"AI" Rejection

[grenefire]

Envelope icon at top right, start new conversation, put Laurel as recipient.

Thanks!

 

[Cali_Love]

Grammarly doesn't change anything. You can accept or reject changes its suggestions.

My rant overwrote my intended point and my feeble attempt to help. Surely something is happening as AI flags seem to be on the rise. It is only supposition, but I believe that Easter eggs are being added to a document by proofreading tools whether they admit it or not. Especially if the tool has been upgraded recently, these tools have come out of the closet and embrace being AI. Maybe you can turn that part off maybe it doesn't turn all the way off like one thinks. It used to be simpler, like Grammarly would just add a hidden html tag in the background for all suggestions, taken or not. Maybe it's more sinister now.

It's currently a real problem for professors (probably more angst for students). I read about a prof who wrote a test paper, ran it through his filter and after getting a 0% AI score, ran it through Grammarly, where it then filtered to 99% AI.

So here's what the academics are suggesting to students and this technique could work here on Lit. Save a pre-proofreading-tool version of your story and if the edited final gets kicked back for AI, you can prove it is not.

CL

 

[jaF0]

Might be an interesting question to ask any and all that post this kind of rejection.

Did you use Grammerly or anything similar in any way during your writing of the rejected story?

 

[alohadave]

Might be an interesting question to ask any and all that post this kind of rejection.

Did you use Grammerly or anything similar in any way during your writing of the rejected story?

I use Grammarly and haven't had anything kicked back. I also copy and paste the text into the submission field, so it's pure text that is being submitted.

Maybe that is something to try. Copy and paste the text instead of uploading a file.

 

[djrip]

My rant overwrote my intended point and my feeble attempt to help. Surely something is happening as AI flags seem to be on the rise. It is only supposition, but I believe that Easter eggs are being added to a document by proofreading tools whether they admit it or not. Especially if the tool has been upgraded recently, these tools have come out of the closet and embrace being AI. Maybe you can turn that part off maybe it doesn't turn all the way off like one thinks. It used to be simpler, like Grammarly would just add a hidden html tag in the background for all suggestions, taken or not. Maybe it's more sinister now.

It's currently a real problem for professors (probably more angst for students). I read about a prof who wrote a test paper, ran it through his filter and after getting a 0% AI score, ran it through Grammarly, where it then filtered to 99% AI.

So here's what the academics are suggesting to students and this technique could work here on Lit. Save a pre-proofreading-tool version of your story and if the edited final gets kicked back for AI, you can prove it is not.

CL

If grammarly or any tool is hiding tags or anything like that in a document, just cut and paste the plain text, that should lose anything hidden in there that isn't actual changes to the text, if I'm understanding the issue correctly.

 

[MillieDynamite]

My rant overwrote my intended point and my feeble attempt to help. Surely something is happening as AI flags seem to be on the rise. It is only supposition, but I believe that Easter eggs are being added to a document by proofreading tools whether they admit it or not. Especially if the tool has been upgraded recently, these tools have come out of the closet and embrace being AI. Maybe you can turn that part off maybe it doesn't turn all the way off like one thinks. It used to be simpler, like Grammarly would just add a hidden html tag in the background for all suggestions, taken or not. Maybe it's more sinister now.

It's currently a real problem for professors (probably more angst for students). I read about a prof who wrote a test paper, ran it through his filter and after getting a 0% AI score, ran it through Grammarly, where it then filtered to 99% AI.

So here's what the academics are suggesting to students and this technique could work here on Lit. Save a pre-proofreading-tool version of your story and if the edited final gets kicked back for AI, you can prove it is not.

CL

That would be a watermark. But if it is there, it should show up if you turn on show markup, so you can delete those markers. I doubt there is any kind of marker for that in Word, whether you're using Grammarly or not. What would be the purpose of such a tool, to highlight whether Grammarly or ProWriting Aid was used? I can't see them doing that. If a writer looks at it, see it, he can assume Grammarly or ProWriting Aid wants to out him for using AI. Wouldn't that be a bone of contention for a writer purchasing the software?

 

[alohadave]

That would be a watermark. But if it is there, it should show up if you turn on show markup, so you can delete those markers. I doubt there is any kind of marker for that in Word, whether you're using Grammarly or not. What would be the purpose of such a tool, to highlight whether Grammarly or ProWriting Aid was used? I can't see them doing that. If a writer looks at it, see it, he can assume Grammarly or ProWriting Aid wants to out him for using AI. Wouldn't that be a bone of contention for a writer purchasing the software?

