ChatGPT disaster: SF mag Clarkesworld closed to submissions. Amazon Kindle flooded.

[highplainsdem]

Welcome to the dystopia of AI being used for "creative" writing by people trying to make a quick buck.

I've been posting about this in GD, but with all the news stories in the past 24 hours, I thought there should be a thread on it here.

This is my latest OP on it in GD: https://www.democraticunderground.com/100217666418 . See reply 17 there for links and excerpts from news stories about Clarkesworld and Amazon Kindle.

This is the OP I posted 6 days ago, after seeing Neil Clarke's first posts on his concern about the flood of AI-generated work, when he was still hoping he wouldn't have to shut off submissions: https://www.democraticunderground.com/100217651861

But 12 days before that, I'd posted about the trouble ChatGPT and other AI would cause: https://www.democraticunderground.com/100217612552 . See the second half of that OP.

I'm not happy that I was right. This AI mess hurts editors and writers, especially aspiring writers. But anyone seeing the news stories about people trying to exploit AI, to try to pretend to be creative when they aren't, could have guessed what would happen.

Amazon currently isn't refusing AI-generated fake writing. I suppose it's fine with them, no matter how bad the books are, if they produce any revenue at all. But there isn't a greatly increased audience for those books, so real writers will be hurt by the flood of AI-generated competition. A lot of which isn't even identified as such. And since ChatGPT plagiarizes, authors whose work is ripped off are also being hurt. AI can be told to write in the style of particular writers.

Neil Clarke had the integrity, and the sense, to ban anyone who submitted AI-generated "work."

Every publisher should do the same.

[John1956PA]

1. Digital pulp.

Maybe some of it make for good reads, but I would rather stay with flesh and blood authors.

[Scrivener7]

6. How can we tell, though?

[OAITW r.2.0]

2. How can you tell what is AI generated vs. something a real human writes?

Understanding how to differentiate the two would help normal people in understanding how this tech works.

[highplainsdem]

3. It isn't always easy to, which is why it's creating huge problems for

teachers as well, so some school systems have banned it or tried to.

OpenAI released ChatGPT and they know it's being used for cheating and fraud. They're trying to come up with something to detect it, but the detector they've produced so far is a miserable failure. OpenAI is also conflicted on this, since their initial free version of ChatGPT has two premium successors, better versions, priced at $20 and $42 a month. Lots of students and wannabe writers who don't want to do the work will think those are essential.

AI is hurting artists, too, maybe especially one fantasy illustrator whose name has been used around 100,000 times in prompts by people trying to copy his work. He found out months ago that a search for his name on Google turned up more crappy imitations than his own work. See https://www.technologyreview.com/2022/09/16/1059598/this-artist-is-dominating-ai-generated-art-and-hes-not-happy-about-it/ .

Google has developed a very sophisticated program to "create" music with all sorts of prompts, even paintings, that will trigger the AI to go through everything it's scraped or ripped off. But they haven't released it yet since they know damn well there are copyright problems.

And Meta released a science AI in November, trained on lots of science data and documents, that had to be withdrawn in three days since it produced authoritative sounding garbage, full of mistakes and imaginary references that sometimes used real scientists' names. It recommended eating crushed glass, for instance.

ChatGPT gets facts and references wrong, too. But with fiction, something that seems bizarre might be meant as surreal, or just an unusual style.

[OAITW r.2.0]

4. I get the problem, I just don't see the tools to detect it. nt

[highplainsdem]

7. Neil Clarke has said there aren't AI detectors that work well enough, but

there are some patterns he's noticed:

http://neil-clarke.com/a-concerning-trend/

I’m not going to detail how I know these stories are “AI” spam or outline any of the data I have collected from these submissions. There are some very obvious patterns and I have no intention of helping those people become less likely to be caught. Furthermore, some of the patterns I’ve observed could be abused and paint legitimate authors with the same brush. Regional trends, for example.

Maybe he's told other editors what to watch out for. I hope editors are sharing this info, but it's best not to make it public.

Here's an article from a month ago on how well various detector tools work:

https://www.digitaltrends.com/computing/how-to-detect-chatgpt-plagiarism/

[highplainsdem]

8. More recent article on detection tools:

https://theconversation.com/we-pitted-chatgpt-against-tools-for-detecting-ai-written-text-and-the-results-are-troubling-199774

[ymetca]

5. More barbarians at the gate

desperate for the gatekeepers to disgorge some cash.

The only reason Amazon exists at all is because the book publishers and retailers consolidated, price-gouged, made only a handful of authors absurdly rich, and churned out endless piles of garbage.

You do realize that the vast, vast majority of Stephen King's published books are decaying in landfills now, right?

I'd rather save the trees.