Debate: Will robots learn the world end to end from massive data, or do they need explicit models of it?

Debate: Will robots learn the world end to end from massive data, or do they need explicit models of it?Davison vs end-to-end scaling. Explicit models. Massive data won't unlock robotics on its own: robots need structured maps of the scene, plus simulation and planning (Andrew Davison). Or: End to end. One model from pixels to actions, scaled with data, as in end-to-end driving andThe Physical AI Voices by Herbert Scale Experts.