Solving Rubik’s Cube with a robot hand

In three linesOpenAI trained two neural networks to solve a Rubik's Cube with a humanoid robot hand. Reinforcement learning in simulation uses OpenAI Five code and a new Automatic Domain Randomization (ADR) technique. The system handles unseen situations during training, such as being prodded by a stuffed giraffe.

Summary generated by Claude — human-verified