Researchers have developed the Video Joint Embedding Predictive Architecture (V-JEPA), an AI model that learns about its environment through videos and exhibits a sense of "surprise" when presented with contradictory information. Unlike traditional pixel-space models, V-JEPA uses higher-level abstractions to focus on essential details, enabling it to understand concepts like object permanence with high accuracy. The model has potential applications in robotics and is being further refined to enhance its capabilities.

Researchers have developed V-JEPA 2, a neural network trained on one million hours of YouTube videos to enhance robotic understanding of physics through video prediction rather than language processing. This model enables robots to perform actions in new environments with impressive accuracy, demonstrating zero-shot generalization and significant efficiency compared to traditional methods. Despite its successes, the model faces challenges with camera sensitivity and long-term planning.