At WAIC 2026, Tencent's Chief Scientist Zhang Zhengyou questioned the intelligence of AI that can reason but lacks understanding of the physical world, referring to it as a 'brain in a jar.' He emphasized that true intelligence integrates language, vision, spatial awareness, body control, and environmental feedback in a closed-loop system.
Tencent unveiled several models, including Hy-Embodied-VLM-1.0, which interprets images and scenes with significantly reduced computational demands. The Hy-Embodied-RxBrain-1.0 combines text reasoning with visual imagination for continuous cognitive processing, while Hy-Embodied-VLA-0.5 translates high-level goals into actionable steps for various robot forms. These innovations are supported by extensive training data and are already being implemented in real-world applications.
The company is also upgrading its Tairos platform for a more open-source approach, facilitating easier development from models to applications. Tencent's Robotics X Lab is collaborating with multiple robotics companies and partners to transition these technologies from theoretical concepts to practical implementations in diverse environments.
Editor's Note
Tencent's advancements in robotics highlight a significant shift towards embodied intelligence, integrating various cognitive functions to enhance robotic capabilities. This approach may influence future developments in AI and robotics, particularly in manufacturing and automation sectors, where efficiency and adaptability are crucial.
Leave a comment