Are we just going to ignore that books are destroyed for AI?
forbes.com
forbes.com
And from all i have read about the settlement the sum was 1.5 billion or -3000 per book. For 500,000 deliberate copyright infringements by a multi-billion-dollar corporation with the intention to generate revenue from the data this is hardly something i would call 'sued to pieces'
For two, while some book scanning projects were done non-destructively, my understanding is that was for preservation purposes (e.g. museums) or because the books were borrowed, not owned (e.g. Google's library scanning project). But generally speaking commercial book digitizing is destructive because that is easier, faster and more likely to obtain a good result.
Of course, if our government were actually functional, it could solve the problem with destructive scanning by allowing those scans to be distributed. For anything actually out of print, that should have pretty much no impact on the market since if it's out of print, likely most copies are second hand at this point anyway. And for stuff that's in print, the destructive nature of the scanning isn't a problem because we can print another one. Allowing the distribution / sharing of those books would also reduce the total amount of destruction because it wouldn't be necessary for each trainer to obtain their own physical copy to scan.
I've pitched it before, but I really think our government should be getting into the business of providing data sets for AI. It would solve concerns about destruction and preservation, vastly cut down on the amount of scraping and bot traffic online, and reduce the amount of energy and resources spent on gathering training data. Never mind creating a whole bunch of jobs for data cleanup, classification, categorization and retention / preservation. It's win win for basically everyone except the people that want the government to stop AI from happening, which is (IMO) about as likely as the government stopping the tides.
Some future historian might like to see what my generation learned in second grade math but for that they just need a sample, not mine specifically.
2. How do you know,
3. How many is countless,
4. So what?
Because the implications that the rarity of said books makes them valuable is entirely unsupported.
> 2. How do you know,
Nearly 2 million books are destroyed each day in the US. Are you asserting not one of them is rare?
How do you know "two million", by the way?
So, does this make the destruction of books to feed AI a non-problem? Is the loss of printed information to landfill completely fine and not regrettable? The loss of books to AI is on top of the loss to landfill, and seeks out books somewhat methodically, which means a bias toward rare books, compared to obtaining books by weight.
The argument goes like this, I suppose: some percentage of the 640 thousand tons of landfill books (or whatever this year's figure is) were rare, but unwanted. Therefore, because some rare books are unwanted, and some unwanted books are not valuable, the rare books that were methodically bought to feed AI were kind of sort of going to go to landfill anyway, and nothing of value was lost.
It makes fatuous declarations of the sanctity of books be obviously absurd.
I think you may be overestimating how many books are being scanned and shredded in addition to being sent to landfill, rather than instead of being sent to landfill.
Having a collection of some few hundred rare books, some of them dating back to the 1700s, the market value of these things are extremely low. They’re bulky, they’re fragile, and there aren’t a lot of buyers. I literally couldn’t give them away to booksellers or libraries (I tried!) except for specific items that had nominal value (in the tens of dollars) and were relatively desirable.
As a book lover, I wish this wasn’t happening as well. But as someone with some knowledge of the book world, these AI companies are only able to buy these things at economical rates because nobody wants them.
They’re also presumably not purchasing many copies of the same book (the goal is to scan them, redundancy beyond a few copies doesn’t help much) so I don’t think this will have some major impact on the overall world knowledge. Anything they’re buying has more copies in circulation or is so niche as to be undesirable by buyers/collectors.
There's a similar issue in archeology. There's no monetary value in preserving the ancient mound with (somewhat) fascinating relics of early culture,* apart from robbing it of any gold or items that could be auctioned. The mound is like your 300-year-old unsaleable books of niche interest.
I say "no value", but that's contradicted by "fascinating". So it's more like there's no readily available path to monetizing the preservation of a rare item. In principle it ought to be worth something, because people are interested, but the difficulty of actually preserving it defeats that interest (and destroys the price).
*including perhaps literal ancient garbage, now of interest. https://en.wikipedia.org/wiki/Oxyrhynchus_Papyri
Finding someone who wants books which may be rare but are not broadly market-desirable is hard. Hard enough that I literally couldn’t give them away.
When Anthropic calls you up and says “I’ll take these off of your hands and I’ll even pay you for them”, you probably take it.
In a capitalist economy the only alternative to rich companies buying these and doing as they please with them is someone else buying them who pledges not to destroy them.
Edit: now a good question here might be why universities, many of them rich, aren’t offering to take these instead. it would be much better if they did.
when I contacted UBC about some of my books, they said they had a backlog and it’d take a lot of time and effort so I’d need to catalogue each in detail and send them the list first. I didn’t have the time so I didn’t, but they do have the resources if they choose to employ them.
I wouldnt.
Why do you think bookstores exist?
Anyone remembers when Google Books project might have provided a digital library after painstakingly non-destructively scanning books?
Now they are following copyright law.
Selling the books afterwards would be additional work and not guaranteed to bring any profit or even any significant reimbursement. A company is likely to find simply pulping the books to be the most cost-effective solutions.
The books being destroyed as so worthless they're bought in bulk and were destined for junking anyway, like the 600 million books that are destroyed each year just in the US. At least as pulp they can contribute something of slight value.
Here's a way these old supposed valuable books can get a new life. Go wild! Prices start at $12.99 per linear foot. This clearly demonstrates the value society assigns to these precious rectangular prisms of inked cellulose.
I have a sneaking suspicion the whole tempest-in-a-teapot was astroturfed by publishers who were miffed at the courts looking favorably on scanning for AI training.
Yes there are people who have this psychology.
As for "old rare books out of copyright", those generally predate ISBNs and so are not being so mercilessly crawled by semi-automated means.
No professional bookseller today is pricing a book that they aren't familiar with without putting the ISBN into a bookseller search and finding out the market price. The hunting booksellers have dedicated PDAs that will scan the UPC on the back, and speak the current price into earbuds for quick searching at charity book sales.
Within my own (non-destructive!) scanning contributions for the shadow libraries, I have gone through innumerable ISBN-issued books from Eastern Europe and the former USSR printed in very small runs, often 50-100 copies. When one of those few copies finds its way to a used-bookseller, it is usually priced very low: it doesn't look attractive to a general audience, there is no agreed "market price from a bookseller search" because maybe no other seller has ever seen it, and the bookseller doesn't know that it holds exclusive information that some scholars would prefer be saved.
It's taking up space. It's not fashionable, and it looks kind of ugly to me. Who's Doctor Gachet anyway? I can't predict that anybody will ever want this. So, I'll throw it away.
But you can also reasonably say:
It's not taking up much space. Fashions change along with culture and ideas. I like paintings. People like paintings generally. I can't predict that nobody will ever want this. So, I'll keep it.
I conclude that if you're willing to be an archivist, you're doing good, and somebody somewhere should be willing to be an archivist. Ideally several people. This doesn't depend on the price of items at time of archiving.
No, I was not asserting that, obviously.
These older books that don't have ISBNs don't exist anywhere else. They aren't digitized.