Story incorrectly rejected due to AI

[DDX84]

A lot of this could be cleared up if Literotica was more transparent with what AI detector they are using. Whichever one it is, it is obviously way to sensitive. Either that or there's a very large AI usage problem on the site.

There's definitely a huge influx of new content, which had the knock-on effect over Christmas of causing publishing delays for some, but not all.



I don't think they need to be transparent over which detector they're using. None of them are 100% accurate. However, I think it's important that they're transparent about the percentage threshold that they consider a positive result to be.

As definitive proof isn't possible for any publisher, a determination has to be made on the balance of probabilities. What percentage from the detector do they classify as a positive result?

51%?
65%?
85%?

65% might get you kicked off Amazon, for example, but not Literotica. Or vice versa.

However, doing nothing isn't an option. They're obviously not happy with the influx of content that's almost certainly AI produced. So, they pick a scanner and choose a threshold.

As long as we're all subject to that in the same way, with no favouritism, I'm in favour of it.

 

[nice90sguy]

View attachment 2304765

Thumbs up for the chart

 

[MajorRewrite]

I ran this entire thread through an AI detection app and it scored 82% fake. So you’re all a bunch of bots debating with each other. I should have known.

 

[MattblackUK]

I decided to test AI out. I asked it to write two 1,000 word feature articles on specific subjects and two different plagiarism checking sites declared both articles as 100% original and not plagiarised.

 

[XerXesXu]

According to your chart, "sapling" is purportedly the most accurate free AI detector, boasting a 68% accuracy rate.

I sampled the following paragraph, I had written years ago, and it returned a 99.9% fake result.
View attachment 2304789
After making slight changes (removing a few commas, contractions, and hyphens, it dropped to 1.9% fake!
View attachment 2304791
After reevaluating the revised text ten minutes later, the result indicated 52.9% fakeness.
View attachment 2304797

You blindly rely on ridiculously unreliable tools, inflicting an intolerable injustice upon your fellow writers

Love your screen name.

 

[CalebZhass]

I don't believe the reliability (or rather the lack of thereof) of these checking services was ever a question around here. I mean, come on. People who have been publishing here for a prolonged period of time are running into this problem. People who are obviously not using AI to write their text. If we change the standard to say, that any grammatically correct and stylistically appropriate text is AI written, as no human can do that on their own, then we might just as well fire and burn down the whole industry and tool chain focusing on increasing production quality of publications.

I always found it ridiculous, that using a software aimed to give your grammar suggestions, synonyms to avoid repetition, phrasing alternatives to make your description more colorful, so a tool like that to be considered a sin. It's just weird.

 

[onehitwanda]

Just going to throw it in here that AI would flag e e cummings as AI-generated. It would also probably flag Poe as AI-generated.

 

[alohadave]

Just going to throw it in here that AI would flag e e cummings as AI-generated. It would also probably flag Poe as AI-generated.

Both of them are probably in the corpus so it stands to reason that the checker would recognize them.

Something interesting that pops up in colleges is when teachers use plagiarism detectors, students get flagged for copying because previous work they've done has been entered and scanned against.

 

[ScrappyPaperDoodler]

With plagiarism detectors you at least get to see where the flagged text previously appeared. None of the AI detectors I've looked at (paid or free) are very transparent about the methodology they use. Because AI is the trendiest thing in Silicone Valley at the moment, I wouldn't be surprised if a lot of these firms are entering the market with half-baked tech and bad science.

 

[alohadave]

Because AI is the trendiest thing in Silicone Valley at the moment, I wouldn't be surprised if a lot of these firms are entering the market with half-baked tech and bad science.

It's the new Blockchain.

 

[onehitwanda]

With plagiarism detectors you at least get to see where the flagged text previously appeared. None of the AI detectors I've looked at (paid or free) are very transparent about the methodology they use. Because AI is the trendiest thing in Silicone Valley at the moment, I wouldn't be surprised if a lot of these firms are entering the market with half-baked tech and bad science.

Wanda's First Theorem: you could average the positions of nine orangoutangs in an enclosure over an hour and throw a dart at the resulting heatmap and then get a blind Soothsayer to read the location of the dart and consult the Auguries and it would STILL be a better system than AI-based AI detection.

 

[lilshymynx]

Something interesting that pops up in colleges is when teachers use plagiarism detectors, students get flagged for copying because previous work they've done has been entered and scanned against.

That's because self-plagiarism is a thing. If you quote someone, even yourself, you have to cite it.

 

[alohadave]

That's because self-plagiarism is a thing. If you quote someone, even yourself, you have to cite it.

It's more that they are getting flagged because their writing style doesn't change. They aren't self-plagiarizing.

 

[lilshymynx]

It's more that they are getting flagged because their writing style doesn't change. They aren't self-plagiarizing.

