With New AI Model, Meta Hopes Imperfect Hands Will be a Thing of the Past

Too Long; Didn't Read

Unlike existing AI systems, Meta's so-called Image Joint Embedding Predictive Architecture, or I-JEPA, is designed to aid users based on what it already knows about the world. While existing models compare nearby pixels to aid in image creation, Meta's I-JEPA is already fed with an internal model of the outside world which it then uses to determine what it needs to produce. The result? A generative model that is less prone to producing people with too many, too few, or impossibly interconnected fingers.