Natura-1 turns natural-language instructions into dexterous robot actions

Natura-1 turns natural-language instructions into dexterous robot actions without task-specific examples, aiming at the costly bottleneck of hardware-specific teleoperation data. The model handles object placement and orientation, tracks what it has moved or attempted, and chooses the next step during longer tasks while adapting its grip and splitting work between hands. It is a 50-billion-parameter mixture-of-experts model trained on nearly one million hours of egocentric human video.