I'll look in some of my Word files to see if Grammarly left something hidden.

 

[MillieDynamite]

I would, but I'm far too lazy right now.

I'll look in some of my Word files to see if Grammarly left something hidden.

 

[SyleusSnow]

ProWritingAid does not leave anything behind in documents. Besides, that's not how these worthless "AI detection" services work. They look at the text itself and compare it to content known to have been generated by ChatGPT and others.

OpenAI launched a detector back in January but have since taken it down for being highly inaccurate (it only guessed correctly 26% of the time). These over-hyped services are causing problems for writers of all kinds:

https://authory.com/blog/how-ai-detectors-are-destroying-livelihoods

 

[jaF0]

Don't use machine generated or assisted text.

It' really just that simple.

 

[SyleusSnow]

"For every complex problem there is an answer that is clear, simple, and wrong."
- H. L. Mencken



The thing is, it isn't just that simple. These AI detectors are unreliable snake oil. If OpenAI, the folks who make ChatGPT can't themselves make an accurate detector, it's unlikely any of the bazillion "me too" detection services (one of which Lit has probably subscribed to) are going to do any better.

 

[XerXesXu]

ProWritingAid does not leave anything behind in documents. Besides, that's not how these worthless "AI detection" services work. They look at the text itself and compare it to content known to have been generated by ChatGPT and others.

OpenAI launched a detector back in January but have since taken it down for being highly inaccurate (it only guessed correctly 26% of the time). These over-hyped services are causing problems for writers of all kinds:

https://authory.com/blog/how-ai-detectors-are-destroying-livelihoods

I'm sorry to hear about that anonymous, experienced content provider who lost his job. It's a chilling warning to us all.

Does that guy sell his 'certificates', per piece? I'm glad to know it's still possible to do righteous work and make money doing it.

 

[Bamagan]

I suppose we all mostly assume that Laurel is using some kind of AI detection software, but it's entirely possible that she's just using the old-fashioned, "This looks like it was written by a middle-schooler" eye test and kicking it back as suspicious. Certainly quicker and cheaper for her.

 

[QueenOfFarts]

I wonder whether Grammarly is part of the problem. While it is a useful tool, it does risk standardising stories; I find I reject about a third of the suggestions because they either make things too formal or actually make things worse.

I don't think so. As an IRL software engineer, I can tell you that Lit is probably just using an AI detector, and those are snake oil. I can throw tons of human-written content into all of the big ones and flag it.

All it's doing is comparing the probability of your choice of the next word in a sentence to the same probability transformer that an LLM uses. If the results are similar, you get flagged. This is pretty silly, because LLMs build those transformers based on what humans have already written, so you're essentially just confirming that your story is something that average person would have come up with. Maybe a minor insult to your creativity if you want to take it that way, but certainly not a hallmark of AI.

Honestly anyone selling these AI "detectors" are just profiting off of fear. They don't work, they essentially can't work, but people are scared enough of AI that they'll pay for them anyways.

 

[djrip]

I ran AI Busted against my story and found some parts that were marked as 100% AI generated. Those were the ones I added or rewrote late to flesh out the sex scenes.

I also asked it to score a text I generated with an AI and used several tools like DeepL against and it scored 100% human written.

I rewrote one of the parts from scratch. But only after I added mistakes like:
"That's my girl," he said with satisfied voice.
instead
"That's my girl," he said with a satisfied voice.

The score reduced the probability of being AI written. This change got it down from 40% AI to 2% AI. That's plain redicolous.

Don't add errors to avoid AI detection. That's ridiculous. Don't change the story at all, if you truly wrote it without AI assistance. Just resubmit with note to editor explaining you didn't use any AI assistance.

 

[SyleusSnow]

Don't add errors to avoid AI detection. That's ridiculous. Don't change the story at all, if you truly wrote it without AI assistance. Just resubmit with note to editor explaining you didn't use any AI assistance.

Exactly. That's the approach to take. Let Lit know about the false positive, and hopefully they'll eventually stop using whichever snake oil AI detection service they've bought into.

When experimenting with these detection services, using tiny snippets of text will give invalid and inconsistent results. Feed them a couple paragraphs at a minimum. But that's also a fool's errand: each one uses different engines behind the scenes, and those engines can change at any time.... text detected as AI-written today may be marked as 100% real the next. Even the detection service OpenAI themselves produced was less accurate than a coin flip.

