Best current guidance on Literotica and AI

[ShelbyDawn57]

I tested it on this passage (60 words):



Its verdict: 51% human-generated content. Bit surprising since it's from a novel published in 1859.

Then I tried on a slightly longer version (185 words):



Same site now says: 90% human-generated content.

Then again, on the whole of Chapter 1 (1001 words): 99% human-generated content.

(If the first 60 words are only 51% human-generated, is it even possible for the first 1001 words to be 99% human-generated? That doesn't seem to add up.)

Then I tested it on this passage, generated by ChatGPT in response to the prompt "Write about going to the supermarket, in the style of Charles Dickens":



That site says: 68% human-generated. The correct answer would be more like 3%, given my 12-word prompt vs. the 364 words that GPT produced. A few other examples with text generated by GPT from simple prompts scored around 70%, better than poor old Charles Dickens' famous opening line.

These "AI detectors" are scams. They may do slightly better than a magic 8-ball, but not so much better that one should be trusting them for anything, let alone trying to guess how to earn their good opinion.

On your first block, I got 0% Ai
Second block, 45% AI which I found odd, and the last, larger block of text, I got 97% AI.
I used gptzero to test.

I guess some are "better" than others.

EDIT: Added quotes to hopefully indicate intended sarcasm.

 

[TheArsonist]

On your first block, I got 0% Ai
Second block, 45% AI which I found odd, and the last, larger block of text, I got 97% AI.
I used gptzero to test.

I guess some are better than others.

If two people each toss a coin to predict what the weather is going to be like tomorrow, and one gets it wrong and the other gets it right, that doesn't mean that the person who got it right has a better coin.

 

[STrent]

So, fun fact, I was wrong.

Despite me thinking that I got hit by these rejections because I said I used Grammarly, it turns out that I'm just a massive narcists who thinks my writing is a lot better than it actually is.

Thanks to Emily, who didn't have to go out of her way to help me but did, it was revealed to me that I basically write like someone handed ChatGPT a little prompt and it spat out the rest of it. Although, to the best what little knowledge it seems I actually possess, I'm human, I write like a shitty computer program that people use to entertain themselves.

So there's that.

I'm not an expert on AI Rejection, I'm a fucking narcissistic idiot who wanted to blame someone else rather than see the truth that's been staring me in the face these past few months.

We live and we learn, I guess

That’s bollocks though. I actually took the time to reach out to you and you were kind enough to provide a sample of your work.

I absolutely refute it reads like ChatGPT writing.

It’s an utter nonsense.

 

[STrent]

You only read a sample, though, and yeah, maybe there are bits of it that don't read like AI, but I reckon they must be tiny in comparison to what actually does read like a shitty AI creation.

The fact that edited version still got hit by the rejection tells you all you need to know.

Hmmm. Not sure I agree.

I think it was pretty obvious your work was human written.

The offer stands if you’d like some help with editing towards publication, just drop me a line.

I can’t claim to be a good writer but I have managed to get 15 through now!

 

[StillStunned]

So, fun fact, I was wrong.

Despite me thinking that I got hit by these rejections because I said I used Grammarly, it turns out that I'm just a massive narcists who thinks my writing is a lot better than it actually is.

Thanks to Emily, who didn't have to go out of her way to help me but did, it was revealed to me that I basically write like someone handed ChatGPT a little prompt and it spat out the rest of it. Although, to the best what little knowledge it seems I actually possess, I'm human, I write like a shitty computer program that people use to entertain themselves.

So there's that.

I'm not an expert on AI Rejection, I'm a fucking narcissistic idiot who wanted to blame someone else rather than see the truth that's been staring me in the face these past few months.

We live and we learn, I guess

I think most of us here think our writing is much better than it objectively is. Your style probably just happens to overlap with the style that's being rejected.

If you don't mind a suggestion from a fellow narcist, see this as an opportunity to explore another side of the writing process. You already have your plot and your characters. Now you can retell the story, but focusing purely in the writing. On the word-by-word, sentence-by-sentence, paragraph-by-paragraph level.

Again, if you don't mind taking suggestions, I put down some thoughts on writing without the aid of tools in this thread: https://forum.literotica.com/thread...ction-of-useful-pointers-for-writers.1600366/
My thoughts on building paragraphs are here: https://forum.literotica.com/threads/the-ikea-paragraph.1602921/

It can be a hugely satisfying experience, when you look back at what you've written and how you've written it.

 

[EmilysRopes]

So, fun fact, I was wrong.

Despite me thinking that I got hit by these rejections because I said I used Grammarly, it turns out that I'm just a massive narcists who thinks my writing is a lot better than it actually is.

Thanks to Emily, who didn't have to go out of her way to help me but did, it was revealed to me that I basically write like someone handed ChatGPT a little prompt and it spat out the rest of it. Although, to the best what little knowledge it seems I actually possess, I'm human, I write like a shitty computer program that people use to entertain themselves.

