OpenAI's Leo Gao questions the conceptual limits of using interpreter models to understand complex AI systems · Digg