(An amusing aside: the original snake oil patent medicine wasn't even real snake oil. It was "mineral oil, a fatty oil believed to be beef fat, red pepper and turpentine". We already know the ingredients of these AI detection services: 100% bullshit )

 

[djrip]

I did that and mine got rejected again xD Resubmitted it a few days ago so I'm just waiting for a response now

I can't guarantee results! I'm just philosophically opposed to letting an AI write your story in the sense of changing it around to satisfy an AI detector which is in fact an AI.

 

[MillieDynamite]

Friend, either you're closing your eyes to a situation you do not wish to acknowledge
Or you are not aware of the caliber of disaster
Indicated by the presence of AI in our community
Well, ya got trouble my friend
Right here I say, trouble right in Literotica, why sure, I'm a writer
Certainly, mighty proud I say, I'm always mighty proud to say it
I consider that the hours I spend with a pen in my hand are golden

Help you cultivate horse sense and a cool head and a keen mind
Did you ever take and try to find and iron-clad ending for yourself
From a three-plot mystery story?
But just as I say it takes judgement, brains, and maturity to write
In a contest a worthy story, I say that any boob
Can take and shove a description in an AI window
And I call that sloth the first big step on the road to the depths of deg-ra-dation
I say, first, a teaspoon of AI enhancement, then a story from the magic window

 

[SyleusSnow]

Friend, either you're closing your eyes to a situation you do not wish to acknowledge
Or you are not aware of the caliber of disaster
Indicated by the presence of AI in our community
Well, ya got trouble my friend
Right here I say, trouble right in Literotica, why sure, I'm a writer
Certainly, mighty proud I say, I'm always mighty proud to say it
I consider that the hours I spend with a pen in my hand are golden

Well, if you submit that to Lit as poetry, ya got trouble:

AI busted: 42% Human Written, 58% Bot Written
Scribbler: 69% Probably generated by AI
AI-detector.net: "There is low risk that your text was written by AI."
www.writecream.com detector: Percentage: 9%
smodin.io: 0% Likelihood of Complete AI content
Me: 100% Meredith Willson homage

 

[MillieDynamite]

It wasn't AI. It was creative plagiarism with a flair and used under parody and fair use exception.

Well, if you submit that to Lit as poetry, ya got trouble:

AI busted: 42% Human Written, 58% Bot Written
Scribbler: 69% Probably generated by AI
AI-detector.net: "There is low risk that your text was written by AI."
www.writecream.com detector: Percentage: 9%
smodin.io: 0% Likelihood of Complete AI content
Me: 100% Meredith Willson homage

 

[SyleusSnow]

It wasn't AI. It was creative plagiarism with a flair and used under parody and fair use exception.

Impossible! Both AI Busted and Scribbler's detection engines say it was largely written by AI. And we all know AI always tells the truth.

Besides, it's way too clever to have been written by a mere human.

 

[FieryLion]

So from what I can gather from this thread: don't submit any stories to Literotica in the near future because it will get rejected for being ai-generated due to AI detection on text being almost as bad as random chance and it's an extremely frustrating experience to go through. Also gonna grab some popcorn to watch all the witch-hunting.

 

[SoCalAshley]

My latest story just got rejected for AI content. Just like many of the previous posters, I used a grammar program to help me with my writing. I am not here to debate whether grammar programs are AI or an extension of the creativity of the writer. I will leave that to minds more legally capable than mine.

I am posting to this thread because I believe a well-written story is in the best interest of not only the reader but also the Literotica site.

If the owners of the site cannot appreciate that, then perhaps it is time to stop writing for this site and go to a place where well-written stories are appreciated.

 

[BelleCanzuto]

If the owners of the site cannot appreciate that, then perhaps it is time to stop writing for this site and go to a place where well-written stories are appreciated.

I think the owners definitely DO appreciate a well written story. This seems to be a very recent issue (just the past few months, really). As others have suggested, I'd recommend resubmitted with a note that you wrote it yourself, and maybe mention the grammar checker thing. And also PM Laurel with the same info.

Most people who have resubmitted (it appears) have gotten approved, and I think the more people she hears from, the more likely it is that she'll tweak whatever process she has for evaluating AI.

 

[SoCalAshley]

I think the owners definitely DO appreciate a well written story. This seems to be a very recent issue (just the past few months, really). As others have suggested, I'd recommend resubmitted with a note that you wrote it yourself, and maybe mention the grammar checker thing. And also PM Laurel with the same info.

Most people who have resubmitted (it appears) have gotten approved, and I think the more people she hears from, the more likely it is that she'll tweak whatever process she has for evaluating AI