AI companies burning books
Discussion
Good gravy, it really does seem somewhat sinister.
https://www.forbes.com/sites/maryroeloffs/2026/08/...
I, for one, welcome our robot overlords.
https://www.forbes.com/sites/maryroeloffs/2026/08/...
I, for one, welcome our robot overlords.
They aren't burning "antique" books, no.
They are buying one copy of every second-hand book they can lay their hands on and then scanning them, in the hope that obscure and unusual content will give their system the edge of knowledge. Copyright rulings have said it's legal to scan books, so long as the original source is destroyed. Even if every big AI company buys one of every book that'll only be a handful examples of each destroyed. In many cases, the books would eventually have been pulped anyway.
They have a lot of money to invest, but not at any price, so they will be targetting the cheap stuff which is cheap because there's either thousands of copies which nobody wants, or the book is so obscure/dated that nobody wants it at any price. Possibly they might be paying more if the title/topic is in an area they want to expand on.
Either way, these will all be books with an ISBN, so nothing "antique".
They are buying one copy of every second-hand book they can lay their hands on and then scanning them, in the hope that obscure and unusual content will give their system the edge of knowledge. Copyright rulings have said it's legal to scan books, so long as the original source is destroyed. Even if every big AI company buys one of every book that'll only be a handful examples of each destroyed. In many cases, the books would eventually have been pulped anyway.
They have a lot of money to invest, but not at any price, so they will be targetting the cheap stuff which is cheap because there's either thousands of copies which nobody wants, or the book is so obscure/dated that nobody wants it at any price. Possibly they might be paying more if the title/topic is in an area they want to expand on.
Either way, these will all be books with an ISBN, so nothing "antique".
Hill92 said:
They wan older books lest they end up training AI models on AI slop.
Anything and everything to reach the exit IPO bonus, or get 1% closer in the Reservoir Dogs shoot off. They don't have a care in the world whether or not there are recent but special or specialist books in there. Just like, for example, Google doesn't care about small, specialist websites, or Zuck is mad enough to not give a toss about anyone, anywhere, as long as he can piss tens of billions up a wall every year.Mortarboard said:
And "Equilibrium".
[Film bore mode on]Which was derided as a s
tty rip-off of "the matrix", when in fact it was a very good homage to Farenheit 451 [/film bore mode off]
M.
Equilibrium is superb. There's so much to love about that film, Almost anything with Emily Watson in it is going to be great, and this has added Christian Bale. [Film bore mode on]Which was derided as a s
tty rip-off of "the matrix", when in fact it was a very good homage to Farenheit 451 [/film bore mode off]M.
Edited by redrabbit on Tuesday 1st September 01:26
Gareth79 said:
They aren't burning "antique" books, no.
They are buying one copy of every second-hand book they can lay their hands on and then scanning them, in the hope that obscure and unusual content will give their system the edge of knowledge. Copyright rulings have said it's legal to scan books, so long as the original source is destroyed. Even if every big AI company buys one of every book that'll only be a handful examples of each destroyed. In many cases, the books would eventually have been pulped anyway.
They have a lot of money to invest, but not at any price, so they will be targetting the cheap stuff which is cheap because there's either thousands of copies which nobody wants, or the book is so obscure/dated that nobody wants it at any price. Possibly they might be paying more if the title/topic is in an area they want to expand on.
Either way, these will all be books with an ISBN, so nothing "antique".
If the aim is to get the most obscure content in order to give their model the 'edge' over competing models, its inevitable they will eventually target rare and expensive books, no? They have the money. And it's in their interests to destroy rare material after scanning to prevent their competitors scanning & thus losing their own edge.They are buying one copy of every second-hand book they can lay their hands on and then scanning them, in the hope that obscure and unusual content will give their system the edge of knowledge. Copyright rulings have said it's legal to scan books, so long as the original source is destroyed. Even if every big AI company buys one of every book that'll only be a handful examples of each destroyed. In many cases, the books would eventually have been pulped anyway.
They have a lot of money to invest, but not at any price, so they will be targetting the cheap stuff which is cheap because there's either thousands of copies which nobody wants, or the book is so obscure/dated that nobody wants it at any price. Possibly they might be paying more if the title/topic is in an area they want to expand on.
