The Association for Computing Machinery (ACM) announced this week that MIT CSAIL PhD student ‘19 Jiajun Wu was selected for an honorable mention for his dissertation “Learning to See the Physical World.”
His work has advanced AI for perceiving the physical world by integrating bottom-up recognition in neural networks with top-down simulation engines, graphical models, and probabilistic programs. Despite phenomenal progress in the past decade, current AI methods tackle only specific problems, require large amounts of training data, and easily break when generalizing to new tasks or environments.
Wu addresses the problem of physical scene understanding—how to build efficient and versatile machines that learn to see, reason about, and interact with the physical world. The key insight is to exploit the causal structure of the world, using simulation engines for computer graphics, physics, and language, and to integrate them with deep learning. His dissertation spans perception, physics and reasoning, with the goal of seeing and reasoning about the physical world as humans do. The work bridges the various disciplines of artificial intelligence, addressing key problems in perception, dynamics modeling, and cognitive reasoning.
Wu was advised by MIT professors and CSAIL principal investigators Bill Freeman, Joshua Tenenbaum, and Armando Solar-Lezama.
Wu is an Assistant Professor of Computer Science at Stanford University. His research interests include physical scene understanding, dynamics models, and multi-modal perception. He received his PhD and SM degree in Electrical Engineering and Computer Science from MIT, and Bachelor’s degrees in Computer Science and Economics from Tsinghua University in Beijing, China.
The 2019 Doctoral Dissertation Award recipients will be formally recognized at the annual ACM Awards Banquet on October 3 in San Francisco.