World Labs acquired SceniX last week, merging two teams focused on spatial intelligence, the problem of teaching machines to understand physical space. Fei-Fei Li, who created ImageNet and now leads World Labs, sits down with Martin Casado and Yunzhu Li, Columbia professor and SceniX co-founder, to explain the strategic logic behind the deal and why this problem is harder than it looks.
The conversation goes deeper than the acquisition announcement. Fei-Fei Li built ImageNet, the dataset that unlocked modern computer vision, and is now betting her next act on spatial intelligence as the core unsolved layer of AI. Yunzhu Li brings simulation expertise that World Labs sees as the key to generating the kind of physical-world training data that cameras alone cannot provide.
Read the full transcript for the technical argument on why simulation is the unlock, not just a shortcut, and for Fei-Fei Li's specific claims about what spatial intelligence enables that current foundation models cannot do. The acquisition framing is context. The real content is the roadmap.
[WATCH ON YOUTUBE →]