this post was submitted on 20 Sep 2023
285 points (95.5% liked)

Technology

58303 readers
9 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. Check for duplicates before posting, duplicates may be removed

Approved Bots


founded 1 year ago
MODERATORS
 

John Grisham, George R.R. Martin and more authors sue OpenAI for copyright infringement::John Grisham, Jodi Picoult and George R

all 36 comments
sorted by: hot top controversial new old
[–] [email protected] 52 points 1 year ago (2 children)

Oh, so that's what GRRM is doing instead of writing the next ASOIAF book!

But seriously, AI proponents seem to think it's the Wild West and they can use whatever they want, for profit, with no repercussions...

[–] [email protected] 21 points 1 year ago (1 children)

Thinking about it, perhaps GRRM should've used AI as a sounding board in order to get his book finished much sooner.

[–] [email protected] 9 points 1 year ago (2 children)

As much as I complain about his extreme slowness, I'd still rather get a book written 100% by GRRM! The computers can get their turn after he, me, you, and everyone else is dead...

[–] [email protected] 7 points 1 year ago (2 children)

I was just thinking the opposite. Maybe we can get AI to give us that last book that we are otherwise may never see. Especially if it actually does end as poorly as the TV series did.

Chat GPT write me a final novel for ASOIF in the style GRRMartin but with a better ending.

[–] [email protected] 2 points 1 year ago

I think I just read somewhere that there was already an AI-produced version of the last two books.

[–] [email protected] 2 points 1 year ago

Speaking of unfinished books before the creator died, I'd really like to know how AI would try to finish Tintin and Alph-Art, the last book before Herge died

[–] [email protected] 4 points 1 year ago* (last edited 1 year ago)

I get your sentiment, and I'm not gonna argue against it. As a freelance illustrator, I've had my own fair share of problems meeting deadlines. I wish I can just make art at my own pace, but my clients want work to be churned out as soon as possible. Which I will relate this conversation with AI. I feel like AI (at its current usage) just perpetuates the vicious cycle in capitalism where quantity trumps quality, or that they have unrealistic expectations for work to be of high quality but with unrealistic deadlines. That's why they turn to AI.

[–] [email protected] 0 points 1 year ago* (last edited 1 year ago) (1 children)

AI proponents seem to think it’s the Wild West

How exactly has OpenAI harmed GRRM? It's not like you can ask it to output a free copy of an entire book.

Last I checked, the author is doing just fine, and I'm sure if he finishes another book before retiring it will sell very well regardless of the outcome of this lawsuit.

Also - I bet if OpenAI's balance sheet were public there wouldn't be any profits to be found. They're losing money hand over fist.

[–] [email protected] 1 points 1 year ago (1 children)

If they used his works to train models without arranging proper licensing prior they've been unjustly enriched.

[–] [email protected] 9 points 1 year ago* (last edited 1 year ago) (1 children)

I get the logic, but I think it is a more complex issue than that.

How many writer's have read his works and been influenced by them? Did they buy a proper license or just buy/borrow the book from somewhere?

[–] [email protected] 1 points 1 year ago (2 children)

There's a fundamental difference between an author inspired by anothers work and an algorithm that was manufactured by a corporation to comb others work and reproduce derivatives, in my mind.

[–] [email protected] 5 points 1 year ago* (last edited 1 year ago) (2 children)

Like the other commenter, I would be genuinely curious to hear your thoughts on that fundamental difference.

I am by no means an AI expert, but my impressions is that AI sill needs to process each book and incorporate the new knowledge into its existing knowledge. Which at least from a surface level sounds a lot like what I do when I read a new book.

The fact that each AI is effectively a non-sapient slave of a person or corporation really doesn't change my opinion.

Have you ever had a reason to read much in a new or developing sub-genre? As a fan of LitRPG, a genre that virtually didn't exist 10 years ago, I can tell you with some certainty that everything is a derivative work of something. It is amazing how as soon as one author pulls in and idea from another genre, the next 30 novels that come out will have some variation of the same idea, and the 300 that follow it will each have variations of those.

