Neil Clarke of Clarkesworld has announced they’re closed to submissions.* The reason, they’ve been swamped by AI-created stories overwhelming them.
This seems to be a real problem for magazines. There is no good solutions. No one wants to restrict submissions to established writers or to charge to submit. Some have suggested going back to mail submissions. Programs to detect AI stories are not reliable.
I hope this can be addressed quickly.
*The magazine is not in trouble and is not thinking of folding.
Why would you want to weed out the AI submissions? Some of them might be good. You want to weed out the bad submissions. If there are too many submissions, either you need an AI that can reasonably reliably assess the quality of a story well enough to act as a gatekeeper, or pass the buck by only accepting material via reliable agents.
As I said in my OP, AI detectors are too unreliable.
And requiring an agent means that perfectly good writers – especially those who are just starting out – who can’t get one are locked out. Note, too, that agents don’t usually rep short stories – not enough money in them.
I’d love to see an AI story that was anything other than slush material.
In one Clarkesworld submission, Clarke writes, someone submitted a story with the following sentence: “Sitting on its three years’ experience, the fittest Shell was originally the size of more android subliminal observations than any other single subject in the Grandma.” The submission was reconstituted from a story published in 1956.
Yeah, I’ve sold stories (and submitted (but not sold) to Clarkesworld), but it’s very unlikely I’ll ever have an agent. I had heard (this weekend) that Clarkesworld was getting a lot of AI submissions, but hadn’t heard that they were closed to submissions (hopefully temporarily).
By the way, in other bad news for short-fiction writers, Ralan.com, which had been compiling lists of markets online for 20-some years, shuttered this month.
How long before someone trains a bespoke AI to grade stories? A publisher already has the corpus of everything they’ve published, which is the perfect training set for what that publisher is looking for. Let the AI go through the slush pile and give the top manuscripts to the editors to make the final decisions.
A publisher is in the business of printing stories that sell. While there might be a market for “organic” stories, a good story stands on its own merit for salability, regardless of its origins.
…in order to “print stories that sell” you need stories that appeal to the Clarkeworld target market.
How exactly would AI stories do that?
I can write an AI story right now if I wanted to. And it would be formulaic and feel like its been copy and pasted because it practically has been.
If you think AI generated stories are gonna sell, then the market is right there for you now. You can build a website quickly on squarespace. Start getting submissions. Start publishing. Good luck!
But it would be unwise for a publisher to adopt overnight a completely new business model, marketing a product that is untested in the marketplace, on their audience that is currently happily consuming the content that is being provided.
What’s a good story? Who’s your favourite author, and why did you enjoy their work over other people in the genre?
I would think plagiarism would be a very good reason to be very cautious about publishing AI generated stories. And I’m not sure they have the capacity to waste checking them over.
If the AI submissions are actually as good as human writing, sure. I think the problem is that they are still ultimately shit, but not so obviously shit that it’s not a huge amount of work to read through a few pages of every one of thousands of submissions and reach that conclusion.
I’m certain they don’t want a hit to their reputation if they accept something that turns out to have been written by an AI. Current detectors might let something slip through that future detectors will catch, and it’s a certainty that some readers will be feeding all kinds of published text into ever-improving detectors. Detectors might become browser plugins that automatically flag AI-generated text. Heck, after thinking of that, I want one right now.
…then why did you say “Why should they refuse a story their buyers want because of how the story was written?”
I’m sure Clarkeworld are more familiar with their customer base than any of us. I haven’t seen anybody crying out for AI science fiction stories.
I think the editors should be making the decision based on the editorial policy on their magazine, and not do anything that could potentially piss off their current readership. It would be utter foolishness, especially considering how bad the technology is at the moment, to start publishing something that hasn’t been tested in the market yet.
Outside of clickbait blogposts and SEO friendly copywriting AI isn’t really good at this stuff yet, if ever. Its formulaic, because its the product of formula.
Confusing (to me) phrasing here. In the original question, are the editors refusing stories because of how they were written, or do the buyers want them because of how they were written?
…I’m not even sure what this means. But I don’t have many posts in this thread, and I’ve used the quote feature to show the posts I’ve responded to in context, so perhaps just start there.
Could it also be that the editors are just swamped and can’t get to many submissions since there are now a lot more than usual?
There’s also the issue of legal rights - can Clarkesworld legally publish the product of an AI owned by a third party who is not identified as the author?