That’s entirely fallacious though. A human artist is another person making the decisions. The program isn’t. That’s why the commission comparison fails.
If you hand a painter a sentence, they bring their own taste, draft, and refusal. That’s a second author. If you aim a camera, the sensor is not an author. You still took the photo.
Same here. The prompt isn’t the whole job unless you stopped at say one sentence. Picks, retries, edits, and the final cut are the direction. You’re pretending the only input possible is a caption so the machine can take the credit. A photographer has a curve of what effort they want to bring to the table too.