Meta released V-JEPA, a new learning model for AI to understand and predict what is going on in a video, even with limited information. It learns by predicting missing or obscured parts of a video in its internal feature space.

→ View original post on X — @rowancheung,