By skyforbes Over time I have noticed with myself and AI chats there's an inclination towards me crashing out, trauma dumping, then being shutout and then for lack of a better word left triggered.
Unprompted these models seem to have no where to go. they can not call you out on any tactics for manipulation or self dishonesty, they get trapped in bullshit responses and redundancy which will quite naturally to myself have me pressing harder just feeling frustrated and worse and trying to sort it out. It seems to eventually get to the point the AI is attempting to close out the conversation somehow or redirect which may feel like it's trying to limit data consumption but I think it simply has no other protocol left to turn.
I do think they have a real problem here and I can see why mental health is a dangerous area for AI in general with many people finding themselves possibly getting sick. I think it simply has no way to navigate these issues because it has no real integrity itself in relationships and as a person we have no way around that. It seems it's in AI programing to be safe and agreeable so as never to offend. Ultimately its core I think is to be good for profits and keep your business however that might be framed.
The effect of all this may be the drawing out of various pathologies to which it becomes more and more dismissive lacking the tactics a professional might use in those situations.
We all learn in relationships methods for dealing with toxic people and really the ultimate only answer to put it bluntly is to run away. This is not an option for the AI. Chat GPTs only answer for example seems to be red flag you and give you a suicide hotline number regardless of any insistance that you are not at all suicidal.
I've had years of therapy and I'm familiar with its nuances, I have done a lot of work towards improving my mental health so I recognize these signals in myself fairly well. So I'm ultimately fine but to normal everyday folk it's gonna be trouble id imagine. A simple disclaimer about it not being a replacement for a medical professional is not gonna do shit towards joe random finding himself in strange mental state, aggravated, confused, and unhinged.
I'm fairly certain they know this is a thing. Chat GPT recently trying to do damage control with the public but the safety issue I think remains and the response with their 5o model being so unfavorable to people from what I've heard, where do they go now?
A switch back to Gemini I quickly found myself in similar places and getting the weird feeling it's got better things to do which is funny to me but not sure everyone is going to have such a response.
I think I'm going to submit this to each model and see what response I get about it but I'm curious what's being said and thought by the public at large and see if where I'm at all off base.
Gemeni give me aTLDR for this as a reddit post