Speaking on Saturday at an industry event in Shanghai, Yao said core capabilities such as “reasoning, planning and control should ultimately be integrated within a unified framework”, noting the importance of world models instead of simulating. He added that “scalable new approaches to data collection” should be explored to solve the scarcity of training data, and robots must evolve from performing isolated tasks to whole-body coordination.
Yao, the 2000 Turing Award winner and the dean of Tsinghua University’s College of AI, also called for the establishment of open industry benchmarks and safety standards instead of “each fighting their own battle”.
Yao was addressing the Global Developer Pioneers Summit and the International Embodied Intelligence Skills Competition, a government-led technology event for developers from December 12 to 14.
The event hosted competitions in scenarios including industrial assembly, home services, flower arranging, healthcare and elderly care, and restaurant services. Teams of developers deployed different robot models to demonstrate their practical capabilities across real-world tasks.
Shanghai also flexed its muscle during the summit, establishing a local industry alliance, with Pudong district rolling out a package of support measures, including free access to computing resources, talent housing and start-up funding.
