I will blame overlong copyright term lengths. 70 years after authors death or 95 years after publication, allowing most recent work to enter the commons effectively after a century, or more, from now [0].
This is the rare case when Europe is even worse. Metropolis, the 1927 Fritz Lang film, is out of copyright in the United States but will still be in copyright in Germany until 2047: 120 fucking years.
It’s preposterous, and offensive to anyone’s intelligence to claim that this is about incentivizing production; does anyone seriously believe there is a potential artist out there who would avoid making their magnum opus if it could only be under copyright for 119 years?
Given the argument over LLMs consuming books illegally, I think publishers could be a little concerned that an LLM that combined partial previews on every modern work on a subject might be a destroyer of the market for the average book on the subject with the license to do so having been properly granted via this feature.
Right, that's what I'm saying. For whatever reason it seems publishers decided they don't want their preview-only books as part of the full-text search across all books. If they decide that, Google has to comply.
This isn't like web search where web pages are publicly available and so Google can return search results across whatever it wants. For books, it relies on publisher cooperation to both supply book contents for indexing under license and give permissions for preview. If publishers say to turn off search, Google turns off search.
Data that you can prove was generated by humans is now exceedingly valuable ...and most of that comes from the days before LLMs. The situation is a bit like how steel manufactured before the nuclear age is valuable.
I just checked and yes, search inside of books with previews is still possible.
(a) when you search books.google.com and find a book with a preview, it opens their new book viewer - the search is at the bottom of the page. You can also click "View All" to see all references of your search in that book.
(b) if you go to the book homepage (clicking X in the top right of the book viewer if that opened), there's still a "Search Inside Book" next to the "Preview" button under the title.
So, if you search for some text that occurs at the end of one chunk, will it then preview a following chunk? And could chaining these chunks give you the entire book?
If so, I could see someone doing this to exfiltrate books.
You're talking about in-book search (TFA is about search across all books), and yes that was indeed once a known technique for extracting whole or nearly whole books.
That's why publishers responded by excluding sections of books from search (it will list the pages but you can't view them), and individual Google accounts became limited in how many extra pages they were ever allowed to see of an individual book beyond the standard preview pages.
But then LibGen, Z-lib, and Anna's Archive became popular and built up their collections...
Since I pretty much only use Google Books for public domain books, old magazines, and newspapers I haven't noticed any problem with it. Maybe it's not as dead as this person thinks.
If search gives you a preview with a few surrounding words, it is fairly simple to abuse search with quotation marks to extract bigger and bigger sections of the books, potentially till you have the whole book.
I’m genuinely curious how you feel about LLMs being trained on pirated material. Not being snarky here.
Your comment reflects the old “information wants to be free” ideals that used to dominate places like HN, Slashdot, and Reddit. But since LLMs arrived, a lot of the loudest voices here argue the opposite position when it comes to training data.
I’ve been trying to understand whether people have actually changed their views, or whether it’s mostly a shift in who is speaking up now.
why would that change anything? copyright is still a tax on the whole of society for the benefit of rich people and corporations. it opposes innovation, evolution and progress
maybe a short copyright would be fine (10 year fixed?) but copyright as-is seems indefensible to me
They don't do full text search anymore esp for copyrighted books. I wonder if this is not a regression but an intent to give them a let up in the AI race.
The YT transcripts are linked to on the YT page itself. If they remove that, it is trivial to use a local STT model to transcribe the video. If they make it impossible to download a video, you could just have a microphone record all of the sound, and so on. Once you have the transcription of anything, summarizing is trivial. I have a local script that does this and I use it all of the time. Also produce diagrams for YT summaries. Hours saved, per day.
The left results are contemporary, the right are decades old. That includes editions of the same book --- surely the newer edition is going to be preferred by most readers.
Google Books is long dead. If you click on the author's name in one of the results, it will search inauthor:"Author's Name" and this search will return garbage because it chokes on double quotes. This has been true for at least a couple of years; Google Books is not compatible with itself. Changing the double quotes to single quotes fixes it. Also, lately, when you filter only for books that have Full View some results that have Full View get dropped for no intelligible reason.
Nobody is looking at it. I wouldn't be surprised if the preview search was switched off by accident.
For me Books is only useful (and it is very useful) for books out of copyright, 100+ years old. Sometimes they aren't at archive.org.
