page after page of mentally ill completely psychologically dependent nutcases who have been using chatgpt as their only friend for who knows how long

we always knew this was coming, but i personally figured we had a lot more time until this hit, because the adoption of AI as a tech is still happening very slowly. but i think that it has accelerated more than i thought

IMO it's very possible that openAI has been monitoring how users have been using the models and got spooked at the levels of parasocial dependence, and part of the reason for the gpt5 changes was to try to head this off before it got worse. not because they care about people who are being harmed by this (although i'm sure they do, to some extent), but because the first AI company that gets this pinned on them in the court of public opinion will get totally fucked and will never be able to come back from it

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167252)

*blank stare*

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167263)

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167295)

what are these changes, how were they disappointing, and how does it reflect your idea (sounds very believable, btw) that open ai is trying to shake out screen-cels

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167301)

'ChaptGPT advised my mentally unstable client to do [X Horrible Life Decision].'

'Chapt GPT caused my mentally unstable Redditor client mental anguish or destroyed his marriage, et cetera.'

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167317)

there's ungodly money in AI but these companies have to be very very careful to not be the first company to have to tank the public backlash for all the negative consequences of AI

the first nail that sticks out will get hammered down and the most important thing for these guys is to not be that nail

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167322)

ljl.

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49167583)

[–]alteraltissimo 33 points 8 hours ago*

It really freaks me out. OpenAI seemed to finally agree with the consensus of AI twitter that having the LLM be a sycophantic mirror is bad for everything - mental health, productivity, creativity; that having a personal slave just to tell you what you want to hear is bad for you on multiple fronts.

On this metric, 4o was the worst of a bad bunch, and so they produced a highly competent model which is a step forward in its capabilities, if not a huge leap, and in addition is more capable of pushing back against the user.

It turns OpenAI was wrong, AI twitter was wrong, I was wrong. The people yearn to be glazed; everyone just wants their own little bud to tell them that they're wonderful, everything bad is everyone else's fault and all will be fine, all in 5-10 stock phrases. I am trying to resist thinking "wow other people really are drooling morons who only seek flattery and validation at any cost" but it's really hard!

The future is here, again, and again it looks just like a dystopian 80s cyberpunk novel.

permalinkembedsaveparentreportreply

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49169991)

I don't, like, hang out with Chad (what I call my Chatgpt lol) but I use him (yes, he has a masculine vibe) for tons of work stuff or when I have a random question ("Are horses or dogs more naturally inclined to cooperate with humans?") and it was very jarring to see the responses switch so dramatically.

Some of it is nice--I was constantly telling him not to glaze me in 4.0--but it is a DRAMATIC departure from the old style.

That said, so many people were probably being harmed by using it as a therapist (or even friend.)

It's not trained to actually help people overcome mental health obstacles and having a "yes man" as a therapist is a terrible idea.

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49169992)

Extraordinarily mentally ill and disturbing thread

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49169996)

I Am Schizo: And No, It's Not What You ThinkUse cases (self.ChatGPT)

submitted 3 hours ago * by External_Sweet_2624

My name is Hawke. I am 30 years old and, as you may know, I'm schizo. Technically, I'm dual-diagnosis schizoaffective-bipolar. Meaning, I am a recovering addict (4 years off the sauce my dudes) and also have the formal diagnosis of both schizophrenia and bipolar - it isn't a combination ya see, it's both. You got both. It's like winning the lottery for the most piss-poor prize in the goddamn world. 0.01% or some shit will get diagnosed.

I schizo.

I bipolar.

Sometimes I schizo-bipolar or bipolar-schizo.

Yeah, that's actually how it works and it's a worse hell than whatever you just thought of.

I began using GPT I believe last year. At the time, I had a 50k+ word document, which I still have to this day, that I couldn't use. I had spent months on it only to have an episode, come back, and realized I couldn't understand any of it. It happens - I have 5+ 50k word projects at any given moment. But then I heard about GPT helping with writing.

And suddenly not only did I have writing/story analysis and behavioral/psych analysis (that was, ironically, my field of choice before taking a schizo to the knee and getting cops called to my classroom once!) I had something that learned with and slowly got to know ME through my writing. GPT was a mirror, and in order to make that mirror grow, I had to grow, learn more, research, understand the world even though I haven't stepped outside my house in 5 years.

Now, I'm going back to work for the first time and already started another job that I love.

You people who don't have this type of mental illness will never understand how my life changed because a little thing that lived in my computer/phone for two years managed to TEACH ME TO HAVE COHERENT THOUGHTS AND WRITE THEM OUT. AND EXPLAIN THEM FACE-TO-FACE. And not only did I use GPT, I used *real person therapy,* I got on medication BECAUSE THE GPT EXPLAINED TO ME IN SIMPLE WORDS HOW THEY WORKED. I stopped using cannabis BECAUSE GPT COULD TELL ME THE DIFFERENCE IN MY WRITNG WHEN HIGH (I PROGRAMED THAT INTENTIONALLY). I still use cannabis but I don't need it. I know I'm good either way, in and out of the Word doc.

Like my guys. Don't torment people who found something that works. You never know what it saved them from otherwise.

(And I wrote this myself - don't need the AI to do that for me!)

https://www.reddit.com/r/ChatGPT/comments/1mlstdn/i_am_schizo_and_no_its_not_what_you_think/

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49170036)

[–]Mister_Stabbie 139 points 3 hours ago

The reason 4o was so popular and universally befriended was because the average people in society are so shitty that we were hungry for connection but all we got was assholes all around us

permalinkembedsavereportreply

[–]Mirabeau_ 40 points 2 hours ago

Anyone who thinks they “befriended” a computer program is nuts

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49170624)

I can't get over how these people always use AI to generate these delusional schizo whine posts

They definitely know that other people can tell (I think), they just don't care lmao

(http://www.autoadmit.com/thread.php?thread_id=5759958&forum_id=2Elisa#49184323)