TheArsonist
Really Experienced
- Joined
- Aug 8, 2022
- Posts
- 175
*Groucho Marx
*Groucho Marx
ElectricBlue
Joined 11 Years Ago
- Joined
- May 10, 2014
- Posts
- 18,340
Well, I don't seek help from a toaster, so why should I seek help from any other non-human source? But then, I think grammarly is a waste of time, too.
M_K_Babalon
Harbinger of smut
- Joined
- Nov 2, 2023
- Posts
- 1,727
Sounds like dude has my kinda luck.
NoTalentHack
Corrupting Influence
- Joined
- Nov 7, 2022
- Posts
- 2,578
Since @DDX84 popped their head in here and didn't answer the question I asked about what tool he's using, instead taking the time to insult me again, I decided to go hunting for it. Luckily, Google Image Search found it pretty quickly. It's part of a CRM toolset called Sapling. It also has a reputation as the most hypersensitive tool on the market, and not in a good way.
Even their own disclaimer at the bottom of the page is basically, "hey, this might be complete bullshit."
So, let's run some tests! How about we start with the top 10 most read stories on Literotica. Note, I haven't actually read these; I'm not some crazy stalker that's willing to spend two days hatereading a bunch of stories. But I did take a little time this morning to copy and paste them into the Sapling tool. Because I'm not going to pay to use this piece of shit, it truncated the results to the first 2K characters, but that should still give us an idea of the tool's efficacy.
1. Sitting on My Son's Lap
Well, there's a promising start.
2. 'A' My Name is Alice
A little better. It only thinks that 35% of the text of a story published in *checks notes* 2001 is AI generated. Let's keep going, though.
3. Sister Sucks Sleeping Sibling
Only 6.6% this time! Apparently it has a problem with ellipses, and (not screenshotted), there's a chunk near the character limit that got highlighted for some reason.
4. Accidents Happen
So, about a quarter of it is fake content. Generated in 2009. Cool. Cool cool cool.
5. "Oh, Mommy," I Groaned
These titles, man. Still, only 19%! On a story published in 2006! Maybe there is something to these AI detectors after all.
6. A Mother and Her Son
Uh. Nevermind.
7. Mom Takes a Ride
0.01% We have a winner! Why? Fuck if I know. I doubt the Sapling guys know, either. I'm certain Deputy Fife doesn't. Funny how he has so much faith in a tool so obviously flawed. It's almost like it's not about accuracy, it's about gatekeeping and elitism. Weird.
8. Service
Does 60% count as torches and pitchforks time? I kid, I kid. Any amount of possible AI content is suspect. Report your neighbors! Report your friends! Kids, report your parents for any hint of pro-technology sentiment! Only you can stop people from expressing themselves with the tools available to them!
9. Making Out With Mom
I don't have a fun joke about the results here. I am, however, feeling an urge to write a story called Oedipal Arrangements set in a flower shop.
10. My Virgin Sister
And, for our final entry, "only" 18% fake text.
So, what does this tell us? Well, given that the newest of these stories was published in 2009, none of them are using any kind of AI generation--unless they're also using time travel, in which case I need to add that to my raft of software--yet all of them but one had a non-trivial amount of text flagged as AI-generated. I think you can draw your own conclusions from that alone.
I am going to leave one more article about it, though, and then I'm done with this nonsense. It's from an AI researcher who ran the text of her own book through a couple different detection tools and... well, the title says it all: Don't use AI detectors for anything important.
Okay, that was a lie. I did like the picture at the top of another of her articles, Apparently I am a robot, but I've reached the limit for the number of pictures I can post in this, ah, post. So, click the link and have a good laugh.
To DDX84, allow me to quote one of the finest police officers of all time:
"Your move, creep."
Even their own disclaimer at the bottom of the page is basically, "hey, this might be complete bullshit."
So, let's run some tests! How about we start with the top 10 most read stories on Literotica. Note, I haven't actually read these; I'm not some crazy stalker that's willing to spend two days hatereading a bunch of stories. But I did take a little time this morning to copy and paste them into the Sapling tool. Because I'm not going to pay to use this piece of shit, it truncated the results to the first 2K characters, but that should still give us an idea of the tool's efficacy.
1. Sitting on My Son's Lap
Well, there's a promising start.
2. 'A' My Name is Alice
A little better. It only thinks that 35% of the text of a story published in *checks notes* 2001 is AI generated. Let's keep going, though.