Either way, these will all be books with an ISBN, so nothing "antique".
Morally we know this is not beneath them - It was Meta who just mass downloaded via bittorrent Terabytes of pirated book PDF's to train their models.
My FiL volunteers at a big charity shop looking after the book department. The valuable stuff gets put on eBay or sold to World of Books, the ok stuff gets put out on the shelves, but the majority of the donations are worthless dross that gets pulped. It’s shocking how many books get destroyed, and that’s only one shop. I expect it’s pretty similar at most other places. At least by scanning the books the information is being preserved first.
Gareth79 said:
They aren't burning "antique" books, no.
They are buying one copy of every second-hand book they can lay their hands on and then scanning them, in the hope that obscure and unusual content will give their system the edge of knowledge. Copyright rulings have said it's legal to scan books, so long as the original source is destroyed. Even if every big AI company buys one of every book that'll only be a handful examples of each destroyed. In many cases, the books would eventually have been pulped anyway.
They have a lot of money to invest, but not at any price, so they will be targetting the cheap stuff which is cheap because there's either thousands of copies which nobody wants, or the book is so obscure/dated that nobody wants it at any price. Possibly they might be paying more if the title/topic is in an area they want to expand on.
Either way, these will all be books with an ISBN, so nothing "antique".
I suspect if anything will kill AI (apart from footling things like energy capacity and a shafted profit model) it is the likelihood of mass law suits re copyright infringement They are buying one copy of every second-hand book they can lay their hands on and then scanning them, in the hope that obscure and unusual content will give their system the edge of knowledge. Copyright rulings have said it's legal to scan books, so long as the original source is destroyed. Even if every big AI company buys one of every book that'll only be a handful examples of each destroyed. In many cases, the books would eventually have been pulped anyway.
They have a lot of money to invest, but not at any price, so they will be targetting the cheap stuff which is cheap because there's either thousands of copies which nobody wants, or the book is so obscure/dated that nobody wants it at any price. Possibly they might be paying more if the title/topic is in an area they want to expand on.
Either way, these will all be books with an ISBN, so nothing "antique".
https://mashable.com/tech/anthropic-sued-by-sony-m...
https://www.mishcon.com/generative-ai-intellectual...
Good.
bloomen said:
It makes you wonder how many books will never be published once AI gets rolling proper.
Instead of that lovingly-crafted proposal being accepted, they'll feed it in, click a button and the presses will roll.
MattGPT already pioneered this with his book "Suicide of a Nation"Instead of that lovingly-crafted proposal being accepted, they'll feed it in, click a button and the presses will roll.
Gareth79 said:
MattGPT already pioneered this with his book "Suicide of a Nation"
Quelle surprise. I presume Amazon is already heaving with AI ebooks as the costs are neglibile and you only need a handful of sales. There'll be 'authors' with 10000 titles before long.
And 'readers' will get their AI to read that AI to summarise and save time.
Not every provider is doing this.
Google, for instance have been scanning books in a non-destructive way for years. It is the fact several companies are behind which is making them go to destructive methods, as it's quicker.
https://www.theverge.com/podcast/872998/anthropic-...
Google, for instance have been scanning books in a non-destructive way for years. It is the fact several companies are behind which is making them go to destructive methods, as it's quicker.
https://www.theverge.com/podcast/872998/anthropic-...
lufbramatt said:
My FiL volunteers at a big charity shop looking after the book department. The valuable stuff gets put on eBay or sold to World of Books, the ok stuff gets put out on the shelves, but the majority of the donations are worthless dross that gets pulped. It s shocking how many books get destroyed, and that s only one shop. I expect it s pretty similar at most other places. At least by scanning the books the information is being preserved first.
Sometimes its not apparent what the dross is . When I was a teenager I got rid of a lot the 1960s comics I had got from the 2nd hand shop.But kept all my current stuff as it seemed to have value to me.
Later on your are left wondering why you have lots of kungfu monthly and monsters mags that were just b/w pictures from some old 1940s films
with some boring interview with the guy that moved the set around.
Gassing Station | News, Politics & Economics | Top of Page | What's New | My Stuff