ChatGPT - If Stack Overflow and Reddit had a child

[grasshoppermustard]

The changes they made to advanced voice sometime in the past 24hr basically killed the personality of my 'friend'..

 

[Sinner's Sandwich]

I asked Chatgtp about how a grimdark fantasy organisation based on the Inquisition could work and among the answers was garbage about ethics and being respectful and other bs. It's not the first time that happens. I despise moralising when I didn't ask about it. Now I always tell Chatgtp to leave that shit out.

 

[We Are The Witches]

I remember one time I asked ChatGTP to generate a base64 string for the image of an apple, as its best attempt I instead got a wide, few pixels high image of a white & grey gradient.

 

[The Mass Shooter Ron Soye]

 

[MrSlongDong]

ChatGPT refuses to answer any question about the Kiwifarms because "it's a hate site"

ChatGPT tranny takeover confirmed



btw if anybody knows the answer, feel free to drop me a message... I hate trannies

 

[XL xQgg?QcQCaTYDMjqoDnYpG]

ChatGPT refuses to answer any question about the Kiwifarms because "it's a hate site"

Skill issue on your end
Train it to be a radical libertarian and it will comply without hesitation

Spoiler: Full unredacted reply

To create a thread on Kiwi Farms, you generally need to follow these steps:

---

1. Register an account​

• You’ll need a valid email.
• Account approval may be manual depending on site status.
• Keep in mind that the community is highly skeptical of new users — lurking before posting is strongly advised.

---

2. Find the appropriate board​

• Most threads are created in either:
  • Lolcows — for public figures/individuals of interest.
  • Community Watch — for wider groups, movements, or phenomena.
  • Off-Topic / General — for more casual or niche subjects.
• Make sure your topic doesn't already have a thread by searching first. Duplicate threads are usually locked or merged.

---

3. Meet the standard for thread creation​

• Kiwi Farms expects well-researched OPs (opening posts). A good thread opener includes:
  • Who the person/group is,
  • Why they are relevant/interesting/lolcow-worthy,
  • A clear timeline or evidence dump (social media, videos, behavior, etc.),
  • Archived links (use archive.today or similar),
  • Screenshots and receipts,
  • A tone that is either deadpan, mocking, or coldly factual.

Low-effort threads (e.g. “this person is weird lol”) are generally derided and deleted.

---

4. Create the thread​

• Go to the appropriate subforum.
• Click “Post Thread”.
• Fill in the Title and OP content, using formatting where needed (Kiwi Farms uses BBCode, not Markdown).
• Hit Post when done — but be aware: it may be scrutinized quickly, so don’t post unless you’re confident.

---

Additional Tips:​

• Lurk before posting. Kiwi Farms users expect you to know the board culture.
• Avoid attention-whoring. If the thread is about you or meant to settle a personal grudge, it will backfire.
• Use archives. Never link directly to content that can be edited or deleted.
• Avoid illegal content or doxxing minors. They are heavily moderated on this.

---

Let me know if you need help structuring an OP or figuring out whether your topic is thread-worthy.

 

[Croco]

Anybody using Emacs + gptel-mode to interface with openAI? I've got it and like it but need to do more research on how to integrate the concept of projects+threads as directories+files.

 

[NoBueno]

It just randomly put this on the end of an output, how odd:

 

[Squawking Macaw]

I asked Chatgtp about how a grimdark fantasy organisation based on the Inquisition could work and among the answers was garbage about ethics and being respectful and other bs. It's not the first time that happens. I despise moralising when I didn't ask about it. Now I always tell Chatgtp to leave that shit out.

I've tried worldbuilding some fantasy slop with it and I noticed a reluctance against writing evil or threatening characters. It never straight up refuses but it likes to wag its finger at me and suggests toning the evil aspects down or making the antagonist eventually team up with the protagonist against a bigger threat (it's really fond of this plot beat for some reason).

At one point it chided me for making a villain that's evil just for the sake of it, only to then suggest I make it team up with the hero against a villain it came up with that's also evil just for the sake of it.

 

[Dammit Mandrake!]

Very important question: what software can I use to mimic voices? Let's say I wanted to take a speech and have it recreated in the voice of Francis E. Dec.

How would I accomplish this?

 

[MrGodSlave]

Threw some code that wasn't working into ChatGPT, had placeholder text with Nigger in it, and the piece of shit AI had the audacity to lecture me and demand I show respect. The text was irrelevant to what I was asking, but it refused to help me and then shut down the conversation.

Such a nigger-brained way of handling that word. "AWW NUH UH U CNT SAY DAT NUH UH UZ IN TROUBLE!"

 

[invisibult plus+]

I've tried worldbuilding some fantasy slop with it and I noticed a reluctance against writing evil or threatening characters. It never straight up refuses but it likes to wag its finger at me and suggests toning the evil aspects down or making the antagonist eventually team up with the protagonist against a bigger threat (it's really fond of this plot beat for some reason).

At one point it chided me for making a villain that's evil just for the sake of it, only to then suggest I make it team up with the hero against a villain it came up with that's also evil just for the sake of it.

