By Yunzhong, from QbitAI
QbitAI | Official account QbitAI
Seventy years ago, in a meeting room at Dartmouth College, the concept of "artificial intelligence" was first put to paper, launching a long exploration of machines simulating human thought; seventy years later today, the question of the AI era has long advanced from "can machines compute" to "can machines understand the world”。
". When algorithms step out of the virtual data space, how to make machines truly see, perceive, and integrate into the real physical world has become a threshold the entire industry must cross.
On October 20, the2026 Physical AI and World Model Conferenceofficially opened in Nanjing, under the theme "Modeling the Physical World · Intelligent Start for a Better Future".
The conference was jointly hosted by Jiangsu Provincial Data Group, Nanjing Big Data Group, and the Chinese Association for Artificial Intelligence. In venues linked both online and offline, explorers of different generations and different tracks gathered together under the same question about the future, exchanging ideas and anchoring direction.
From "computing" to "understanding": AI's next required question
That artificial intelligence has arrived at a crossroads of paradigm shift is the consensus of the entire conference.
Chinese Academy of Engineering academician Dai Qionghai cut straight to the core: physical AI lets AI master objective rules such as mechanics and temporal changes; world models integrate physical laws, spatial relations, and behavioral logic into a dynamic system capable of causal reasoning, driving AI to upgrade from recognizing the world to understanding the world, predicting the world, deducing the world, and transforming the world.
Chinese Academy of Sciences academician Tan Tieniu then pointed out in one sentence the foundation of world models:Modeling the physical world is a technical challenge, and even more a question of the times; enlightening a better future requires global wisdom, and even more requires open exchange.
In fact, it was no accident that this conference about AI's future landed in Jiangsu.
Here, a complete industrial system, massive application scenarios, and active data elements, together with an industrial ecosystem that encourages innovation and tolerates trial and error, have jointly paved the most suitable soil for physical AI to walk out of the laboratory and into the real world.
The same chain of questions: from Turing Award masters to young entrepreneurs
On the conference's keynote stage ran a chain of questions spanning half a century.
From master-level Turing Award winners to young entrepreneurs at the industry frontier, the answers varied in emphasis but pointed consistently in the same direction: letting AI truly enter the real world.
AMI Labs Executive ChairmanYann LeCun, Chinese Academy of Sciences academicianAndrew Yao, and Chinese Academy of Sciences foreign academicianJohn Hopcroft, the three Turing Award winners, each delivered deep thinking spanning the globe via video.
The on-site presentations likewise went layer by layer deeper along this chain of questions.
Foreign member of the US National Academy of Engineering and Chairman of the Council of the Hong Kong University of Science and Technology Harry ShumProposition: The dexterous hand is humanity's last bastion of dignity, the Mount Everest of embodied AI technical complexity, and the hardest machine-bound reproduction of humanity's core capability to transform the world.
Yu Haibin, academician of the Chinese Academy of Engineering and director of the Institute of Industrial AI of the Chinese Academy of Sciences, focused on intelligent manufacturing:A world model is AI's worldview, reflecting AI's understanding of the laws governing how the physical world operates. With a world model, AI can rehearse scenarios in its 'brain' to make better decisions.
Entrepreneurs from the front lines of industry brought vivid practices of technology implementation.
Chen Boyuan, founder and chief R&D officer of Inverse Matrix Technology and head of the Behavioral World Model Innovation Center at the Beijing Academy of Artificial Intelligence, believes:Physical AI must give machines the true ability to extrapolate from limited examples in a real-world environment full of unknowns; the mission of physical intelligence must ultimately be to return human value to humanity.
Momenta CEO Cao Xudong proposed:The world model is accelerating the Moore's Law of intelligent driving, and he predicts that 2028 will become a critical inflection point for the large-scale deployment of L4.
Gao Haichuan, founder and CEO of Qianjue Technology, focused instead on the general-purpose robot brain for embodied intelligence:Deployment in the physical world demands far higher fault tolerance and reliability than the digital world.
Consensus takes root: data, standards, and industry all sprout
The value of this conference lies not only in the collision and exchange of ideas, but even more in allowing industry consensus to take root step by step, moving from concept to practice.
At the data foundation level, the National Data Labeling Pilot SU-Wan City Alliance (Yangtze River Delta Data Labeling Joint Innovation Center) was formally established at the opening ceremony.
Ten cities—Nanjing, Wuxi, Suzhou, Huai'an, Yangzhou, Zhenjiang, Suqian, together with Wuhu, Ma'anshan, and Chuzhou—joined hands, marking the data labeling industry's entry into a new integrated development stage of 'ten cities in coordination' from 'one city piloting first'.
Jiangsu simultaneously released the 'Physical AI and World Model High-Quality Dataset Action Plan', centrally launched nine typical 'Physical AI and World Model' innovation achievements, and published 50 typical 'Physical AI and World Model' application scenarios for 2026, covering five major areas—industrial development, social governance, scientific research, livelihood services, and consumption upgrading—providing a clear path for technology deployment.
Standards system building also took a key step forward. The National Information Technology Standardization Technical Committee approved the establishment of the Embodied Intelligence Standards Working Group, and Zhu Xiumei, Party Secretary and Vice President of the China Electronics Standardization Institute, read out the founding announcement on site; the pre-research work on the national standard 'Artificial Intelligence World Models' officially launched, laying a solid institutional foundation for the industry's regulated development.
Industry and capital also moved quickly to connect on-site. The nation's first "Physical AI and World Model Investment Fund" strategic cooperation agreement was signed, with the fund's total scale at5 billion yuan, primarily investing in cutting-edgeworld modeltechnologies, focusing on areas such as medical artificial intelligence, industrial world models, and autonomous driving.
Notably, the event also linked the "Physical AI and World Models" academic journal with an AI youth community: the journal is openly soliciting academic papers, technical case studies, and industrial innovation achievements from across society, welcoming experts, scholars, industry practitioners, and university faculty and students to submit contributions; the youth community provides a one-stop portal for reading news, submitting resumes, and interactive exchange—here, you can be both an observer of new technologies and a creator of them.
The development of physical AI and world models is not a sprint to be completed overnight, but a long-distance race requiring sustained effort; time will test today's explorations and will also bring more unexpected innovative answers.
May every young seed carrying a dream of innovation find its own direction of growth in this open soil.
*This article is republished by QbitAI with authorization; the views are those of the original author alone.