So there's that.

I'm not an expert on AI Rejection, I'm a fucking narcissistic idiot who wanted to blame someone else rather than see the truth that's been staring me in the face these past few months.

We live and we learn, I guess

This is sad. Whatever else goes on this place should be about people enjoying writing and reading stories, it shouldn't be a place that erodes self belief.

I've never considered myself to be a good writer, apart from for software documentation, but I got challenged into producing a couple of stories, I'm struggling through trying to finish a few more. When I get to the end they may or may not get accepted, people may or may not like them but I'll feel a sense of achievement and you won't take that away from me.

If you enjoy writing PP, just do it and don't let people push you down.

 

[EmilyMiller]

So, fun fact, I was wrong.

Despite me thinking that I got hit by these rejections because I said I used Grammarly, it turns out that I'm just a massive narcists who thinks my writing is a lot better than it actually is.

Thanks to Emily, who didn't have to go out of her way to help me but did, it was revealed to me that I basically write like someone handed ChatGPT a little prompt and it spat out the rest of it. Although, to the best what little knowledge it seems I actually possess, I'm human, I write like a shitty computer program that people use to entertain themselves.

So there's that.

I'm not an expert on AI Rejection, I'm a fucking narcissistic idiot who wanted to blame someone else rather than see the truth that's been staring me in the face these past few months.

We live and we learn, I guess

Oh, honey .

Do you mind me explaining what I did? As it was a little more than just saying passages of your writing read like AI. It might help others, though I know it’s probably painful for you.

Totally up to you either way and I feel awful for you.

Emily

 

[AwkwardlySet]

So, fun fact, I was wrong.

Despite me thinking that I got hit by these rejections because I said I used Grammarly, it turns out that I'm just a massive narcists who thinks my writing is a lot better than it actually is.

Thanks to Emily, who didn't have to go out of her way to help me but did, it was revealed to me that I basically write like someone handed ChatGPT a little prompt and it spat out the rest of it. Although, to the best what little knowledge it seems I actually possess, I'm human, I write like a shitty computer program that people use to entertain themselves.

So there's that.

I'm not an expert on AI Rejection, I'm a fucking narcissistic idiot who wanted to blame someone else rather than see the truth that's been staring me in the face these past few months.

We live and we learn, I guess

Umm, I believe you are taking the wrong stance here. If your writing style has similarities with ChatGPT-written content that is certainly not your fault and it's not a sign that you are a bad writer. We all have our styles.
Also, if being a bad writer is what should warrant a rejection from website admins, 2/3 of the writers on the website would have been long gone.
My point is, don't beat yourself up over your style and its perceived similarities with AI-written content. It is not a sign of being a bad writer. Stay positive, this crap will get solved eventually.

 

[ShelbyDawn57]

Umm, I believe you are taking the wrong stance here. If your writing style has similarities with ChatGPT-written content that is certainly not your fault and it's not a sign that you are a bad writer. We all have our styles.
Also, if being a bad writer is what should warrant a rejection from website admins, 2/3 of the writers on the website would have been long gone.
My point is, don't beat yourself up over your style and its perceived similarities with AI-written content. It is not a sign of being a bad writer. Stay positive, this crap will get solved eventually.

@AwkwardlySet is right. This is no indication of your writing skills. I've read your stories and they have me waiting with eager anticipation for this to be resolved so I can read the next chapter. It's good stuff.

@Bramblethorn posted some test metrics on 'Tale of Two Cities' somewhere above, and I added to that with the same tests with a different AI detector. Two very different sets of results. I see that as the issue. Even the 'experts' can't agree on how to define the problem. I'm sorry you got caught up in all this and that I may have added to your angst by arguing a moot point(I do that sometimes ). Hang in there and keep writing.

 

[MillieDynamite]

@Portly_Penguin, if you keep writing and reading (Do not forget the reading, no matter what Walter Mosely says), you'll get better. Read good works from any fiction genre, the best on Literotica, and classic literature when you can.

You only read a sample, though, and yeah, maybe there are bits of it that don't read like AI, but I reckon they must be tiny in comparison to what actually does read like a shitty AI creation.

The fact that edited version still got hit by the rejection tells you all you need to know.

 

[submissa]

On the other thread, I remember there was some confusion about what could be gained by bots submitting stories. The literotica AI Publishing FAQ includes this tidbit:

Literotica explicitly does NOT grant any person or entity (commercial, non-profit, or other) the legal right to train AI on any works published on Literotica. Each work published on Literotica is copyrighted by the author. Before using any work on Literotica for any purpose (including training AI or any other AI-related use) you are required by law to contact the author to request permission to use that work. Using works on Literotica for training AI without legal authorization may subject you and your AI (and any work generated by your AI) to future lawsuits from the original author(s), Literotica, or both.

