AI written books

[j.p.s]

Originally Posted by Hitch

after all, Jon's still here, even if I kick the bucket.

Isn't usually being correct a necessary, although not sufficient or primary, requirement to be a pedant?

[Hitch]

Originally Posted by j.p.s

Isn't usually being correct a necessary, although not sufficient or primary, requirement to be a pedant?

CUTE, there, @j.p.s,but if memory serves, according to my buddy the OED, no, actually, accuracy and correctness are not the primary driving forces--it's the concern with being correct and formal, whether or not they are merited, overblown, etc. :-)

So, I know I can not worry about all y'alls if I kick it.

Hitch

[Quoth]

No-one likes a pedant, but they hate ones that are always 100% correct. Especially when it seems crazy, but they are right about the city burning down. See also Cassandra and "Clever Dicks".

[Quoth]

https://www.theguardian.com/books/20...ng-their-books

Two authors have filed a lawsuit against OpenAI, the company behind the artificial intelligence tool ChatGPT, claiming that the organisation breached copyright law by “training” its model on novels without the permission of authors.

Mona Awad, whose books include Bunny and 13 Ways of Looking at a Fat Girl, and Paul Tremblay, author of The Cabin at the End of the World, filed the class action complaint to a San Francisco federal court last week.

[ZodWallop]

Originally Posted by Quoth

https://www.theguardian.com/books/20...ng-their-books

I hope that lawsuit doesn't go anywhere.

[Quoth]

Why? The content was likely scraped from a pirate site, but hopefully the court case will answer the questions.

Also why should corporations building "AI systems" get all the training data for free and why should they scrape without curating the truth/accuracy or copyright status of content?

Edit:
These are not opinions but questions that need answered. MS funded Open AI refuses to say exactly how they obtain or curate content.

[DiapDealer]

I also hope the case goes nowhere. A "win" would only open things up for copyright holders to sue any one for studying authors, or genres (without "permission") and using what they learned to write their own works that are influenced by what they read.

I know you like to hang your hat on the differences between machine "training" and human "influence/study" But I don't think those differences amount to actual legal distinctions. Could be wrong, but I don't believe I am.

If they had any real evidence that ChatGPT used pirated copies of works for its training, then that should be the legal issue that should be being pursued. Not the silly, overreaching accusation of merely accessing copyrighted works in order to train a system to write its own fiction. Humans have done that for centuries.

I don't see any legal relevance in the difference between influence/knowledge being gained by words being scraped by machine "eyes" and influence/knowledge being gained by words being scraped by human eyes.

Time will tell.

[nabsltd]

Originally Posted by DiapDealer

I don't see any legal relevance in the difference between influence/knowledge being gained by words being scraped by machine "eyes" and influence/knowledge being gained by words being scraped by human eyes.

Unfortunately, there is a huge legal difference if the machine-readable form (e-book) was never released without DRM. The only way a machine learning system could be trained on such works is by defeating the DRM, and that's a copyright violation all by itself.

Humans don't have to defeat the DRM to "scrape" the data in an e-book, because there are authorized devices and software that the human can use to get the data into their eyes.

[DiapDealer]

Originally Posted by nabsltd

Unfortunately, there is a huge legal difference if the machine-readable form (e-book) was never released without DRM. The only way a machine learning system could be trained on such works is by defeating the DRM, and that's a copyright violation all by itself.

First off: no, defeating DRM is NOT a copyright violation all by itself. And second: even if it were, there are countless ways that a machine learning system could be trained on DRMed content (without defeating it).

And we're right back to no legal difference in my book. Not to mention being right back to needing to sue people over actual legal violations (piracy, DRM removal, etc...) rather than inventing legally irrelevant (in my opinion) differences between human and machine consumption/study.

[Quoth]

When you read a DRM'd or not book you are not making a copy. The ML process makes a copy and keeps it, though some disingenuously claim it doesn't.

I'm happy for the court to decide. Open AI will have to divulge how it is and what on the "training" is done or no-one can make an informed decision.

