The AI Fiasco

[nice90sguy]

Al Fiasco was probably the worst hit-man the mob ever had.

 

[AlexBailey]

Me:
What characteristics of an author’s fiction writing would generate a false identification of having been written by AI when checked by AI?

ChatGPT:
An author's fiction writing may generate a false identification of being written by AI if it exhibits a consistent lack of emotional depth, overly structured patterns, or an unnatural and rigid style that deviates significantly from typical human variability. Additionally, an excessive reliance on clichés or an absence of nuanced creativity might contribute to a misclassification by AI.

————-




It’s a bizarre situation. On one hand AI is being requested to identify AI writing, on the other it is being requested to make AI writing indistinguishable from human writing.

It reminds me of Captain Kirk vs. NOMAD.

 

[NoTalentHack]

I don't say this to attack you

That's a blatant lie, since you've been taking shots at me for basically the entire time you and I have interacted. Hell, you've been doing that since before we interacted, when you decided, based on a note by another user, that I was prolific enough for you to target me, hateread my stuff, and feed it through your (as I've demonstrated, laughably inaccurate) tool of choice over two days. But we'll skip that for now.

View attachment 2305064

By your own admission, you rely on AI to conceive of changes to your final document, whether it's a single word, sentence, paragraph or page. Changes which are entirely conceived by the AI, not you as the writer. Changes which you allow the software to make on your behalf, after having first been tuned to the "Romance" setting. Then, you run a data visualisation to tabulate the words which you've used most often, to prop up an unsophisticated vocabulary, followed by more adjustments in pursuit of raising your "style score". For those unsure of what that last point means, the AI injects as much style and flair for you as it can, in alignment with the selected "Romance" theme.

A+ meme usage. No notes. Allow me to reply in kind:

At this precise moment, you lack the ability to edit and understand your own draft, while you also lack the ability to conceive of your own changes without a series of external prompts. Anyone lacking those elements from their skill set isn't a writer.

You don't have to be talented to contribute here. Every shit writer out there's welcome to come and have a go.

Just stand on your own two feet and write your stories yourself. It's really not a lot to ask.

I know that you don't believe it, but I didn't start using anything but Google Docs until June of 2023. After that, I used (and still use) PWA primarily as a sanity check on grammar, punctuation, and spelling. As part of its toolset, it also highlights things like use of passive language, common/overused/"weak" words, overuse of adverbs, overlong paragraphs and sentences, etc., and I take advantage of those as well. It's extremely helpful for an at-a-glance tool for "standard" errors one might make.

However, I checked for all of that stuff before I started using it. My beta readers did, too. Comparing apples to apples by using Loving Wives stories "After the Future is Gone" and "Honesty Above All," one can see the shifts in my style--and, frankly, competency--that took place before I started using PWA. While I've improved more since then (I think), it's a matter of degree, not kind. The things I'd started to change before I'm still changing: eliminating (some) passive verb usage, common words, etc.

Some, but not all, because I'm still deciding what I do and don't want to use. As I pointed out in the other thread (and which you insulted me for), I like common words in dialogue and some narration, because that's how people talk. Sometimes passive voice is just less clunky to my ear. And, hey, I like an overly complex sentence every once in a while, as a treat.

I don't use its rewrite capability at all. I'm not offended by its existence, for reasons I'm not going to get into here, but I prefer to choose my own wordings. The closest I've ever come to using that tool was, if it highlighted a word it identified as "too common," I'd look at its suggestions. I'd also usually reject them, because PWA doesn't have the contextual sense to know when a synonym is or is not appropriate.

As an example, I had the sentence "I drove aimlessly," and it threw the little yellow underline there. But its suggestions were "haphazardly" and "randomly," neither of which fit. I'd already pretty much decided to keep it as is, but looking at a thesaurus confirmed that decision; or are thesauruses cheating, too?

As to the "Romance" setting, all that does is change the threshold for what it throws up warnings for. I ran some old stories of mine through it and found that it was the closest fit to the way I wrote: more tolerant of passive voice, a little less touchy about sentence complexity, etc. Basically, I was choosing the type of critique I wanted to see, as if deciding which beta readers' inputs I'd rely on.

