AI written books

[John F]

Originally Posted by John F

My local channel (WFSB) had an interview with author Luanne Rice, and the anchor said that "... They (the AI developers) are taking your books and in this case pirated. So an ebook version that nobody has even paid for, inputting it into A.I. ..."

I had assumed that they were using legal purchased (or public domain) material. I was surprised buy this in that they would cheap out and expose themselves to litigation on this front.

...?

Someone PM'd me that it was proven that they are using copyrighted books. Any one have link(s) for the proof?

The PM also said that even using purchased copyrighted material was a copyright violation. I have my doubts about this, and will wait for the courts opinion/clarification on this.

[Hitch]

Originally Posted by John F

Someone PM'd me that it was proven that they are using copyrighted books. Any one have link(s) for the proof?

The PM also said that even using purchased copyrighted material was a copyright violation. I have my doubts about this, and will wait for the courts opinion/clarification on this.

This, AFAIK, is that "list" of reading materials--I don't think that the genesis of the books being used for learning is at dispute, is it? Or...anybody know?

I finally left a comment on the YT version of that video (with Rice). Again...we really have entered the age of unreason, in which FACTS have become something akin to "your mummy wears army boots." It's...discouraging, it really is.

Hitch

[graycyn]

Originally Posted by Hitch

The result? The same shitty books that I've seen now, on KDP, without any benefit of AI, for the last 15, coming up on 16 years.

Hitch

Sometimes bad is just, you know, bad.

But sometimes, sometimes bad is bad
Ah doo doo doo doo doo doo ah
I said cool is a rule, but, sometimes, bad is bad

--Bad Is Bad
Song by Huey Lewis and the News

[Hitch]

Originally Posted by graycyn

Sometimes bad is just, you know, bad.

Yes, yes, yes! Amen! That's the truth. I've seen enough BAD, long 'ere the whispers of AI, that I already know that it doesn't take ChatGPT or Claude or Bard or (whoever the hell) to WRITE bad stuff.

For that matter, most of the stuff that these BURN THE WITCHERS are espousing as being AI-written, I would quite literally bet very good-sized amounts of my own cash moolah, were not. Because in my experience, which is now fairly substantial, the damned AI writes BETTER. I can promise you--99% of what's being decried is not AI. It's just crap writing. An d translating and...all the other realities around writing!

Is it something *I* want to read? No, it can be a bit florid, or too vanilla or too this or that, but it's better generally than most of what is being bandied about as "AI" out there, and I would love to see these Henhouse Hens stand up and proclaim, I really would. And brothers and sisters, I would LOVE to see them put their goddamned money where their mouths are about this too.

Hitch

[Quoth]

The Tsunami of rubbish ebooks started a few years before the AI books. The AI detection tools are useless. You can decide a book is poor, or plagiaristic, but that doesn't make it AI. If you are very familiar with the writing from as student you'll spot when they are either using AI or outright copying (maybe search & replace of names, places, car models and more work for she<->he). You might not know which it is if it's poor.

If you are very familiar with an author you might spot someone pinching their name. if it's very good and not plagiarism it might not be AI.

It takes expertise, not just some SW tools to detect AI or human plagiarism. Even then you can be mistaken; big publishers have been fooled and had to pulp. Plagiarism is as old as writing.

Perhaps even before the advent of AI written text the idea of Self Publishing ebooks at zero cost and no "gate keeper" had become broken as people churned out and published junk.

[Hitch]

Originally Posted by Quoth

The Tsunami of rubbish ebooks started a few years before the AI books. The AI detection tools are useless. You can decide a book is poor, or plagiaristic, but that doesn't make it AI. If you are very familiar with the writing from as student you'll spot when they are either using AI or outright copying (maybe search & replace of names, places, car models and more work for she<->he). You might not know which it is if it's poor.

If you are very familiar with an author you might spot someone pinching their name. if it's very good and not plagiarism it might not be AI.

It takes expertise, not just some SW tools to detect AI or human plagiarism. Even then you can be mistaken; big publishers have been fooled and had to pulp. Plagiarism is as old as writing.

Perhaps even before the advent of AI written text the idea of Self Publishing ebooks at zero cost and no "gate keeper" had become broken as people churned out and published junk.

From my first perusal of Smashwords in the mid-to-late aughts....the torrent (ha!) of garbage was obvious and again, back then, nobody was yammering and repeating and parroting "AI AI AI!" as if the WitchFinder General was running a contest for the most obsequious.

It's...it is what it is. There was a time when we had gatekeepers; if naught else, lack of money "helped" in filtering out some of the dreck. (When I first started in the print side of this biz, in this century, I mean, a would-be self-publisher had to be prepared to pay thousands--thousands--just for a print layout for their novel and cover, ignoring printing costs, etc.)

That alone kept out a lotta the woulda-shoulda's. Now? Now there's naught blocking the door and those selfsame gatekeepers that used to protect us from the onslaught--now, we are they.

Hitch

[DNSB]

Originally Posted by Hitch

That alone kept out a lotta the woulda-shoulda's. Now? Now there's naught blocking the door and those selfsame gatekeepers that used to protect us from the onslaught--now, we are they.

