Quote by paddlingincognito
I am curious… Why?
If you feel certain that some stories are AI written (or heavily AI assisted), why are mods letting them slide through? Why don’t you shut those stories down and/or ban those authors from the site?
I'll answer your first question first.
I get that too. It's something I also went through. When I write a story, I usually do 3-5 drafts, and I heavily edit each one. The story usually doesn't feel right until I get to that 4th draft, and then when I upload it to Lush, always end up doing some more minor changes as I read through it one last time before hitting submit.
Our stories don't have to be perfect, but the grammar and syntax should be at a level where they don't distract from the enjoyment of reading your story. We're all amateurs here, no one is setting an expectation of perfection. A hard lesson to learn (from a fellow perfectionist!) - over-polishing a story will ruin it more often than it will help it.
As for the AI issue...
It's a difficult question to answer, but I want to be as transparent as possible. The short answer is that it's an ongoing and ever-evolving conversation. There are a lot of angles to consider. We're also aware of the situation that is happening with Literotica and their writers, and we're taking measures to not repeat their mistakes. So while we're taking a more conservative approach, we're keenly aware of the problem and have discussed many different possible courses of action, trying to determine how to best deal with the problem. Our goal is to find an objective way to avoid confrontations ie provide objective evidence, not make accusations.
I've done A LOT of research in terms of using different AI detectors, and we haven't been happy or confident with any of the results. I don't want to say anything that will help AI writers, but the publicly available AI detectors are flawed and based on "x,y, and z" can give false positives or false negatives, so they don't set the standard that we want to have.
I have access to 2 different private AI detectors that are MUCH more reliable.
One was trained from the ground up by a group of professional Dungeons & Dragons writers because their jobs are being threatened by AI writing. That detector is incredibly accurate, but we don't have constant access to it. I can send them stories one at a time to check, but that's not a practical solution when we publish as many stories as we do.
I've also recently gained access to a detector that is being developed by the University of Waterloo. For those that don't know, U of W is sort of like Canada's version of MIT in that it's a university that is known for math, engineering, computer science etc.
What they have is incredible and funny. It's being made for students to scan their papers before they submit them, and because its target audience is students, the feedback it gives is really funny. I've dubbed it the 'Snarky AI detector' because it always says the funniest shit.
I don't know what the timeline is for that going public, or if it will actually go public, but right now I have a friend who is a librarian at a local college, and he's been given access to it as he's part of the stress-testing team. Its accuracy is outstanding.
And finally - and I realize how unconvincing this will sound to a lot of people, and few would accept it as being grounds to ban someone, especially a paid account - but the honest truth is that we read so many AI stories, a lot of us can tell just by reading them. We don't need fancy detectors, the AI tells come through in the writing. When you know what to look for, it's easy to spot and hard to ignore. I don't want to name or list things publicly, but these things are present in all AI stories, even ones that have been heavily edited by a human. The AI 'imprint' runs much deeper than most people realize, so just removing em dashes and other common features of AI writing doesn't actually hide anything from us.
I won't say what, but there are 3 things that I can look for, and if all three are present in a story, there's an 80+% chance (and I'm being generous, it's probabbly a 99% chance) the story was AI-written or AI-assisted.
And I keep saying this, but it's true - people who think that they're getting away with it because we're currently publishing those stories aren't fooling us. We're just talking a conservative approach to dealing with the issue.
You are FAR better off writing your own stories, developing your own skills, and building a following slowly over time, than using AI to leap ahead and then letting down your readers when they come to realize that you didn't actually write your own stories.
