Nature Will releases Natura-1 model, demonstrating dexterous tasks such as robot hands wrapping wontons

Nature Will recently released a demonstration video of the Natura-1 model, in which robotic hands, without human assistance, completed tasks such as wrapping wontons, packing express deliveries, brewing tea, applying phone screen protectors, and writing Chinese calligraphy with a brush. Natura-1 is a 50-billion-parameter mixture-of-experts (MoE) model, with data processing covering nearly 1 million hours of first-person human interaction data. It can drive a 20-degree-of-freedom dexterous hand within a 100-millisecond control loop and follow complex natural language instructions to generalize in unseen scenarios.