Apple, in collaboration with top universities, has developed a new method for training humanoid robots using first-person footage from devices like the Apple Vision Pro. Their system, powered by a model called the Human Action Transformer (HAT), learns from over 25,000 human and 1,500 robot demonstrations. By using headsets to capture motion and video, the team created a scalable, low-cost way to train robots. The approach improves generalization and allows robots to perform complex tasks—even ones they haven’t seen before—more efficiently than traditional training methods.
Source: 9to5Mac