[–] [email protected] 7 points 1 year ago* (last edited 1 year ago) (1 children)

I replied to the other guy with an answer on the difference. The TLDR IMO there's a difference between someone who has been inspired by another work and someone who has taken a work with the intent creating a product based on it.

It's one thing for those 30 LitRPG books to come out from authors who loved the concepts of the first book. It'd be another thing for a company to analyze what made the first book successful (for lack of a better word the author's style) and then create a tool that can release hundreds of LitRPG books a day flooding the market and making it harder for the original author to sell his work.

Also the authors of those inspired novels have the capacity to add their own creativity to their works an AI can only add the creativity from other's works.

Thanks for the civil response.

[–] [email protected] 3 points 1 year ago* (last edited 1 year ago) (1 children)

Thank you for the response. I am not sure I agree with your exact stance, but you make several compelling points along the way.

Using the Fair Use doctrine is definitely a good way to narrow down where the dividing line is. I think we can easily agree that making a GRRM specific AI to make derivative, non-parody, commercial works would definitely be on the wrong side of the line.

When I was picturing the bots, I was picturing something more along the lines of AI bots that had consumed all human literary works, or at the very least all modern English literary works.

ChatGPT write me a short story where the Main Character is a Magical Golem that follows the Three Laws of Isaac Asimov. It should be written in the style of a Greek Tragedy but set in Feudal Japan. The Main Character should be able to gain in magical power until he eventually attempts to break into the Heavens. There should be gods trying to interfere in his ascension but not in ways the MC cannot resolve. Base the gods off of archetypes from Norse Mythology, but name them after characters from GRRM's game of thrones based on similar personality types.

Such a work would both be wholly derivative and yet wholly unique. Despite swiping GRRM's unique names this work should be perfectly fine in my mind. Edit: Even if it was commercialized.

[–] [email protected] 4 points 1 year ago

I see what you're getting at but I feel that once we get past the GRRM specific example some of the same issues still remain.

For example if it was instead a GRRM & JK Rowling and wrote in a style of the two in my eyes it hasn't escaped the problem of the GRRM AI it's compounded them by adding one more author who's works is being used. Then we could extrapolate that out to it's true scale where it uses most written literature (although some presumably pub domain).

I just see it as if doing it to one person's work is wrong doing to two people's work aught to be twice as bad.

[–] [email protected] 2 points 1 year ago

That's pretty much my opinion about this, too. It's not like GRRM invented dragons that were hatched from eggs or anything like that. Having said that, I do think it's problematic if the AI model belongs to a company and it's not transparent about what data is being used to train their model.

[–] [email protected] 1 points 1 year ago (1 children)

Please, articulate that difference, focusing specifically on what copyright actually protects.

That's not a flippant statement. That's an open, honest request.

I do not see any relevant difference between hiring a person to read and discuss a bunch of books, and a program built to read and discuss a bunch of books.

[–] [email protected] 6 points 1 year ago (1 children)

I'd argue that the scenario where someone is paying someone to intentionally replicate others work shares the same problems I see with a program doing the same. There's a difference IMO from the man who wants to write a fantasy story because he loved GRRM's work and the man who is being payed to analyze the works of GRRM to create a product.

As for copyright protections, I'd recommend Tom Scott's video on the subject. There's a number of factors considered when deciding fair use and there's no clear line in the sand. The important things to consider is how and why a source material was used. If this program was made to solely produce a parody of GRRM's world or writing style that'd be a strong fair use argue IMO.

Here's the factors for fair use:

  1. the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes; 2)the nature of the copyrighted work;
  2. the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and 4) the effect of the use upon the potential market for or value of the copyrighted work.

