Are Stories Being Pulled for Using AI?

As everyone knows by now, you can't just simply produce erotica written by an AI since they have restrictions around those. So you first have to jailbreak it, then guide it along to write the story you have in mind. If you leave the AI to write an erotic story, even with guidelines, the end result is a very short story with bog standard sex.

Jailbreak? What are you talking about? You can just download GPT-2 and train it yourself with writing prompts taken from Lit stories. Though, admittedly, that would completely defeat the purpose of having an easier time creating stories, since it would take A LOT of work preparing those writing prompts.

But, that's not even the problem with AI "assistance".

As I see it, writing porn isn't hard. If I take a look at the daily submission in any category and on any given day, at least half of it is masturbation fantasies without a story or plot. They, more or less, go straight to the sex, not bothering to establish characters beyond describing the D-Cup breasts and fit bodies the women got through yoga, and the ten-inch schlongs and six-pack stomachs the guys got from playing high school football (or some other cliché). So, the hard thing to write seems to be what happens in between the romps. And for that, you can use GPT-4 without a problem.
 
If you are having to use an AI to come up with parts of your story, then should you be writing? Exactly how much of your story needs to be your creation rather than that of an AI and still be yours.
You don't get to tell me how I write my stories.

Bye bye.
 
You don't get to tell me how I write my stories.

Bye bye.

I mean... honestly? You clearly said that you only used AI to come up with a chant. As long as the rest of the story is yours, that's not really that different from using AI to check your grammar (Grammarly), or MS Word's thesaurus function. But it is still a valid question in general. How much AI-generated text IS usable to still call it YOUR story? Keep in mind, AI can't come up with anything new yet. It can only reword stuff it was fed beforehand. So... even that chant is something it took from someone else, albeit not verbatim.

Especially if you immediately go all snitty over it, making it seem like he hit a nerve or something.
 
Last edited:
Especially if you immediately go all snitty over it, making it seem like he hit a nerve or something.
Should I have gone back and forth with him for 7 pages before blocking him? I'm not going to waste my time with someone who comes out of the gate with a personal attack.

If you are having to use an AI to come up with parts of your story, then should you be writing?
This right here is the bullshit I responded to. I'm not a songwriter, and I feel no shame or embarrassment about using it to make that part of my story. If this person, who has a history of shitstirring, has a problem with it, it's their problem not mine.
 
I tried to get AI to write a poem for me. I tried to give some specific instructions about the style, which I thought it ignored until I read it. Here are a few extracts.

With trochaic verses, I ensnare their minds,
In anapestic verses, promises that binds,
They yield to me, surrender to my snare,
In trochaic rhythm, their lives I tear.

With trochaic rhythm, I cast my siren's spell,
A dance of whispers, a tale to weave and tell,

Anapestic whispers wrap around their souls,
Entwining dreams and bending wills like coals,

In stanzas long and short, the tale unfolds,
Of how their destiny, my grasp enfolds,

Anapestic chants, they hear within the wind,
As trochaic notes of longing, dreams unpinned,

I'll let you all guess what instructions I gave it :ROFLMAO::ROFLMAO::ROFLMAO:
 
Jailbreak? What are you talking about? You can just download GPT-2 and train it yourself with writing prompts taken from Lit stories.

Training a large language model is not a casual DIY kind of thing. OpenAI have been cagey about exactly what it cost to train GPT-2, but it's likely to have been in the tens of thousands of dollars just for computing time.

You'd also need a LOT of training data. GPT-2 was trained on a data set of about 10 billion tokens, which would be approximately the size of the entire Literotica story archive.

(In case of confusion, "prompting" and "training" are not the same thing.)
 
ChatGPT wrote this for me. I'm serious.

Velvet touch on your face,
Smooth and fresh, quick pace,
In every shave, embrace,
The silky grace,
Burma Shave.
 
Should I have gone back and forth with him for 7 pages before blocking him? I'm not going to waste my time with someone who comes out of the gate with a personal attack.


This right here is the bullshit I responded to. I'm not a songwriter, and I feel no shame or embarrassment about using it to make that part of my story. If this person, who has a history of shitstirring, has a problem with it, it's their problem not mine.

I wouldn't be comfortable doing this myself, even for just a few lines of chanting. But I think you're very likely right that that particular post/er didn't bode well for chances of productive discussion, and you did the right thing by being open with Laurel about it.

Every story ever published draws on things the author didn't invent, whether it's a Shakespearean turn of phrase or a historical figure who seemed interesting enough to write about or whatever. Some of those uses are generally accepted as okay, some generally taken as unacceptable, some like this one are much more grey. Approaching this with a mindset of "stories are 100% pure original or they're nothing" isn't helpful in resolving that.
 
