As a researcher deeply involved in the development of AI technologies, I strongly believe that AI systems above a certain capability threshold must have interpretable decision-making processes. It is crucial for ensuring accountability, safety, and fairness. We need to understand how these systems make decisions to trust them in critical applications.