A journey into Kindle AI slop hell

Royal Road is one of the platforms for indie authors to publish serialized web novels. Many people use that to jump into a full time writing career -- successful new serials get on a ladder, paetrons get set up, and when enough chapters are written, they get edited and published to Kindle. The publication to Kindle Unlimited ends up with the books getting stubbed. In return though, you have a kind of presale so you know people like reading the work, and a sufficient fan base to appeal for support for a successful launch on KU.

One of the things I've been seeing the authors reporting to their subscribers are people stealing that work by scraping the chapters, running it through LLM, and then publishing it on KU. Authors have been adding watermarks into the text, though I don't know how successful that is.

(On the other hand, many authors use generative image AI to create covers, which has angered artists whose work has been sucked up by the generative AI machine).

I'm getting more and more convinced that tech people are living in a dangerous bubble re: the generative AI boom. I've been talking to people outside this bubble, and the unanimous response to all the "exciting" AI developments of the last ~3 years has been that their experience of generative AI has been purely negative with no upsides whatsoever. All they see is the content slop, the shitty search results, degraded experiences in e.g. customer service, and so on.

And it's starting to alarm me that nobody in tech appears to care about this and is just going "damn the torpedoes, full speed ahead". That kind of arrogant dismissal of popular mood and forcing unwanted change on people is how resentment and revolution happen.

> Is Amazon cutting out the author middle man and using A.I. and user data to generate books on its own?

Amazon has done just this elsewhere: https://techcrunch.com/2022/11/29/amazons-alexa-ai-animated-...

By the by, I’ve enjoyed my Kindle a whole lot more since turning on its airplane mode. I connect to WiFi to sync new books to it, but don’t give my Amazon overlords other opportunities to present AI slop from the vaults.

> This week’s edition is a guest post about spooky Kindle A.I. slop from Leah Beckmann, an L.A.-based screenwriter and journalist and Chief Kindle Bullshit Correspondent for Read Max.

I had to reread this a couple times before understanding that the guest post - not the A.I. slop - was from Leah Beckmann.

I've been getting these too — they've largely supplanted the plausibly-human-written glurge I used to get (mil-SF and disturbingly specific romance subgenres). I don't think I've given Amazon any reason to believe I'd want them, though. My guess is that these are just the highest-expected-profit books that Kindle has, and it's what you get if Kindle doesn't have a more specific signal for you.

Hit up support and ask them to remove the lock screen ads, even if you didn't buy the "ad free" kindle support usually does it.

It worked for me.

> What else could this sick and twisted mind have cooked up? I wanted more. I clicked on Bette’s author page; this was the only title listed. Through the penumbral fog of half-sleep, squinting one-eyed at my phone, I googled, “a girls.quest for healthy watinf bette santinir.” Even the politely auto-corrected version yielded no meaningful results

> And then things got spooky. When I opened my Kindle again, there, illuminated by the inoffensive whitish glow of my device, was an ad for A Girl’s Quest for Healthy Eating. Only, it was vaguely distorted. Like a spot-the-difference Highlights game, here was the same book with minor discrepancies. This cover featured several little girls, presumably all on healthy eating quests of their own. The page count was slightly different. And most distressing of all, the author's name: Bette Santinir. In other words, Santini plus R. Where had I seen this name before? Oh right: from one minute ago when I misspelled Santini, Santinir.

> As the very editor of this Substack texted me, “If you didn’t have pics of this, I would think you were schizophrenic.”

I really wish the supposed pictures had been included in the piece.

I've tried to use generative text LLMs myself for story ideas, filling out backstory in characters, etc. No matter how good the model is, no matter how much context you supply or repeat, it will inevitably spiral down into some degree repetition and forgotten story beats.

But hey, at least it's all technically grammatically correct. Most of the time.

What disturbs me more is the thought of lost context in things like code, medical notes, and actually critical workflows.

AI was a mistake

A company grows big and then turns into shit and this cycle keeps repeating I guess. I love playing games. Ubisoft, EA (in particular) etc are now shit. How long do we have to enjoy Steam (and other Valve products) until it suffers the same fate? Or will it remain good until some influential boss leaves?

I think while praising a business and getting excited we should have their average age in back of our minds.

Today I drifted in AI generated dread feeling that half of what I read (including this article) was AI generated -- maybe more since it included AI generated music. Was it due to my recent return from some number of days in the Oregon outback off the grid with no interweb pipes? I knew it was too short a time. I feel a great panicky need to run from an impending Matrix moment.

I can't help thinking there's a huge opportunity for discussing this side of AI in a more principled way. When something (i.e. AI) is this viscerally annoying, it's tempting to slip into this style of discussion, I get it. But it's......hard to read even when you agree and smile along the way.

"The brief bio read: A little girl with blue eyes and blonde hair leads her friends on a healthy eating mission, culminating in the creation of a neighborhood farmer’s market. Okay. Sort of Triumph of the Will: Bedtime Story for Kids and Adults." (for context, Triumph of the Will is the name of a famous Nazi Germany-produced propaganda documentary)

To flesh it out a bit more with examples, to make it clear it's not cherry-picking:

- "What generative models are being used? What kinds of prompts are generating these texts? I don’t know. I’m not a virgin. As I said, I have a child." (virgin really is a complete non-sequitor here, unless the idea is only virgins use ChatGPT?)

- "But I did want to read about my nice Nazi Youth friend and her farmer’s market journey"

- "More importantly, what is a bedtime story for kids and adults? I am an adult. A bedtime story for an adult is just a real book. I read real books. Or at least, I used to? Basically, did Kindle think I was stupid now?"

amazon has another battle on ai-generated listings on their marketplace as it is (https://news.ycombinator.com/item?id=38969417).

while highly unethical and a nuisance to navigate as a user, i respect the hustle compared to drop shipping bs people speak about on social media.

the bottom line is that economically these are the parties most likely to find a way to filter out ai-generated content going forward. or they will die trying.

AI is still pretty terrible at maintaining coherence over long texts.

So after experiencing this nightmare she will still give her son a reading tablet someday?

>What kinds of prompts are generating these texts?

>I don’t know. I’m not a virgin. As I said, I have a child.

>But I did want to read about my nice Nazi Youth friend and her farmer’s market journey

Are they capable of writing without snark?

The intro section before the main article was worse than slop.

Loved Leah’s writing otherwise.