Meta has introduced V-JEPA 2, a new video-based world model designed to improve machine understanding, prediction, and planning in physical environments. The model extends the Joint Embedding Predicti ...