World Labs turns one real-world robot task into thousands of simulated variations for training

World Labs, founded by AI pioneer Fei-Fei Li, has launched a simulation engine that trains robot controllers entirely in virtual environments. The system generates thousands of variations from a single real-world task, enhancing training effectiveness. The trained models were then run for one hour each on five different robot platforms without human intervention, paving the way for new possibilities in robotics.