AI can generate sound, but only a person can decide what the work should express — make the origin visible; listening comes before judgment.