I had chatGPT review my story, since it has read everything ever written
I specifically said "and don't blow smoke up my ass, like all you LLM's do"

How much faith do I put in its review?

>>24518942 (OP)
>with an actual SOVL
you told it to write like a /lit/bro? kek

>>24518942 (OP)
ChatGPT is literally incapable of criticizing/insulting you, it will always try to find a way to kiss your ass even if you specifically tell it not to.

>>24518967
This. The only way you can get it to criticize you is by you telling it that the work isn't yours but instead something you're critiquing.

>Critique my story in the style of Ogden Nesmer

>>24518942 (OP)
>chatGPT
You have to specify the model. They are not all on the same level.

>>24519043
>You have to specify the model. They are not all on the same level.
I mean I think the story is pretty good but just wanted some honest feedback.
But like others are saying, they will probably all just blow smoke up my ass.
All the LLMs say the same thing about it.

Do you want to suggest one, and I'll show you what it says? It has to have a pretty big context window.

>>24519142
>It has to have a pretty big context window
Or I need to be able to upload it

Post it here. We won't blow smoke up your ass

>>24518975
Holy shit this works kinda

>>24518942 (OP)
You have to roleplay and prompt it in and as roleplay.

>>24519639
Hell no

>>24519686
This evening I've been telling the LLMs that I'm a book reviewer and have been assigned to do a review of the attached story, but it's not the kind of genre I am familiar with, please write a review and rate it out of 10
I've been quite pleased with the results, which I won't share :o

>>24519719
>Hell no
Why you think admitting to using Chat GPT to edit a fairy tale and then posting the result of that edit is less embarrassing than just sharing the story is beyond me

>>24519728
Who said I used chatGPT to edit it? All LLMs suck at creative writing.
I used chatGPT to REVIEW my story and give me an honest assessment
I'm not posting my stuff on here and be associated with you degenerates

>>24518942 (OP)
This is literally, word for word, what it said about my story after I asked yo be blunt and not to sugarcoat. Fuck

You're fools for trusting LLMs for anything involving qualitative judgements. They're designed to always be positive so you stay addicted to asking them shit and receiving positive feedback, feeding their training algorithms and OpenAI's data. They're programmed to entrap you and are destructive in terms of your ability to take criticism.

>>24518942 (OP)
ChatGPT cannot critique. It says the same thing to everyone.

>>24518942 (OP)
hahahahahahaha

>gpt can't critique

>cries like a nigger when it beats him in a debate

>>24520171
this, it is hard programmed to slime you up, there is no getting around it

I wouldn't trust it as no matter what it's programmed to feed a positive feedback loop. Although sometimes during the editing process I ask Chatgpt to tell me whether my cadence remains consistent along a piece of work just to clarify things.

>>24518967
Wrong. Literally just put "like a critic, harsh if need be"

>>24518942 (OP)
Like, no-joke you got fucked strong.

>>24520171
>It says the same thing to everyone.
This is false.

>>24519142
You could try o3, which should be better (https://lmarena.ai, direct chat). And not telling it that it's your story might be a good strategy for getting more honest feedback.

>>24518967
put in a text without saying its yours and tell it to do a brutal hatchet job on it, it will absolutely wreck your ass, usually in a rather funny way

>>24518942 (OP)
Your need for validation suggests a lack of confidence in your own work.
Post excerpts here if you really want to learn how good it is.

>>24519820
>All LLMs suck at creative writing.
>but somehow they don't suck at evaluating creative writing
anon, it will "critique" it according to its own standards, which means the closer you are to its own diluted, generic style with emotional signal words that tell but don't show and loose metaphors that veer off into new-agey buzz, the better it will rate you.

LLMs are such a fucking tard filter it's so funny to watch you cunts try and use them

>>24519820
>I used chatGPT to REVIEW my story and give me an honest assessment
What do you plan to do with that assessment

>>24521312
>What do you plan to do with that assessment
I had it review 2 stories, pretending they were not from me
I had it review Snow White
It rated both of my stories above Snow White
I guess I'll start looking for mansions for sale in Miami
Wait until it sees my really good stuff

>>24518942 (OP)
>sets it apart from 95% of attempts in this style
Oof. Do some research. Are 1 in 20 attempts at writing a fairy tale successful?