Hitch

And sadly, we are not keeping the glut of books that would not have made it past the slush pile from being published—often accompanied by reviews that suggest the person writing the review has never read the book but they *LOVE* it anyways.

[Fiat_Lux]

Originally Posted by John F

Someone PM'd me that it was proven that they are using copyrighted books. Any one have link(s) for the proof?

The PM also said that even using purchased copyrighted material was a copyright violation. I have my doubts about this, and will wait for the courts opinion/clarification on this.

Very specific vocabulary was used by the AI that is found only in books of an extremely specific sub-niche of an extremely obscure sub-genre of a niche, of an obscure sub-genre of an RPG (I think) that nobody has heard of it.

I don't have links.
At least one if the items in question was both age-restricted and geographically restricted.

^1: Age restriction is due to sexual content. Geographic restrictions due to mere discussion of said act is illegal in some jurisdictions.

[Fiat_Lux]

Originally Posted by Hitch

If they have to, they'll find a way to hide it, to hide the servers, etc.
They'll create AI-beards, if needed, but...unless the AIs are literally shut down, unless the plug is pulled on Claude, et al, this is never unwinding. I don't give a flying rat's what the AG suit does. IT IS TOO LATE.

You can't shut the AIs down.
* Source code for one of the prototypes of ChatAI (i think. might have been the competitors) was inadvertantly leaked by the developers. Said source code is available on a website blocked by 75+ countries.
* At least one AI has been released with a GPL licence.
* Tom Swan _Pascal Programs For Business_ has an early version of code that generates text that sounds like the human that it was "trained on" writes.

If one is willing to the manual proofreading & OCR work, 10,000,000+ public domain/CC0 works are available for the asking, to use for training purposes.

[Hitch]

Originally Posted by Fiat_Lux

You can't shut the AIs down.
* Source code for one of the prototypes of ChatAI (i think. might have been the competitors) was inadvertantly leaked by the developers. Said source code is available on a website blocked by 75+ countries.
* At least one AI has been released with a GPL licence.
* Tom Swan _Pascal Programs For Business_ has an early version of code that generates text that sounds like the human that it was "trained on" writes.

If one is willing to the manual proofreading & OCR work, 10,000,000+ public domain/CC0 works are available for the asking, to use for training purposes.

Yes, exactly.

Vis the allegedly infamously hidden, sub-genre of sub-genre, in forbidden areas...yeah, well. My gut on that is a big "yeah, sure it was/is/does." I'm not critiquing you; just...I mean, we have this broadbrush claim that 79 or 89 out of 100 books--HARDLY some sub-genre of a sub-genre, etc. in YA novels were AI written. Eighty percent, basically. That comment that claim has no dataset intersection, that I can see, what what you are discussing.

That claim, made during that video, is just (IMHO) made-up and repeated and regurgitated drivel. Which is my rather vociferous objection to it. Being flung around as "fact" by people who can barely spell fact, much less recognize one when it bites them in the ass.

Hitch

[John F]

Originally Posted by Fiat_Lux

Very specific vocabulary was used by the AI that is found only in books of an extremely specific sub-niche of an extremely obscure sub-genre of a niche, of an obscure sub-genre of an RPG (I think) that nobody has heard of it.

I don't have links.
At least one if the items in question was both age-restricted and geographically restricted.

^1: Age restriction is due to sexual content. Geographic restrictions due to mere discussion of said act is illegal in some jurisdictions.

My apologies, I meant to ask if it was proven that they were using pirated copyrighted material.

I don't doubt that they are using copyrighted material.

[Hitch]

Originally Posted by John F

My apologies, I meant to ask if it was proven that they were using pirated copyrighted material.

I don't doubt that they are using copyrighted material.

Yup, I'm with him, question-wise ^^^. What, specifically, has actually been proven, rather than simply alleged? Anybody know?

Hitch

[kacir]

Originally Posted by Fiat_Lux

You can't shut the AIs down.
* Source code for one of the prototypes

Source code is not the problem. Training it properly is. And also having enough computing power to run ChatGPT. At the beginning one answer that ChatGPT provided to an astonished public testing it was backed up by immense server capacity and ChatGPT was burning through One Thousand Milion(*) of dollars in investment. This is why after a while people started to complain it was being dumbed down. It was way too expensive to operate at full tilt. There was also a scandal about the abuse of people that were testing the first versions, because you have to teach an AI to answer properly - without racism, without inciting violence, without using graphic sexual language, without telling the user to go and kill themselves. It was "trained" not only on books but also on huge amounts of text scraped from questionable Internet discussion forums and chat-boards.

(*) I do not want to use word billion because it means 10^9 in USA and 10^12 in the rest of the world.

[Quoth]

And electricity and cooling water.

[John F]

Originally Posted by kacir

...

(*) I do not want to use word billion because it means 10^9 in USA and 10^12 in the rest of the world.

Are we sure about this?

According to wiki, billion is used differently around the world.

https://en.m.wikipedia.org/wiki/Billion