The woman who taught machines to see is now teaching a generation of AI builders what rigor actually looks like.
The Summary
- Fei-Fei Li, who created ImageNet and co-founded World Labs in 2024, says she runs her startup "like a tiger mom" with demanding standards for her young team of engineers and researchers.
- World Labs raised $1 billion from AMD, Autodesk, and Nvidia to build "world models" that help AI understand and interact with 3D environments, positioning the company for what comes after ChatGPT.
- Li warns that growing tech backlash in the US poses risks for global AI progress, saying technologists need to better communicate AI's benefits.
- The Stanford professor, by far the most senior person at World Labs, has shifted from academic mentor to startup operator running what she calls a "high standard" shop.
The Signal
Fei-Fei Li didn't leave Stanford to build just another AI company. World Labs is developing "world models", AI systems designed to understand and interact with 3D environments. Think beyond chatbots that process text. Think agents that navigate physical and digital spaces, understand spatial relationships, and build things in three dimensions. This is infrastructure for Web4.
The $1 billion raise from AMD, Autodesk, and Nvidia tells you who sees the upside. Nvidia wants chips in every AI system. Autodesk needs AI that understands 3D design workflows. AMD needs a counter-narrative to Nvidia's dominance. World Labs sits at the intersection of all three bets.
"I'm by far the most senior in the company. From my co-founders to my engineers, they're so talented, but they're also young."
Here's what makes the "tiger mom" framing interesting. Li isn't just managing a team. She's operating with a level of discipline that's rare in the move-fast-break-things culture of Silicon Valley. She previously said she feels like a mentor even in job interviews. Now she's running a shop where the pressure is on and the standards are non-negotiable. In a field flooding with capital and hype, that matters.
But there's tension here. Li is publicly warning about growing tech backlash in the US and saying technologists need to communicate better about AI's benefits. Translation: public trust is eroding and it's starting to affect the operating environment. When someone raising billion-dollar rounds starts talking about communication strategy, they're seeing regulatory or reputational headwinds.
Three pressures converging:
- Building genuinely new AI infrastructure (3D world models, not chatbot iteration 47)
- Managing a young, high-talent team under the scrutiny that comes with her reputation
- Navigating a public conversation about AI that's souring in the country funding most of the innovation
The ImageNet dataset Li created in 2009 became the foundation for modern computer vision. Every image recognition system you use today traces back to that work. She's not theorizing about what AI could be. She's built the rails it runs on. Now she's saying the future lies beyond chatbots, and she's got a billion dollars to prove it.
The Implication
Watch how World Labs deploys this capital. If they're building foundational infrastructure for spatial AI, they're building for agents, not users. Agents that can navigate real and virtual environments. Agents that understand physics, depth, occlusion, and movement. That's the scaffolding for autonomous systems that operate in warehouses, design software, manufacturing floors, and eventually physical robotics.
For anyone building in the agent economy, the question is whether your system will integrate with spatial reasoning or stay stuck in flat text interfaces. Li is betting the next platform shift is about dimension. If she's right, the companies that can't make that jump will look like they're still building for Web2 when everyone else moved on.