NEO AI Model Learns Causal Explanations from Observation

A new AI model named NEO is being developed, which takes a distinct approach compared to traditional world models. While models like Sora focus on predicting future pixels and JEPA on compressing observations, NEO aims to uncover the underlying causes of events. Instead of merely forecasting outcomes, it endeavors to understand the "why" behind observed phenomena. NEO achieves this by autonomously discovering and learning reusable "program" building blocks from raw observational data. For instance, if shown an object moving left and then down, it identifies "left" and "down" as separate, fundamental actions that can be recombined. This method allows NEO to construct concise explanatory programs from simple, reusable steps, moving beyond black-box predictions to provide interpretable causal insights without explicit labels or symbolic programming.