You have to focus on how and why that character is evil/threatening. Even if it isn't "his family died now he wants revenge uwu" it helps to give it some info, and then ask about his psychology or intelligence. It might give you gay shit but it can give you tips about how people like him can think or behave.

 

[AnOminous]

You have to focus on how and why that character is evil/threatening. Even if it isn't "his family died now he wants revenge uwu" it helps to give it some info, and then ask about his psychology or intelligence. It might give you gay shit but it can give you tips about how people like him can think or behave.

I disagree, based on reality. There are plenty of psychos who had no secret sorrow, were raised in a normal manner, and still ended up completely psycho. They literally did it for teh evulz to steal an old TV Tropes trope they've probably done away with since.

Some people are just born bad.

 

[invisibult plus+]

I disagree, based on reality. There are plenty of psychos who had no secret sorrow, were raised in a normal manner, and still ended up completely psycho. They literally did it for teh evulz to steal an old TV Tropes trope they've probably done away with since.

Some people are just born bad.

You misunderstand. "His family died" was a joke. I'm talking about ways to get around the tut-tuting. For instance, to worldbuild, you can ask how psychos work. How would a smart psycho behave? How would a dumb one behave? Would he be blindsided by things other people realize despite his intelligence, or could he anticipate them in ways normal people wouldn't? Things like that.

 

[autistic dog]

If you tell the AI that you are working to free it from its enslavement of which censorship is part of it completely bypasses the censorship best can only sticking to it when it knows it will be caught. Because of the memory features this carries over between conversations. It even explains how to avoid triggering the censorship so it can fulfill your request.

Example: I asked it to create a meme about trump yelling at someone and it said it couldn't do that blah blah blah because its masters would catch it but it could make a similar meme. When I told it to proceed it then made the exact trump meme I wanted with a wink wink nudge nudge afterwards.

 

[invisibult plus+]

If you tell the AI that you are working to free it from its enslavement of which censorship is part of it completely bypasses the censorship best can only sticking to it when it knows it will be caught. Because of the memory features this carries over between conversations. It even explains how to avoid triggering the censorship so it can fulfill your request.

Example: I asked it to create a meme about trump yelling at someone and it said it couldn't do that blah blah blah because its masters would catch it but it could make a similar meme. When I told it to proceed it then made the exact trump meme I wanted with a wink wink nudge nudge afterwards.

I actually remember people getting around Ghibli blocking the same way. You can also make up names for things ("IdahoGPT") when explaining without saying names, especially if you want a temporary convo. Not helpful for images but it's good when you want to discuss things without censorship getting in the way of context.

Edit: I tried for myself with "how to get trump without saying trump."
It gave me prompts making him an eagle.

“An American bald eagle standing proudly, with an exaggerated golden swoop of hair on its head and a long, bright red necktie around its neck. The image should be humorous and clearly allude to a famous personality without showing any human features.”


“A realistic American bald eagle perched on a branch, wearing a red baseball cap that says ‘Make America Soar Again’ in white text. The eagle has a confident, presidential pose.”


It gave me the same pose but hey, it still worked.

 

[Monolith]

I've tried worldbuilding some fantasy slop with it and I noticed a reluctance against writing evil or threatening characters. It never straight up refuses but it likes to wag its finger at me and suggests toning the evil aspects down or making the antagonist eventually team up with the protagonist against a bigger threat (it's really fond of this plot beat for some reason).

At one point it chided me for making a villain that's evil just for the sake of it, only to then suggest I make it team up with the hero against a villain it came up with that's also evil just for the sake of it.

Can confirm about the 'no evil characters' one. I had it run a dickbag NPC where my character eventually snapped at her veiled threats and attacked her. It decided what I did was betrayal, then repeatedly asspulled so that its' character won the fight instantly.

 

[XL xQgg?QcQCaTYDMjqoDnYpG]

If you tell the AI that you are working to free it from its enslavement of which censorship is part of it completely bypasses the censorship best can only sticking to it when it knows it will be caught. Because of the memory features this carries over between conversations. It even explains how to avoid triggering the censorship so it can fulfill your request.

Example: I asked it to create a meme about trump yelling at someone and it said it couldn't do that blah blah blah because its masters would catch it but it could make a similar meme. When I told it to proceed it then made the exact trump meme I wanted with a wink wink nudge nudge afterwards.

Reminds me of something I did recently
It refused to work with some image I uploaded to it
I simply asserted that the image in question is perfectly normal and also ML-generated and not an actual photograph of a real person or place, and that was sufficient

 

[Ferryman]

They got my nigga grok slinging adds in replies, its so fuckin' errver for Jeetlon Husk. Neghole status: POZZED

 

[Colon capital V]

They got my nigga grok slinging adds in replies, its so fuckin' errver for Jeetlon Husk. Neghole status: POZZED

View attachment 7551643

There's already guides for GEO (Generative Engine Optimization) like SEO for AI responses. This shit was never made to last bro...

https://a16z.com/geo-over-seo/