Visual Model-Based Reinforcement Learning as a Path Towards Generalist Robots

Those videos of the computer's prediction of events that _haven't happened yet_ are fascinating.