Well, it doesn't make sense to use this exact model - this is just demonstration that it can learn world model from pixels. An obvious next step towards a more

Well, it doesn't make sense to use this exact model - this is just demonstration that it can learn world model from pixels. An obvious next step towards a more playable game is to add state vector to the inputs of the model: it is easier to learn to render the world from pixels + state vectors than from pixels alone. Then it depends what we want to do. If we want normal Counter Strike gameplay but with new graphics, we can keep existing CS game server and train only the rendering part. If you want to make Dream-Counter-Strike where rules are more bendable then you might want to train state update model...