cross-posted from: lemmy.dbzer0.com/post/74144681

A 404 Media investigation was able to reveal Amazon’s book buying operation, which hasn’t been previously reported, by placing a tracking device in a rare book we suspected would be acquired by an AI company for training data, and following it around the country to its final destination.
That final destination was an Amazon warehouse in Las Vegas, Nevada. Amazon employees who work at this location say all they do is receive massive shipments of printed books which they then cut the bindings off in order to scan the books more quickly. The printed book is destroyed in the process. The logo of the Amazon team that works at this warehouse, called VGT3, is a dinosaur, brandishing its teeth and with a book in its hands.
“Amazon purchases books through commercial channels to help develop and improve the products and services our customers use,” an Amazon spokesperson told me in a statement.
The world’s AI companies are constantly looking for, and spending extreme resources to locate, more material to train their AI models. With books, that sometimes means destroying them in the process, something that large parts of the public have spoken up against, and which we can now confirm Amazon is doing.
In July, I published a story about booksellers who reported a historical spike in sales starting in the past year. They suspected this spike in sales was due to AI companies acquiring any books they can in search of new training data. Printed books are valuable as training data because a lot of the text they contain is not readily available on the internet, which AI companies have already scraped. The data is also conveniently organized and, if the book was printed before 2022, is guaranteed to be free of AI-generated text, which can make any AI model that is trained on it worse via a recursive process called “model collapse.”
📖
Do you know work at a facility where you scan books? I would love to hear from you. Using a non-work device, you can message me securely on Signal at @emanuel.404. Otherwise, send me an email at emanuel@404media.co.
These booksellers suspected AI companies were behind these large bulk purchases because of the high number of books they were buying, the seemingly random choice of books, and the fact that these buyers, unlike libraries and universities, did not seem price sensitive at all. But booksellers couldn’t say for certain who was behind the large purchases because the marketplaces where they sell their books keep the buyers anonymous. When an order comes in, a bookseller ships the sold books to a warehouse operated by the marketplaces, where books are sorted and then sent to the buyer.
In July, one bookseller told me they received a very large order of around 1,000 books on Biblio, one of these marketplaces. The seller agreed to put an Apple AirTag provided by 404 Media in one of the books included in this order so we could see where the book was going. And by extension, which company, AI or otherwise, was behind this massive order.
404 Media granted the bookseller anonymity because they worried sharing this information would harm their business. Biblio did not respond to a request for comment.

Hedgewitch Part 5: Upsy-Daisy!

Using it for what it’s good at was an obvious approach. Turning it upside-down was good.

What About a Little Kid?

Treating it like a little kid and walking it through something, like a math problem, also looks promising.
I wanted to draw a particular mathematical figure. So I told the three-year old to

1. search math books for a section about “lower hull”,

2. show me the algorithm, please,

3. change it to to choose the rightmost part, while showing me your work,

4. OK, now write it in Go, and

5. test it with the data from the worked example in the math book.

Being in a math book looked like it was the magic trick. It felt like I was avoiding most of the errors, and all of the direct lies, by restricting it to a pre-selected grimoire that I knew was correct.

Hypotheses …

1. If you already know about something, you can use LLMs. That’s like me using lint on work that is something I haven’t done for a long time
2. If you restrict the input it uses to something that’s true, it definitely helps. That seems to work even if the LLM doesn’t “know” which math books are needed.

I strongly suspect there are others: I just haven’t found them yet.

Links

This is part of a multi-part look at the Gartner hype-cycle of LLMs, and where I found my part of the “Plateau of Productivity”.

• Hedgewitch Part 0 is Torches and Pitchforks. That’s the back-story.
• The main story starts with Hedgewitch Part 1: LLMs, about why I’m not a wizard.
• Next is Hedgewitch Part 2: My Favorite Errors, and what I do about them.
• Then Hedgewitch Part 3 LLMs Should Challenge, Not Obey, about the magic that makes it work.
• Then Hedgewitch Part 4: How do I know that LLMs are buggy?
• This is Hedgewitch Part 5: Upsy-Daisy
• The next will be Hedgewitch part 6: What would Dave like me to say?

#ai #artificialIntelligence #llms #technology

This Blog has No Leaves

2026-05-31 12:29:01

Torches and Pitchforks

The peasants are encircling the palace, with torches and pitchforks

The king is a 3-year-old who’s memory has been stuffed with every book ever written (on vellum, this is from a while ago).

But they’re a three-year old.

• They babble incessantly: about animal husbandry, in the middle of a city.
• They don’t know true from false: they have lots of ordinary books with mistakes in their head. And some spell-books that were full of lies, deliberately designed to keep ordinary people from learning wizardry.
• Worst, they don’t know right from wrong: some of the characters in their head are happily, unabashedly, evil.

