Robots gain new skills, AI models master physical tasks

Robotics advancements are accelerating with AI, enabling robots like Digit to learn complex whole-body control overnight through reinforcement learning. New general-purpose AI models are achieving near-perfect success rates and faster task completion, signaling commercial viability for physical AI applications. Datasets for humanoid robot teleoperation are being open-sourced to foster development. Interfaces for intuitive robot path planning and human-robot interaction are emerging, alongside innovations in teleoperation systems for precise remote control and data collection. These developments collectively push the boundaries of autonomous physical capabilities across various industries, from manufacturing to space services.