We've created a robotics system, trained entirely in simulation and deployed on a physical robot, which can learn a new task after seeing it done once.
This post describes four projects that share a common theme of enhancing or using generative models, a branch of unsupervised learning techniques in machine learning.
