AI Research Reframes Intelligence With JEPA And AGI

In a walkthrough, Pourya Kordi outlines emerging AI research directions, highlighting Meta’s JEPA architecture and DeepMind’s ‘‘minimal AGI’’ target for 2028. The piece contrasts Yann LeCun’s skepticism with Demis Hassabis’s gradualist view, critiques LLM limitations, and emphasizes a shift toward abstraction, counterfactual reasoning, and integrated world-language-vision systems. These trends may reshape model design and research priorities.