On the afternoon of May 21st, Lin Jiawei, Vice President of Cross dimensional Intelligence Market, gave a speech at the 2025 Global Spatio Temporal Intelligence Conference (WGDC2025), stating that the biggest difference between embodied intelligence and the previous generation of industrial automation, Industry 4.0, or Industry 2025 is its intelligence, better generalization ability, and universality. It can make intelligent decisions, know the mutual position relationship between its hardware ontology, space-time, environment, and object tasks, and thus better judge what kind of action it should take in the next step. The core of achieving universal embodied intelligence with high usability lies in its perception of the size of the brain. This requires intelligent capabilities from the underlying data to the middle brain, and even strong coupling at the hardware level.
Compared to the decades long accumulation of text, images, videos, and other data in big language models, Lin Jiawei frankly stated that based on the current development stage of embodied intelligence, data scarcity has become a major bottleneck in the industry. Robots are a three-dimensional world, including three-dimensional coordinates and even 6D degrees of freedom poses, involving issues in the physical world space. Without such a large amount of data and modalities, it is not sufficient to train embodied abilities under the so-called Scailing Law paradigm. ”
It is reported that the cross dimensional intelligent team has developed the only self-developed intelligent engine at the bottom level in China, which constructs environmental assets. For example, corresponding robot hardware assets will be built for environments such as venues, homes, factories, and intelligent manufacturing, integrating various modal data. The environment will be generated through automated patterns, objects will be generated, operation tasks will be generated, model training will be automated, and model SDK files will be generated, which will ultimately be deployed to the real world. For enterprises, the most important thing is to find the necessary scenarios when landing, so that technology can truly bring value to the scenarios. At present, cross dimensional intelligence has achieved 100% synthetic data in semi-structured scenarios, and even for synthetic data in simulation engines, millimeter level robot operations can achieve a success rate of over 99.9%.
