Clojurians Log v2
ai-skeptics 2026-07-27
> AI companies are bulk-buying rare books, scanning them through high-speed machines that cut the spines off, and shredding the originals. > Anthropic hired the former head of Google Books partnerships to obtain “all the books in the world.” https://xcancel.com/HedgieMarkets/status/2081534588485296565
If true, that's disgusting. But why doesn't this person link to anything, or references any of the court cases, or what else.
And why did they add an AI generated image to their tweet?
Which in turn I found here: https://futurism.com/anthropic-shredded-millions-of-physical-books
(I had to follow a chain of linked Futurism articles to find that—they really want you to read more of their articles!)
FWIW, according to this article https://www.ibtimes.co.uk/ai-companies-demand-printed-books-copyright-concerns-1810831 a judge ruled that this was not illegal: Anthropic bought the books legally and therefore owned them, and so could legally digitize them and keep those digital copies in-house while "throwing away" the physical books.
Obligatory Krazam: https://www.youtube.com/watch?v=IU4ByUbDKNc
My understanding is that it’s that ruling specifically that has resulted in the destruction. That they’re allowed one copy, so if they destroy the original, they are technically living up to that. If so, then this presumably means that they don’t have backups of the scans as that would result in more copies.
And either way, it’s a genius slopmaxxing move. You can generate slop, sure, but if you also destroy existing information, the ratio goes up way faster. I wouldn’t have thought of that.
They destroyed the books as part of the scanning process: they cut the spines off so they could run the pages through scanning machines.
So, no, they were destroying books before the recent ruling.
Aha, including the shredder at the tail end? Then I had misunderstood.
Well, they didn't want all those now-detached pages just lying around and they no longer needed them so... into the shredder they went.
They’re too young to understand the backup value of hard copies I guess.
It's also where the AIs learned the em-dashes from. Pretty common in written works
When I moved house, I didn't want to pay for a lot of old technical books to be packed and moved. The local library didn't want them since they were "outdated". So they all went into the recycling.
The shredder is just a shareholder value optimization, nothing to see here. In seriousness, I think that the physical book destruction (if it's only done 1 copy for each title) is not so criminal. But the fact that they are only looking for pre-2022 books is hilarious.
That included some books I bought in the '70s and early '80s that were probably long out-of-print.
That depends on how many copies of the book is in existence. I guess there’s a non-zero chance that the only copy in existence of some book is on an Anthropic or Open AI server right now.
Strictly speaking, it's not much different from the unique copy of the book sitting at their shelf forever and then eventually catching fire.
If you read the articles, they talked to some rare book dealers who said this wasn't widespread enough to "damage" the market for the books or threaten all copies of any book. But, yeah, I guess it's possible Anthropic bought a one-and-only copy... very, very unlikely, I'd say.
But they could indeed make the scanned versions public (for stuff that reached the public domain age). Which of course they wouldn't do.
But anyway: • Neuromancer • Snow crash • ... • Fahrenheit 451 <--- YOU ARE HERE • ... • Mad Max • Terminator
Politically ~impossible, but you could centralize it. Ingest the books, give anyone who wants it access to the results, subject to some agreement that gets the publishers off your back. Level the playing field. If we’re serious about advancing AI, it would even things out for minor and major players and provide value to the ecosystem rather than a particular org.
The goal of advancing AI, if it was ever a goal, stopped the moment they turned from non-profit to for-profit.
Or at least the goal of advancing ai for the greater good of humanity. They are of course still advancing ai for the greater good of their share holders.
@henrik digitization as a service? Sounds good but you'd have to fight the copyright cartel (I'll do it for free)
The issue is it's not like a single entity in the whole world is digitizing one book by destroying it. So many competitors whose moat is digital book database are now competing and each can destroy their copy to digitize it, etc. It might not be illegal, but clearly problematic and likely something you'd want the government to regulate a little bit. I reckon, isn't there some big Christian conglomerate that's been similarly buying up rare manuscripts and artifacts of anything that could disagree with their version of history? I remember reading about that.
I'm also surprised, it's 2026, we have humanoid robots, AI, robotaxis, we were able to read a scripture that's basically ash without unrolling it, and we can't scan a book without destroying it?
We can, it’s just not as efficient as destroying it in the process.