A reinforcement-learning algorithm was open-sourced this week by Facebook that can train AI bots to navigate simulations, with each droid armed with just a camera, GPS, and a compass and no map. Dubbed decentralized distributed proximal policy optimization, …
Read More