Been seeing all these headlines about people getting sent into chat-gpt-induced psychosis. Hard to tell if those stories come out of pearl clutching akin to the DnD scare, or if there is actually something nasty going on. I’m leaning towards the latter.
I think it’s mostly people with existing symptoms that get exacerbated by an affirming chat bot that they believe is real AI. So when it says “yeah, there are people behind the curtains watching you” or whatnot they believe it even harder. I imagine it’s very easy to slip into psychosis if you’re prone to it and have supernatural beliefs in general, but LLMs would absolutely reinforce all kinds of ideas.
I’m not so sure it’s nasty so much as a natural consequence of the way LLMs are tuned to be agreeable and the same people who believe q anon think it’s the real deal AI so of course they just run with it when it agrees a lot.
I guess it could be nefarious still, but idk it just seems like people are feeding the LLM nonsense and getting it back?
I bought new running shoes recently and the high schooler helping me out literally (and I have never felt older/more depressed in my life) told me I should leave chatgpt running to record my sleep. All so it can provide better feedback on my running form. It was incredibly difficult to keep a straight face
Been seeing all these headlines about people getting sent into chat-gpt-induced psychosis. Hard to tell if those stories come out of pearl clutching akin to the DnD scare, or if there is actually something nasty going on. I’m leaning towards the latter.
I think it’s mostly people with existing symptoms that get exacerbated by an affirming chat bot that they believe is real AI. So when it says “yeah, there are people behind the curtains watching you” or whatnot they believe it even harder. I imagine it’s very easy to slip into psychosis if you’re prone to it and have supernatural beliefs in general, but LLMs would absolutely reinforce all kinds of ideas.
I’m not so sure it’s nasty so much as a natural consequence of the way LLMs are tuned to be agreeable and the same people who believe q anon think it’s the real deal AI so of course they just run with it when it agrees a lot.
I guess it could be nefarious still, but idk it just seems like people are feeding the LLM nonsense and getting it back?
I bought new running shoes recently and the high schooler helping me out literally (and I have never felt older/more depressed in my life) told me I should leave chatgpt running to record my sleep. All so it can provide better feedback on my running form. It was incredibly difficult to keep a straight face