I hate Google, but I think it's a bit absurd to criticize them on this if somehow it's over AI. The only reason Google created Books may even have been AI, but they were hoping to have the books open to everyone, and the publishers and authors whose full text is being blocked are literally the people who stopped it from happening. Maybe they spoke up about AI, too. I find it even hard to even criticize that Google doesn't take care of Books - it has no purpose or profit potential for them anymore, it's obviously charity that they don't take it down completely.
I suspect it's actually the opposite. Standard inverted index text search is incredibly cheap and mature. Vector search requires generating embeddings and running approximate nearest neighbor queries, which is significantly more compute intensive than simple keyword matching. If they switched, it wasn't to save on compute costs.
Remember that preview functionality is granted by contract with the publishers. Which is why some books have it and some books don't.
Almost certainly, this is something that publishers requested the removal of, under threat of requiring previews to be removed entirely.
Books that are out of copyright still have full search and display enabled.
So blame publishers, not Google.
I will blame overlong copyright term lengths. 70 years after authors death or 95 years after publication, allowing most recent work to enter the commons effectively after a century, or more, from now [0].
[0] https://en.wikipedia.org/wiki/Copyright_law_of_the_United_St...
This is the rare case when Europe is even worse. Metropolis, the 1927 Fritz Lang film, is out of copyright in the United States but will still be in copyright in Germany until 2047: 120 fucking years.
It’s preposterous, and offensive to anyone’s intelligence to claim that this is about incentivizing production; does anyone seriously believe there is a potential artist out there who would avoid making their magnum opus if it could only be under copyright for 119 years?
Given the argument over LLMs consuming books illegally, I think publishers could be a little concerned that an LLM that combined partial previews on every modern work on a subject might be a destroyer of the market for the average book on the subject with the license to do so having been properly granted via this feature.
The previews are still there though, they just don't rank.
Right, that's what I'm saying. For whatever reason it seems publishers decided they don't want their preview-only books as part of the full-text search across all books. If they decide that, Google has to comply.
This isn't like web search where web pages are publicly available and so Google can return search results across whatever it wants. For books, it relies on publisher cooperation to both supply book contents for indexing under license and give permissions for preview. If publishers say to turn off search, Google turns off search.
Anna's Archive [0]:
> The largest truly open library in human history
[0] https://en.wikipedia.org/wiki/Anna%27s_Archive
Mirrors https://open-slum.org/
Open-slum currently experiencing heavy traffic, but here's an additional mirror: https://open-slum.pages.dev/
How funny. They have a DMCA Takedown Requests link...
[dead]
It might be time to update the mission statement.
“Our mission is to organize the world’s information and make it universally accessible and useful”
https://about.google/company-info/
Why it's almost certainly not by choice.
* for us, advertisers and our AI models
My guess is that AI training is the main issue.
Data that you can prove was generated by humans is now exceedingly valuable ...and most of that comes from the days before LLMs. The situation is a bit like how steel manufactured before the nuclear age is valuable.
4 replies →
I just checked and yes, search inside of books with previews is still possible.
(a) when you search books.google.com and find a book with a preview, it opens their new book viewer - the search is at the bottom of the page. You can also click "View All" to see all references of your search in that book.
(b) if you go to the book homepage (clicking X in the top right of the book viewer if that opened), there's still a "Search Inside Book" next to the "Preview" button under the title.
But you have to know what book you are looking for.
So, if you search for some text that occurs at the end of one chunk, will it then preview a following chunk? And could chaining these chunks give you the entire book?
If so, I could see someone doing this to exfiltrate books.
You're talking about in-book search (TFA is about search across all books), and yes that was indeed once a known technique for extracting whole or nearly whole books.
That's why publishers responded by excluding sections of books from search (it will list the pages but you can't view them), and individual Google accounts became limited in how many extra pages they were ever allowed to see of an individual book beyond the standard preview pages.
But then LibGen, Z-lib, and Anna's Archive became popular and built up their collections...
Google Books could have been a subscription service ala Netflix.
Then it would have been hella useful.
Since I pretty much only use Google Books for public domain books, old magazines, and newspapers I haven't noticed any problem with it. Maybe it's not as dead as this person thinks.
This was addressed in the post, I'm sure you just missed it when you read it:
"But a few days ago they removed ALL search functions for any books with previews, which are disproportionately modern books." <emphasis mine>
right, my point was just because what they use it for is now useless mine isn't and personally I think mine is more useful.
