Data Collection via Agent Play: Since we cannot collect human gameplay at scale, as a first stage we train an automatic RL-agent to play the game, …
Read more on gamengen.github.io
Stuart Bruce flipped this story into PR Futurist on Flipboard•44d
