Chinese robot makers lack data and 'brain' for embodied AI, WAIC experts say
During the World Artificial Intelligence Conference (WAIC) held in Shanghai, experts pointed out significant obstacles hindering the advancement of embodied AI in enhancing robots' interactions in real-life situations. Wang Xiaogang, co-founder of SenseTime and chairman of Ace Robotics, emphasized that the primary issue lies in integrating hardware, data, models, and real-world applications into a cohesive iterative system. He mentioned that training data obtained from human demonstrations needs to be optimized alongside hardware design and data collection techniques. Nonetheless, the deployment of robots in practical environments remains insufficient. Yao Maoqing, partner and senior vice-president of AgiBot, noted that the existing multi-modal data regarding the physical world is inadequate compared to that utilized in large language models, posing a challenge for training future humanoid robots.
Key facts
- WAIC concluded on Monday in Shanghai.
- Wang Xiaogang is co-founder of SenseTime and chairman of Ace Robotics.
- Wang said the key is to link hardware, data, models, and real-world scenarios into a closed-loop iterative system.
- Training data is collected from human demonstrations.
- Hardware design must be jointly optimized with data-collection methods and physical structure.
- Robot deployment across real-world scenarios remains too limited.
- Yao Maoqing is partner and senior vice-president of AgiBot.
- Multi-modal data about the physical world is far from adequate compared to that used in large language models.
Entities
Institutions
- World Artificial Intelligence Conference (WAIC)
- SenseTime
- Ace Robotics
- South China Morning Post
- AgiBot
Locations
- Shanghai
- China