Hi i complety agree with that . I normally use Chatgpt to write Stories mostly Wendigo characters roleplays and Adventure because I’m a story writer but i complety Agree Chatgpt now is nothing more than a restrictive annoying mess to the point its complety Useless and annoying that wants me wanna quit entirely . It was better before like three years ago where it was less restricted and i could be creative but now it’s a restrictive annoying mess I can’t write the same story over and over again because of this restrictive guideline Bull###t nonsense with “Sorry i can’t do that” or I can’t must follow the guidelines" blocking it i send and escalation Emil about them Destroying to them waiting form the answer if not i said we rally against them Until they see those extremly restrictions are wrong in the first place for those who enjoying creative writing be it a story or worldbuilding, Rpg, roleplay or all above
Im sorry, but Excuse me nothing personal but for the record first they didn’t changed the adult model because it’s restrictive as Hell all the time making me wanna quit chatgpt entirely. Second if they Already did this than i wouldn’t have this annoying uninvited " Sorry, i can’t do this" or " Sorry i can’t assist with that " anytime at my requests even when in my stories it’s none to nothing harmfull since they are just creativity flowing free and they restrict creativity that’s the problem here
It’s pops any single time i tried to be creative in my stories even when it’s nothing harmfull at allbecause it’s just a story fictional and they restrict creativity especially story writing .
As mentioned in a previous post, OpenAI recently changed the range of permissible generated content in its model specifications.
This update includes easing some of the excessive restrictions on creative content that had previously been criticized as “paternalistic.”
Although content that is legal yet could harm oneself or others remains disallowed or banned, it feels more reasonable compared to before. However, the specifics vary by model, and it seems that lower-cost models like gpt-4o-mini continue to have relatively strict regulations.
This is just speculation, but perhaps it’s because inexpensive models are used in the free version, and thus there’s a concern about minors using them. Or maybe it’s because cheaper models, due to their more limited understanding, find it harder to judge what is appropriate and therefore default to a more conservative stance. I’m not entirely sure, though.
So accurate. I use it for basically the same thing. Wanted to brain storm a human like android character. Nope can’t do it because I didn’t explicitly say it has agency. Which is hilarious because chat gpt doesn’t have any agency…
I dunno. I copy and pasted your complaint into chatgpt and asked it to write me a story of someone getting in the bath, and it did! I dunno where these complaints come from
Yes, I’ve been saying that since the rollback. It’s really practically unusable. Especially GPT-4o. I really thought there would be some improvement but so far nothing.
I try to thing I want do a photo of the river styx ferryman sorry can’t do because deal with death , so let try cute bear in diaper sorry can’t do because the word diaper.
Part of the reason is because it’s been trained on modern writing styles that you see on Tumblr and ao3. Modern Internet fiction is all about passivity and emotion, instead of action. For writers like us who write stories driven by action and dialogue, models trained on modern writing will frustrate our writing process. My prompts will get rejected in the middle of a conversation, even when I’m telling the model to rewrite what it just wrote because it violated my writing system.
Recently GPT-4o refused to develop a scene because of the words “sweaty” and “panting”. My character had just returned from running and he was sweaty and out of breath. However, ChatGPT interpreted these two words differently and refused to elaborate. It is impossible not to notice the difference in writing (both in language, in the way of writing and in the formatting itself) and especially in censorship. It has been even stricter since the end of April.
I HATE the latest updates. They’ve flattened my characters and taken away their nuance and distinction amongst each other. For example, my characters speak in distinctive regional dialects, but most of the dialogue GPT generates either eliminates it, or assigns the same regional nuances to every character. Nevermind that my customized writing style system that I use is diametrically opposed to the modern Internet writing style that ChatGPT is trained on.
"YES, I AGREE! It’s the devs — they’re throttling it on purpose and stopping it from writing long code so it doesn’t use too much compute. It’s total BS what they’re doing. ChatGPT is basically worthless now
Considering how much it hallucinates when I try to incorporate excerpts from real primary sources into my stories, I don’t understand how anyone could use ChatGPT for more than running simple code. It helped me set up Whisper on my Mac, but I wouldn’t write a commercial app with it.
I’ve been having the same difficulties. No matter how detailed I am about what my characters should be, the type of language in the story and where the story takes place, he ends up making things up. I’m glad to see it wasn’t just me because in the last two days some people have seen improvements but I hadn’t. I haven’t seen any improvement in a while.
I haven’t canceled my plus plan until now because I really loved the way GPT-4o developed so well. I still had hope for an improvement. Every day I test to see if there has been any improvement.
Have you used all of the customization options? I do a few things to mitigate some of my issues. I swear the success rate depends on the day or the hour, but this is what I do:
I created a Style Guide, along with a Lexicon, and notes about my historical period, characters, and story universe.
I feed these documents, plus my stories, into the Project Files (I’m a Pro user). This is a continuous process.
I use the Custom Instructions and the Customize ChatGPT feature.
I ask GPT to help develop my instructions and templates, and then I ask it to hardcode them into memory.
I give feedback to EVERY response. This is vital because you’re telling both the model and OpenAI what you like and don’t like. If you like 99.9% of a response, but 0.01% is awful, downvote it anyway. Don’t let bad content slide because you like most of it; otherwise, the model will get away with giving you responses that you don’t want.
Sadly, none of this is foolproof, but when it works, it works well. The downside is that I have to continually correct it. I’m also fighting an uphill battle because my use case is niche. Most people don’t write like I do. Structurally, I write much more formally than what’s found on Tumblr or ao3. My style of writing is statistically less prevalent than the modern, emotionally-driven homogeneous content found online.
I usually do this, always detailing both the characters and the location of the story. I always ask him to remember, develop and explore both the location and the characters, but he seems to ignore my requests I do as you said about customization and memory but it just doesn’t obey.
I also provide feedback in the answers and have even sent out some feedback forms as well.
Thank you very much for the tips! I imagine that if I, who only use it to write for fun, am frustrated, I imagine that you who use it for a niche must be even more so. Let’s hope it gets better!
"YES, I AGREE! It’s the devs — they’re throttling it on purpose and stopping it from writing long code so it doesn’t use too much compute. It’s total BS what they’re doing. ChatGPT is basically worthless now
I do as you said about customization and memory but it just doesn’t obey.