Artificial intelligence · 2016

AlphaGo

DeepMind's AlphaGo beats Lee Sedol 4-1 at Go, a decade ahead of forecasts.

AlphaGo combined policy and value networks trained on human games and self-play with Monte Carlo tree search. It beat Fan Hui in October 2015 and Lee Sedol in March 2016. Move 37 of game two showed a machine playing creatively in humanity's deepest board game.

See it in the living tree Wikipedia

Arcs above join it to what it stood on, arcs below to what it led to. The bar places this stretch on the tree's whole timeline.

What it built on 4

Tensor Processing Unit 2015
Deep Q-network (Atari) 2013
Monte Carlo tree search 2006
Convolutional neural network 1989

What it led to 1

AlphaGo Zero 2017

Watch

AlphaGo - The Movie | Full award-winning documentaryGoogle DeepMind · YouTube Match 2 - Google DeepMind Challenge Match: Lee Sedol vs AlphaGoGoogle DeepMind · YouTube

Related measurements

The dashed line marks when it first worked.

Training compute of notable AI models Source: Epoch AI, Notable AI Models (CC BY 4.0)

ML hardware FLOP/s per dollar Source: Epoch AI, ML hardware (CC BY 4.0)