No the search results went from pretty good to absolute garbage https://bsky.app/profile/adamnemecek.bsky.social/post/3mdbup...
My guess is they detected being scraped and did this as preventive measure.
My guess is they're cozier with publishers now than 20 years ago when they fought all the way to SCOTUS.
"Hey, remove search?"
"OK, it was costing money anyways."
If search gives you a preview with a few surrounding words, it is fairly simple to abuse search with quotation marks to extract bigger and bigger sections of the books, potentially till you have the whole book.
my guess is that the copyright landscape changed due to AI training, and these publishers won't let Google use that data anymore
The books are still there, it seems like the rankings have changed though.
Thats easy.
Check out library genesis, Anna's archive, and scihub for content.
Piracy isnt theft if buying isnt ownership.
I’m genuinely curious how you feel about LLMs being trained on pirated material. Not being snarky here.
Your comment reflects the old “information wants to be free” ideals that used to dominate places like HN, Slashdot, and Reddit. But since LLMs arrived, a lot of the loudest voices here argue the opposite position when it comes to training data.
I’ve been trying to understand whether people have actually changed their views, or whether it’s mostly a shift in who is speaking up now.
why would that change anything? copyright is still a tax on the whole of society for the benefit of rich people and corporations. it opposes innovation, evolution and progress
maybe a short copyright would be fine (10 year fixed?) but copyright as-is seems indefensible to me
Ironic those doing the most for making information open and accessible are the criminals.
Of course. When it's criminal to make information open and accessible, only criminals will make information open and accessible.
A centuries old problem. Early translations of the Bible to English were illegal or required licenses.
William Tyndale was put to death for translating the Bible into English, which would have been an act to make information open and accessible.
1 reply →
None of these does full text search.
And they are under constant threat by nation states. sci-hub hasn't seen new papers in ages.
Build a local index
1 reply →
zlibrary does
https://en.wikipedia.org/wiki/Z-Library
4 replies →
Title is: Google has seemingly entirely removed search functionality from most books on Google Books
The change happened on or around Jan 21. Overnight the results went from pretty good to absolute trash.
Here are two screenshots taken on Jan 20 and Jan 23 https://bsky.app/profile/adamnemecek.bsky.social/post/3mdbup...
They don't do full text search anymore esp for copyrighted books. I wonder if this is not a regression but an intent to give them a let up in the AI race.
Yup, it's for AI.
Similarly, a year ago or so ChatGPT could summarize YouTube videos. Google put a stop to that so now only Gemini can summarize YouTube videos.
The YT transcripts are linked to on the YT page itself. If they remove that, it is trivial to use a local STT model to transcribe the video. If they make it impossible to download a video, you could just have a microphone record all of the sound, and so on. Once you have the transcription of anything, summarizing is trivial. I have a local script that does this and I use it all of the time. Also produce diagrams for YT summaries. Hours saved, per day.
It isn't obvious why the left results are preferred over the right results.
The left results are contemporary, the right are decades old. That includes editions of the same book --- surely the newer edition is going to be preferred by most readers.
4 replies →
Google Books is long dead. If you click on the author's name in one of the results, it will search inauthor:"Author's Name" and this search will return garbage because it chokes on double quotes. This has been true for at least a couple of years; Google Books is not compatible with itself. Changing the double quotes to single quotes fixes it. Also, lately, when you filter only for books that have Full View some results that have Full View get dropped for no intelligible reason.
Nobody is looking at it. I wouldn't be surprised if the preview search was switched off by accident.
For me Books is only useful (and it is very useful) for books out of copyright, 100+ years old. Sometimes they aren't at archive.org.
I hate Google, but I think it's a bit absurd to criticize them on this if somehow it's over AI. The only reason Google created Books may even have been AI, but they were hoping to have the books open to everyone, and the publishers and authors whose full text is being blocked are literally the people who stopped it from happening. Maybe they spoke up about AI, too. I find it even hard to even criticize that Google doesn't take care of Books - it has no purpose or profit potential for them anymore, it's obviously charity that they don't take it down completely.
My guess: Text search and indexing is expensive. And you are getting some kind of AI vector search instead.
Which tends to be kind of poop compared to true text search.
I suspect it's actually the opposite. Standard inverted index text search is incredibly cheap and mature. Vector search requires generating embeddings and running approximate nearest neighbor queries, which is significantly more compute intensive than simple keyword matching. If they switched, it wasn't to save on compute costs.