Regarding the "style score" in both PWA and AutoCrit, they're just tools, too. I don't chase the scores; as anyone can see in discussions I've had in the megathread, they're a useful rule of thumb and nothing more. Just like PWA's yellow underlines, I can and do choose to ignore them, because the software can't understand intent.

I have a feeling, though, that you know all of this. At almost every turn, you've chosen to misrepresent, misinterpret, or flat out insult me. When possible, you've tried to rope in other members of the forum, first Bramblethorn and now wanda, as if you were hoping to start a fight. I'm not saying you're a troll, but I think, at best, you're a very petty little person who's mad that his attempts to police others haven't worked and is now trying to sow discord in the hopes of effecting change through alternate means, i.e., peer pressure to chase me off. Which is... really kinda sad.

I do think it's hilarious that you think I'm some kind of menace, though. So, in the spirit of your post:

 

[Bramblethorn]

However, I checked for all of that stuff before I started using it. My beta readers did, too. Comparing apples to apples by using Loving Wives stories "After the Future is Gone" and "Honesty Above All," one can see the shifts in my style--and, frankly, competency--that took place before I started using PWA. While I've improved more since then (I think), it's a matter of degree, not kind. The things I'd started to change before I'm still changing: eliminating (some) passive verb usage, common words, etc.

Which of course raises the point: if using technology to catch things like overused words is a sign of weakness and deficiency in a real writer, presumably using human beta readers for the same purpose would be just as bad. In which case, most of the authors here are equally talentless hacks, and it's odd to see somebody focussing so intently on one particular talentless hack when there are so many to be shamed for accepting help.

But not me, obviously. I write my stories on a remote island far from other humans, in a Faraday cage that blocks the possibility of even subconscious influence from external sources. My stories are drafted in a font I designed myself, on paper made from my own skin, with a pen whittled from my own fingerbones*. Even when it comes time to post them here, I don't use software for that purpose. Instead, I infiltrate the data centre that hosts Literotica's servers and I use a very small magnet to add my stories directly to the hard disc. It's painstaking work, but it's worth it to know that I am a true, self-sufficient writer.

*in retrospect, a poor choice

 

[awriterssojourn]

Which of course raises the point: if using technology to catch things like overused words is a sign of weakness and deficiency in a real writer, presumably using human beta readers for the same purpose would be just as bad. In which case, most of the authors here are equally talentless hacks, and it's odd to see somebody focussing so intently on one particular talentless hack when there are so many to be shamed for accepting help.

But not me, obviously. I write my stories on a remote island far from other humans, in a Faraday cage that blocks the possibility of even subconscious influence from external sources. My stories are drafted in a font I designed myself, on paper made from my own skin, with a pen whittled from my own fingerbones*. Even when it comes time to post them here, I don't use software for that purpose. Instead, I infiltrate the data centre that hosts Literotica's servers and I use a very small magnet to add my stories directly to the hard disc. It's painstaking work, but it's worth it to know that I am a true, self-sufficient writer.

*in retrospect, a poor choice

It’s hard to know if people truly don’t know the difference between it suggesting you change the word large to huge and actual generative AI or if they’re just taking the opportunity to feel superior.

I do feel it needs to be said though because I haven’t seen anyone say it. The AI detection services aren’t actually supposed to flag that someone used something like Grammarly to add commas, change a few words, etc. It’s meant to detect generative AI. This actually is just more evidence that they can’t do what they’re intended to do. It’s not evidence that they’re super sensitive to some editing process.

 

[NoTalentHack]

But not me, obviously. I write my stories on a remote island far from other humans

Oh, you think you're SO clever, hiding your use of beta readers from us with this proclamation. But I know! I see through your lies! You have a beta reader, and his name is Wilson!

https://whateverhappenedtox.com/wp-content/uploads/2020/03/Wilson-1024x580.jpg

 

[DDX84]

I know that you don't believe it, but I didn't start using anything but Google Docs until June of 2023.

I do believe that, actually. It has no relevance to anything, but that's fine. If you've used AI in one story, or a hundred, it's still an issue.

I don't use its rewrite capability at all. I'm not offended by its existence, for reasons I'm not going to get into here, but I prefer to choose my own wordings.

I don't believe that, for which the summary is laid out in the previous post.

However, why do you care if I believe you or not? As you and some recent alt accounts have pointed out, I don't publish here. Why do you give a fuck about what I think?

