Robots Teach Themselves Tricks

Reinforcement learning, used to create more realistic video game characters, could be the latest tool to train the next generation of ninja robots. We think that given current limitations, this might take longer than the hype would suggest, but recent work in metalearning is particularly promising since it provides a data-driven way to generate reasonable priors. For a fascinating deep dive on this topic (with a healthy dose of really cool videos), check out Alex Irpan's post here


Want to receive more content like this in your inbox?