ChatGPT Works — Do We Really Need To Know Why?

Will interpretability grow more important as LLMs take on more decisive roles?