A version of this story appeared in Total Annarchy, my fortnightly newsletter that helps you be a better writer, storyteller, marketer. Get it in your inbox; you’ll love it.
***
*** Trigger warning: This piece contains a graphic description of animal death. ***
I wasn’t going to write about AI this week because I’m tired of thinking about it. Plus I’m feeling better and it’s late summer in New England and I want to be in this low-humidity-beautiful world and not feel both existential and actual dread. In other words, you and I need a break.
But then Amazon went and bought and destroyed cases of books just to feed their AI and… welp. I’m triggered. Guess we’re talking AI again.
To catch you up: Amazon has done something so extraordinarily violent that it’s gutting to anyone who loves books.
Amazon bought cases of obscure, out-of-print books. Then they cut off the spines with something that resembles an industrial meat slicer, scanned the pages with a high-speed scanner (sucking all the intelligence and sentence structure and information out of each page), and then gave the old books a burial straight out of The Sopranos: They dumped them in a dumpster behind a nondescript warehouse under the punishing Nevada sun.
The building bears the logo of a T-rex holding a book in its scaly claws and chomping a book to smithereens. Cute.
It’s not clear what VGT in that logo is short for. But I can only guess “Vanishing Gutenberg Texts” or “Voices Ground to Tokens.”
404 Media discovered the operation by hiding an AirTag in a book shipment. (Free advice to the AirTag marketing team: You should capitalize on this moment: “Helping you keep the precious safe.” Or something.)
* * *
I read that story and my heart aches. One time as a child I stood by horrified, watching the way my uncle skinned a catfish. The fish was still alive (this was the horrifying part), and he gripped the wriggling fish’s jaws with a set of greasy pliers in one hand and with a second pair of pliers yanked the skin off the fish’s body. The fish thrashed violently as my uncle moved down its body, tugging at its skin, working it an inch or two at a time.
At the time, I felt the visceral horror of that scene: the way the fish almost seemed to arch its back as its skin was peeled off and the way I reflexively arched my own back. I wanted to throw up. It was all so senseless and gross.
When I read about the books, I thought of the fish that day. One naïve minute you’re swimming in a lake. The next you’re being held aloft and skinned alive.
I haven’t thought of that scene for years. I’m sure I thought of it now for no reason at all.
* * *
Why destroy old books to extract their words?
Mostly anything written before 2022 was written without AI assistance. (I say “mostly,” because gen AI existed before then, but it was less loquacious than it is now.) By now, generative AI models are as ravenous for human-only written text the way zombies are for brains. They need it to keep their training data fresh.
AI models learn the beats of language and sentences and how they are built from their training data—”data” being words that you and I have written, of course.
As more and more writing is created by or with AI models, original human writing is harder to identify. It’s also more valuable as the source material that keeps models freshly tuned to the entire scope of how people think and write and put down words that create sentences that make up the paragraphs that make someone else feel something. The surprising and playful oddities of human language—the things that make my writing distinguishable from yours—start to fade.
What happens then? AI models can start to decay. The output (gross) starts to flatten. Sentences sound more repetitive. Boring. The zombie eats its own brain.
* * *
So who cares? These are old, obscure books. You and I aren’t exactly clamoring to own a musty volume from a long-forgotten academic.
Maybe not. But we don’t really know what’s in those boxes, do we? They may have scholarly or historical value despite not having commercial value. We can’t really know.
What’s more, once the book is slurped up by Amazon, that doesn’t make it publicly available. An AI is not a library. Instead, we get it filtered and extruded and plopped onto the AI conveyor belt.
(There are conspiracy theories about why AI companies are racing to disappear all these texts. You can go down that rabbit hole if you’d like, but fair warning: That rabbit hole is dark and deep and more upsetting even than my fish story.)
Anyway, the bottom line is that human-written language reliably untainted by AI is becoming valuable. Right now, companies are literally destroying physical books to get to it. Some suspect they are systematically working their way through every ISBN in the universe.
And by the way, there are ways to slurp up the words without destroying the books themselves. Without the spine-slicing and dumpster-dump. But that requires careful page-by-page scanning, and that’s too slow for an industry caught up in some kind of words arms race.
* * *
What’s this all mean for the future of writing?
To me, there’s never been more clear evidence: There’s never been a better time to realize the value of your most human words. At work, there might be pressure to adopt AI. But if human-written words are valuable enough for AI companies to destroy books to get to them, maybe we should think twice before we stop making our own.
There’s never been a better time to do two things:
>>> Pause & consider what the moment is asking of you before you reach for any tool, AI or otherwise.
Not “Can AI do this?” (It probably can.)
But instead: What do I lose if I hand this over? Is the value here in the output—or in the thinking required to get someplace new and wholly our own?
>>> Understand that NO is a choice, too.
Leaders: Give your teams permission to say: “Yo, I did not use AI here. And here’s why.”
No is a choice. Right now a lot of organizations reward visible adoption: Who’s using it? How much time are we saving? How many workflows have we automated?
Those are the wrong questions.
Instead: Where did AI genuinely improve the work? Where did human attention improve it? That turns AI literacy from tool proficiency into discernment.
And another thing:
>>> Write Your Own Words! This one is a rallying cry for writers everywhere… whether you write books or blog posts. Do we need WYOW T-shirts? Maybe. You know I love merch.
* * *
The Amazon situation is gutting for people who love books. Especially people who, like me, love the physicality of the book, the smell of the paper, the whisper of the page-turn, the way you get lost in the mind of another. There’s nothing like it.
Gutting… and supremely ironic. Amazon launched just over 30 years ago with a vision of being a bookstore that could offer almost every one of the millions of titles in print. If you’ve been around the internet for a while, you might remember Amazon’s pivot from a place to buy any book you wanted to a place where you could get paper towels and Murphy’s Oil Soap overnighted with the urgency of a donor kidney.
There’s something profoundly dystopian about it all: The company that began by figuring out how to sell books at scale is now destroying books at scale to turn them into data… to sell to us again.
And side note: despite its book-focused beginnings, Amazon’s scale, convenience, and pricing also made life exponentially harder for independent bookstores and the book people who generally own them.
So the irony is double, I suppose. Or maybe triple. I can’t even count the layers anymore.