The "so called Libertarians" want money making corporations scraping the internet to be treated the same as people reading it.

Also ML /Chat bots absolutely do not produce works the same way as humans influenced and inspired by what they consumed. That's a corporate lie.

Let's have some transparency. We don't know what content they are using or how they process it.

[DiapDealer]

Originally Posted by Quoth

I'm happy for the court to decide. Open AI will have to divulge how it is and what on the "training" is done or no-one can make an informed decision.

That's fair. However ...

Originally Posted by Quoth

The "so called Libertarians" want money making corporations scraping the internet to be treated the same as people reading it.

It has nothing to do with WANT in my case. No one who knows anything about me would mistake me for a Libertarian. Yet I also believe scraping the internet to be the same as people reading it (regardless of whether the scraping done by a corporation or not).

Originally Posted by Quoth

Also ML /Chat bots absolutely do not produce works the same way as humans influenced and inspired by what they consumed. That's a corporate lie.

That statement is at complete odds with your initial claim of being "happy for the court to decide." You've offered no proof of corporate lies. Only opinions. Why not let the courts decide? That seems to be the gist of the law suit in question afterall.

[Quoth]

No-one called you a "so called Libertarian". That was a general comment

That statement is at complete odds with your initial claim of being "happy for the court to decide." You've offered no proof of corporate lies. Only opinions. Why not let the courts decide? That seems to be the gist of the law suit in question afterall.

No, it's a fact they won't say what they scrape (copy) and how they process it. Various companies with "AI" have been caught lying.

Also ML /Chat bots absolutely do not produce works the same way as humans influenced and inspired by what they consumed. That's a corporate lie.

Completely separate issue to this court case and fact.

[DiapDealer]

Completely separate issue to this court case and fact.

Is it though? Both separate AND fact? I'm thinking it's neither. Otherwise, there would not be the amount of handwringing, legal debate and/or armchair litigation we have over it right now.

What I'm hearing is a lot of opinions on what constitutes intelligence or learning being presented as facts (some by people who offer great opinions). But I am not hearing facts. Only noise.

[Hitch]

Originally Posted by DiapDealer

I also hope the case goes nowhere. A "win" would only open things up for copyright holders to sue any one for studying authors, or genres (without "permission") and using what they learned to write their own works that are influenced by what they read.

I know you like to hang your hat on the differences between machine "training" and human "influence/study" But I don't think those differences amount to actual legal distinctions. Could be wrong, but I don't believe I am.

If they had any real evidence that ChatGPT used pirated copies of works for its training, then that should be the legal issue that should be being pursued. Not the silly, overreaching accusation of merely accessing copyrighted works in order to train a system to write its own fiction. Humans have done that for centuries.

I don't see any legal relevance in the difference between influence/knowledge being gained by words being scraped by machine "eyes" and influence/knowledge being gained by words being scraped by human eyes.

Time will tell.

There is definitely plenty of space for sliding slopes in here. Not sure yet where they all are, but...they are hard to miss, looming in the distance like that, like the next iteration of the Himalayas. Yup, unmissable.

You should HEAR the dirging (it's piteous, truly) on the KDP forums already! Good god, it's like they're all being driven on by the lamentations of their women and all that...

Hitch

[nabsltd]

Originally Posted by DiapDealer

First off: no, defeating DRM is NOT a copyright violation all by itself.

17 USC 1201(a)(1) says it is. For an easy-to-understand analysis, see https://en.wikipedia.org/wiki/Anti-c...#United_States

And, just because you "bought" a book from Amazon, you do not have Amazon's to use something like deDRM_tools to remove it so you can read the book on you Kobo.

Nobody will ever get sued for removing DRM for personal use, but that does not change the fact that it could happen, because the law is on the side of the copyright holder.

It also may never get tested directly in court, because any time it comes up, it is part of a larger case, where somebody broke DRM then distributed the copies. Proving the distribution is all that is needed to win, so the DRM breaking usually isn't litigated.

That said, I don't think there is anything morally wrong with removing DRM for personal use. I know I'm committing copyright infringement when I do it...I just don't care.