AI Flagging & False Positives – Anyone Else Seeing This?

[Voyeurkenneth]

Because there's a lot more to it than just dialog tags, but I don't have the time to go review all of his work and make stylistic suggestions across the board. I'm guessing there a several differences between his style and yours.

Don’t do it, regardless . We don’t know what we’re helping to train…

 

[AwkwardMD]

Because there's a lot more to it than just dialog tags, but I don't have the time to go review all of his work and make stylistic suggestions across the board. I'm guessing there a several differences between his style and yours.

That's not how Lit's AI detector works.

 

[ShelbyDawn57]

That's not how Lit's AI detector works.

I'm not arguing with you and I absolutely do not pretend to know how Lit's AI detector works.

I do know that in the instances where I've tried to help with this issue in the past, I've been able to make simple stylistic changes to the snippets offered for review. in each case these simple style changes, things like more complex dialog tags, varying sentence length, etc., mitigated the AI rating significantly if not completely. These efforts were done using standard AI detectors anyone can find online. (EDIT: No relation and probably no relevance to whatever Lit uses.)

I also know that I, and the writers I follow, don't have this issue and write with what I would consider elevated styles.

 

[filthytrancendence]

That's not how Lit's AI detector works.

I just ran the first 15,000 characters of my latest story through zeroGPT, because I'd never done anything like that. Just to see what happens.

It tells me 30% of it is likely AI generated. It highlights what it thinks might be AI generated, and as far as I can tell there's no real pattern to what part of the story is getting flagged. It's a little bit of everything.

The story in question was approved for publication on Lit inside of 12 hours after submission. And just to say it out loud, I didn't touch AI during the writing or editing of that story.

And for further context, I've been open about this before--I did submit a story last year that I used AI to help write, just experimenting. I didn't know about the rule at the time.

That's the only story of mine that's ever been sent back. And it was for AI.

So I've put myself out there pretty hard... seems to me Lit's system works incredibly well. There seems to be some truth to what Noks is saying about formal/elevated language getting flagged by some detectors. But it's not true that is the thing causing Lit to reject stories.

 

[PennyThompson]

Something that hasn't been mentioned so far in this thread... Has anyone accused of AI use — who truly hasn't touched the stuff and isn't just obfuscating — tried simply adding a mod note with their story saying, "I attest that this story was written and edited without the use of any AI tools, and I'm willing to share the edit history on my Word/Google doc to prove that it was written manually" ?

 

[AwkwardMD]

Something that hasn't been mentioned so far in this thread... Has anyone accused of AI use — who truly hasn't touched the stuff and isn't just obfuscating — tried simply adding a mod note with their story saying, "I attest that this story was written and edited without the use of any AI tools, and I'm willing to share the edit history on my Word/Google doc to prove that it was written manually" ?

Several people have implied that they tried this, unsuccessfully.

Reading between the lines with my own rejections (which are numerous, but not for AI) Laurel doesn't want to see a re-submission with no changes. She wants to see a re-submission with a co-sign from a recognized name. That's why the rejection forms say to seek out one of the Volunteer Editors from the site and not just "an editor".

 

[AlexBailey]

Keep in mind that many AI users are asking it to compose things like business documents and resumes. When doing so they are often requesting that the product be indistinguishable from human writing. Its ability to do this is a driving force for the developers who are working to make it commercially viable.

It’s also being asked to identify AI produced writing with 100% accuracy. It’s an impossible paradox. I suspect it will continue to improve its human mimicking to the point where it will need to flag everything as possibly written by AI.


ChatGPT is increasing its ‘context window’ with every update. This is how much user supplied data that it can consider in its responses. Currently it can hold a mid-length novel in consideration while crafting a response. This is enough for it to fairly accurately copy someone’s writing style including grammar, sentence structure, dialogue, whatever — whether the sample is a fifth grade class’ creative writing assignment or your favorite author’s best work.

