World Labs Builds Worlds for Robot Training with Spatial Intelligence

World Labs, co-founded by Fei-Fei Li, has shared early results of using spatial intelligence to build worlds for training robots. - The work aligns with the vision that spatial intelligence includes not just perceiving and generating virtual/physical worlds, but also interacting with them. - A video demonstration shows the system in action.
Key insights
World Labs Advances Robot Training Through World-Building and Interaction
The demonstration from World Labs shows a practical application of spatial intelligence: creating interactive worlds that can be used to train robots. This moves beyond static world generation to environments where agents can interact, aligning with Fei-Fei Li's earlier statements about the full scope of spatial intelligence. The early results suggest that simulated interactive worlds may become a key component of robotics training pipelines.