Clarkesworld sci-fi journal overwhelmed by tons of of AI-generated tales

Clarkesworld, a science fiction journal, has banned new story submissions after receiving tons of of low-quality AI-generated items

By Matthew Sparkes

Science fiction journal Clarkesworld has halted story submissions after receiving a rising deluge of AI-generated items. The journal’s founding editor, Neil Clarke, says the issue has been created by individuals selling surprisingly capable AI language models resembling ChatGPT as a solution to earn cash from fiction publishing – regardless of the poor high quality of the AI tales.

“The machine-written submissions we’ve obtained are removed from publishable high quality,” he says. “I’m positive there are some which can be much less detectable, however the majority we’ve obtained have been simple for me to establish.”

Clarke says that he has talked to different journal editors who presently have the identical drawback, though he says they’ve been reluctant to talk to the press – as he was, till the issue grew to unsustainable ranges.

The journal usually has an open submission coverage to encourage new writers, however took the choice to shut submissions on 20 February after receiving 50 AI-generated story submissions that day. At that time, Clarkesworld had obtained 700 reliable submissions for the reason that begin of the month and 500 machine-generated ones, with the speed of enhance that means that AI-generated tales would quickly take over.

Clarke says the rise in “spammy” submissions started in the direction of the top of 2022, the identical time that accessible massive language AI fashions had been launched, and have elevated month on month. “[Those submitting AI-generated stories] don’t care about their status within the area. That makes it extra like malware or bank card fraud and faces comparable challenges. We’ll attempt to minimise these cases and they’re going to attempt to get round it,” he says.

Not all fee-paying science fiction magazines have seen the identical drawback, nevertheless. Djibril al-Ayad, editor at The Future Fire, says he has observed no spike in spam submissions, however which may be all the way down to being a smaller title. “I suppose the issue is that if your submissions pile triples in dimension and all of the AI stuff is abysmal however nonetheless must be learn, then it turns into a form of denial-of-service assault state of affairs,” he says, referring to attackers who knock servers offline by submitting enormous numbers of requests for information.

“My take could be to not be significantly frightened in regards to the hazard of by accident publishing AI-generated fiction,” he says, as most will likely be dangerous. “If it isn’t, then effectively… cool.”