World foundation models (WFMs) are neural networks that simulate real-world environments as videos and predict accurate outcomes based on text, image, or video input. Physical AI developers use WFMs to generate custom synthetic data or downstream AI models for training robots and autonomous vehicles
https://www.nvidia.com/en-us/glossary/world-models/
https://www.reddit.com/r/MachineLearning/comments/1kf3pes/discussion_what_exactly_are_world_models_in_ai/