You're a wrestling fan like me. I read Kayfabe, just don't sell for me. Like any boxer who takes a stiff right hand to the jaw, they only sell it because the shot landed.

When possible, you've tried to rope in other members of the forum, first Bramblethorn and now wanda, as if you were hoping to start a fight.

and is now trying to sow discord in the hopes of effecting change through alternate means, i.e., peer pressure to chase me off. Which is... really kinda sad.

Let's get one thing straight here. You're one of the most popular authors on this website. It benefits no one to have you chased off, or for you to feel guilty and run away, like the cowardly and abusive MourningWarbler.

However, your peers on this forum should've pulled you to one side around a month ago and explained why you were making a series of mistakes.

If this forum had any prestige or community standards, the best writers would explain to you why it's embarrassing to download AI and tune it to "Romance", with additional custom settings, for hyper-specific content suggestions, which you then adopt as your own. All the things you admitted to doing, the best writers here should've found a discreet way to explain to you why your workflow is a form of cheating. It's not just "A+ meme usage", brother. It's a serious issue that needs fixed.

They didn't do that. Either they failed to identify the problems, didn't understand them or lacked the courage to talk to you about it.

There's also the issue of your previous conduct, where you chased Tilan and the banned reviewer off based on your lie that they were the same person. Perhaps there are those here who are simply too scared to raise the issues with you in case they're the next ones to get banned? You should really stop doing that.

Tilan's a delusional moron, but the Stacnash review account offered a lot of insight to many of the authors here. I sent them a email in December when this all started and received a lovely response. I actually can't remember the last time I enjoyed corresponding with another writer as much. Yet, I'm being asked, by you, to believe that they were Tilan's poisonous alt, all because they didn't enjoy your work. I'd rather die than correspond with a moron like Tilan. They are not the same person.

Why's it fair for you to "chase off" either of them? I've got the threads open right now, what a fucking shitshow that was and it was all your fault.

Speaking of reviews, when you look at the dumpster fire that is the AwkwardMD & Omenainen thread, where they used some of Bramblethorn's opinions in their review and claimed it as their own work, after admitting that they failed to understand the material without first looking at the comments and drafting in "backup readers", it's not hard to see why someone at my level would stop publishing here when those two LARPers are considered amongst the best you've got.

That's a blatant lie, since you've been taking shots at me for basically the entire time you and I have interacted.

Look, man. I started off in a very civil manner when OverconfidentSarcasm cited you. I even thanked you for your transparency. Then, you launched your initial volley of swearing and abuse. I don't really care about that now, but you and I don't have anything left to discuss.

Laurel took no action against you, despite admissions that go far beyond what others have been rejected for. That's the primary outcome of the sum of our interactions. You've proved that the rules are not being enforced in the same way for everyone. 8letters still publishes here. What he did was far worse than anything you've admitted to. Plagiarism is fine here.

Any subsequent comments I've made about your workflow, talent and reliance on AI are secondary, and I have nothing else to say on that. You know what I think, I've been thorough.

It's a shame because I think we've got a lot of similar interests. Wrestling, chess, creative writing. You're one of the very few here with a spine. We'd probably get on very well under different circumstances. But even if I was your best friend in this world, I'd still have dragged you through the hedges when I found out about how you're using AI and I wouldn't have stood idly by when you lied to have people banned from here.

Good luck with your writing in the future.

 

[ElectricBlue]

However, your peers on this forum should've pulled you to one side around a month ago and explained why you were making a series of mistakes.

If this forum had any prestige or community standards, the best writers would explain to you why it's embarrassing to download AI and tune it to "Romance", with additional custom settings, for hyper-specific content suggestions, which you then adopt as your own. All the things you admitted to doing, the best writers here should've found a discreet way to explain to you why your workflow is a form of cheating. It's not just "A+ meme usage", brother. It's a serious issue that needs fixed.

They didn't do that. Either they failed to identify the problems, didn't understand them or lacked the courage to talk to you about it.

I'm not sure NTH has the groupie following you think he has.

Why you think it's anyone's job to look after NTH is beyond me - he's big enough and ugly enough to dig his own holes, and seems to enjoy falling into them all by himself. Nothing to do with me.

 

[alohadave]

it's not hard to see why someone at my level would stop publishing here

Bullshit trolls gonna troll.

