Review of Physical Artificial Intelligence
组织者
演讲者
赵卓阳
时间
2026年05月22日 15:00 至 16:00
地点
A3-1-301
线上
Zoom 204 323 0165
(BIMSA)
摘要
The deployment of large generative models often encounters a "physics consistency" crisis, as pure data-driven correlation learning fails to capture the fundamental conservation laws constrained by the physics manifold. Based on a featured review from the Journal of Computer Science and Technology (JCST), this talk introduces the paradigm shift and technological frameworks of physical artificial intelligence designed to embed natural laws into machine learning. I will describe the three components of physical AI: the physics-informed AI, the generative physical AI, and the embodied AI. A "five-dimensional red line" evaluation protocol are presented to illustrate how physical laws can be quantitatively verified beyond traditional statistical metrics, and how core challenges regarding gradient stability and unified frameworks can be addressed.
演讲者介绍
Zhao Zhuoyang is a first-year Ph.D. student in a joint Program between BIMSA and Renmin University of China, majoring in Mathematics, under the supervision of Professor Zhang Xiaoming.