Wang Xingxing: The core challenge of Physical AI lies in aligning models with the physical world.

At the “Humanoid Robots and Embodied Intelligence: Innovation, Development, and Collaborative Opportunities” sub-forum of the 9th Hongqiao International Economic Forum, Wang Xingxing, Founder and Chairman of Unitree Robotics, pointed out that the core challenge facing Physical AI today is alignment between models and the physical world. He explained that while multimodal models’ intrinsic understanding capabilities are already relatively mature, the key bottleneck lies in alignment errors at both the input and output ends: on the input side, these manifest as discrepancies between sensor measurements and actual physical states; on the output side, they appear as gaps between action commands and real-world executed motions. Wang Xingxing expects breakthrough solutions to such issues within the next one to two years.