Learning to See the Physical World: An Interview with Jiajun Wu

Summary

In this interview, Jiajun Wu discusses his long‑standing focus on physical scene understanding—building AI that can see, reason about, and interact with the real world. He explains his hybrid methodology that combines bottom‑up deep recognition, top‑down graphical models, and differentiable simulators, and describes two current research paths: using physical structure as inductive bias and grounding large vision foundation models in physical reality. Wu highlights applications ranging from robotics to game design and reflects on the field’s evolution amid AI hype, emphasizing the need for data‑efficient, fundamental research. Looking ahead, he is excited about co‑evolving loops between continual interactive learning and foundation models that refine each other.