Fei-Fei Li: Simulation is Key for Robotics as Real-World Data Falls Short

YunzhuLiYZ · x · 2026-07-30

In a recent discussion, Fei-Fei Li highlighted the critical role of simulation in robotic learning. She explained that simulation allows us to play out events that haven't happened or cannot happen in the real world, enabling robots to learn how to act in those situations. Emphasizing that real-world data alone is insufficient for robotics, she pointed to self-driving cars—specifically Waymo's use of billions of hours of simulation—as proof that even the simplest robots rely heavily on simulated training.

Original post →

Fei-Fei Li: Simulation is Key for Robotics as Real-World Data Falls Short

More from Embodied