Learning to run - an example of reinforcement learning deepsense.ai

Turns out a walk in the park is not so simple after all. In fact, it is a complex process done by controlling multiple muscles and coordinating who knows how many motions. If carbon-based lifeforms have been developing these aspects of walking for millions of years, can AI recreate it? Moving by controlling the muscles attached to bones, as humans do it, is way more complicated and harder to recreate than building a robot that can move with engines and hydraulic cylinders. Building a model that can run by controlling human muscles recreated in a simulated environment was the goal of a competition organized at the NIPS 2017 conference.