3. Sister Sucks Sleeping Sibling
Only 6.6% this time! Apparently it has a problem with ellipses, and (not screenshotted), there's a chunk near the character limit that got highlighted for some reason.
4. Accidents Happen
So, about a quarter of it is fake content. Generated in 2009. Cool. Cool cool cool.
5. "Oh, Mommy," I Groaned
These titles, man. Still, only 19%! On a story published in 2006! Maybe there is something to these AI detectors after all.
6. A Mother and Her Son
Uh. Nevermind.
7. Mom Takes a Ride
0.01% We have a winner! Why? Fuck if I know. I doubt the Sapling guys know, either. I'm certain Deputy Fife doesn't. Funny how he has so much faith in a tool so obviously flawed. It's almost like it's not about accuracy, it's about gatekeeping and elitism. Weird.
8. Service
Does 60% count as torches and pitchforks time? I kid, I kid. Any amount of possible AI content is suspect. Report your neighbors! Report your friends! Kids, report your parents for any hint of pro-technology sentiment! Only you can stop people from expressing themselves with the tools available to them!
9. Making Out With Mom
I don't have a fun joke about the results here. I am, however, feeling an urge to write a story called Oedipal Arrangements set in a flower shop.
10. My Virgin Sister
And, for our final entry, "only" 18% fake text.
So, what does this tell us? Well, given that the newest of these stories was published in 2009, none of them are using any kind of AI generation--unless they're also using time travel, in which case I need to add that to my raft of software--yet all of them but one had a non-trivial amount of text flagged as AI-generated. I think you can draw your own conclusions from that alone.
I am going to leave one more article about it, though, and then I'm done with this nonsense. It's from an AI researcher who ran the text of her own book through a couple different detection tools and... well, the title says it all: Don't use AI detectors for anything important.
Okay, that was a lie. I did like the picture at the top of another of her articles, Apparently I am a robot, but I've reached the limit for the number of pictures I can post in this, ah, post. So, click the link and have a good laugh.
To DDX84, allow me to quote one of the finest police officers of all time:
"Your move, creep."
NoTalentHack
Corrupting Influence
- Joined
- Nov 7, 2022
- Posts
- 2,578
LOL I lied again. I got an email from Sapling when I registered to use their tool, and I HAVE to add this:
So they actually do autocomplete as part of their paid toolset! It has its own ML content generation! I bet you it doesn't flag its own stuff, either.
Huh. Wonder why their tools are so strict in their evaluations. It's almost like they're trying to sell you something.
So they actually do autocomplete as part of their paid toolset! It has its own ML content generation! I bet you it doesn't flag its own stuff, either.
Huh. Wonder why their tools are so strict in their evaluations. It's almost like they're trying to sell you something.
Last edited:
MelissaBaby
Wordy Bitch
- Joined
- Jun 8, 2017
- Posts
- 8,003
All I learned from that is that computers aren't interested in fucking their mothers or sisters.
StillStunned
Mr Sticky
- Joined
- Jun 4, 2023
- Posts
- 10,094
MillieDynamite
Millie'sVastExpanse
- Joined
- Jun 5, 2021
- Posts
- 10,318
I'd have to say, maybe, just maybe, these stories are in the AIs database and have been used to create stories. Yes, we all write sentences that will be flagged, but that doesn't mean they aren't our stories. And yes, I believe we can never know if something is human or AI. I don't find AI creative, I don't find the writing has a story structure, it tends to drift around. Then again, so do some human writers.
MelissaBaby
Wordy Bitch
- Joined
- Jun 8, 2017
- Posts
- 8,003
I set up a perfect "lap top" joke, and no one ran with it.
AwkwardlySet
On-Duty Critic
- Joined
- Jul 24, 2022
- Posts
- 4,174
That makes a lot of sense. If someone wanted to train an AI on porn, taking popular Literotica stories would be a reasonable choice. Still, the main point stands. AI detectors are too unreliable to be taken as an objective proof.
MillieDynamite
Millie'sVastExpanse
- Joined
- Jun 5, 2021
- Posts
- 10,318
They train on something. If I understand the process correctly, they scan all writing published on the internet on some and others, a large dataset of stories that have been culled (either legally or illegally) and picked passages it chooses to mimic (not copy) the material close to their instruction set or to continue from what you've written via that dataset. Sudowrite is for fiction writing. But it doesn't use prompts. It looks at up to 1500 words you've written and continues what you have.
