Artificial intelligence · 2013
Deep Q-network (Atari)
One network learns dozens of Atari games from raw pixels and the score.
DeepMind's DQN combined Q-learning with a convolutional network, experience replay and a target network. The 2015 Nature version reached human-level play on 29 of 49 games with the same hyperparameters. Google bought DeepMind in 2014 for a reported 400 million pounds.
What it built on 2
What it led to 2
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)