Edge 320: Meet I-JEPA: Meta AI’s First Super Model Based on their Theory of Autonomous Intelligence
The new model is able to make predictions based on abstract predictions instead of pixel-level detail.
Last year Meta AI’s Chief AI Scientist, Yann LeCun, presented an architecture that could enable a new foundation for autonomous intelligence. The core idea of the architecture was to emulate human and animal’s ability to develop cognitive models of the world without having to accumulate huge amounts of data or go through painful experimentation cycles. The architecture relies on self-supervised learning (SSL) methods as its core building block and incorporates several components that match human cognitive abilities, such as memory or perception. Last week, Meta AI unveiled I-JEPA, the first model based on Mr. LeCun’s ambitious architecture.