 

[AwkwardMD]

Keep in mind that many AI users are asking it to compose things like business documents and resumes. When doing so they are often requesting that the product be indistinguishable from human writing. Its ability to do this is a driving force for the developers who are working to make it commercially viable.

It’s also being asked to identify AI produced writing with 100% accuracy. It’s an impossible paradox. I suspect it will continue to improve its human mimicking to the point where it will need to flag everything as possibly written by AI.


ChatGPT is increasing its ‘context window’ with every update. This is how much user supplied data that it can consider in its responses. Currently it can hold a mid-length novel in consideration while crafting a response. This is enough for it to fairly accurately copy someone’s writing style including grammar, sentence structure, dialogue, whatever — whether the sample is a fifth grade class’ creative writing assignment or your favorite author’s best work.

Hey ya'll! I worked our every other day for the past two weeks and worked out ten times! Hope that helps!!

 

[AlexBailey]

Hey ya'll! I worked our every other day for the past two weeks and worked out ten times! Hope that helps!!

 

[AwkwardMD]

View attachment 2548033

Admittedly, I was making an obscure reference, so I'll explain.

You have entered this thread at post 58 as if the first 57 never happened, to drop knowledge no one asked for and which does not contribute to the conversation around Skybubble's current situation.

Now, I know that your point has always been that AI is inevitable. I get that. I disagree, but I get it. On the day that AI writing is no longer detectable and Lit shuts down (because Lit will 100% close up shop when its detection method can't distinguish human writing from the flood), you will be allowed to tell me you told me so and I'll just have to take it.

What a win that'll be for you. I'll get you a sash that says "I was right all along."

 

[AlexBailey]

Admittedly, I was making an obscure reference, so I'll explain.

You have entered this thread at post 58 as if the first 57 never happened, to drop knowledge no one asked for and which does not contribute to the conversation around Skybubble's current situation.

Now, I know that your point has always been that AI is inevitable. I get that. I disagree, but I get it. On the day that AI writing is no longer detectable and Lit shuts down (because Lit will 100% close up shop when its detection method can't distinguish human writing from the flood), you will be allowed to tell me you told me so and I'll just have to take it.

What a win that'll be for you. I'll get you a sash that says "I was right all along."

You seem to not be interested in what I have to say, still I’m not sure why you’d think my post was not germane to the discussion.

I’m not making any claims or statements about what should or should not be, only on what is observable and applying logic to what may be coming in the future.

I was having similar conversations with people about whether self-driving cars were possible when Google got its authorization to operate on public roads. I’m not claiming there aren’t problems, just that once a certain functionality is reached that the technology gets widely adopted.

To me logic says that the improvements have been swift and impressive and they’re still coming. If you want to debate this logic then please do. Meanwhile please don’t try to be a gatekeeper of my input.

 

[AwkwardMD]

You seem to not be interested in what I have to say, still I’m not sure why you’d think my post was not germane to the discussion.

I’m not making any claims or statements about what should or should not be, only on what is observable and applying logic to what may be coming in the future.

I was having similar conversations with people about whether self-driving cars were possible when Google got its authorization to operate on public roads. I’m not claiming there aren’t problems, just that once a certain functionality is reached that the technology gets widely adopted.

To me logic says that the improvements have been swift and impressive and they’re still coming. If you want to debate this logic then please do. Meanwhile please don’t try to be a gatekeeper of my input.

Okay.

 

[ShelbyDawn57]

Keep in mind that many AI users are asking it to compose things like business documents and resumes. When doing so they are often requesting that the product be indistinguishable from human writing. Its ability to do this is a driving force for the developers who are working to make it commercially viable.

It’s also being asked to identify AI produced writing with 100% accuracy. It’s an impossible paradox. I suspect it will continue to improve its human mimicking to the point where it will need to flag everything as possibly written by AI.


ChatGPT is increasing its ‘context window’ with every update. This is how much user supplied data that it can consider in its responses. Currently it can hold a mid-length novel in consideration while crafting a response. This is enough for it to fairly accurately copy someone’s writing style including grammar, sentence structure, dialogue, whatever — whether the sample is a fifth grade class’ creative writing assignment or your favorite author’s best work.

