1humouress
I'm halfway through reading one of my Early Reviewer wins and trying to be lenient about the writing style. I've jotted down notes as I'm going along and I was curious to see if other reviewers agreed with me so I took a peek at the reviews.
One reviewer claimed that the book was AI generated. I can't find evidence for or against but googling took me to Goodreads and the first review on there lays out a pretty convincing case, with details, for it being AI. I have no expertise in the subject though so I still don't really know if it is.
I was wondering if LibraryThing has a policy for Early Reviewer books being AI generated: whether authors are supposed to be allowed to submit such; if they should state that they've used AI, if so?
One reviewer claimed that the book was AI generated. I can't find evidence for or against but googling took me to Goodreads and the first review on there lays out a pretty convincing case, with details, for it being AI. I have no expertise in the subject though so I still don't really know if it is.
I was wondering if LibraryThing has a policy for Early Reviewer books being AI generated: whether authors are supposed to be allowed to submit such; if they should state that they've used AI, if so?
2lilithcat
We do not accept AI-authored titles in the Early Reviewers program.: https://www.librarything.com/ner/howitworksofferer
3humouress
>2 lilithcat: Ah, good. Thanks.
I wonder if LT has a way of confirming that books are not AI authored?
I wonder if LT has a way of confirming that books are not AI authored?
4keristars
There was a bit of disgust going around social media yesterday about at least 3 novels published recently that had chatGPT prompts left in the text!
iirc, they were all self-pub romance novels. Lena McDonald, KC Crowne and Rania Faris were the names I saw. I haven't looked to see if any of them offered books through ER, but i guess fyi that genAI "assistance" is a real problem.
iirc, they were all self-pub romance novels. Lena McDonald, KC Crowne and Rania Faris were the names I saw. I haven't looked to see if any of them offered books through ER, but i guess fyi that genAI "assistance" is a real problem.
5vwinsloe
Did everyone see the summer reading list that was published in Hearst newspapers that was full of fake book titles with real author's names that was generated by AI?
https://www.wbur.org/npr/nx-s1-5405022/fake-summer-reading-list-ai?utm_source=sf...
AI slop is really getting out of hand.
https://www.wbur.org/npr/nx-s1-5405022/fake-summer-reading-list-ai?utm_source=sf...
AI slop is really getting out of hand.
6lilithcat
>5 vwinsloe:
Yes, there was a discussion about it over here: https://www.librarything.com/topic/371058
Yes, there was a discussion about it over here: https://www.librarything.com/topic/371058
7AbigailAdams26
>1 humouress: The short answer to this question, as lilithcat notes, is that we do not accept books written with AI into the Early Reviewers program. We specify this on the How It Works page for authors and publishers: https://www.librarything.com/ner/howitworksofferer
That being said, as the staffer who approves giveaways for each batch, I do not read selections from each book, to try to determine whether AI was involved (I do check for other things, like publication date), so it's likely some will slip by me.
You can always reach out to me with concerns about specific books, either by PM-ing me here on the site, or via email, at: abigailadams@librarything.com
That being said, as the staffer who approves giveaways for each batch, I do not read selections from each book, to try to determine whether AI was involved (I do check for other things, like publication date), so it's likely some will slip by me.
You can always reach out to me with concerns about specific books, either by PM-ing me here on the site, or via email, at: abigailadams@librarything.com
8hipdeep
>3 humouress: It's not clear to me that *anyone* has a way of confirming that writing is not AI authored.
There is some interesting work being done in this area, including some important discussions of what a "good enough" tool would look like, but there are some fundamental logical problems. The biggest, IMO, is that the existing tools don't actually *address the question* of whether a piece of writing *was* AI authored. They address the question of, given a piece of writing, *how hard would it be* to make an AI say the same thing. This is profoundly different than the "this text is exactly the same as that other text" approach to anti-plagiarism software.
My crystal ball is notoriously terrible, but I'll say it's an open question at best whether AI detection is even possible, absent some kind of massively intrusive registry of AI-generated work.
There is some interesting work being done in this area, including some important discussions of what a "good enough" tool would look like, but there are some fundamental logical problems. The biggest, IMO, is that the existing tools don't actually *address the question* of whether a piece of writing *was* AI authored. They address the question of, given a piece of writing, *how hard would it be* to make an AI say the same thing. This is profoundly different than the "this text is exactly the same as that other text" approach to anti-plagiarism software.
My crystal ball is notoriously terrible, but I'll say it's an open question at best whether AI detection is even possible, absent some kind of massively intrusive registry of AI-generated work.
9timspalding
>3 humouress: I wonder if LT has a way of confirming that books are not AI authored?
Yes. We really can't confirm that. We couldn't even if we had the text, but we do not get the text of books before we offer them. So this has to be about our having a rule, members getting annoyed and authors fearing blow-back.
FWIW, I think it's clear that LLMs are going to be involved in most books in some way. But there is a world of difference between getting suggestions from an AI, or using a grammar checker, and having an AI write your book. If, for example, some author said that, as part of their process, they ask an LLM to read over their book and make bulleted comments and suggestions, we would not yank their book. I myself do this when I write a blog post. In my experience 1/3 of the LLM's suggestions are dumb, 1/3 are wrong but put the finger on something that could be improved, and 1/3 are good.
Yes. We really can't confirm that. We couldn't even if we had the text, but we do not get the text of books before we offer them. So this has to be about our having a rule, members getting annoyed and authors fearing blow-back.
FWIW, I think it's clear that LLMs are going to be involved in most books in some way. But there is a world of difference between getting suggestions from an AI, or using a grammar checker, and having an AI write your book. If, for example, some author said that, as part of their process, they ask an LLM to read over their book and make bulleted comments and suggestions, we would not yank their book. I myself do this when I write a blog post. In my experience 1/3 of the LLM's suggestions are dumb, 1/3 are wrong but put the finger on something that could be improved, and 1/3 are good.
10norabelle414
>8 hipdeep: Yeah, AI-detection-systems use AI to detect AI and are just as good at that as AI is at anything else (not very).
11timspalding
>10 norabelle414:
People who use AI a lot, and see a lot of AI writing, get good at recognizing it. Teachers and professors have become experts here. So I have some hope better automated detectors are possible.
People who use AI a lot, and see a lot of AI writing, get good at recognizing it. Teachers and professors have become experts here. So I have some hope better automated detectors are possible.
12keristars
re my comment at >4 keristars:, 404 Media did a write up.
imo it's pretty egregious for writers to be using genAI, which has been trained using books without petmission, and threatens those authors' future viability. It's more obvious when it's the image generators creating near-identical pictures based on movie frames, I think.
anyway https://www.404media.co/authors-are-accidentally-leaving-ai-prompts-in-their-nov...
imo it's pretty egregious for writers to be using genAI, which has been trained using books without petmission, and threatens those authors' future viability. It's more obvious when it's the image generators creating near-identical pictures based on movie frames, I think.
anyway https://www.404media.co/authors-are-accidentally-leaving-ai-prompts-in-their-nov...
These three authors are just the most recent examples. 404 Media previously reported that many more people publishing books right now are using some form of AI assistance. Amazon is filled with AI-generated slop. Even local libraries are starting to fill with AI-generated books written by authors who do not exist.
14.mau.
>12 keristars: Even local libraries are starting to fill with AI-generated books written by authors who do not exist
I thought that an author gets such a thin slice of the cost of a book that it was not worth to use AI for saving money :-)
I thought that an author gets such a thin slice of the cost of a book that it was not worth to use AI for saving money :-)
