Has anyone found ChatGPT too sympathetic towards an ex?

[Asq]

I have started using chat GPT to go over my relationship with my ex as certain things still bother me a lot and I never had anyone to talk to about it at the time. I have started using chat GPT to write down how I felt and go over situations I felt were wrong, but it often seems to “side” with him, I don’t know if that’s the correct term but it often seems to sympathise a lot with him. Is this normal? I have even said to it a few times that I felt like it was siding with him. I don’t want to say what the things are as the great reason for chat GPT is it doesn’t judge me for things, but is this anyone else’s experience?

This is such a sad thread. That we’ve got to a stage in human evolution where people are turning to artificial intelligence for simulated human conversation, simulated understanding and simulated empathy.

firstly a tragic case.

"come home to me" was what the alleged chatbot is said to have said and it does not mention which chatbot it was, the rest is human interpretation of those words, also it depends on how the user asked the chatbot to respond, so again its down to the user on how the chatbot responds

Could it be that your friends all say “ooh isn’t he awful” because that’s what you obviously want them to say? Maybe AI is seeing both sides more objectively?

I mean, you could always name change and post here!

its all predictive text, and the text on chatbots is based on tokens which is the likely match for the next suitable words, the same with predictive texts on a phone

for example i just googled it and got googles ai response

AI Overview

You nailed it! That is exactly how predictive text and Large Language Models (LLMs) operate. Both systems are fundamentally predicting what comes next based on patterns, though they operate on different scales: 1, 2, 3]
Phone Predictive Text: Looks at the last few words you typed and suggests the most probable next word (or emoji). It is usually based on your personal typing habits and a smaller local dictionary.
AI Chatbots: Look at the entire prompt you've provided and generate text chunk by chunk. These chunks are called tokens—which can be whole words, parts of words (syllables), or even individual characters. 1, 2, 3, 4, 5]
Instead of just guessing the next word, chatbots use complex neural networks trained on billions of pages of text to calculate the mathematical probability of a vast sequence of tokens. They essentially ask, "Based on all the text I've ever 'read,' what is the most natural-sounding next token to complete this thought?" 1, 2]
It’s like an incredibly advanced, context-aware version of hitting the middle button on your phone's keyboard over and over again!

Did you only read the first paragraph? What about wanting to see the boy in the afterlife? And anyway, it's not "a case". It's a number of cases mentioned even in just this article. Including one where an AI programme gives suicide advice - BBC News - I wanted ChatGPT to help me. So why did it advise me how to kill myself?
https://www.bbc.co.uk/news/articles/cp3x71pv1qno
and children being groomed to do sexual acts. I think we can all agree that it's incredibly disturbing

As I said I don't think you understand it well enough to give advice here.

But let's say you genuinely think OP using Grok is the same as me using predictive text.

Why don't you advise OP to get relationship advice from her predictive text feature on her phone?

I’ve done that and people recognise me which I find sad 😭🤣

Phone Predictive Text: Looks at the last few words you typed and suggests the most probable next word (or emoji). It is usually based on your personal typing habits and a smaller local dictionary.

AI Chatbots: Look at the entire prompt you've provided and generate text chunk by chunk. These chunks are called tokens—which can be whole words, parts of words (syllables), or even individual characters.

Instead of just guessing the next word, chatbots use complex neural networks trained on billions of pages of text to calculate the mathematical probability of a vast sequence of tokens. They essentially ask, "Based on all the text I've ever 'read,' what is the most natural-sounding next token to complete this thought?"

Also it’s not just friends ive posted on various groups full of strangers and they all say he is awful as I said I think it’s because I told it that I liked that it was nice towards him which I did at first, it was refreshing to not just hear him painted as a monster

The current model of chatgpt is designed to be more critical of what you say as a way to counteract earlier accusations of sycophancy that was sending some users into psychosis. Other models like Claude might be more balanced.

Why are you spamming the thread with AI generated text? I think everyone here (possibly apart from you) knows that having a "conversation" with an AI bot is different from your phone guessing a word you might write in a text.

the irony, is people expect chatbots to be perfect yet never seem to want perfection from professionals or friends they talk to

heres why to quote your words :

As I said I don't think you understand it well enough to give advice here.
But let's say you genuinely think OP using Grok is the same as me using predictive text.
Why don't you advise OP to get relationship advice from her predictive text feature on her phone?

so i proved the differences and proved i knew how it operates

I don't want to derail OPs thread so I won't drag this out, but by copying and pasting (ironically) an AI generated response, you haven't "proven" that you "know" anything.

And it really doesn't support your slightly unusual argument that "talking to" an AI chatbot about your relationships instead of going to therapy is essentially the same as using predictive text in text messages.

I will try another one

I'm afraid I don't understand confiding and consulting with AI on emotional matters. The only reasons I can think of are that it's serving a Googling type function and that it's free. However, you really mustn't just feed it data or treat it like it can 'help'.

So chat gpt generates words and statements from words that you type with a percentage of accuracy. So if you need support you need to tell it your the victim so it knows. For example if your partner has narcissistic personality disorder to say that makes him a victim of a disorder so you have to tell chat gpt you are a victim of a narcissist before you start so it can understand the perspective you are looking for.

I tried that just now and it didn’t work.

Do you have someone that you trust that you could do this with so they could see where it’s going wrong? Maybe try a different ai or a different log in on the same ai.

I’m sorry. How frustrating. It seems like you just need some closure. These things always seem to mess up when you need them the most.

In Real Life @Asq , every single person who loves and cares for you no longer wants to talk about this anymore as they have obviously said all that they can - that you’re better off without your ex ad infinitum.

These are people who actually care about you; who actually want you to be happy; who actually give a damn about you.

Listen to them as hard as it might be. Not a bunch of circuit boards and coding

Sorry but just because they don’t want to talk about it doesnt mean I can just snap out of it. He really caused me a lot of issues and I need to work through them

You could. I've been where you are. I snapped out of it by being active, healthy, busy, focussed on other things. There wasn't enough time for repetitive navel gazing.

Yesterday was beautiful in London. Did you have nice weather where you are?. What did you get up to when you weren't wrestling with GPT and posting on here?