China’s Robotics Riddle: Can Better Brains and Data Solve the Industry’s Biggest Headaches?

The recent World Artificial Intelligence Conference in Shanghai highlighted a pressing issue plaguing Chinese robotics companies: the lack of a robust “brain” to drive innovation, combined with insufficient data to fuel it. This seemingly simple combination – better brains and more data – has become an industry-wide fixation for those trying to push the boundaries of embodied AI.

Wang Xiaogang, co-founder of SenseTime and chairman of its robotics spin-off Ace Robotics, pinpointed the problem as one of creating a seamless loop between hardware design, data collection, physical structure, and real-world scenarios. Chinese robot makers need to get their priorities straight: should they focus on designing better robots or gathering more data? Currently, they rely heavily on human demonstrations for training data, which is not enough.

The crux of the issue lies in replicating real-world scenarios at scale. This task may seem daunting, but it’s not new territory. Similar concerns have been raised by experts in computer vision and natural language processing. These researchers have long grappled with collecting high-quality data that accurately represents human experience.

Compared to large language models, embodied AI has a significant disparity in training data availability. Yao Maoqing, partner and senior vice-president at Shanghai-based humanoid robot maker AgiBot, noted this discrepancy during an interview with the South China Morning Post. This bottleneck hinders developing world models capable of navigating their surroundings – a critical step towards creating next-generation humanoid robots.

While better brains and more data are essential for advancing embodied AI, it’s unclear if Chinese companies are putting too much emphasis on mimicking human behavior without first understanding what makes humans tick. Or perhaps the issue lies in how they collect and utilize their data: do they truly need more, or is it just a matter of better analysis?

The World Artificial Intelligence Conference was filled with promises from many companies touting their latest advancements in embodied AI. However, addressing these fundamental issues is crucial for making meaningful progress. Without it, the industry’s biggest headaches will persist.

Chinese robot makers would do well to remember that their products will be judged by how well they serve people – not just by how advanced their AI is. They need to strike a balance between technological innovation and practical applications. This means re-evaluating priorities: better brains and more data are indeed necessary, but so too is a deeper understanding of what makes humans tick.

The next few months will be crucial in determining the direction China’s robotics industry takes. Will they continue down the path of incremental innovation or take bold steps towards creating truly groundbreaking products? Only time – and a healthy dose of skepticism – will tell.