition. It cemented DeepMind’s reputation and energized the field’s ambitions.