Training a large language model is not a casual DIY kind of thing. OpenAI have been cagey about exactly what it cost to train GPT-2, but it's likely to have been in the tens of thousands of dollars just for computing time.

You'd also need a LOT of training data. GPT-2 was trained on a data set of about 10 billion tokens, which would be approximately the size of the entire Literotica story archive.

(In case of confusion, "prompting" and "training" are not the same thing.)

Yes, but you don't have to train it from scratch. You can download GPT-2, fully trained, as it was when they switched over to the next version. You "only" have to train it to... expand its capabilities towards literary porn.

And, yes, I also agree that it is not a casual DIY project. This is why I wrote before that it would completely defeat the purpose of having an easier time writing books if you first have to create all the needed writing prompts.
 
Lots of flavors of ai writing assistants out there, some with few content restrictions. Personally, I love writing erotica up to the point of actual boinking, which is excruciatingly boring for me to write. Having a junior writer cover that part would be grand. Fun to write about how the protagonists went from strangers to naked in the sheets, after that, blech. But readers demand the 'after that' portion of the story, where primal urges combine in a symphony of senses, as a (checks notes) testament to their love
This is interesting -- I personally enjoy the sex part the most. And that's the part that AI is going to be the worst at, because it's never done the things it's writing about. Sure, most of the stories on here are fantasies that people haven't ever experienced, but IMO the best sex scenes on here are written from experience. Or, at least, inspired by experience.

And then there's the aspect of drawing from visual porn. Unless a human were to write out a description of each and every sex act that is depicted, AI wouldn't be able to draw inspiration from that. Not to mention, as human authors, we naturally focus on the parts that are most interesting to us, rather than each and every detail in a scene.
 
I wrote the first paragraph, then ran it through the AI and got the second paragraph. In-terrresting –

I tested some descriptive-non-erotic paragraphs with CHAT-gpt3, asking for improvement of various sorts. What came back was flat and lifeless, for the most part, formulaic, but the vocabulary was occasionally better. Admittedly, it was to dabble and see what it was about. I don't have a handle on what type of prompts to tell it in order for it to work 'effectively.' I should have asked it what prompts it would like to get. ;) When I left the session, I found my own versions more satisfying. I didn't have the heart to tell the AI that ... just told it 'good job' and left it at that.:ROFLMAO:

I experimented with some non-explicit descriptive paragraphs using CHAT-gpt3, seeking various improvements. The responses I received were mostly dull, often following a predictable pattern, although the vocabulary occasionally showed improvement. I must admit, my intention was more to explore and understand its capabilities rather than to provide it with optimal prompts. Perhaps I should have inquired about the type of prompts it preferred. When I concluded the session, I found my own versions to be more satisfying. I couldn't bring myself to inform the AI of this; instead, I simply praised its effort and left it at that. 😂

Now, back to me - I'm laughing now because Grammarly wants to 'reword' three of the AI's sentences as I look at this!
 
Hum, I don't think it actually plagiarizes, though. Supposedly, it learns by the examples of their writing, as do most authors. We learn from what we read.
I’ve played around with such a tool and it is 100% plagiarism. In the middle of the story, it drops the content of an ad, for example. I also received a “thank you for reading my story” note of an author as output. And a hint to purchase a subscription to read the full story 🤣
 
I don't know when, but the sitewide content guidelines have been updated and now include this:



To that end, we DO NOT publish works of any type featuring the following content:

Works generated by artificial intelligence (AI), large language models (LLM), or other non-human automated systems. Literotica publishes works by human for humans. For more information, see our AI policy.


https://www.literotica.com/resources/content-guidelines




There appear to be a number of other wording changes that I don't remember seeing before.
 
So apparently this place is officially the "Literotica Author Support Forum". Who knew we were doing such important work?
 
There appear to be a number of other wording changes that I don't remember seeing before.
The new page on the "not quite an AI Policy but it will do" went up earlier this year, I reckon.. The other content policies seem to be a little more precise, too.

The "no overt political content policy" has been around for a long while now.
 
ChatGPT wrote this for me. I'm serious.

Velvet touch on your face,
Smooth and fresh, quick pace,
In every shave, embrace,
The silky grace,
Burma Shave.
That reads like the kind of poem you'd see in an episode of Family Guy where Brian and Stewy end up in a basement cafe full of Beatniks and Shatner is on stage - reading that - rather than what an actual poet would write.
 
Back
Top