Astra shows any 3D/4D prior can be distilled into VLMs, a new embodied AI paradigm

mariyaivasileva · x · 2026-09-21

Jitendra Malik highlights that 3D/4D reconstruction — from multi-view geometry to VGGT and SAM 3D — is directly relevant to robotics, arguing that learning-era robotics papers ignoring 3D structure waste valuable signal. Commentators note Astra's key takeaway: virtually any 3D/4D prior can be distilled into VLMs, a potential new paradigm for embodied AI built on a real2sim2real tradition.

Original post →

More from Embodied

Embodied channel →