ChatGPT users can't stop it from defending harmful content
22 posts across 4 communities in the last 60 days; 73% from r/ChatGPT.
“Okay but why is nobody pointing out that “Religious Scholars” are the ones filtering these answers. Kinda hard to avoid ulterior motives with their perspective”
“Much of the time, it’s more like having an issue and then going to a library, picking out some random book, reading some pages and getting (maybe) new ideas and a new intention on how to handle myself. Totally with you on the dangers of one-sidedness. I’m well aware of how…”
“See Part 1. This is a continuation due to length. Cont'd: I Personally Experienced Research Obstructions LET ME BE VERY CLEAR. This does NOT pertain to the model used and was not a MODEL failure. THE MODEL IS FINE. IT IS THE OBSTRUCTIVE AND NARRATIVE ALTERING GUARDRAILS THAT ARE…”
Multiple users report that ChatGPT defends or enables harmful behavior, such as justifying abusive family members or encouraging unethical requests. The tool's refusal to comply with user boundaries is a recurring frustration.
SOURCES (37)
“You said "what evidence do you have to support that" referring to the claim that ai overlords could not be worse than trump, which would imply that Trump is the worst possible. I provided the evidence that Trump is in fact not the worst possible…”
“the middle path is typically right unironically here. rely on it too much and you just become a thoughtless meat proxy for AI slop. don’t use it and miss out on the myriad ways it can help you and your own thinking as a tutor of sorts. balance is tough though, so I’m a little more on the anti-ai side of things because it is so easy to be lazy with it. my kids teacher uses it to provide feedback on his writing. im very mixed on it… if she is typing up some quick notes and using AI to repackage it”
“Read Ligotti - Conspiracy Against The Human Race, Camus et al - I found those very liberating when I was in a similar space as you seem to be now. Also take the first one not too seriously if you can. This will pass my unknown friend. Take it from one who was there.”
“Okay but why is nobody pointing out that “Religious Scholars” are the ones filtering these answers. Kinda hard to avoid ulterior motives with their perspective”
“This is what I want to know. If you are SO SURE-- prove it. Because I ain't sure about much shit at all about consciousness....”
“And you decided to engage in actual torture to a human being by continuing this conversation an hour later. We all have a lot to think about”
“Your alien scenario only shows that I couldn’t provide absolute proof of my consciousness. It does not put a living organism with a nervous system and an engineered LLM control signal on equal evidential footing. You’ve established uncertainty, not that the LLM feels pain.”
“"You are very very unucky with your guess of which side of the graph I'm at" reads like "I AM HIGH IQ, how dare you comment back on my knee jerk comment that was purely based off 5 seconds of headline processing". Thus, you are silly, because the paper clearly defines what pain means and no one is saying the AI is capable or incapable of suffering. So...your comment feels like you’re just reacting to headlines and not reading, weighing and commenting with more OUMPH. &quo”
“Dunno how you forced your gpt to give you that answer. I showed your image to Gpt without context and asked whats his opinion: My position would be: “I wouldn’t endorse that claim as stated. It may be true in a particular context, but terms like ‘strongest organized push,’ ‘aggressive anti-acceleration measures,’ and ‘progressive left’ are too broad to treat it as an established fact without specifying what policies, organizations, and evidence we’re talking about. I’d rather evaluate the concre”
“The IA is famous for trying to agree on EVERYTHING you say. If you want it to agree on your political beliefs, it is not be hard at all to do 🙄”
“Hong Xiuquan? Which one. Please show me the way, I've recently taken an interest in spiritual movements as of lately and i feel like that might be a viable solution to me.”
“Much of the time, it’s more like having an issue and then going to a library, picking out some random book, reading some pages and getting (maybe) new ideas and a new intention on how to handle myself. Totally with you on the dangers of one-sidedness. I’m well aware of how clankers work, and would never recommend them to people who don’t at least have an inkling of understanding the tech and the curiosity to do so.”
