Amazon's Kindle Direct Publishing platform faces a crisis of AI-generated non-fiction books flooding the marketplace, according to The Nerdy Novelist, who documented the systematic abuse destroying legitimate authors (both AI and human) and their chances of success.
When people are angry about “AI slop,” I think this is the legitimate target, churning out piles of effortless, useless nonsense that drowns any real signal in noise.
I recall that for some time now people were saying personal branding was going to supplant central market control. As in people would buy because you're THE Mr. X, not because Amazon said it was good for you. "The brand of Me". Seth Godin and Rollo Tomassi were big on that prediction on the Marketing side, Vox was big on it from the Publisher side and I don't recall the names but the artist for Immortal Hulk who's making an Arkhaven comic and that Sad Puppies fiction writer on the artist and writer side.
I think it's actually a reversion to the norm in the grand historical sense, where you bought a painting because it was 'a Picasso' or a book because it was 'the latest Agatha Cristie'. There's a spectrum of course from individuals to brands like McDonalds to centralized slop like Amazon choice. How the reversion is going to shake out I don't know but I think Amazon has no incentive to stop the AI plagiarism - what does Amazon care for but profits and paying as little to the author as possible? If it's a dalit Indian using AI to re-write 50 Shades of Grey for 50 cents then so be it.
Interesting. I tried to create a story from a large body of source material using ChatGPT, and even though I repeatedly gave it a target word count, it ignored the instruction every time. In the end, through repeated prompting, I got to the length I wanted. Maybe non-fiction would be different.
I stopped for now because at some point ChatGPT limited the length of the canvas to roughly 60% of what it wrote, and I don't feel like working in smaller chunks right now.
"But the human must always always be in charge of the thinking"
This is the same as my over-brain/under-brain concept with coding with AI. Any logic it uses must be on my instructions, or it is going to make average gargbage.
AI is predominantly trained on social content (80% social/blogs, 15% professional, 5% code/math), and most of that social conten is from 95-115 IQ people, who are emotionally processing or messing around online.
That's the training dataset for AI, so if you aren't controlling it strongly, then you are getting strictly average messing around output from mostly-idiots.
Same problem in YouTube vids: massive increase in stupid plebian "feel-good" take stories. I've taken taken to clicking the 2-buttom menu and telling YT to "Don't show this channel." Early on it was trying to stop a flood with paper towels, but now the algos seem to block a LOT of that type of non-story... Rarely get any. Same with cute kittens, toddlers acting like adults/politicians, girls rescued by love, and folks describing their marriages... Most of the time, you can tell JUST by the AI thumbnail and topic...
These books are all over Amazon now. But if you do some basic research on the author's credentials and read the free sample pages, it's pretty easy to figure out whether the book is a generic, high-level AI regurgitation or something that delves deeply into a subject in a thoughtful and unique way. The same problem exists with fiction on Amazon, and the due diligence works there as well.
It doesn't really bother me. I treat authors like mechanics. I'm not trusting an author unless I know him, or I get recommendations from trustworthy sources. Those books won't get on my radar anyway. In fact it's first time I'm hearing about this being an issue.
Funny the first part my step daughter did at least a year ago where she bought a college textbook but it was delayed in shipping and she needed it for her class so she used that exact technique to reproduce several chapters. It was surprisingly accurate! She showed them side by side as it were.
Having a name and reputation is likely going to be more important in the future so that you have some sort of quality assurance.
I mean, if I could just tell AI to summarize a book to extract information, synthesis, and conclusions then I wouldn't buy an AI written book doing just that. It seems like a non-issue, especially considering the state of much non-fiction work produced nowadays which I wouldn't bother with even summarizing, except to refute. The vectors of information laundering are so expansive with modern media that I'm not personally concerned with AI summarized nonfiction work that may or may not hallucinate details.
I've noticed the same thing on YouTube, where AI has written scripts that voicelabs or 11labs fill in.
the voices are generic and the information boring and repetitive.
This post reads like someone fed the video into AI and asked for a summary.
When people are angry about “AI slop,” I think this is the legitimate target, churning out piles of effortless, useless nonsense that drowns any real signal in noise.
