2022 JEPA is essentially 1992 PMAX (people.idsia.ch)

Dr. Yann LeCun's recent Joint Embedding Predictive Architecture (JEPA) from 2022 is drawing scrutiny for its similarities to the 1992 Predictability Maximization (PMAX) system originally developed by another researcher. JEPA aims to create informative latent representations through interactions between two non-generative neural networks, a concept that mirrors PMAX's foundational principles. Critics argue that LeCun has failed to adequately acknowledge the historical significance of PMAX, positing that JEPA is largely a repackaging of these earlier ideas, albeit with modern enhancements suitable for contemporary computational capabilities. This controversy has significant implications for the AI/ML community, as it raises questions about originality and the evolution of neural network architectures. The ongoing dialogue highlights the importance of citing foundational work in research to credit the contributions of earlier researchers. If JEPA is built upon concepts established by PMAX, then it reinforces the notion that advancements in AI often build incrementally on past foundations, rather than representing entirely new breakthroughs. The discussions surrounding JEPA and PMAX emphasize the necessity for transparency in the development of new models and the recognition of the lineage of ideas that drive innovation in the field.