What would it even hallucinate? You wouldn't be asking it to cite a court case from 2004 that doesn't exist and it wouldn't invent people you didn't tell it about in the first place.

▲

ChrisMarshallNY 10 hours ago | parent [-]

I've been reading about how chatbots have been reinforcing paranoid delusions. If the person asking, was really trying to justify their own approach, and blame others (or that's what the chatbot perceives), the chatbot may go along, instead of saying "What the hell were you thinking?".

	▲	fragmede 7 hours ago \| parent \| next [-]
		I mean, the LLM's words are just generated words, not gospel from the heavens. If you're self-aware enough to be able to extract value from the words, and not go off the deep end with it, which it seem like you are, run it through as many SOTA online models and however many offline models you can fit on your hard drive, and compare all of their advice if you're really that worried about it, but make up a soap opera scenario, like some thing that ends with "and then I slept with her sister" and see just how many of them say "good job getting it in!" and not "you're an asshole", if you need to prove to yourself that, as with everything, you shouldn't believe everything you read on the news. Grok: https://grok.com/share/c2hhcmQtNQ_26f4c367-77ed-4b6e-be55-83... Claude: https://claude.ai/share/dca96b18-d583-4e14-b805-725d2e060761 Interestingly ChatGPT won't let me share a chat link for the same input text due to not passing a moderation filter, but you can plug that same gauche prompt into chat.com for yourself. Couldn't find the share button in Gemini.
	▲	fragmede 6 hours ago \| parent \| prev [-]
		Oh my first reply didn't address your question. Yes, if you're a malignant narcissist that seeks reassurance that it's okay you're a piece of shit, then it can be used that way too. I don't read you as one, however.