Embodied AI needs messier data. Physics validated edge cases for robots before they meet us.
Infinite digital commodity creation for RL training data. Physics validated scenarios, dynamic novelty threshold, solving the $10B humanoid robotics problem of self evaluation by 2030.