Know What's Happening in AI

World Models, Architectures, and the Next Phase of AI

At the Spring School AI For Impact, Yann LeCun and Eric Xing sparred over the core design of world models: LeCun’s Joint Embedding Predictive Architecture (JEPA) predicts future latent representations without reconstruction, while Xing’s Generative Latent Prediction (GLP) re‑attaches a decoder to validate predictions against pixel‑level output. The debate mirrors the classic rationalist‑empiricist split and influences every major AI direction—from diffusion versus autoregressive video generators to transformer‑based planners and state‑space models. Recent research shows that pure approaches struggle with either computational efficiency or fidelity, prompting a wave of hybrid architectures that learn when to trust abstraction and when to demand reconstruction. The next decade is likely to be defined by models that dynamically blend JEPA‑style efficiency with GLP‑style grounding, shaping the path toward safe, self‑improving agentic AI.