Fei-Fei Li
Fei-Fei Li is discussed in 133. 对谢赛宁的7小时马拉松访谈：世界模型、逃出硅谷、AMI Labs、两次拒绝Ilya、杨立昆、李飞飞和42 as a major influence on Xie Saining. The source says her personal story and autobiography gave him comfort, but the more technical lesson is her ability to define a problem clearly enough for a field to move.
Problem Definition
Xie argues that ImageNet should not be reduced to “making a dataset”. In his reading, Fei-Fei Li defined the image-classification problem in a form that made deep learning progress visible, comparable, and scalable. That makes her a central example for Problem Definition In Research.
A case for AI models that understand, not just predict, the way the world works mentions Fei-Fei Li as one of the researchers pursuing World Models in some form. The Marketplace Tech episode does not detail her route, but it reinforces the wiki’s existing link between visual intelligence, representation, and model architectures that go beyond language-only prediction.
宇树上市暴涨，但人形机器人的钱到底从哪里赚？｜S10E26 adds World Labs as the robotics-adjacent branch associated with Li in the source. The episode uses World Labs to discuss simulated-world data for robots, connecting her page from image and representation problem definition into Robotics Simulation Evaluation and Physical AI.
Connections
- ImageNet — dataset and problem definition associated with her in the source.
- Xie Saining — researcher influenced by her story and research framing.
- Representation Learning, Multimodal Intelligence, and World Models — surrounding technical themes in the interview.
- Research Taste and Problem Definition In Research — methodological lessons the source draws from her work.
- Gary Marcus and LLM World Model Gap — Marketplace Tech context where she is named as part of the world-model wave.
- World Labs, Robotics Simulation Evaluation, and Physical AI — simulated-world robotics branch added by What’s Next S10E26.