The store is in Newtown, Wellington, and is called "Bookhaven." It stocks mostly New Zealand books, many of them non-fiction dating from the 1970s and 1980s, and some of them Maori editions.
Business was as usual until suddenly massive, multiple-copy orders for outdated non-fiction books came racing through from a Canada-based operation. The owner, Annemarie Thorby, smelled a rat, as the cost of shipping alone would make the purchases cost-ineffective. So, was it a AI enterprise, aimed at data-mining? Where the actual books were destroyed? She feared so, and cancelled the orders.
Interviewed, she said:
"One order was for 71 books, and we physically can't handle that.”
The buyer wanted books with ISBNs (a unique code used to identify a specific book, edition, or format) from the 1970s and 1980s, she told RNZ’s Midday Report.
“They were local non-fiction niche books, so New Zealand local history, a tourist pamphlet on Waitomo Caves, Māori language books, New Zealand cookbooks, biographies, political cartoons, very niche - the local anglers club.”
With postage and tariffs there was no way the buyer could turn a profit, she says.
“A $10 book could end up costing $40, so ridiculous. It had to be something else. That's when we did some sleuthing and contacted AbeBooks, their little online forum.”
These kinds of odd orders were happening around the world, she says.
“Tens of thousands of books being bought by this one company to go to a warehouse in the USA. It was India, it was Europe, it was us, and it wasn't just us in New Zealand, there were other bookshops.”
AI companies are buying these books and training their large language models, sometimes destroying books in the process. It's the only way for them to do it, she says.
“They're not recycling, they're devouring them, they're destroying them. One of the quotes from the Anthropic court case was, Project Panama - so this is from the company, from Anthropic - Project Panama is our effort to destructively scan all the books in the world.”
Annemarie Thorby, according to Techspot, is perfectly correct. This massive buying of non-fiction books from that era is going on all over the world. AI firms are buying up these older books -- ones that can be guaranteed free of any AI content, which apparently degrades over time -- to train their models, and destroying the books in the process.
The outfit that is creating this immense database is known as "ISBNdb" and is employing firms such as ABE books to make purchases ranging from 1000 to a million books from secondhand bookstores.
And these books are often rare. Yet they are treated with no respect at all. The spines are ripped off and the loose pages run through scanners, then shredded.
What are the AI companies carrying out this wholesale process? The names are kept secret in an atmosphere of strict confidentiality. But the name of Anthropic crops up all the time
No comments:
Post a Comment