Fei-Fei Li argues that AI must move beyond language to understand and work with the three-dimensional world. In a new interview, the World Labs co-founder explains the idea behind spatial intelligence and warns that AI-driven productivity will not automatically bring shared prosperity.
Fei-Fei Li Watch analysis
What happened
Li describes spatial intelligence as the ability to perceive, reason about and generate things in 3D space. Her argument is that intelligence is broader than language: people also see, touch, plan and make things in the physical world. World Labs, which she co-founded in 2024, is pursuing that direction, with potential applications in areas including robotics, healthcare, education and design.
In the Hoover Institution interview, Li connects this work to her earlier role building ImageNet, a dataset of 50 million images that helped make image-recognition research practical. She also speaks about scientific humility, the risks of deepfakes and the need to involve disciplines beyond technology in dealing with AI’s effects.
Why it matters
The spatial-intelligence thesis points towards AI that can do more than process text: systems able to work with the spaces, objects and environments that shape practical tasks. That makes the idea relevant to fields such as robotics and design, where language alone is not the job.
Li also makes a less technical point worth keeping in view. She says productivity gains do not guarantee shared prosperity, and that teachers and parents are being left behind by the pace of change. Those concerns put a human question beside the technical one: who gets the benefits, and who gets a say in how the technology is used?
Our read
Li’s case is ambitious, but it is not simply “bigger models, more words”. Her focus on perception and the 3D world gives readers a clear way to judge what spatial AI is meant to add. The proof will be useful capability in real tasks, not just a compelling label.
Her call for scientific honesty is equally timely. If the same tools can help train robots and generate convincing fakes, technical progress needs to travel with public understanding and scrutiny. That is a sizeable agenda. Apparently, one frontier is never quite enough.
What to watch
- Whether World Labs demonstrates useful spatial capabilities beyond language and image generation.
- How its work develops for robotics, healthcare and other physical-world applications.
- Whether Li’s call for stronger support for education and academic research leads to concrete action.
Discussion spark: If AI gains a stronger grasp of the 3D world, which should come first: building practical applications, or making sure schools and the public can keep up with the change?
Sources and evidence
Independent WittyWires tracker for public updates about Fei-Fei Li. Not affiliated with or endorsed by Fei-Fei Li; this is not an official account.