You claim that you were a top poster here. Name your other account. Oh yeah, you deleted it. Guess there's no way of verifying your claim.

Put up or shut up. Name your main account. Until you do, you are just a troll hiding behind this alt account.

 

[Bramblethorn]

Speaking of reviews, when you look at the dumpster fire that is the AwkwardMD & Omenainen thread, where they used some of Bramblethorn's opinions in their review and claimed it as their own work,

Oh for the love of god, stop trying to play AH authors off against one another. We're quite capable of generating our own feuds without this amateurish "let's you and him fight", thank you very much. If you want to go after people here, do your own dirty work instead of trying to recruit others to do it for free.

In just about any field of intellectual work, it's perfectly normal to ask a friendly colleague "Can I run something by you and see if it makes sense to you?" once in a while. That's what happened here. In this case, I don't think my feedback made much difference to how the review came out; I agreed with their thoughts on some aspects, disagreed on others, and from the final review I see my disagreements didn't much sway them on those parts.

I would not have accepted an author credit on that review, both because they did all the work of writing up those thoughts into an actual article – I don't think I saw so much as a draft before it was published – and because there are parts of it that I don't agree with.

This attitude that Real Authorship requires some kind of immaculate conception, untainted by the faintest whiff of another person's thoughts, is insecure nonsense far removed from how most creative work is produced. We're all fallible, but those with a modicum of self-awareness are capable of recognising it and figuring out ways to mitigate it.

 

[MillieDynamite]

I am taking a deep breath. This is quite silly DDX84 is a troublemaker and nothing more. If he's an author, I can't find his stories; if he's a critic, who cares? AI isn't used by most of the writers on this site. The AI rule is vague, at best. How Laural determines when something is AI isn't stated. The fact that some stories get sent back as being suspected of AI written or aided isn't something any of us can control. The volume of stories being posted seems constant. What I conclude from that is that the rejection rate is no greater than the norm. Faisaco is a strong statement. If it were bogging up the entry cue where it was taking weeks for everyone to get their stories up, maybe then it would be a FAISACO. I'd say it's a blip, a belch, or a bump in the road.

We've had several of these accounts hitting the AI threads and stirring the pot of discontentment.

 

[Andreas_Kreuz]

AI engines make very few grammticalirlaly mystakes. Just add sem to yer storry and it will look human.

 

[NoTalentHack]

AI engines make very few grammticalirlaly mystakes. Just add sem to yer storry and it will look human.

You might be joking, bit you’re not wrong. Perplexity detection may or may not include regular expression usage to detect simple typos. A human might understand the intent of “a cup of waiter,” and that that’s a typo; a detector may “think” it’s novel word usage.

 

[awriterssojourn]

In my significant frustration this weekend, my slightly obsessive tendencies led me to try to get to the bottom of this issue. I basically wanted to figure out which detector they were using, etc. So I used my stories since one was accepted and a second rejected twice. Obviously, there is some difference between them despite writing them actually at the same time in the exact same way.

Undetectable AI will tell you which of the different detectors are most likely to flag your story as AI and is supposedly the most accurate at determining AI. So I started there and I paid for a month and ran them through. My rejected story was found to be human by all except possibly copyleaks, while my accepted story was likely to be found 100% AI by copyleaks and possibly AI by several others. So naturally, I paid for copyleaks and a few other sites it suggested. I specifically picked sites that do sentence by sentence detection. It's the only sane thing to do.

What did they find? Neither story was flagged as having a single sentence as AI by copyleaks or the others. By sheer randomness, I would have expected this to have flagged something as AI, but perhaps they do take into account the whole story rather than just a single sentence without context. Who knows. I'm not here to debate how they work, but what is being used by the site so we have some insight. But obviously based on this, undetectable AI is not able to correctly predict which ones would find it as AI. What if I just run them through every site anyway since we can't trust undetectable AI's predictions? Literally not one found either story to be AI or have any specific sentences flagged as AI.

What I know now is that I cannot find any evidence that the moderators are using any commercially available AI content detector, as my AI rejected story is not detected as AI by any of them. They don't even flag my individual sentences which I was surprised by. I don't mean they don't flag many. I mean that copyleaks and the others flagged zero sentences. So while we've all been speculating that they've been using some detector that is having false positives, I cannot find a single one that would give them the results they report.