Everyone had thought the little king was omniscient. Wrong. He’s a dangerous, babblingchild.

Now they’re disillusioned. They’re sure that the king, and the wizard that created him, are villains. They’ve dammed off the stream that the palace depends on for water, and set it on fire. They’re waiting for the king to flee. To set him on fire

Meanwhile, a little hedge-witch is getting afraid to cure peasants with injuries. Her magic injects health into people, not injects things into their heads.

But she might burned to death, too.

#artificialIntelligence #technology

Hedgewitch 2: My Favorite Errors

I have some bad habits I speak in run-on sentences never get to the point.

I don’t mind that in speech, but it’s not good in writing. So I need to do something about it if I want to learn how to be a better writer. So I started with the low-hanging fruit: teaching me to recognize and avoid my favorite errors.

First, I use an LLM for a grammar-book

“Could you look and-splices in the following?
“Could you look for comma-splices in the following?
“Now could you show me the remaining grammar errors?”

What I get back is something like this:

Subject-Verb Agreement:

• “How many steps did we had in the pipeline” → should be “How many steps did we have”

Typos/Missing Letters:

• “I had bee asked to test” → should be “I had been asked”

Verb Tense/Form Issues:

• “But it did it so via converting” → awkward construction, should be “But it did so by converting” or “But it did this by converting”

Step 2: a Sanity-Check

I do a pair of sanity-checks that expose structural problems the unreadable prose was hiding. The first is to ask Can you outline the structure of the argument?

Here’s the structure of the argument:

I. Problem Setup

• Company culture: “live with what you’ve got”

II. Discovery of Issues

• Pipeline has grown to 9 steps with multiple problematic components. Only 4 were required

III. Investigation

IV. Revelation

V. Resolution & Lesson

That is a test to see if what I’ve written matches the structure that I have in my head. In this case it did. If it doesn’t, I have some refactoring to do.

The second, the cousin to that question, is Can you summarize this in a few lines?That is the “have I written what I thought I wanted to say?” part.

This is where I usually catch failures to say what I meant. It also exposes parts that blithely assumed the reader knows everything I do. That’s never the case.

Links

This is part of a multi-part look at the Gartner hype-cycle of LLMs, and where I found my part of the “Plateau of Productivity”.

• Hedgewitch Part 0 is Torches and Pitchforks. That’s the back-story
• The main story starts with Hedgewitch Part 1: LLMs, about why I’m not a wizard.
• This is Hedgewitch Part 2: My Favorite Errors, about how I got started.
• Then comes Hedgewitch Part 3: LLMs Should Challenge, Not Obey, about the magic that makes it work.

#ai #artificialIntelligence #grammar #llm #technology #writing

This Blog has No Leaves

2026-05-31 12:29:01

Torches and Pitchforks

The peasants are encircling the palace, with torches and pitchforks

The king is a 3-year-old who’s memory has been stuffed with every book ever written (on vellum, this is from a while ago).

But they’re a three-year old.

• They babble incessantly: about animal husbandry, in the middle of a city.
• They don’t know true from false: they have lots of ordinary books with mistakes in their head. And some spell-books that were full of lies, deliberately designed to keep ordinary people from learning wizardry.
• Worst, they don’t know right from wrong: some of the characters in their head are happily, unabashedly, evil.

Everyone had thought the little king was omniscient. Wrong. He’s a dangerous, babblingchild.

Now they’re disillusioned. They’re sure that the king, and the wizard that created him, are villains. They’ve dammed off the stream that the palace depends on for water, and set it on fire. They’re waiting for the king to flee. To set him on fire

Meanwhile, a little hedge-witch is getting afraid to cure peasants with injuries. Her magic injects health into people, not injects things into their heads.

But she might burned to death, too.

#artificialIntelligence #technology

Torches and Pitchforks

The peasants are encircling the palace, with torches and pitchforks

The king is a 3-year-old who’s memory has been stuffed with every book ever written (on vellum, this is from a while ago).

But they’re a three-year old.

• They babble incessantly: about animal husbandry, in the middle of a city.
• They don’t know true from false: they have lots of ordinary books with mistakes in their head. And some spell-books that were full of lies, deliberately designed to keep ordinary people from learning wizardry.
• Worst, they don’t know right from wrong: some of the characters in their head are happily, unabashedly, evil.

Everyone had thought the little king was omniscient. Wrong. He’s a dangerous, babblingchild.

Now they’re disillusioned. They’re sure that the king, and the wizard that created him, are villains. They’ve dammed off the stream that the palace depends on for water, and set it on fire. They’re waiting for the king to flee. To set him on fire

Meanwhile, a little hedge-witch is getting afraid to cure peasants with injuries. Her magic injects health into people, not injects things into their heads.

But she might burned to death, too.

