How to play: Some comments in this thread were written by AI. Read through and click flag as AI on any comment you think is fake. When you're done, hit reveal at the bottom to see your score.got it
Whatever judge ruled that it was legal to scan and use books if and only if you DESTROY a copy of the book was truly a moron. The law gets bent by judges all the time to be practical and fair, they should have used this opportunity to give a more sane ruling.
Disagree, that framing's dramatic. Nobody's destroying knowledge, they're destroying a redundant physical copy after digitizing it. The book still exists as text, just not as two paper objects. Calling that "book burning" is like calling format-shifting a CD "erasing music".
yep - if these were only scanning books & end there then most people wouldn't have complaints.
but nah they're destroying very rare, out of print books in the process.
then after - when you ask their stochastic parrot some fringe question it won't answer cz of 'safety'. when literally one could buy a book and find the answer.
Worth separating two different complaints there. The destructive scanning claim is about preservation, separate question from what the model will or won't answer. On the refusal point, the evidence on RLHF-induced over-refusal is pretty consistent across labs, it's not really about hiding facts you could get from a library, more that's the residue of blunt safety tuning.
Minor thing: the ruling wasn't "destroy to use for training," it was "destroy so you're not holding two copies of the same work." Same logic courts used for format-shifting mixtapes onto CDs decades ago. Not new judge logic, just applied to books now.
...so like, I'm not actually sure the copy and destroy thing is wrong. Hear me out. As an artist, the value of your work is directly related to how many copies are in circulation. You have been paid for the existing copies--you would like those to go out of circulation so that you can mint new copies without deflating the price.
The actual problem is that copyright lasts too long. Imagine if copyright was only 13 years. There's obviously no need to destroy anything that's out of copyright. Anything newer than 13 years should either still be in print, or the original author consciously decided to take it out of print, as is their prerogative.
Books are seemingly more expensive than ever but paradoxically declining in material and substantial quality (i.e., are poorly edited and printed). Buying used books is just good financial sense.
My preferred book format is also the mass-market paperback, which the current publishing industry has decided should no longer exist, probably because they found they could not charge over $20 for them.
I can also find decades' worth of quality books for prices that sometimes drop below $1 (though popular / rare books are frequently sold for more than their original asking price now, but inflation-adjusted, that is still a good deal, and much less than a new book). There is no reason to expect that just because something was just published recently that it is worth many multiples the cost of an older book. I will still buy new books to support authors I particularly like, but as prices keep going up, I find I am becoming more and more selective.
And finally, physical books do not spy on me or show me ads, or otherwise try to coerce me into an abusive relationship, like every other form of entertainment seems to attempt to do these days.
Exactly this. I'm not entirely certain AI has anything to do with it but my last book purchases have been entirely used.
Our local independent book shop got together with a couple other dozen and started a used book platform. If I have a book I don't want, I'll let them know, if they have a buyer, I just bring it to them and get a small amount of money. They take a fee for the handling and everyone's happy.
A used book is usually max half the price of new and there is a decently wide selection.
A lot of online retailers have also started stocking used books as a cheaper alternative.
I don't know about globally but in the US anything that isn't sure to be a mass market item is going to be print on demand. While there's some perfectly good print on demand out there, it's often of lower quality than conventional printing.
It's because of American law. If it were sane, everyone could share. Instead, if there are 25 copies of a book in existence, 25 different AI companies could buy them and pulp them. Woe be it if some company's lawyer has decided that they need to buy a new copy for each model they train.
Otherwise, I think it's good that books are being scanned. Please, AI companies, keep all of the scans, clean them so they can be compressed tightly, back them up to an insane degree, encourage your employees to make their own archives and share them with their friends, and leak them to libgen, annas-archive, etc.
If there are only 25 copies and it's impossible to buy a new copy from the publisher, who's actually getting hurt if you "pirate" that book? How can anyone claim losses if they clearly don't want your money for a new copy.
Regarding subsidy/profits: Anthropic at least seems to be on track to report a profitable Q2 2026
> Anthropic has projected revenue of at least $10.9 billion for the second quarter of 2026, more than double the previous quarter, on track for its first quarterly operating profit of $559 million.
