Speaking at the 2026 World Robot Conference forum, held in Beijing from August 19-22, Feng Yi, Vice President of AMD's Embedded Solutions Group in Greater China, delivered a keynote address on the evolving role of physical artificial intelligence. He emphasized that the core objective of this technology is never to substitute human workers, but rather to unlock and magnify their inherent capabilities.
Feng illustrated this vision by highlighting several practical applications where physical AI is already delivering transformative value. In the medical field, AI-powered systems assist surgeons in performing procedures with greater precision. In manufacturing, these intelligent technologies are creating safer production environments and enabling technicians to conduct inspections of high-risk facilities without direct exposure to danger. Furthermore, in agriculture, AI empowers farmers with farming methods that are both high-yield and sustainable, while also being easily replicable across different regions.
The executive elaborated on the broader societal impact of this technological shift. Across every industry, intelligent machines are increasingly taking over the tasks that are repetitive, physically demanding, or hazardous. This transition allows human workers to redirect their focus toward complex challenges that demand creativity, deep industry knowledge, and comprehensive judgment. According to Feng, this future is not a distant prospect; it is arriving at an unprecedented speed in the present day.
Feng also offered a forward-looking perspective on the next major milestone for robotics: full autonomy. The robotics sector is currently undergoing a significant transformation, moving away from script-based automation and toward agentic AI systems that possess independent decision-making capabilities. This evolution imposes new and demanding requirements on the internal computing architecture of robots. Whether it is an industrial robotic arm, a surgical robot, or a general-purpose mobile robot, each must now run dozens, or even hundreds, of concurrent agent threads on its central processing unit to handle the complexity of autonomous operation.
Comments