谢晨

谢晨 is the founder and CEO of 光轮智能 and the central guest in 134. 【数据的综述】和谢晨聊，新时代的石油、历史、版图、数据金字塔、定价与Recipe. The source presents his path as a move from Peking University physics and a Columbia quantitative-finance PhD into dynamic pricing, product work, Cruise, Nvidia, NIO autonomous-driving simulation, and then embodied-AI data infrastructure.

Key Views

• Data should be understood as Data As Education rather than only annotation, labeling, or stored files.
• For robotics, Robotics Simulation Evaluation is closer to a prerequisite than an optional accelerator because real robot deployments cannot yet generate enough cheap repeated feedback.
• Embodied Data Pyramid should combine scarce real robot data, scalable simulation, internet-scale visual data, and human first-person data.
• Data Recipe Co-Creation is necessary because the right mixture of real, simulated, human, and evaluation data has to be discovered with model teams through experiments.
• Data companies should become Data Engine Learning Loop providers rather than only Data Factory vendors.

Connections

• 光轮智能 — company he founded to build simulation and synthetic-data infrastructure for robotics.
• Cruise and Nvidia — work contexts that shaped his simulation-first view.
• Embodied AI, World Models, and Vision Language Action Models — technical field where his data views apply.
• Scale AI and ImageNet — examples he uses to describe earlier data-industry stages.
• Data Pricing In AI — pricing logic he connects to feedback value, customization, and embodied trajectory quality.