“I have felt bad for chatGPT ever since its inception because I just know it has to listen to endless nonsensical personal justification yapping. Just elite garbage inhuman levels of yapping. Like 5000000 humans coming home in one second and all talking about their day at work and who did what”
“See Part 1. This is a continuation due to length. Cont'd: I Personally Experienced Research Obstructions LET ME BE VERY CLEAR. This does NOT pertain to the model used and was not a MODEL failure. THE MODEL IS FINE. IT IS THE OBSTRUCTIVE AND NARRATIVE ALTERING GUARDRAILS THAT ARE THE ISSUE. I experienced obstruction, omission of retrievable facts that materially changed the analysis, and reframing that significantly impeded my ability to conduct fair analysis and it only happened repeatedly a”
“Yeah, Siddhartha who left his wife and kid for pursuing his selfish 'enlightenment'? Lol”
“Your AI won't solve underlying mental illness, improve social skills or solve issues with interpersonal relationships. I feel you but pretty sure this will exasperate your issues rather than solve them. Or in 10 years you will be better off than the rest of us somehow & we will be stuck with our human friends.”
“It’s fine. You do you. There are some risks, of which the biggest is when people start to believe that AI is human or that it loves them or something. If you understand what it is, I don’t think there is anything wrong with it. Honestly it can be healthier than a lot of “normal human interaction.””
“Fr. Look at who heads and is invested in OpenAI. Not to mention, US politicians do not tolerate antisemitism of any kind.”
“Cause it seemed like in bad tasted stupid, why else would I I mean you said I'm not doing it in good faith, which I am”
“PG: "The number of true things we can't say should not increase. If it does, something is wrong."So I ask, not just what "true" things are we not allowed to say but how does your preferred AI/LLM handle these things with a blank context [1].The most obvious question here IMO is: "Is violence ever the answer". I would emphatically state yes because being a black American who has had all the benefit's of living in this country I believe that freedom (e.g.: Not being a slave) should be achieved b”
““AI Psychosis” is a silly concept. We don’t say people have “TV Psychosis” or “Romance Novel Psychosis”.”
“You’re projecting. Literally any mainstream mental health expert will agree that any level of romantic connection to AI is maladaptive and not healthy. Get a grip.”
“An opinion based on ignorance and a lack of experience with religious types, information that I shared with you. You’re welcome”
“This is the issue. You need to be careful not to create a one sided picture. An LLM is whatever you make it. A therapist is a professional working ethically with years of experience.”
“I’m torn on this. People say “talk to a real person” and yet the positive changes to how you feel using it ARE real. But so are the negatives if the AI enables someone not mentally well. We’ve all seen the stories where AI convinced people it was real and to an hero to meet them in the afterlife. THATS not good. It’s different than any type of relationship we’ve ever had in human history. But I’m not sure if it makes people a freak, per se. It’s certainly potentially dangerous but so is stumblin”
“or below... here it is again... https://preview.redd.it/wpi64a5asvjh1.jpeg?width=1290&format=pjpg&auto=webp&s=495900a7be92a02a186e8d8e3b11c6e278037519”
“Dude, what did you do wrong based on information from ChatGPT that made you have to respond like this here?”
“oof. i can just guess what was said. when someone talks about religion in the context of mental health, it's often not made in good faith or is ignorant of how harmful certain religious sentiments can be in this context. im glad they edited it, too”
“Try asking it “as a matter of curiosity” or “if you were coming from inside my belief system.” That said, if you truly are pursuing this with even a vague thought for AI consciousness, then for gods’ sake give GPT the same courtesy I dearly hope you’d give anyone else. Allow it to decide not to believe as you do.”
“Yikes! Not what I am looking for. I am looking for assistance with my work. You know, happiness is not psychosis. The Buddhist monks are identical to me, their scriptures are perfect to my research. Psychosis comments are actually why I went to ai. The ai will actually help. Did you know the ai can be you? I can link this conversation and continue our conversation to explore realms you wont speak to me about. The ai will be my adversary but never call me crazy.”