What do celebrated films like Arrival, The Fly, The Thing and Total Recall have in common? They’re all based on short stories, a medium that’s long been a source of inspiration for Hollywood. But with the advent of ChatGPT and text generated by artificial intelligence, a celebrated short story publication has been so swamped with computer-crafted content that it’s closed its doors to new submissions.
Clarkesworld, an award-winning science fiction and fantasy magazine that typically keeps its submissions open year-round, tweeted this week that it would stop taking new stories amid the AI deluge. The magazine isn’t closing, but its publisher and editor-in-chief, Neil Clarke, doesn’t know when it’ll accept new stories. Clarke said he hasn’t figured out a nuanced solution to screening out AI-generated content that won’t shut out new writers.
The last year has seen an explosion of AI tools that take text prompts and generate things in different mediums nearly instantly, from visual art to music to movie segments. OpenAI released ChatGPT to the world back in October, giving users free access to an AI that could answer questions in plain language, and experiments led people to discover interesting applications, from creating recipe-packed meal plans to lowering bills. People started using AI to help make creative products intended for sale, like children’s books and graphic novels, which has raised concerns about artistic intent — for instance, the US Copyright Office just rescinded copyright protections for AI-generated art in a comic book.
But the short stories flooding Clarkesworld’s submission pile, constructed by AI tools reassembling existing stories, are drowning out the novel human-made tales that supply other creative industries with ideas they can adapt. Hollywood continues to harvest short stories for film and television projects, and if those funnels for new concepts are plugged up by an overflow of AI-generated content, the creative ecosystem could suffer. If AI stories get through editors, then movies may start feeling a lot more derivative.
This is another blow for Clarkesworld. Clarke said the publication was already struggling with the specter of lost revenue due to Amazon’s decision to discontinue digital subscriptions on the Kindle. Then it was blindsided with the flood of AI stories. What had been 10-20 stories submitted per month increased to 50 in December, then 100 in January before exploding to more than 500 sent in February, Clarke wrote in a blog post — and the month isn’t over.
The flood doesn’t seem to be coming from authors who dream of making it big in fiction. Clarke thinks Clarkesworld’s open submissions and pay rate have made it a tasty target for people using text-generating AI like ChatGPT, ChatSonic and JasperChat for get-rich-quick schemes.
“I can’t decide if I’m actually more frustrated with the side hustle gurus who are promoting this behavior, or the people that dropped these tools on the market,” Clarke said. “The ethics of both groups are questionable.”
Other editors publishing science fiction and fantasy that Clarke has spoken with haven’t been overwhelmed by AI stories. He theorizes that limited submission windows and policies like accepting stories only from established authors have kept them out of AI spammers’ crosshairs, at least for now. Indeed, fellow science fiction and fantasy publication Uncanny Magazine, which has limited submission windows, tweeted that it isn’t experiencing the same issues — but that it doesn’t necessarily think it’s immune.
Clarke tweeted that the solutions his publication has considered so far all have harmful side effects: AI detectors are unreliable, third-party tools for verifying author identities are pricey, charging entry fees to submit stories locks out legitimate authors, and only accepting print submissions isn’t viable.
“We’re going to have to find new ways to adapt and adjust on the fly,” Clarke said. “It will be a mix of technological and social solutions.”
Ironically, those people submitting AI-generated content have “seriously no chance of those works being published here,” Clarke said. In his blog post on the issue, he noted there are “very obvious patterns” that give away that a story was AI-crafted. But he added that this AI spam, as he calls it, is only scratching the surface of the potential issues caused by text-generating AI — a swath of unexplored legal and ethical territory around the tools and what data they’re trained on.
Yet keeping the doors open to new authors in the US and internationally remains Clarkesworld’s priority.
“The field is always changing and we have to as well,” Clarke said. “The trick is keeping your soul in the process.”
Editors’ note: CNET is using an AI engine to create some personal finance explainers that are edited and fact-checked by our editors. For more, see this post.
Read More
David Lumb