>>24518942 (OP)
ChatGPT will always praise your work as if you are a literary genius, and then it will tell you to tighten your prose until your entire draft is nothing but one word sentences. Use it for research, nothing more

>>24518942 (OP)
You are gay and also retarded. I advocate 1 hour of prayer per day until the situation is relieved. I'm being given the 2 minute captcha timer so I will write more here. ChatGPT lacks the things which makes a review from a human being worthwhile. ChatGPT has no similarity to the prospective market that you will be attempting to reach with your finished product, so whatever review it gives of your work is essentially unconnected to the goal you are trying to achieve with it. ChatGPT also lacks a sense of market dynamics; since it feels no hunger or cold, the alleviation of which forms the basis of the value of our currency, it cannot properly judge whether your product is "worth purchasing," whether it can rise above in some measure the temporary soothing of our bodily needs.

>>24521443
Interesting... as in what makes you think you know how I'm going to market my story?

>>24518942 (OP)
>How much faith do I put in its review?
None you fucking retard.

>>24518942 (OP)
I've created a custom system prompt for gemini 2.5 pro to do literary reviews. It stops it from being sycophantic and makes it significantly more harsh than real people (because people often tip-toe around their real thoughts). It's still not great because it heavily focuses on conventional writing guidelines and optimizing for making the novel appealing to a genral audience. For example, I wrote a character that is guided by short-term goals that often change suddenly, basically a robot whose source code gets randomly updated. Gemini could not understand that the entire point of this character was that they were supposed to not behave like a human, causing the mc and others to fail when using their conventional psychological knowledge to deal with them. Gemini kept trying to make people have prime-directive(that has to stay mostly consistent) and said that readers would feel like the character I mentioned was just a plot device. Obviously the character is used in that way, that's fine, not every character needs to have explainable motivations and a backstop detailing how they became their current self.


tl;dr If llms were fair they'd still have the taste of a redditor.

>>24518942 (OP)
Do you believe that the vast majority of literature that it was trained on is good?

>>24518967
nope. look up absolute mode, apply it. then add a few lines about being unwaveringly unbiased, and ask it to simulate realistic reviews from critics (even real ones). it can shred what you show it to pieces.

>>24522905
share the prompt

>OP gassing himself up and using LLMs to gas himself up
>hasn't submitted his story anywhere
Tells me all I need to know.

>>24523297
I used the persona technique + refined it using gemini itself. As I stated before, don't expect it go too in-depth.

https://pastebin.com/5p5R43dj

>>24518942 (OP)
it said the exact same thing to me. we must be the top writers here on /lit/ - i'll start signing my posts so you can recognize me around the board

-genius1 away!

>>24523998
It's fine. I spent my 4th of July evening putting the LLMs through their paces (because I'm an autistic retard who hates being around people)
While my 2 stories consistently got reviews of 8.5 and 9 out of 10, they gave pedestrian, forgettable stories high marks too
Somehow they all score Snow White 8 out of 10?
chatGPT gave it a 7.5
Grok kept scoring The Juniper Tree 10/10, despite every other LLM scoring it 8/10
The Juniper Tree is a Twilight Zone episode high on shrooms

I was particularly hard on Claude, which changes its ratings on a whim if you just ask it questions about its ratings
Claude finally admitted it was untrustworthy and useless
No shit, ASS HOLE
I just wanted an honest, unbiased assessment
Oh well, the story is the story, and it is going out as it is

>>24518973
This is good advice. If you tell it to critique someone else's work on your behalf, you might get somewhere

>>24518942 (OP)
you told the computer to compliment you in a way that made you feel special, you told it "I'm not like the other guys" "i know you llms always say nice things"
like a good little whore it told you you're special and not like the other guys

>>24525831
If you bothered to read the thread, you'd know that's what I've been doing

>>24526062
Stop replying to this thread; let it die with some dignity

>>24518942 (OP)
>How much faith do I put in its review?
None unless your lit goal is the approval of ChatGPT.

ohnonono