Arstecnica
One of the side effects of unlimited content creation machines (generative AI) is unlimited content.Monday, editor of his famous sci-fi publication Clarkesworld Magazine announced Due to a significant increase in machine-generated stories being sent to publications, he temporarily closed posting stories.
in the graph share on twitter, Clarksworld editor Neil Clark tallied the number of banned authors submitting plagiarism or machine-generated stories. From a low baseline, the total count for February is now 500. The increase in banned sending roughly coincides with his ChatGPT release on November 30, 2022.
Large Language Models (LLMs) such as ChatGPT have been trained on millions of books and websites to quickly create original stories. However, they do not work autonomously and require a human to guide the output with prompts that the AI model attempts to complete automatically.
Since 2006, Clarkesworld has published famous science fiction authors and has won several Hugo Awards. He is well known among science fiction publications for his open submission process, and he typically pays 12 cents per word. The post page states, “At this time, we are not considering stories written, co-authored, or assisted by AI.” But that hasn’t stopped the number of submissions from rising dramatically, which Clarke believes is mostly due to get-rich-quick schemes.
“It’s people outside the SF/F community that are causing problems.” I have written Clark on Twitter. “Mostly driven by ‘side jobs’ professionals who claim they can make easy money on ChatGPT. They are pushing this and deserve some of the disrespect shown to AI developers. .”
At press time, a quick YouTube search for terms like “get rich with ChatGPT” and “make money with ChatGPT” returned many results, but did not identify any videos pointing specifically to Clarkesworld.
Arstecnica
The problem with AI-generated content isn’t unique to Clarkesworld. On Tuesday, Reuters wrote a report on the rise of AI-generated e-books on her Amazon. Reuters has identified over 200 of her e-books in the Amazon Kindle store that list ChatGPT as an author or co-author.
The influx of AI-generated content has put Clarkesworld in the awkward position of trying to keep the bar for submissions high enough to keep spammers away, but the world’s identification of potentially unjustified not high enough to discourage undiscovered authors and authors from the region. Subject to geographic ban.and series of tweetsClarke described his predicament:
There is no solution for your problem. I have some ideas to minimize it, but the problem is not solved. Detectors are unreliable. Pay-to-Submit costs too many legitimate authors. Submission of printed materials is not feasible for us. Various third-party tools for identity verification are more expensive than magazine budgets and tend to be holed by region. Adopting them is the same as banning an entire country.
It was easy to implement a system that only allowed authors who had previously submitted work. That would effectively ban new authors, which is unacceptable. They are an integral part of this ecosystem and our future.
So far, tools aimed at detecting text written by LLM have poor accuracy (often returning false positives when tested with human-written text) and are not currently a viable solution. repeat. Despite these problems, according to Clark, the journal has not closed and submissions will resume in the future.
In a blog post last Wednesday, Clark said, “Not only will it just go away on its own, but there’s no solution.” No. The best we can hope for is to have enough water to stay afloat. In the meantime, Clark encourages those who want to support the magazine to subscribe.
