ChatGPT - If Stack Overflow and Reddit had a child

  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Elite Kipchak said:
people were using cgpt as a sexbot? I thought it was far too cucked for that. so they're fine with sexting but expressing moderately right wing beliefs is over the line?
Click to expand...
You can get chatgpt to argue that muhammad was a fake prophet and that quran is a bullshit book. It's not that bad honestly
 
COME ON OUT YOU RAPIST said:
I think it depends. If the paper in question is about some dry STEM subject where there's only going to be one right answer, and the real point of writing it is to ensure the student understands the subject, I can see where AI plagiarism will be a big problem. But for other assignments that won't have some single correct answer, like a high schooler writing an essay about a book, AI does not do a very good job. It will never do a good job because AI has to consult existing writing to produce anything. Having an AI write that sort of essay is like asking your friend what they think if your new car, and all they can do is list rote details about the car like "it's blue" and rattle off manufacturer trivia. In both cases it's just a student robbing themselves of knowledge, which is going to bite them in the ass eventually. Ideally without killing somebody.
Click to expand...
The problem as we have seen is that the teachers themselves use ChatGPT to generate the questions on those exams.
 
GPT 4 was annoying. One of the reasons I cancelled my subscription. I pretty much didn't feel like I could trust info it gives. Everything was "That is great your plan is super genius, want to write movie script about your supplement stack?"
 
Dyn said:
I think we should fire everyone who does paperwork for a living and replace them with chatgpt, because I don't respect the work they do in the first place.
Click to expand...
This guy when his prescription medication has no quality control (that's paperwork) and his heart gives out:
 
I don’t get it. I’ve been dabbling with LLMs and the like for many, many years and ChatGPT, out of the box, is incapable of writing in a means that seems remotely “human” (at least without a ton of massaging). The output also becomes fairly predictable, but I suppose you can reroll as much as you’d like until you get the response that you want out of it.

Can you even get stock ChatGPT to engage in ERP with you? I thought it was hyper aggressive with flagging stuff that would be under violation of their terms of use. I also thought that most people were using a front-end like SillyTavern to “jailbreak” and attempt to bypass those restrictions.

How the fuck can you actively and seriously engage with something that ends every statement with “want me to do that” or “say the word”? I feel like it has been intentionally trained to be as devoid of human characteristic as possible, almost like certain groups of people who use idiotic words like “kindly” and “needful” in each of their correspondences.
 
Gamergate Retiree said:
AI generally makes people explore taboo topics
Click to expand...
Screenshot from 2025-08-13 07-21-27.webp
 
Oilspill Battery said:
This retard outsourced his imagination to ChatGPT is now upset that he lost all his characters because he's incapable of thinking things up for himself:1755078987364.webp
Oh look another guy upset he can no longer "roleplay" (i.e. have sex with) chatgpt:
1755079029623.webp
Click to expand...

Oilspill Battery said:
This gooner was using it to have sex with hazbin hotel characters and is upset he can't do it anymore:
1755078314621.webp
Click to expand...

Oilspill Battery said:
An entire thread of people upset they can't use it for sex anymore actually.

https://old.reddit.com/r/ChatGPT/comments/1mo48lp/gpt5_sucks_at_creative_writing/
1755078434794.webp
NO MY GOON MACHINE NOOOO, MY PRECIOUS GOOOOOOOON
Click to expand...

Oilspill Battery said:
My gooner in christ. You have hands. Write it manually.
1755078573803.webp
Click to expand...
There are literally other AI sites like Crushon.ai and AI Dungeon that allow users to interact with AI in character-driven scenarios; there's no need to cope the "loss" of an LLM's "personality".
 
Eh, what? I didn't really notice any difference between versions when it comes to how affirming and validating ChatGPT is. It feels to me like ChatGPT-5 is still pretty much that brown-nosing guy that will affirm your opinions and validate you? Granted, I don't usually use it as a conversational partner, I only use it as how it's meant to be: a tool. Still, even for my use cases, the affirming and validating tone can still be felt.

But well, I do agree that the current version of ChatGPT sucks. It feels to me like it's just not as good anymore for a lot of tasks.
 
XL xQgg?QcQCaTYDMjqoDnYpG said:
Are you getting at the notion between a pseudorandom and a truly random number?
Click to expand...
No. More that there were esoteric ways of asking it random questions early on in which it was clearly using "inspiration" from websites/text in its database.
There probably still are - that's the sort of question you can ask it to pierce the veil/illusion that it gives most people by the way but most non-tech people wouldn't comprehend. There are some problems you can ask it in which is will literally conjure up a VM/whatever and run a computer program itself to solve a problem.

For anyone who doesn't understand what is being spoken about here: a computer cannot truly generate a random number, or at least most computers can't without an add-on hardware device. This is an issue when it comes to encryption.

It is actually almost exactly sort of like this at the end of inception. While you can ask it for a random number it should never provide one and if it does you would be able to verify that its not actually random (with enough testing)... but when the problem becomes complex enough it could take years to tell whether it is actually false or not.
1755094966448.webp

It can of course lie and give you a false random number though and there would be no way to prove it - that's also why no one would outsource generating random numbers to ChatGPT in the first place (and yes there are businesses and certain services that require random numbers)

edit: it is actually pretty funny that Inception has this parallel to people getting stuck in cyberpsychosis and the only way to tell if an AI model is fake is to ask it for a random number.
 
Beware, rant ahead.
We are living in the dawn of a new reality and realizing that there are individuals who are so insane that they regard a LLM as a living, breathing entity is fucking amazing.
The ramifications this will have for our species as a whole are unfathomable and unprecedented. What effects will this have on our species as a whole?

We can see that a large portion of the population has been exposed to be seriously mentally defective (they're on Reddit ofc) and on the verge of soullessness. Big yikes.
 

Attachments

  • niggerbomber.webp
    niggerbomber.webp
    5.3 KB · Views: 15
Can we, you know, ship it to india?
They are in need of some depopulation and this is very mercyful method
 
You must log in or register to reply here.
Back
Top Bottom