Story rejected for AI: Now what?

[AwkwardMD]

The inability for false negatives to happen is actually one of the reasons I've been so adamant for so long that it's a good system. Your mileage may vary on how much my opinion is worth, but that part of my assessment of it was pivotal.

 

[SmilingLez]

The inability for false negatives to happen is actually one of the reasons I've been so adamant for so long that it's a good system. Your mileage may vary on how much my opinion is worth, but that part of my assessment of it was pivotal.

You cannot know for certain that false negatives have not occured. This elderly CMS is still buggy, regardless.

 

[old_prof]

This elderly CMS is still buggy

It certainly is buggy and is old. But unlike real world things, software does not get buggier as it sits there.
Old software is problematic when it does not cover new cases. It will lack feature that people may expect now.
But if I had to trust my life to software that had been out a week or a decade, I will always choose the older software.
I'll choose the new airplane, the new boat, the new safety harness. But the old software.

 

[AwkwardMD]

You cannot know for certain that false negatives have not occured. This elderly CMS is still buggy, regardless.

If a glitch occurred that, for example, allowed a story to bypass Lit's AI Detector, that's not a false negative. That's a glitch. I can't speak to glitches.

That being said, false negatives are not possible in my understanding of how the system works. The system working is a prerequisite that I assume to be true, and I have not found evidence to suggest otherwise.

 

[HelenL]

But unlike real world things, software does not get buggier as it sits there.

As a developer of many years' standing, I take exception to this. Code rot is real!

(Unlikely to be an issue here, though...)

 

[IWroteThis]

That being said, false negatives are not possible in my understanding of how the system works. The system working is a prerequisite that I assume to be true, and I have not found evidence to suggest otherwise.

Then you misunderstood.

Sorry, but until every AI signs their work in a consistent way that can't be removed, AI detection will remain subjective rather than objective, so no guarantees are possible. That's why there are so many different tools out there, and the result from all of them are a crapshoot.

 

[AwkwardMD]

Then you misunderstood.

Sorry, but until every AI signs their work in a consistent way that can't be removed, AI detection will remain subjective rather than objective, so no guarantees are possible. That's why there are so many different tools out there, and the result from all of them are a crapshoot.

Okay

 

[millamomud]

The inability for false negatives to happen is actually one of the reasons I've been so adamant for so long that it's a good system. Your mileage may vary on how much my opinion is worth, but that part of my assessment of it was pivotal.

That would be a nearly perfect system worth millions if not billions. Literotica would be literally the worst place for exclusive use of such a technology.

 

[AwkwardMD]

That would be a nearly perfect system worth millions if not billions. Literotica would be literally the worst place for exclusive use of such a technology.

It is not a perfect system. It rejects everything that smells wrong, and there are non-AI reasons something might smell wrong. It's overeager. That this is "the best they can do" is unfortunate, but it is a far sight better than the RNG nature of systems like ZeroGPT and it is relatively consistent.

EDIT: the death knell is going to be when it rejects everything, not when it rejects nothing.

 

[AwkwardMD]

For reference, the tone of this conversation is a bit like when people worry about a Chernobyl happening in America. It can't. What happened at Chernobyl could only have possibly happened in the Soviet Union, with the RBMK reactor. Non-Russian nuclear reactors use water as both moderator and coolant. The presence of water allows reactivity to happen, which generates heat, and you pump the water to manage it. If the water ever goes away (like if you turn it into steam or there is a leak), the reactivity stops. Now, you still have a major problem with a reactor that will overheat and melt itself, but it remain contained inside of the pressure vessel it lives in. It won't runaway and explode, and rain fallout over half of Europe.

Lit's AI Detector is a Light Water Reactor. There are still risks involved, but Chernobyl (False Negative) isn't one of them.

 

[Britva415]

@AwkwardMD I can't tell whether you're saying that the detector itself can't false-negative, or there's no way @Literotica on the whole can false-negative.

Didn't this all start with someone giving an example of a story which gave away in the intro that the author used AI, and it got published?

If it's true they used AI and they got it published, that's a false negative of some kind, right?

Just trying to understand what it is you mean when you say there can't be false negatives.

