the irony of AI interpretability is that AI might actually solve its own black box problem. “What we did not always have is many automated scientists to work in it, far faster and in far greater numbers than we can. Now we do. They may be able to do in months in frontier language models what took humans most of a decade for vision models”