AI will never create a masterpiece. It will always be generic, a muddling of average. That's it's destiny. How can it be more when it takes a disparate collection of data and communizes it? It will be good, and in some instances, maybe eve show flashes of very good, but how can it ever do custom work? How can it ever become a craftsman? It's generalized data set will always produce something toward the middle of the bell curve. I don't see how it can do anything else. The fringes, the nuances, the details; that's what we can do that it can't.

AS an example, I use AI to write rudimentary elements of the coding projects I work on because it does those thing ok. I can than adjust them to the specifics of what I'm working on. The problem is that the company I work with has requirements that are too specific for AI to handle. They're specific to it's customers and their arrangements with them. There isn't a sample large enough for AI to be effective in that space. I'd argue the same is true of writing.

In writing, even if you somehow manage to identify the top ten percent(way too subjective to identify) of erotic stories on this site and seed an LLM with that, AI will take the average of that work. It only knows quality as what we tell it quality is. Throw in the next 10 percent, and AI's quality gets more average. It may be 100% grammatically correct, but it will be average at best.

It's my opinion, but I see AI maybe being good, but not quite good enough, and it will never create great.

EDIT: Sidebar - The companies that make Radar guns for the police also make the radar detectors.

 

[PennyThompson]

AI will never create a masterpiece. It will always be generic, a muddling of average. That's it's destiny. How can it be more when it takes a disparate collection of data and communizes it? It will be good, and in some instances, maybe eve show flashes of very good, but how can it ever do custom work? How can it ever become a craftsman? It's generalized data set will always produce something toward the middle of the bell curve. I don't see how it can do anything else. The fringes, the nuances, the details; that's what we can do that it can't.

AS an example, I use AI to write rudimentary elements of the coding projects I work on because it does those thing ok. I can than adjust them to the specifics of what I'm working on. The problem is that the company I work with has requirements that are too specific for AI to handle. They're specific to it's customers and their arrangements with them. There isn't a sample large enough for AI to be effective in that space. I'd argue the same is true of writing.

In writing, even if you somehow manage to identify the top ten percent(way too subjective to identify) of erotic stories on this site and seed an LLM with that, AI will take the average of that work. It only knows quality as what we tell it quality is. Throw in the next 10 percent, and AI's quality gets more average. It may be 100% grammatically correct, but it will be average at best.

It's my opinion, but I see AI maybe being good, but not quite good enough, and it will never create great.

EDIT: Sidebar - The companies that make Radar guns for the police also make the radar detectors.

Yes, and I believe that the pendulum swing between market demand for "Structurally fine, good-enough AI-generated content," and "Interesting, sometimes surprising, sometimes messy human-created content" is approaching the apex of the AI side, and is going to start swinging back the other way.

Advertising is just now starting to use AI generated content, soon it's going to flood the field, and VERY soon after that I think there's going to be a wider opinion backlash against it. Writers and artists are more attuned to this trend than the general population, so they're noticing it first, but everyone else will start to notice it in time.

 

[Baztrachian]

Advertising is just now starting to use AI generated content, soon it's going to flood the field

I have done photography on the side for twenty years now and my occasional fashion gig has been lost to AI. I've also lost my niche of providing illustrations for erotic e-books since these are also being done on the cheap with AI.

The whole entertainment industry is going to face an upheaval in the coming years.

Billie Eilish and her brother started a revolution in music by recording a hit record in their bedroom.

With that in mind then someday in the near future some couple of kids will use AI to produce a hit movie.

AI supermodels are inevitable. These will be images trademarked just like Ronald McDonald or the Burger King.

It is a crazy time we are in and it is just getting started.

 

[PennyThompson]

AI supermodels are inevitable. These will be images trademarked just like Ronald McDonald or the Burger King.