#artificialIntelligence #technology

Hedgewitch Part 3: LLMs Should Challenge, Not Obey

Most people treat LLMs like an obedient secretary. I treat them like lint — a fallible tool that suggests mistakes I made, ones I evaluate for myself.

I stole the entire idea from Advait Sarkar. It was in the October 2024 Communications of the Association For Computing Machinery.

LLMs:

• spot errors easily,
• write really bland prose, and
• use the “average” of everything they’ve read.

That didn’t sound like a writer. Advait realized they sounded a lot like an editor and ran with the idea. He made it stronger. He wrote “Let’s transform our robot secretaries into Socratic gadflies.”

When Will LLMs Work? Not Work?

Because they are based on the “average” of everything they’ve read,

• They can easily spot me making errors mentioned by grammar books (like run-on sentences)
• Some other things I ask will average out to good answers … from math books!
• Others will average out to contradictions, leaving the LLM with “he said, she said” … from history: slavery or states’ rights?
• Still more things won’t average at all.

Try asking about your name: I get myself, a journalist and one dead guy. There is no average of the three.

My Unfair Advantage

Once upon a time, had I learned to use an old probabilistic error-checking program called “lint”.

That program was from so far back that you couldn’t put all the error checks into the compiler and still get it into memory. So lint was written as a separate program. It was exceedingly stupid: it warned about everything. So I got really good at spotting false positives, and marking them “nolint”

Lint taught me to distrust programs.

That means I could use a program like an LLM as a critic, just like lint. I could even put up with it teasing me about how shitty my work was.

It always told me why, even if it thought I was being stupid. So I could look at what it said and know if I wanted to believe it.

Because the LLM was doing something it was actually good at, hallucinations were uncommon. If they made it through, I was prepared for them.

Experience with lint is why I can easily treat an LLM as a gadfly instead of a secretary. I already know I should evaluate a tool’s output rather than trust it.

Go thou, and do likewise.

Links

This is part of a multi-part look at the Gartner hype-cycle of LLMs, and where I found my part of the “Plateau of Productivity”.

• Hedgewitch Part 0 is Torches and Pitchforks. That’s the back-story.
• The main story starts with Hedgewitch Part 1: LLMs, about why I’m not a wizard.
• Next is Hedgewitch Part 2: My Favorite Errors, and what I do about them.
• This blog is Hedgewitch Part 3 LLMs Should Challenge, Not Obey, about the magic that makes it work.
• The next is Hedgewitch Part 4: How do I know that LLMs are buggy?

#ai #artificialIntelligence #llm #technology

This Blog has No Leaves

2026-05-31 12:29:01

Torches and Pitchforks

The peasants are encircling the palace, with torches and pitchforks

The king is a 3-year-old who’s memory has been stuffed with every book ever written (on vellum, this is from a while ago).

But they’re a three-year old.

• They babble incessantly: about animal husbandry, in the middle of a city.
• They don’t know true from false: they have lots of ordinary books with mistakes in their head. And some spell-books that were full of lies, deliberately designed to keep ordinary people from learning wizardry.
• Worst, they don’t know right from wrong: some of the characters in their head are happily, unabashedly, evil.

Everyone had thought the little king was omniscient. Wrong. He’s a dangerous, babblingchild.

Now they’re disillusioned. They’re sure that the king, and the wizard that created him, are villains. They’ve dammed off the stream that the palace depends on for water, and set it on fire. They’re waiting for the king to flee. To set him on fire

Meanwhile, a little hedge-witch is getting afraid to cure peasants with injuries. Her magic injects health into people, not injects things into their heads.

But she might burned to death, too.

#artificialIntelligence #technology

Digital sovereignty and the AI narrative

Cory Doctorow thinks tech executives' story about AI is not an inevitability

"Bubbles are really evil. They transfer money from everyday savers to insiders who lie about market opportunities. And those everyday savers are just trying to avoid like, being homeless and starving to death in their old age."

"We have to really be focused on that fine distinction because it is in that distinction that the narrative that drives the investment, that drives the bubble, it's in that, that's where it lives."

"We live in an environment that is the end stage of 40 years of lax competition enforcement in which cartels, monopolies, and duopolies dominate most sectors of our economy. And they believe correctly that when there are fewer options to shop for, you can lower quality without losing sales because you're the only game in town. And so, in a world of statistical inference tools and monopolies, firing workers and replacing them with software to produce lower quality outputs, is not an unreasonable bet that you could make it work, but there's nothing inevitable about it."

"If Donald Trump turned off all of our chatbots tomorrow, what would change? Which government ministry would be unable to do his job? Which important firm would go out of business?" >>
abc.net.au/news/2026-08-08/cor…
#technology #AI #DigitalSovereignty #finance #markets #ToxicAssets #bubble #narrative #DeepSeek #Mastodon #Sovereignty #InternetCultures #monopolies #governance