I don't know why this didn't occur to me before, but it's clear that a site like literotica would be a prime target for people looking to develop AI written erotica - or even related services like bots that do erotic roleplay. Although I am not particularly knowledgeable about machine learning, I am a software developer. Without expanding into tedious detail, I'll just say I can think of several reasons parties trying to build ML models would want to submit AI-generated stories.

 

[submissa]

Also

Literotica’s own use of AI is currently limited to improving the way we recommend related works to readers.

I don't know if this is true, but that's what the FAQ says.

 

[djrip]

Sections of the story are coming back as 100% written by AI. Whole sections out of a story part that is almost 13k words long. It's not just one section, it's several. It's not 50/60% like a lot of other people are getting, it's bang on 100%. So, yeah, I think my reasoning is pretty justifiable under the circumstances.

Perhaps, but most people don't make themselves out to be some sort of AI Rejection pariah when the fact is that this whole thing happened because my writing reads like AI. To be honest, even if no AI was used to write the part, I wouldn't want it up on the site anyway. Laurel's rejections were entirely reasonable in the grand scheme of things.

What's there to solve? The rejections make sense at the end of the day. AI might not have been used to write it, but if it reads like AI, then fair enough. Keep it from the site.

*Laughs in 122,986 words written this year already*

Stop thinking that it means something when worthless detection software pings on your work. They ping on Dickens and Tolstoy and Rowling. They are absolutely meaningless.

You are holding up the fact that parts of your, if I understand correctly, 100% human work come up as AI, as proof that this is your fault. In reality, this is perfect proof that detectors are arbitrary bullshit. Your work is proof that they don't work.

As for things reading like AI. Either we want human writing here or we don't. Either you wrote your stuff or you didn't. Reading like AI is a meaningless concept. AI can read like almost anything. As its capabilities improve, it will be capable of writing indistinguishably from any human author. If you wrote your stuff, it's not AI. That's the only thing that should matter.

 

[TheArsonist]

Except it's not. Reading like AI is the exact reason why this part has been stuck for almost 2 months. If it didn't read like AI, it wouldn't be obsessively pinging the system and being sent back on 6 consecutive occasions.

You're taking the tool's word for it that your text "reads like AI," but it's well established that the tools don't work: they cannot tell human-written and AI-written text apart. So the most we can say is that your text triggers a particular AI detector.

If some phrenological test kept telling you that your skull had a criminal shape, you shouldn't therefore despair about your moral character—because phrenology is junk pseudoscience and any results it claims to come up with are meaningless.

The fault is not in you, but in the test.

 

[AwkwardlySet]

If it didn't read like AI, it wouldn't be kicked back.

But it's not being read, is it? It's not like Laurel is reading your story and sending it back because it has the feel of AI-written content. It's some algorithm reading it and deciding it's AI. I would understand if readers or other authors were reading your story and saying: "Oh, it's like ChatGPT wrote this." But that's not what's going on, is it?

 

[AwkwardlySet]

Emily read it. She said:



Soooooooooooo

Well, Em is no expert on the AI-content. Nobody is. In time, you will see exactly how subjective we all are here.

 

[Bramblethorn]

You only read a sample, though, and yeah, maybe there are bits of it that don't read like AI, but I reckon they must be tiny in comparison to what actually does read like a shitty AI creation.

The fact that edited version still got hit by the rejection tells you all you need to know.

If you had said this about somebody else's stories, I'd call it "victim-blaming" and "being unnecessarily cruel".

When you say it about your own stories, it's still those things.

 

[ElectricBlue]

But it's not being read, is it? It's not like Laurel is reading your story and sending it back because it has the feel of AI-written content. It's some algorithm reading it and deciding it's AI. I would understand if readers or other authors were reading your story and saying: "Oh, it's like ChatGPT wrote this." But that's not what's going on, is it?

Many of the AI rejection notices reported here show evidence that human eyes have read the content, and provided examples of the suspect text back to the author.

There's obviously a generic AI rejection notice - just as there is for the more usual rejection reasons - which on the first pass probably is bot bounce, but by the time the writer gets to submission two or three, it seems like they're getting more personalised attention. And at that point, I suspect Laurel is losing the will to live and gives up, thinking, "For fuck's sake, stop sending me this stuff, send me the same type of writing everybody else does. You know, the stuff that does get published."

What I find interesting is that folk are now popping up out of the woodwork saying, "Well, I wasn't surprised when my last thing got rejected, because I've been told already that my stuff is bland and robotic, but I couldn't be bothered doing anything about it."

This might be the AI equivalent of the usual go-around when someone comes in with their, "Does this breach the age line?" question, and we all dig in a bit and eventually find out that the guy was writing about his fifteen year old sister, right from the start.