You may be referring to something else, but that's not what plagiarism detectors are for, and not what they do. They detect actual instances of directly copied work from other papers in their database and flag it.

That info typically gets used for 2 things:

1) Detection of plagiarism - and self plagiarism is a thing, and has nothing to do with writing style.
2) Having too high of a ratio of cited content vs. original content

 

[PandorasBachs]

I don't think they need to be transparent over which detector they're using. None of them are 100% accurate.

None of them are even 25% accurate at the moment.

 

[PandorasBachs]

Both of them are probably in the corpus so it stands to reason that the checker would recognize them.

Something interesting that pops up in colleges is when teachers use plagiarism detectors, students get flagged for copying because previous work they've done has been entered and scanned against.

Only if the teachers don't know what they're doing. Any plagiarism detector worth having allows a ton of flexibility about what actually gets reported as plagiarism and they're always adjustable after the fact if one of the initial settings proves to be problematic.

 

[DDX84]

I sampled the following paragraph,

Your paragraph is far too short.

Longer-form content, like erotic fiction, is easier for the models to test as it can have sight of the various sections authored by the human and by the software.

You blindly rely on ridiculously unreliable tools, inflicting an intolerable injustice upon your fellow writers

Your one-day old account hasn't had the opportunity to follow all the discussions in all the AI threads, so you can be excused for that remark.

Whether you like it or not, Literotica and all other publishers are relying on the tools right now. Everyone knows they aren't 100% accurate, but when used properly for testing, they aren't 0% accurate, as you've asserted.

That chart isn't mine, I linked it from an independent study, but Sapling had a 68% accuracy with no false positives when testing content of sufficient length and complexity. Erotic fiction comes into that category. Your tests don't meet that criteria.

I don't know what detector Literotica uses, but they confirmed they definitely use one. All they need to do is decide for themselves what percentage constitutes a positive result.

Then, those who disagree are free to leave this website.

 

[AG31]

I would be curious to know if you ever use Stockfish to get alternate ideas on how to deal with specific situations when practicing? You play chess online. You practice...
Same concept.

What is Stockfish? When I tried to go to Stockfish.com after a long delay I got "This site can't be reached."

 

[ShelbyDawn57]

What is Stockfish? When I tried to go to Stockfish.com after a long delay I got "This site can't be reached."

Stockfish is, or was, a chess site where you could play against an AI.

 

[AwkwardlySet]

What is Stockfish? When I tried to go to Stockfish.com after a long delay I got "This site can't be reached."

It is a chess engine, or to put it even simpler, it is the most advanced chess program there is, currently in its 16th version, if I am not mistaken. The program is a much stronger player than even the strongest among human players and the use of that program or any other chess program in human chess games is strictly prohibited. Sadly, cheaters often find a way to cheat, especially since those who should be caring about fair play are very lax about it.

 

[samhasstories]

It is a chess engine, or to put it even simpler, it is the most advanced chess program there is, currently in its 16th version, if I am not mistaken. The program is a much stronger player than even the strongest among human players and the use of that program or any other chess program in human chess games is strictly prohibited. Sadly, cheaters often find a way to cheat, especially since those who should be caring about fair play are very lax about it.

Was it somewhere in this forum that someone posted a story about a player cheating using vibrating anal beads? I guess the idea being that someone looked up a move - maybe on this Stockfish - and then communicated that to the player using a series of anal vibrations.

 

[AwkwardlySet]

Was it somewhere in this forum that someone posted a story about a player cheating using vibrating anal beads? I guess the idea being that someone looked up a move - maybe on this Stockfish - and then communicated that to the player using a series of anal vibrations.

It's not really a story. When the strongest chess player in the world, Magnus Carlsen, accused the American player Hans Niemann of cheating, Elon Musk tweeted a joke about Hans using anal beads as a way for his accomplice to transmit computer moves to Hans during a game. Some internet comedians actually tested the theory, coming up with a code, and with the help of the beads, one of them beat a chess grandmaster, even though he barely knew how to play chess. The chess world these days is full of drama

 

[DDX84]

https://www.theguardian.com/world/2...ed-of-title-after-defecating-in-hotel-bathtub

This guy too, but he took it a bit far.

 

[DDX84]

AlphaZero is by far the most advanced. Wanna get adopted in a 3-minute blitz match?

@AwkwardlySet is right. Stockfish 16 is the most advanced chess engine in the world at the moment.

 

[AwkwardlySet]

AlphaZero is by far the most advanced. Wanna get adopted in a 3-minute blitz match?

AlphaZero isn't available to the public and is set up to run on a supercomputer, while Stockfish is open source, uses an ordinary computer, and is available to the public for free. There is a reason why AlphaZero isn't ranked along with the rest of the chess engines. It would be unfair to compare the performance of two programs that are running on very different hardware in the sense of price, performance, and availability.

And no, Nakamura wannabe, I don't want to play.