Real machines doing real work, integrated production and training: Why is the next piece of infrastructure for embodied intelligence the "real-world training platform"?
The next piece of infrastructure for embodied intelligence is being pointed toward the "real-world training platform": putting training into real workstations, allowing collection, annotation, training, and the return of failure samples to form a data pipeline. A CAICT report shows that as of the end of June 2026, more than 70 embodied intelligence training grounds have been built and put into use nationwide, with another 46 under construction or planned. However, general embodied models require at least tens of millions of hours of real interaction data, while the world currently has only on the order of one million hours of high-quality real-machine corpus. "Nengzai No. 1," jointly developed by SAIC-GM and AgiBot, has already taken on cell grabbing and loading on the Buick Electra E7 battery mass-production line, having cumulatively completed nearly 80,000 workpiece operations, with its operating rate stable at 98%.
Body
The English body is being prepared. You can open the original source.