I recall that for some time now people were saying personal branding was going to supplant central market control. As in people would buy because you're THE Mr. X, not because Amazon said it was good for you. "The brand of Me". Seth Godin and Rollo Tomassi were big on that prediction on the Marketing side, Vox was big on it from the Publisher side and I don't recall the names but the artist for Immortal Hulk who's making an Arkhaven comic and that Sad Puppies fiction writer on the artist and writer side.
I think it's actually a reversion to the norm in the grand historical sense, where you bought a painting because it was 'a Picasso' or a book because it was 'the latest Agatha Cristie'. There's a spectrum of course from individuals to brands like McDonalds to centralized slop like Amazon choice. How the reversion is going to shake out I don't know but I think Amazon has no incentive to stop the AI plagiarism - what does Amazon care for but profits and paying as little to the author as possible? If it's a dalit Indian using AI to re-write 50 Shades of Grey for 50 cents then so be it.
Interesting. I tried to create a story from a large body of source material using ChatGPT, and even though I repeatedly gave it a target word count, it ignored the instruction every time. In the end, through repeated prompting, I got to the length I wanted. Maybe non-fiction would be different.
I stopped for now because at some point ChatGPT limited the length of the canvas to roughly 60% of what it wrote, and I don't feel like working in smaller chunks right now.
"But the human must always always be in charge of the thinking"
This is the same as my over-brain/under-brain concept with coding with AI. Any logic it uses must be on my instructions, or it is going to make average gargbage.
AI is predominantly trained on social content (80% social/blogs, 15% professional, 5% code/math), and most of that social conten is from 95-115 IQ people, who are emotionally processing or messing around online.
That's the training dataset for AI, so if you aren't controlling it strongly, then you are getting strictly average messing around output from mostly-idiots.
Same problem in YouTube vids: massive increase in stupid plebian "feel-good" take stories. I've taken taken to clicking the 2-buttom menu and telling YT to "Don't show this channel." Early on it was trying to stop a flood with paper towels, but now the algos seem to block a LOT of that type of non-story... Rarely get any. Same with cute kittens, toddlers acting like adults/politicians, girls rescued by love, and folks describing their marriages... Most of the time, you can tell JUST by the AI thumbnail and topic...
Who is buying random books like this? The very uninformed.
They likely would never think to ask an LLM.
This stuff doesnt look like history, but something like self help and such.
At some point, the AI will use the AI written books to make more low quality again until it's just a black hole of garbage.
These books are all over Amazon now. But if you do some basic research on the author's credentials and read the free sample pages, it's pretty easy to figure out whether the book is a generic, high-level AI regurgitation or something that delves deeply into a subject in a thoughtful and unique way. The same problem exists with fiction on Amazon, and the due diligence works there as well.
A few law suits will sort this out.
Oooh, you mean, like, trust the COURTS!?
Yeah, mebbe not.
It doesn't really bother me. I treat authors like mechanics. I'm not trusting an author unless I know him, or I get recommendations from trustworthy sources. Those books won't get on my radar anyway. In fact it's first time I'm hearing about this being an issue.
Funny the first part my step daughter did at least a year ago where she bought a college textbook but it was delayed in shipping and she needed it for her class so she used that exact technique to reproduce several chapters. It was surprisingly accurate! She showed them side by side as it were.
Having a name and reputation is likely going to be more important in the future so that you have some sort of quality assurance.
I'm also thinking being a "known human creator" is about to be a very big differentiator.
Process videos are the next thing to be faked. "Watch me type this 5000 page novel, with 10.5 fingers"
I mean, if I could just tell AI to summarize a book to extract information, synthesis, and conclusions then I wouldn't buy an AI written book doing just that. It seems like a non-issue, especially considering the state of much non-fiction work produced nowadays which I wouldn't bother with even summarizing, except to refute. The vectors of information laundering are so expansive with modern media that I'm not personally concerned with AI summarized nonfiction work that may or may not hallucinate details.
If anything, generative AI would probably create better work than modern historians if asked to synthesize from the same primary sources