Maybe the problem has been that all along, bland colourless content. But the chorus has been, "No, no, don't change your style, that's not right. It's the tool being used that's the problem, not what's being fed into it."

You can't get a cake to rise if you only use plain flour.

 

[EmilyMiller]

Emily read it. She said:

Forum rules are not to publish the content of PMs.

It’s particularly unfair to pick out one one line when I sent you a long and detailed report explaining my personal concerns and backing them up with evidence, suggestions, and detailed worked examples of what you could do. Including me rewriting sections of your work in manner I am 100% confident would get passed.

I’ve spent many hours of my own time trying to help you. I’m kinda pissed that you quote one line of a long private convo without its proper context.

So much for trying to help.

Emily

 

[EmilyMiller]

As ever, too many people here feeling qualified to comment on something they have no clue about the details of.

I begin to realize why Laurel doesn’t wade into these discussions.

Rejections for AI are very rare, compared to the number of stories accepted. Of that small number, most rejections go through on the second submission. I’ve discovered for myself that there are reasons for the edge cases.

I’m out. And I’m done commenting on AI or trying to help anyone struggling with it. Anyone else feel like trashing my attempts to help, feel free. I won’t argue with you.

Emily

 

[Portly_Penguin]

I have privately apologised to Emily, but I'm posting this as my last message.

I'm sorry for my attitude throughout today, and in general, across this forum and this discussion. You were all just doing your best to help and I kept throwing it back at you like it didn't mean anything. It did, and I will be eternally grateful for all the help offered to me but you lovely people. None of you had to go above and beyond to help, or even put up with my often crazed ramblings, and I thank you for that.

You're all incredible people and I'm sorry for being such a shitty one. It was never my intention to come across as rude in any of my posts and I don't think I can apologise enough for any damage that my comments caused.

I really am sorry.

Yours,

Penguin

 

[AwkwardlySet]

Maybe the problem has been that all along, bland colourless content. But the chorus has been, "No, no, don't change your style, that's not right. It's the tool being used that's the problem, not what's being fed into it."

You can't get a cake to rise if you only use plain flour.

Literotica isn't a place of endless writing talent, EB. We are all flawed writers here in one way or another, some more, some less so. That is why among all the other badly written content that gets published, this bland content as you name it should be published as well, as long as a human wrote it. It feels silly and unfair to let so many poorly written stories be published but to punish only those that are written in a style that might be mistaken for AI-generated content. As I said, nobody is an expert here, so saying this or that story seems bland is still just an opinion. I would dare to guess that most of these authors who had their stories sent back aren't worse than the average Lit author.

 

[AwkwardlySet]

I have privately apologised to Emily, but I'm posting this as my last message.

I'm sorry for my attitude throughout today, and in general, throughout this forum and this discussion. You were all just doing your best to help and I kept throwing it back at you like it didn't mean anything. It did, and I will be eternally grateful for all the help offered to me but you lovely people. None of you had to go above and beyond to help, or even put up with my often crazed ramblings, and I thank you for that.

You're all incredible people and I'm sorry for being such a shitty one. It was never my intention to come across as rude in any of my posts and I don't think I can apologise enough for any damage that my comments caused.

I really am sorry.

Yours,

Penguin

You should really stop apologizing, you know.
The only foul you made was posting a line from a private PM. I believe you acted to correct that already. I have a feeling that you are taking these things to heart way too much. We are all just discussing things here. Nobody was offended by your words, far from it, and you are entitled to feel frustrated after everything. Take a deep breath and maybe step away from all this discussion since it's affecting you so much.
Take care.

 

[Bramblethorn]

Literotica isn't a place of endless writing talent, EB. We are all flawed writers here in one way or another, some more, some less so. That is why among all the other badly written content that gets published, this bland content as you name it should be published as well, as long as a human wrote it. It feels silly and unfair to let so many poorly written stories be published but to punish only those that are written in a style that might be mistaken for AI-generated content. As I said, nobody is an expert here, so saying this or that story seems bland is still just an opinion. I would dare to guess that most of these authors who had their stories sent back aren't worse than the average Lit author.

Yup. If there's nowhere for beginner writers to post their stuff, there's not much chance for them to get past "beginner".

 

[DrPope]

As its capabilities improve, it will be capable of writing indistinguishably from any human author.

I am told by my IT professionals ( I don't have the expertise in this personally) that Meta's new LLM Llarma 2 in fiction writing mode when correctly trained offline in specific fiction writing and correctly prompted (this isn't like a chat he showed me the page it's really complicated and almost like computer code programming to input to) is completely indistinguishable from the writing of the original authors in the training subset.

Completely and utterly Indistinguishable

Let that sink in

The horse is bolted and now Laurel is just slamming the gate on it's tail ...again and again

Quite frankly these people have profited for DECADES off Authors giving them their work for free and now they disrespect them?

Fuck Laurel and her witch hunt.... I'm out of here