After burning some money and and time, my main question is, are they actually using any AI detector at all? Because after going through them one after another, I cannot find any that would even remotely give them the results they're getting. I am unhappy with myself that I've let myself go down this rabbit hole, but I figured I would at least share with anyone who is curious. My advice is this, if you didn't use AI to write your story, don't waste your time trying to figure out why they say it's AI. Just move on and enjoy writing for the sake of it. Perhaps it'll be fixed in the future, but based on the lack of responsiveness, I am not personally optimistic.

 

[Andreas_Kreuz]

I did similar experiments and I came to the same conclusion. I actually generated stories with ai and ran them through the ai detectors. Also random results. With gpt and bard you can actually request the story to be written in a specific style. And then the detectors think it is human. So we have the situation, that human stories are rejected as ai and ai stories not recognised.

 

[aworthyopponent]

In my significant frustration this weekend, my slightly obsessive tendencies led me to try to get to the bottom of this issue. I basically wanted to figure out which detector they were using, etc. So I used my stories since one was accepted and a second rejected twice. Obviously, there is some difference between them despite writing them actually at the same time in the exact same way.

Undetectable AI will tell you which of the different detectors are most likely to flag your story as AI and is supposedly the most accurate at determining AI. So I started there and I paid for a month and ran them through. My rejected story was found to be human by all except possibly copyleaks, while my accepted story was likely to be found 100% AI by copyleaks and possibly AI by several others. So naturally, I paid for copyleaks and a few other sites it suggested. I specifically picked sites that do sentence by sentence detection. It's the only sane thing to do.

What did they find? Neither story was flagged as having a single sentence as AI by copyleaks or the others. By sheer randomness, I would have expected this to have flagged something as AI, but perhaps they do take into account the whole story rather than just a single sentence without context. Who knows. I'm not here to debate how they work, but what is being used by the site so we have some insight. But obviously based on this, undetectable AI is not able to correctly predict which ones would find it as AI. What if I just run them through every site anyway since we can't trust undetectable AI's predictions? Literally not one found either story to be AI or have any specific sentences flagged as AI.

What I know now is that I cannot find any evidence that the moderators are using any commercially available AI content detector, as my AI rejected story is not detected as AI by any of them. They don't even flag my individual sentences which I was surprised by. I don't mean they don't flag many. I mean that copyleaks and the others flagged zero sentences. So while we've all been speculating that they've been using some detector that is having false positives, I cannot find a single one that would give them the results they report.

After burning some money and and time, my main question is, are they actually using any AI detector at all? Because after going through them one after another, I cannot find any that would even remotely give them the results they're getting. I am unhappy with myself that I've let myself go down this rabbit hole, but I figured I would at least share with anyone who is curious. My advice is this, if you didn't use AI to write your story, don't waste your time trying to figure out why they say it's AI. Just move on and enjoy writing for the sake of it. Perhaps it'll be fixed in the future, but based on the lack of responsiveness, I am not personally optimistic.

I for one appreciate the legwork you've done on this!

Given the lack of response and the seeming indifference of the site owners to the issue, I'm moving all of my stuff to ao3. It's been a slice, literotica! (here is the link)

 

[regularguy13]

FYI I had two stories for the Valentine's Day contest rejected as AI. I use Google Doc to and the free version of Grammarly to check for spelling and grammatical errors. Some of their suggestions I accept. Some I don't. I find they have trouble/get confused with dialogue. One phrase they always object to is 'She gave great head'. They want it to be 'She had a great head.'

I resubmitted both stories saying the works were mine, not AI. Days passed. Nothing. I reworked one story, trying to make it more human. I made sure to have a variety of sentences and paragraphs. Some short and some long. Whenever I review a story, I always change something...this could be clearer...less passive voice, etc.

I resubmitted it yesterday. It is pending. I was at a lost as what to do the other one. I was going to review it today, and I was surprised to see that it had been accepted.

I'm not sure what's the lesson? Protest the AI accusation and eventually your work will be published? Maybe my story was 'could be AI' but had a low score? I could use the Literotica spell checker and forego the grammar check. My understanding is the basic Grammarly is an AI tool that edits. It is an AI assister, but since it doesn’t create wholly new content from a prompt, it isn't an AI generator.

