I suspect that our ungrounded (eg dreaming) mental representations have less accuracy and detail on the level of textures and geometry, and more coherence in the structure of scene and general semantics.

→ View original post on X — @plinz