Harrison Kensley and Daniel Kukiela:
We are playing entirely within a neural network’s representation of Grand Theft Auto V. We’ve seen AI play within an environment, but here the AI is the environment.
Nifty.
To collect training data for their neural network, they first created a rules-based AI and let twelve instances of it drive around in the same GTA V game instance on the same stretch of road.
And there’s more:
This model learned to model GTA’s modeled physics. But could it have just as well learned, maybe, some basic real-life physics?
Stay tuned.