I made a post on a local Photograpghy Club Facebook page showing an old photo restoration I did using ChatGPT. It did not appear for a long time, and then I noticed a "pending admin approval" tag. This FB admin is famous for taking very long winter naps. So, curious, I gave ChatGPT the post and asked why Facebook flagged it for approval. I got a long answer, but in short, it said Facebook's AI probably keyed on the terms "Google", "ChatGPT", "Photo restoration" in close proximity, and thought the post could be fake news, or porn, or other forms of fakery, and human eyes should be used to verify.
So Basically: AI says AI does not trust AI.
Reminds me of a recent news article about some researchers who fed a bunch of 30+ year old master thesis papers to a leading AI detection program. It flagged over 60% of them as containing AI content. Their conclusion: Those 30+ year old papers had likely been used to build the AI detectors LLM, so now they look plagiarized.
