Training a robot these days is effectively to train a neural network - we don't give robots 'programs' any more. So it's no surprise that there's a lot to be learned about learning from the way researchers teach robots. This article lists six of them (there's nothing quite so alluring as a listicle). Overall, what you'll learn is that I've taken some of my best ideas from how neural networks are trained: learning is iterative, takes practice, uses models and examples from experts, and can involve a lot of play.
Today: 5 Total: 1102 [Direct link] [Share]