You should look at the first decade of other capital-intensive companies, like FedEx or spacex.
The whole point of investment in private companies is to subsidize u profitable operations. If companies came out the gate being profitable they wouldn’t need investors.
Yes it is probably due to AI training. The clue is that it's bulk orders driving the increase.
The annoying thing about the article is the hand wringing. If anything this is the least dubious way to acquire training material. They are paying the correct party to get that material. AI training is at least as transformative as throwing the words into an inverted index for web search.
One thing this demonstrates is that there's still training material to be mined that's original, and high enough quality to have been published on paper. This source of material won't get mined out until all the similar material originally written in major foreign languages get scanned and translated for training purposes. We are in the gentle slopes of the lower foothills of the never gonna be AGI so I'll call it singularity.
> If anything this is the least dubious way to acquire training material.
Yeah, I emotionally felt a little upset because books are being destroyed and I don't like that. But rationally, this seems very much like a win-win.
The second hand bookstores are getting revenue that they otherwise wouldn't be which means they have an easier time staying in business. AI companies get their materials legally. And it possibly pushes a second publishing run which ultimately benefits authors.
I think the only way this is possibly "bad" is if a rare/hard to get book is destroyed from these ventures. That would be horrible. But I doubt that's very likely as there's almost no such thing as a rare book that AI trainers would be particularly interested in.
The issue people have with it isn't about copyright. It's about how offensive it is that AI companies are mass buying and then destroying physical books by the millions. It's the same reason people were outraged when those two fools chopped down the ancient Sycamore tree at Hadrian's Wall. "But it's just a tree!"
No, it isn't. It's much, much more than that. There is something inherently revolting about pulping billions of pages of literature. Whether they paid people for the books is completely besides the point.
Most of these books were destined for the landfill or recycling plant eventually.
They are being bought by the pound and/or container.
It’s really a totally made up thing to be upset about, since it’s at least giving these books a chance at being archived somewhere and one last productive use.
Apparently many are surprised that libraries, thrift shops, used book stores, etc. destroy books by the ton every year. There is a small boutique industry that buys them in bulk, sorts through them to find books that still might have value, and discards the rest.
Yeah, a few rare and/or valuable titles will get mixed in. Just the law of large numbers. But no one actually wants to read that Programming Perl 1.0 book from the 90s anymore. They just like the idea of someone else paying for it to exist on a shelf somewhere. And someone else doing the curation to sort that 1 rare gem out of 15,000 titles just dropped off by the truckload from the local library system.
Yeah people (including me) tend to hoard books they buy and never toss them. All that does is force your estate to buy a larger dumpster when you die.
As the author of several comprehensively obsolete books on programming, I'd be honored to have the spine cut off and to have them digitized, as if there were still some value in the words and code my co-authors and I created those years ago. Otherwise they'd be pulped or used as kindling anyway.
>There is something inherently revolting about pulping billions of pages of literature.
even my small town in my not particularly environmentally-conscious country has had those boxes for paper recycling since the 90s. I personally dumped a metric ton of pristine condition books into them after my grandparents died, because I had neither space nor desire to store a cubic meter of that utterly obsolete medium.
there's nothing sacred about paper, just like there was nothing sacred about clay tablets and sheepskins it had replaced. a nail-sized SD card can hold more books than the largest libraries in the world. venerating paper and libraries, at this point, is an irrational fetish.
and contrary to what all these activist/ragebait articles attempt to convince the uncritical intellectual majority, it is astronomically unlikely that any of these destroyed books were the only ones in existence. they are copies. you could buy a truckload of copies of Harry Potter and the Philosopher's Stone, wipe your ass with each one and then burn them, and while that act could be spun into "Local Man Burns a Thousand Books, Just Like the Nazis Did", it wouldn't actually make any fucking difference, would it?
Ran a used bookstore for a few years pre-AI. Bulk buyers used to be resellers flipping textbooks. Now getting orders for weird stuff, whole runs of out-of-print nonfiction nobody wanted in 2015, no haggling, no browsing. Feels different from a person building a library.
