I'm not sure why people think these answers are any more trustworthy than any other it gives - it's not like it has a memory of those inputs different from othe

I'm not sure why people think these answers are any more trustworthy than any other it gives - it's not like it has a memory of those inputs different from others. It could be hallucinating them, like it hallucinates literally everything else. >On Thursday, a university student named Marvin von Hagen independently confirmed that the list of prompts Liu obtained was not a hallucination by obtaining it through a different prompt injection method: by posing as a developer at OpenAI. So that's what passes for proof/fact now, huh? I admit it increases the chance, but that's not a confirmation.