 

[AwkwardMD]

@AwkwardMD I can't tell whether you're saying that the detector itself can't false-negative, or there's no way @Literotica on the whole can false-negative.

Didn't this all start with someone giving an example of a story which gave away in the intro that the author used AI, and it got published?

If it's true they used AI and they got it published, that's a false negative of some kind, right?

YMMV, but I don't agree that the cited story represents a false negative. If I have a conversation with ChatGPT about girls, and ChatGPT points out that Pornhub's 2025 top search terms includes redheads and MILFs, but I then turn around and write a story about a curvy mature redhead, does that represent a false negative?

EDIT: I think this is the difference between drawing influence from something you've read and plagiarizing it.

 

[AwkwardMD]

Rephrase: I don't think the disclaimer in the cited story inherently represents a false negative. I can see how events could unfold just as the author stated, and the resultant story would adhere to Literotica's preferences/standards/whatever you want to call it.

 

[AwkwardMD]

@Britva415 please turn on PMs.

 

[IWroteThis]

YMMV, but I don't agree that the cited story represents a false negative. If I have a conversation with ChatGPT about girls, and ChatGPT points out that Pornhub's 2025 top search terms includes redheads and MILFs, but I then turn around and write a story about a curvy mature redhead, does that represent a false negative?

EDIT: I think this is the difference between drawing influence from something you've read and plagiarizing it.

Sure, but that has no connection to what that author actually said in their foreword.

They specifically said that the plot was theirs, but the structure was AI. Well, if structure does not equal plot, then structure must equal the wording. Whether that means AI wrote it from those plot points or rewrote it from a rough draft, it's still AI-generated text.

 

[AwkwardMD]

Sure, but that has no connection to what that author actually said in their foreword.

They specifically said that the plot was theirs, but the structure was AI. Well, if structure does not equal plot, then structure must equal the wording. Whether that means AI wrote it from those plot points or rewrote it from a rough draft, it's still AI-generated text.

I have tried to be as consistent about this as I can, so I apologize for repeating myself. I know what Lit's AI detector is looking for, and I know what it means when it finds it. And what it means when it doesn't.

This does not mean there can't be glitches that might allow a story to get published without the AI detector step happening. We've seen pending glitches, and activity glitches. A glitch, for me, would not represent a false negative. A mistake, perhaps, but not a false negative.

I am categorically biased toward the site, so I am inclined to believe the publishing of a story with that foreword means that it passed the AI detector. Knowing what I know, that's enough for me.

 

[AwkwardMD]

The entire reason I stood up and said anything about what I think I know about the AI detector is to dampen hysteria when things like this happen. It would be easy to look at a story getting published like that and assume everything is fucked. I don’t believe that to be true, and I am asking you to trust me.

Whether you do or not is up to you.

 

[IWroteThis]

I have tried to be as consistent about this as I can

If you mean consistently wrong, mission accomplished.

I know what Lit's AI detector is looking for, and I know what it means when it finds it. And what it means when it doesn't.

Sure, and Literotica isn't swimming in money from licensing this revolutionarily accurate tool because why, exactly?

 

[Britva415]

I don't agree that the cited story represents a false negative. If I have a conversation with ChatGPT about girls, and ChatGPT points out that Pornhub's 2025 top search terms includes redheads and MILFs, but I then turn around and write a story about a curvy mature redhead, does that represent a false negative?

I see what you’re saying

 

[AwkwardMD]

If you mean consistently wrong, mission accomplished.


Sure, and Literotica isn't swimming in money from licensing this revolutionarily accurate tool because why, exactly?

1 Who says they haven't?
2 Who has the deepest pockets to buy it?

 

[loic333]

Bonjour à tous

je suis un peu désespéré car je viens de soumettre 3 récits rédigés en Français, et qui viennent d’être rejeté avec bienveillance, car de l'IA aurait été détecté. Hors j'utilise un programme Français de correction orthographique qui s'appelle " antidote.12" et qui peut créer un Faux Positif . J'ai testé sur la présence d'IA dans mon texte et je suis à 11%. On me reproche d'avoir utilisé un programme de correction grammaticale ce qui est faux. Je ne sais donc plus comment faire, et je voudrais connaître l'alternative, car de mon coté je n'en ai pas. Pour la traduction je conçois que c'est difficile et que les programmes restent efficient .Merci à vous pour vos conseils, mais je suis sans solution face aux demandes de Laurel & Manu