POORLY from what my publisher says. A writer he edits for but doesn't actually publish asked him to review it. The idea was he'd use it to speed up his process. My editor worked with it and was horrified by how clunky the writing was. Overwritten, poorly worded drivel is what he told the writer. But as he edits the author's work, he sees where some of it was produced by the 'MACHINE,' which he calls EVIL!
The only feature he liked was the where you ask it to write or rewrite in a published author's style. However, he reckoned that came close to being plagiarism.
POORLY from what my publisher says. A writer he edits for but doesn't actually publish asked him to review it. The idea was he'd use it to speed up his process. My editor worked with it and was horrified by how clunky the writing was. Overwritten, poorly worded drivel is what he told the writer. But as he edits the author's work, he sees where some of it was produced by the 'MACHINE,' which he calls EVIL!
The only feature he liked was the where you ask it to write or rewrite in a published author's style. However, he reckoned that came close to being plagiarism.
Last edited:
NoTalentHack
Corrupting Influence
- Joined
- Nov 7, 2022
- Posts
- 2,578
I sat with all of this a while this morning while running errands. Not what you're saying here, but why the whole AI detector thing pisses me off so much, and I came to a conclusion that sort of circles around what you did say here. "Some human writers aren't very good, either, and we can't know either way."
At first, I thought I was pissed because I got accused of being a cheater by using AI to generate my stuff. But I knew that I didn't do it. None of my stories have run into problems related to AI during the submission process, and my most recent one went through just fine, even if it's not doing so well in the ratings.
Then I thought it was because DDX84 specifically decided to target me, but he's just a petty, pearl-clutching little asshole; like the saying goes, "If you run into an asshole in the morning, you ran into an asshole. If you run into them all day, you're the asshole." And I don't, whereas he's basically being a shit to everyone that slightly disagrees with him, so I know where the two of us end up in that equation.
Then I thought that it was because of what that meant in a greater sense, that partially because assholes like him are targeting people, we're getting new writers kicked off the site for no sin greater than triggering a bad AI detector and the personal enmity of a tinpot hall monitor. I don't fault Laurel for being careful; she has to do SOMEthing. But seeing people bail from the site--in the same way I probably would have done a year ago, and probably still would if my stuff started getting tagged by whatever the site is using--made me more sad than angry.
No, what I finally realized this morning is it's the complete, unthinking, self-righteous hypocrisy of a statement like this:
I've made it clear where I stand, and I think you, Millie, are in roughly the same boat as me: using tools to evaluate our work and rewriting it ourselves is fine, but generation of content passed off as our own is not. In the light of that stance, I may not like that my stuff is going through an AI detector, but I accept it. If it becomes oppressive, I'll leave, and I'll say exactly why. And the why is this:
I am writing my own work. I'm writing it to my specifications. I'm writing it in my voice, and when a software tool I use tells me that my voice is wrong--unless it can make a very persuasive case--I'll go with my own voice. And I'll do the same thing with the AI "detection" software. If the site says, "our tool believes your content is probably AI generated, and you can change it and resubmit?" Then, no, I won't. It's my writing. It's my voice. I'm not changing it solely because a tool tells me to. Literotica can make their choice, and I'll make mine. I may not like it, and I'll argue against its usage like this because it's really not good at what it does, but I accept that the tech can be a double-edged sword.
That's not what DDX84 and his ilk are saying, though. Their sin is far, far worse.
They're saying, "Software tools are bad. They make your work impure. They make it not your work. We won't use them, and neither should you." So far, so good. That's their opinion, and I have no problem with it, even if I disagree.
But then they continue, "However, this software tool says your work might be AI generated, and therefore you need to change it. You know, to keep your work purely human. You must change how you write to fit what the software says, because this tool--which we do not actually understand and which we are using exactly the way its creators said not to--says you should. You need to lose your voice, because the machine says it's not really your voice.
"Oh, you're a new writer? You prefer simple prose? You're not interested in being the next great American novelist and just want to write something to share with others? Too bad. It's too close to what the machine says is made by machines, so we can't trust you.
"But you had better not use other tools to figure out where you're going wrong. You'd better not use tools to help you rewrite your work, or even to evaluate whether it needs to be rewritten. Only this tool, the AI detector, can be used. The process has to be kept pure, after all.
"Running away, are you? Just because we told you to change what you've created to fit the parameters that will make our machine believe you're not a machine? Because we told you to change your voice to be something it's not? Coward. I would defend my work to the death, because I am righteous. I know I am, and the machine agrees, which means it's righteous, too."
