His response is confusing to me as well. I didn't edit it that drastically. I moved one or two of the clauses around but the fundamental thrust was: if you think it's an AI image because of the whole-milk/low-fat discordance, but real-world designs for the show also have that discordance, then you should consider that the discordance does not mean it's AI.

That would still be true if:

* it was indeed an AI poster

* it was an AI image poster made to look that way

* it was a human-made poster accidentally made that way

* it was a human-made poster intentionally made to look that way

The truth of the show itself could have no bearing on what I was saying. The only thing it does rely on is whether or not the real-world designs did not correspond to the poster image.

The reality is that humans suck at telling AI. Sure, there are obvious tells for certain things, but if one really tries, they can make AI generated content indistinguishable from human made. You even see this on Twitter, where actual human artists are sometimes subjected to a modern day witch hunt by others saying their art is AI, and the artists literally have to prove that it was made by them, sometimes by pulling up various stages of the drawing in progress (and what is even funnier is that now Google's Nano Banana Pro can even generate that sort of progress compilation images).

My point was that the mistake didn't happen during prop creation. Those aren't milk cartons, those are HDP cartons, so the props are correct.

As to the content of your post: It doesn't make sense. Thinking something is not human created when it turns out that the real reason was that it wasn't created by a human in the show is not a valid reason to stop applying that as a useful discriminator between AI and human art. It's a Gettier case, but the J part of JTB knowledge still stands, and there's a reason grappling with the Gettier problem is so gnarly in epistemology.