The value of books
There's been a lot of discourse on my remaining social media this week about certain AI companies purchasing what used to be called 'long tail' books when I worked in a book data company (admittedly it was a long time ago). The purpose being, that these books of all types are a form of undigested, pre AI information/sentence structure ripe to be absorbed.
https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-companies-are-reportedly-shredding-millions-of-books-to-train-models-tech-giants-outsource-to-middlemen-to-secretly-buy-up-books-for-training-material link to Toms Hardware article on the subject.
The voices largely seemed to fall into two camps; the 'this is a crime against all that is sacred' camp and the 'its just old copies of Windows 3.11 for beginners that never sold, what’s the issue' camp.
I am, obviously, much more on the former, but do think that the later camp is missing some of the point. Sure I know that libraries, book dealers etc all pulp massive amounts of books that are long since considered defunct like ancient computer texts, obscure fiction self published and scientific texts that predate the discovery of the atom. These are essentially just word collections to help AI continue in destroying human intellectual curiosity. However, my contention is that among the dirt there must lurk some gems, but these may only be useful to very very small numbers of people and once lost, take with them something for ever.
Raiding my own collection of books, I was able to find some examples. I have a book on tree folk law published by a long since gone university imprint that I only have due to my mothers old job. I have a book on bridge players from the 1950s, for which only one other copy is for sale online (cheaply, its not a valuable book). I have a book published by a East Asian ex-pat group in the US on an obscure cultural reference that I paid to import, quite possibly the only one in the UK and perhaps one of only very very few worldwide. I have a book I purchased from the estate of a book collector. Its a small text on chess and again I can only find one instance online, but my copy is absolutely pristine despite being over 50 years old.
To be clear, non of these are worth any monetary value. Most would of be little interest to anyone other than me, but hold some value to anyone interested in their esoteric contents. These are the books that don't deserve to be sacrificed at the alter of AI as worthless, but to live happily somewhere in the hope that one day someone may seek a small piece of information contained within. The idea of cutting their spines off, absorbing the data and removing one remaining precious copy from circulation when there may be very few left means something irreplaceable is being lost.
To prove that AI can't be trusted with the information contained within books - I offer the following anecdote from my job. A network engineer on my team Googled the end of life date for two of our firewalls. Statement came back confidently stating it was July 2026... now this caused all sorts of concern, one I means we're running out of service firewalls, two the cost to renew does bad things to the end of life predictions of that datacenter. There is a webpage, its top of Google rankings, written by AI, repeated by Googles AI and then written to other pages to further be referenced by AI.
Further reading of the providers documentation offered no supporting evidence, but equally nothing to disprove it either, but its still repeated in a number of places as fact. In the end we contacted our account manager. She, I suspect, had answered this question before and told us there are no plans to end of life that firewall model at all and she's not sure why there is information out there to the contrary. Amazing example of how AI posts then proves its own lie until its shown as fact. Imagine now what happens with all that information from those pre AI books when the real texts don't exist any more. The AI presents an absolute fact that is totally wrong. Who is now able to prove it otherwise, when AI is the last remaining source of data without showing the original. This is the future we're building for ourselves. I suppose my only point is the one I started with, value your books!