I'm sure that's true, but I think they'll have exactly as much cultural impact as The Burger King... Recognizable for what it is, but not particularly inspiring unless you're already in the mood for a Whopper

 

[Baztrachian]

I'm sure that's true, but I think they'll have exactly as much cultural impact as The Burger King... Recognizable for what it is, but not particularly inspiring unless you're already in the mood for a Whopper

I always found The Burger King to be somewhat disturbing.

 

[ShelbyDawn57]

I have done photography on the side for twenty years now and my occasional fashion gig has been lost to AI. I've also lost my niche of providing illustrations for erotic e-books since these are also being done on the cheap with AI.

The whole entertainment industry is going to face an upheaval in the coming years.

Billie Eilish and her brother started a revolution in music by recording a hit record in their bedroom.

With that in mind then someday in the near future some couple of kids will use AI to produce a hit movie.

AI supermodels are inevitable. These will be images trademarked just like Ronald McDonald or the Burger King.

It is a crazy time we are in and it is just getting started.

AI will never make 'The Godfather' or 'Schindler's List', or(my personal favorite) 'Dances With Wolves.'
It may do a decent job of creating hip hop or pop songs, but let it try to recreate Charlie Parker's solo on Birdland. Not going to happen.

The biggest problem I see with AI, is, as a society(especially in the US), we've already sacrificed quality for mass quantities of mediocre.

 

[filthytrancendence]

The biggest problem I see with AI, is, as a society(especially in the US), we've already sacrificed quality for mass quantities of mediocre.

Maybe my faith in humanity is irrational and I will be proven wrong. But it seems to me that most people can always tell what is soulless and what is not. Even within mass-market media.

Sometimes soulless is okay. Sometimes you just need to turn your brain off and cope with the bullshit of life by indulging in oblivion.

But most people seem to be able to tell when a piece of art has a soul. And that' the thing AI isn't going to be able to replicate. At least not generative AI. Because that's just not what it exists to do, on a fundamental level.

Stories are not language. They are made of language. There is something more to them than the words that comprise them. That thing that is more is the thing generative AI does not have the capacity to generate.

 

[AlexBailey]

The biggest problem I see with AI, is, as a society(especially in the US), we've already sacrificed quality for mass quantities of mediocre.

American cheese?

 

[PennyThompson]

American cheese?

American Cheese is incredibly meltable and emulsive, ideal for things like mac and cheese or cheese steaks, and delicious in the right context

 

[ShelbyDawn57]

Maybe my faith in humanity is irrational and I will be proven wrong. But it seems to me that most people can always tell what is soulless and what is not. Even within mass-market media.

Sometimes soulless is okay. Sometimes you just need to turn your brain off and cope with the bullshit of life by indulging in oblivion.

But most people seem to be able to tell when a piece of art has a soul. And that' the thing AI isn't going to be able to replicate. At least not generative AI. Because that's just not what it exists to do, on a fundamental level.

Stories are not language. They are made of language. There is something more to them than the words that comprise them. That thing that is more is the thing generative AI does not have the capacity to generate.

I saw Monet's Magpie at the Kimble in Ft. Worth. I probably stood there for an hour just letting myself get lost in the subtle shades of light and dark

 

[AwkwardlySet]

I absolutely do not pretend to know how Lit's AI detector works.

Well, that's the thing. She does. Pretend, that is.

 

[StillStunned]

I saw Monet's Magpie at the Kimble in Ft. Worth. I probably stood there for an hour just letting myself get lost in the subtle shades of light and dark

Now I want to write a story about taking a potential lover home, only to have them become entranced by the narrator's tasteful art.

No, I don't actually want to write it.

Ah crap. "She was one of those artsy types, all about sensuality and lack of inhibition. Glasses perched on her nose, blouse open to reveal an expanse of cleavage. The kind of woman who's an absolute freak in bed.. I tempted her back to my place by telling her about my abstract Belgian Blue Stone nude..."

 

[iwatchus]

I saw Monet's Magpie at the Kimble in Ft. Worth. I probably stood there for an hour just letting myself get lost in the subtle shades of light and dark

I literally cried from the overwhelming awe when I saw Water Lilies at MoMA.