How it works The artificial intelligence relied on a technique called "reinforcement learning," which rewards the system for taking actions that improves its score.

Google DeepMind AI finds its way through a 3D maze by 'sight' We're one step closer to AI machines that can navigate the real world as humans do.