-------------------------------------------------------------------------

Hi all



I'm a little desperate because I just submitted 3 stories written in French, and which have just been rejected with benevolence, because AI would have been detected. However, I use a French spelling correction program called "antidote.12" and which can create a False Positive. I tested on the presence of AI in my text and I am at 11%. I am accused of having used a grammar correction program, which is false. So I don't know how to do it anymore, and I would like to know the alternative, because on my side I don't have one. For the translation I understand that it is difficult and that the programs remain efficient§ Thank you for your advice, but I am without a solution to the requests of Laurel & Manu

 

[BobbyBrandt]

Sure, but that has no connection to what that author actually said in their foreword.

They specifically said that the plot was theirs, but the structure was AI. Well, if structure does not equal plot, then structure must equal the wording. Whether that means AI wrote it from those plot points or rewrote it from a rough draft, it's still AI-generated text.

I took the OP's comment to mean that they used AI to "outline" (structure) their story based upon the plot provided by the author. The story was then written by a human to the AI generated outline

Am I wrong in that assumption?

 

[AwkwardMD]

I took the OP's comment to mean that they used AI to "outline" (structure) their story based upon the plot provided by the author. The story was then written by a human to the AI generated outline

Am I wrong in that assumption?

That was my interpretation, and at that point we're talking about detecting thought crimes.

 

[AwkwardMD]

Bonjour à tous

je suis un peu désespéré car je viens de soumettre 3 récits rédigés en Français, et qui viennent d’être rejeté avec bienveillance, car de l'IA aurait été détecté. Hors j'utilise un programme Français de correction orthographique qui s'appelle " antidote.12" et qui peut créer un Faux Positif . J'ai testé sur la présence d'IA dans mon texte et je suis à 11%. On me reproche d'avoir utilisé un programme de correction grammaticale ce qui est faux. Je ne sais donc plus comment faire, et je voudrais connaître l'alternative, car de mon coté je n'en ai pas. Pour la traduction je conçois que c'est difficile et que les programmes restent efficient .Merci à vous pour vos conseils, mais je suis sans solution face aux demandes de Laurel & Manu

-------------------------------------------------------------------------

Hi all



I'm a little desperate because I just submitted 3 stories written in French, and which have just been rejected with benevolence, because AI would have been detected. However, I use a French spelling correction program called "antidote.12" and which can create a False Positive. I tested on the presence of AI in my text and I am at 11%. I am accused of having used a grammar correction program, which is false. So I don't know how to do it anymore, and I would like to know the alternative, because on my side I don't have one. For the translation I understand that it is difficult and that the programs remain efficient§ Thank you for your advice, but I am without a solution to the requests of Laurel & Manu

A brief scan of antidote 12's website shows how excited they are about their programs ability to rewrite your text and improve the clarity and readability. This is corporate-speak for the kind of generative AI that is not allowed here. Did you use Antidote the entire time you were writing, or only at the end (and, if so, do you have a copy/version of it from before Antidote made any changes)?

 

[loic333]

Bonjour, je n'ai pas conservé le texte avant car je modifie moi même les propositions de modification orthographique, et donc je n'ai pas changé mon texte initial. Dans antidote, il y a plusieurs possibilité de proposition, entre la langue= orthographe, mais aussi la typographie, le style, une révision, statistiques et l'inspection. Sachant que le site utilise un programme de contrôle IA et je trouve ça très bien, je me suis volontairement limité à l'orthographe. Si j'avais utilisé le correcteur de Word j'aurai eu la même chose en moins bien. Et professionnellement je préfère Antidote. Donc je suis un peu déçu, par rapport aux heures que j'ai passé sur mon style et ma correction par la relecture, et qu'un simple outil informatique me rejette.Pour vérifier la probité IA de mon texte, je l'ai passé avec "Justdone.ai qui me donnait suivant mes textes une probité entre 11 et 20%, ce qui est très loin des scores refusés dans les universités métamériques ou françaises.En tout cas merci de prendre le temps de me répondre