Spatial intelligence is underappreciated compared to large language models, yet it underpins many future applications
While the media focuses on LLMs, Fei‑Fei argues that world‑modeling of pixels and 3D environments is essential for storytelling, robotics, and simulation.
Spatial AI enables machines to understand and interact with the physical world, a prerequisite for embodied AI.
The lack of attention risks under‑investment in the hardware and research needed to advance 3D perception.
Recognizing spatial intelligence as a core pillar can shift funding toward sensors, simulation platforms, and multimodal models.
This perspective reframes the AI roadmap: after mastering language, the next frontier is perception‑action loops.
“Spatial intelligence is underappreciated in the sense that everybody's still now talking about language large language models but really world modeling of pixels of 3D worlds is underappreciated.” — Dr. Fei‑Fei Li
Beyond large language models, spatial intelligence, AI-enhanced education, and labor-market transformation remain overlooked yet hold massive potential for societal change.