7h agoWorld Labs CEO Fei-Fei Li argues text-trained models cannot grasp physical reality and true intelligence requires world modelsWorld models learn spatial, temporal, and physical structures.