At the book scanning facilities they slice the spines off of the books so they can be fed into the scanner faster.
-
At the book scanning facilities they slice the spines off of the books so they can be fed into the scanner faster. These books, often obscure, often, books that have never been read by any person for decades are then thrown away.
The symbolism in incinerating them is obvious, but perhaps the paper is recycled not burned. Perhaps it is packed into a landfill.
Fresh human written text untainted by the AI generated content that fills the internet now is very valuable.
1/
-
At the book scanning facilities they slice the spines off of the books so they can be fed into the scanner faster. These books, often obscure, often, books that have never been read by any person for decades are then thrown away.
The symbolism in incinerating them is obvious, but perhaps the paper is recycled not burned. Perhaps it is packed into a landfill.
Fresh human written text untainted by the AI generated content that fills the internet now is very valuable.
1/
Last year my mother wrote a book. I helped her with it when I could but I found the subject extremely boring. Part of this is going back to arguments we've had since I was little. I never wanted to go to the graveyard to take rubbing of the gravestones or listen to the old people we'd meet on these trips talk and talk.
I can recognize a kind of immaturity in that so I wanted to help my mom complete her very boring book on the Black Civil War veterans of this tiny county in PA.
2/
-
Last year my mother wrote a book. I helped her with it when I could but I found the subject extremely boring. Part of this is going back to arguments we've had since I was little. I never wanted to go to the graveyard to take rubbing of the gravestones or listen to the old people we'd meet on these trips talk and talk.
I can recognize a kind of immaturity in that so I wanted to help my mom complete her very boring book on the Black Civil War veterans of this tiny county in PA.
2/
The book she made is exactly the kind of book that would be "shucked and scanned" in one of these operations. There are fewer than 100 copies. I have one, and the local historical society, and a few libraries. But no one is looking for this book. Not now at least. But in the future someone may be desperate to find it.
"Local family histories" are a pain for independent bookstores. They take up space on the shelves and never ever sell. But I do think they are important.
3/
-
Last year my mother wrote a book. I helped her with it when I could but I found the subject extremely boring. Part of this is going back to arguments we've had since I was little. I never wanted to go to the graveyard to take rubbing of the gravestones or listen to the old people we'd meet on these trips talk and talk.
I can recognize a kind of immaturity in that so I wanted to help my mom complete her very boring book on the Black Civil War veterans of this tiny county in PA.
2/
@futurebird My mother wants me to help her write exactly this kind of book. She is a genealogist, among other things, and has spent extensive time in graveyards and courthouses reserching out family back to the revolution, and helps applicants to the DAR do their geneology. She wants to write a history focussed around Bucks County, PA.
I am dragging my feet on agreeing to help with this project, have been for years, and will probably try to do so until she dies.
-
The book she made is exactly the kind of book that would be "shucked and scanned" in one of these operations. There are fewer than 100 copies. I have one, and the local historical society, and a few libraries. But no one is looking for this book. Not now at least. But in the future someone may be desperate to find it.
"Local family histories" are a pain for independent bookstores. They take up space on the shelves and never ever sell. But I do think they are important.
3/
How many have been bought for pennies and ... whops! That was the last copy of that book. It's gone forever?
"But at least if they scan it it will still be around forever."
This is totally false. And we need to let people know, since most people assume that when a computer scans something it's being saved forever.
4/
-
How many have been bought for pennies and ... whops! That was the last copy of that book. It's gone forever?
"But at least if they scan it it will still be around forever."
This is totally false. And we need to let people know, since most people assume that when a computer scans something it's being saved forever.
4/
"At least if they scan it it will still be around forever."
That is not how training data for an LLM works. The LLM use the data to find patterns. Things that do not fit into patterns? They are lost forever.
If the AI scans my mom's book on Black Civil War veterans it will NOT be able to name them, or tell you what it said on the slate gravestones she made me take a rubbings of as a child. But it may reuse some of her choices of words if they have also been used by other authors.
5/
-
How many have been bought for pennies and ... whops! That was the last copy of that book. It's gone forever?
"But at least if they scan it it will still be around forever."
This is totally false. And we need to let people know, since most people assume that when a computer scans something it's being saved forever.
4/
@futurebird if this is still the AI scanning story: these books have ISBNs, so they are *very* unlikely to get lost, because they’re available at a state library!?
-
"At least if they scan it it will still be around forever."
That is not how training data for an LLM works. The LLM use the data to find patterns. Things that do not fit into patterns? They are lost forever.
If the AI scans my mom's book on Black Civil War veterans it will NOT be able to name them, or tell you what it said on the slate gravestones she made me take a rubbings of as a child. But it may reuse some of her choices of words if they have also been used by other authors.
5/
If you asked the LLM generate text about Civil War it might be a little more likely to mention that Black veterans existed. Maybe, at best.
This scanning isn't not like the projects we have heard about in the past that made many old books searchable online. It's destructive.
It is likely that the original scans are kept for future training. But these are valuable assets that will not be shared with the public.
6/
-
@futurebird if this is still the AI scanning story: these books have ISBNs, so they are *very* unlikely to get lost, because they’re available at a state library!?
Not all of them have ISBNs (my mom's book has one though... I helped with that) but some of this is like... yearbooks.
-
If you asked the LLM generate text about Civil War it might be a little more likely to mention that Black veterans existed. Maybe, at best.
This scanning isn't not like the projects we have heard about in the past that made many old books searchable online. It's destructive.
It is likely that the original scans are kept for future training. But these are valuable assets that will not be shared with the public.
6/
Maybe a solution would be to make ALL training data public? I don't know the correct framework for this but this whole "we are burning the library and keeping it secret" is intolerable. It's villainous.
7/7
-
J jwcph@helvede.net shared this topic
