Neuromorphic algorithm that learns and imagines plays ping-pong

Starting with no knowledge of ping-pong or trajectories, the algorithm learns and imagines its way to wins, beating an opponent built with standard algorithms in a demo shared on r/artificial.