Global Markets
Home›Global Markets›Asia›Chinese robotics firms face data and systems bottlenec…
Chinese robotics firms face data and systems bottlenecks for embodied AI
Industry insiders cited a need to better connect hardware, data, and real-world scenarios into closed-loop systems, and said current robot deployments cover too few environments.
Chinese robotics companies are struggling to improve how embodied AI interacts with the physical world due to gaps in available training data and the difficulty of building a system that can learn in a closed loop, according to interviews conducted around the World Artificial Intelligence Conference (WAIC) in Shanghai, as described by the South China Morning Post.
The outlet reported that Wang Xiaogang, co-founder of SenseTime and chairman of its robotics spin-off Ace Robotics, said the most critical challenge is to link hardware, data, models, and real-world scenarios into an iterative closed-loop system. He also argued that while training data is often collected from human demonstrations, translating that information into better embodied AIs requires jointly optimizing robot hardware design with data collection methods and the physical structure of the robots.
Wang further said robot deployment across real-world scenarios where data originates remains limited, and he questioned how to unlock those scenarios and replicate them at scale.
The South China Morning Post also reported that Yao Maoqing, partner and senior vice-president at Shanghai-based humanoid robot maker AgiBot, said multi-modal data about the physical world is not adequate compared with the data used to train large language models, creating a bottleneck for training the world models expected to support next-generation humanoid robots.