Berkeley's Angjoo Kanazawa teaches machines to understand the physical world for the robotics era

jfiance · x · 2026-09-17

Angjoo Kanazawa, assistant professor at Berkeley EECS and head of the Kanazawa AI Research Lab within BAIR, works at the intersection of computer vision, graphics, and machine learning: teaching machines to understand people, animals, and 3D environments from ordinary images and video.

Questions humans solve automatically — where a person is positioned, how their body moves, handling occlusion, tracking dynamic scenes — remain hard technical problems. As AI moves into robotics, autonomous systems, and mixed reality, these perception skills become critical: robots get no structured prompt and must infer depth, motion, and uncertainty while continuously updating their understanding. The author notes the current robotics/physical-AI startup wave rests on years of foundational research at Berkeley, with Kanazawa among those pushing it forward.

Original post →

More from Companies & People

Companies & People channel →