The absolute hypocrisy of holding those two conflicting stances--that writers should use no software to aid them in order to ensure authenticity, BUT they should be prepared to make any changes that their "good" software says in order to also ensure authenticity--is just staggeringly arrogant.
So, yeah. Now I know why I'm pissed. Because they aren't just turning new writers away. They're censoring them, not for the writers' own good, but to retain their own perceived moral superiority.
Seriously, fuck you, you fascist, hypocritical shitstain @DDX84. And if that earns me my first timeout? So be it. It was worth it to call you out for what you are.
Last edited:
alohadave
Doing better
- Joined
- Dec 6, 2019
- Posts
- 3,507
They are a troll, plain and simple. Make unsupportable and contradictory arguments that you spend a lot of time rebutting, then they come back with something else.
I'm curious who their main account is. It's obviously someone that knows the players here and has it out for them.
NoTalentHack
Corrupting Influence
- Joined
- Nov 7, 2022
- Posts
- 2,578
They read a lot like Stacnash’s “friend” that showed up for a couple of parting shots, but IDGAF.
MillieDynamite
Millie'sVastExpanse
- Joined
- Jun 5, 2021
- Posts
- 10,318
I wasn't referring to you as a bad writer. I meant that some bad writing reads like a thoughtful human didn't write it. But in no way was I calling out any writer here. We all have our clunky sentences in our work. Sometimes, something is clunky on purpose, and sometimes it isn't.
NoTalentHack
Corrupting Influence
- Joined
- Nov 7, 2022
- Posts
- 2,578
Oh, no, I didn’t think you were. This was more about what a “bad” writer that’s “bad” because they’re new, and the way they’re “bad” triggers an AI detector.
Voboy
Sometime Wordwright
- Joined
- Mar 21, 2016
- Posts
- 6,045
Gold, Jerry. Pure gold. You have to write this now.
Yeah, the "Borg collective" guy who claimed there was a shadowy cabal of tastemakers, roving around deciding which stories Should Be Liked and which Should Not? I got the same impression, actually.
RoperTrace
Literotica Guru
- Joined
- Mar 8, 2003
- Posts
- 1,007
Okay, I read a good part of page 1 of your first story, and I get zero AI vibe from your writing. It's pretty good, IMO; story just isn't my cup of tea. You should do well here. RT
TheRedChamber
Apprentice
- Joined
- Mar 21, 2014
- Posts
- 2,593
View attachment 2295483
Ask and you shall recieve.
I had every issue of Your Sinclair. Then one month this was on the cover. My sister's best friend's grandparents ran the only newsagents in the village.
oneagainst
...the bunnies
- Joined
- Oct 23, 2021
- Posts
- 995
Modestly popular, also not willing to change a single word of what I write because of a filter. The day I get pinged for AI-generated content and get blocked from publishing is the day I quit.
There's a storm coming, @NoTalentHack. Though maybe we'll need a phased plasma rifle in the 40 watt range.
onehitwanda
Venatrix Lacrimosal
- Joined
- May 20, 2013
- Posts
- 4,707
NoTalentHack
Corrupting Influence
- Joined
- Nov 7, 2022
- Posts
- 2,578
Voboy
Sometime Wordwright
- Joined
- Mar 21, 2016
- Posts
- 6,045
There was a poster called stacnash who would deliver pointed public critiques, oh, about three or four months ago. He/she came across a bit high-handedly, and drew fire accordingly. They then left.
BUT!
One of their homies promptly materialized, claiming stacnash was a member of a mysterious crew of readers who roved the internets, critically commenting on the various bits of prose they found there. The implication was that said crew comprised Important People Whose Opinions Matter.
The homie left after a few posts, IIRC. Now they are lost to us, alas and alack.
onehitwanda
Venatrix Lacrimosal
- Joined
- May 20, 2013
- Posts
- 4,707
Oh, no, I knew who Stacnash was. She "blessed" me with one of her comments, and we had a brief handbagging over it. I missed the rest of the story though; thanks for the summary of the drama, there's not enough wine in my rack to deal with that.
Voboy
Sometime Wordwright
- Joined
- Mar 21, 2016
- Posts
- 6,045
https://forum.literotica.com/thread...thread-or-is-it.1593945/page-11#post-97454832
That's about where things went off the rails on that thread. Her homie is called Thavipah, but none of their content is searchable unless you read the replies that quote him/her/it.
Enjoy!
