My impression on DeepMind's Deep-Q RL for learning Atari games paper is that it uses the same model to learn to play multiple different games at the same time. I wonder how did the RL agent learn in such a setting. In particular, when the game environment changed, how does the agent know the corresponding objective to achieve in that specific game? Is there a prompt to tell the agent what game it is currently in before training/evaluation? Or the agent can infer what is the goal simply by observing the feature from the environment (e.g. by observing blocks of bricks it can infer it needs to break all of them, can by observing there is a controllable vehicle its knows it is probably a racing game)?