2026-09-04 15:04 UTC
It was posted like some kind of gotcha but…this book was in the dataset.
I’m aware that LLMs don’t keep the dataset in memory but it “knows” that this is an existing work but these sites aren’t doing any wild calculations to figure out if it was AI-generated. They basically just check to see if the sentences exist elsewhere and they do. In the original dataset.
So it was wrong to say it’s AI-generated but it correctly identified that it’s not original.
Replies (3)
-
@RampantParanoia2365@lemmy.world 2026-09-04 18:44
I’m sorry, what? It’s correctly identified that it’s what, now?
-
@Grimy@lemmy.world 2026-09-04 15:31
They analyze statistical patterns, they don’t cross reference the training data. It’s picking up the text as generated because it’s mostly guess work and constantly spits out false positives, especially with non native speakers.
-
@michaelmrose@lemmy.world 2026-09-05 06:13
Everything you said was hallucinated. Nothing you said was even slightly coherent or connected with reality. Absolutely nothing was correct. If you let your cat dance on the keyboard more sense would come out.