Automated robots that perform functions normally done by human beings are nothing new. But these robots, named James and Rosie from Technical University of Munich, aren't programmed. They use visual-detection systems from an Xbox Kinect as well as perception algorithms to assess their surroundings, make inferences about what they see and then perform tasks based on that information. In other words, they break down complex actions, like making a sandwich, into steps and then move through those steps. This kind of robot doesn't need to be given instructions for every single step in the action. Pretty cool.
Via IEEE Spectrum