A new move in an ancient game.

DeepMind’s AlphaGo used deep neural networks alongside advanced search to play Go. The five-game match in Seoul concluded on March 15, 2016.

Why it mattered: Go’s enormous range of possible moves had made strong computer play particularly difficult. Learning helped the system decide which possibilities deserved attention.

The perspective: the result was a milestone in a specific domain, with methods that helped inspire further research.