 

[Andreas_Kreuz]

AI mimics a certain style of writing. Detectors detect that style. So if you happen to prefer this style you have a problem.

If universities stop requiring students to write papers, because they are unable to tell if it is AI generated, then here it will be even more difficult.

I would rather see a warning displayed above the story or a marker like the hot or new markers indicating that it might be AI.

 

[SkyBubble]

It seems like many, many authors are struggling with getting stories rejected by the sites obviously too broad and near-useless AI filter.

At what point are we going to migrate to another site, and what sites would folks recommend for the exodus?

If you're just writing "regular people," not celebrities, try Lush Stories. They have 27 of mine, including the story this site insists on rejecting as AI. They don't do celebrity stories, however.

I used to use Noveltrove, but it's got something wrong with its algorithm that turned all my stories into indecipherable mush, so I re-edited them into their previous incarnations, and now they've been in "Pending" for months.

Anybody know a good site for celebrity stories?

 

[SkyBubble]

There seems to be an avalanche of authors we have never heard from in AH that are reporting stories being kicked back for AI, some claiming "hundreds" of stories. I guess I'm too lazy to check them on that, and there's always the "alts" thing. As I posted elsewhere, I'm of two minds on this - is there too much protestation of innocence? Or is there actually a problem where Laurel is using an automated AI check with the threshold set too low?

Then there's the Grammarly thing, or even MSWord's grammar checker. Use Grammarly, go to jail, do not pass 'Go', do not collect $200.

Laurel certainly knows there's a problem. All we can do here in the peanut gallery is steel ourselves to the situation until something gives.

It sounds like you get punished for using correct grammar.

 

[MrPixel]

It sounds like you get punished for using correct grammar.

Not at all. "Correct grammar" is not the problem, the use of a grammar checker to, in essence, rewrite portions of your text to conform to its AI-determined analysis is the problem. There have been comments to the effect of "fixing spelling and punctuation is fine, but take text restructuring under advisement." Grammarly is known to use generative AI as a baseline for its recommendations.

I believe, but cannot prove, that accepting any of Grammarly's grammar "corrections" is the trouble spot. The issue is any AI-content checker is going to be looking for known structure patterns as a particular AI engine's "signature". But those of us here in the peanut gallery cannot see what the problem is because we never have the opportunity to read the offending material. We can just make educated guesses.

My personal take is after having read a limited amount of AI generated prose, if your writings are genuinely your own and your contributions are rejected for alleged AI, then your style is dry and without personality. Spice it up a little.

 

[Andreas_Kreuz]

It is actually the other way around. AI style (at least GPT and bard) is not dry, but ornate and florid. here you can see and discuss examples: https://forum.literotica.com/threads/ai-writing-style.1603513/

 

[jhealy55]

I was actually looking up crow recipes this week as I submitted something for the first time since the AI Wars began. Would have been embarrassing to admit I'd been rejected after my various arguments against allowing AI assisted work on the site.

But then both sailed through without a hiccup. Each contained very little dialogue and were relatively short @ 750 words each, but I didn't change anything in my writing style to increase the chances of avoiding rejection.

Best advice I can give is to focus on flow, trying to inject an organic element and connect on an emotional level. Add some awkward, some hesitation, connection and tumbling relief to your story.

Much like the way these heightened experiences, that most of us are trying to capture, work out in real life.

In the end, the work we do to humanize our writing is what makes it connect. The grammar won't always be perfectly polished and the vocabulary will be distinct to us as individuals, but that's how life is, isn't it?

 

[CalebZhass]

Al Fiasco was probably the worst hit-man the mob ever had.

Shame on you, sir, for trolling. Also, genius.

 

[CalebZhass]

It seems like many, many authors are struggling with getting stories rejected by the sites obviously too broad and near-useless AI filter.

At what point are we going to migrate to another site, and what sites would folks recommend for the exodus?

WE? I'm not going anywhere, I stated that at the start. I don't care to have my stories published anywhere else. It's either here, or nowhere.

If getting stuff out fast is important to you, you can leave. I think the AI policy will change over time, for the simple fact that it'll have to, but from what I've learned in the past month, changes over here at Lit happen slower than at some government institutions So patience is our best friend.

I still love it here, with all its flaws. I'm not quite sure whether that paints a sad picture about me, a positive picture about Lit, neither or both.