I would say that the kind of use they engage in is commercial. Their source material here is a commercial fiction property, which doesn't help them as much as it would in the case of say something non-commercial and nonfiction. I'll assume they used all his written work to train the algorithms but little shows in the output product, this does help the fair use case. But doing well on one pillar of the fair use test doesn't mean the work qualifies. Lastly IMO again an algorithm can produce fantasy books at a rate which GRRM can not compete and might impact the larger market for fantasy literature.

https://youtu.be/1Jwo5qc78QU?si=OKqxTjJ7gavCs2N4

[–] [email protected] 0 points 1 year ago* (last edited 1 year ago)

I'd argue that the scenario where someone is paying someone to intentionally replicate others work

I'm going to stop you right there. Your argument seems to rest on the idea that knowledge and discussion of a book constitutes "replication". I don't concede this. I don't think we even get to the question of "fair use" because I don't see how knowledge and discussion can be considered copying that book.

If we were taking direct quotes from the book, we could get into copyright and fair use issues, sure. But I have never seen these AIs produce a direct quote from any source.

I don't think what the AI is doing with its knowledge of the book is sufficiently close enough to the book to be considered a derivation or replication. Can you show me otherwise?

Lastly IMO again an algorithm can produce fantasy books at a rate which GRRM can not compete and might impact the larger market for fantasy literature.

Having the capacity to produce a derivative work is not the same thing as actually producing a derivative work. Copyright may prohibit the derivation itself, but it certainly does not prohibit the capability of creating such a derivation.

[–] [email protected] 29 points 1 year ago (1 children)

This probably won't be successful. Enough time has passed that GRRM's books moved into the public domain while we were waiting for Winds of Winter

[–] [email protected] 4 points 1 year ago (1 children)

Maybe AI can finally finish the series.

[–] [email protected] 10 points 1 year ago

It would probably come up with something nonsensical and stupid like Dani goes nuts and destroys King's Landing and Tyrion talks about how Bran has the most interesting story of all and should be king

[–] [email protected] 11 points 1 year ago (1 children)

Are we just going to keep posting this changing which famous author is listed?

[–] [email protected] 7 points 1 year ago

Probably, but it needs to happen. Even this stupid comment belongs to me. If you respond, that belongs to you.

I'm classically trained in machine learning. There's not much different from what we did when I first started my career and now except for the quantity of data and the computing power that gpus have gotten us.

Hal 9000 from 1968 worked on a neural network. I've had to explain that a number of times.

[–] [email protected] 5 points 1 year ago (3 children)

This is going to keep happening until one side wins (I’m with the Authors 100% here)

AI is a complete legal gray area, and will continue to be until precedent is set. Nobody has found what is and isn’t allowed under copyright law yet, and unfortunately I’m pretty sure the judges ruling on this aren’t going to be completely informed

[–] [email protected] 7 points 1 year ago

It's actually not that hard to figure out. It's not the same thing, but here's an article by Kit Walsh, who’s a senior staff attorney at the EFF explains how image generators work within the law. The two aren't exactly the same, but you can see how the concepts would apply. The EFF is a digital rights group who most recently won a historic case: border guards now need a warrant to search your phone.

[–] [email protected] 3 points 1 year ago

I am with the authors more like 80% here.

Authors read each others works and are influenced by them and we don't expect them to go back and buy special licenses for each work that might have influenced their current novels. Art as much as any field stands on the shoulders of mice and giants alike. Pretending that AI language models shouldn't "read" as many novels as possible to assist their own growth is a preposterous idea.

Should they have to buy a copy of the book like everyone else? Sure. Should they get bent over without lube by publishing companies? Well that is a bit more complex.

In my opinion there is no "right" answer right now. We as a society need to decide what we are okay with.

Furthermore, there are a lot of really good books out there that would be truly great "except". Except what? That depends. Maybe it has annoying side characters, or maybe it is littered with plot holes, maybe there are outdated social norms that distract from the real point of the book, or maybe the fact that not one character in the book looks or talks like you and your friends.

It would be wonderful if we could use AI to adjust or even personalize those books.

Can you imagine a Harry Potter that isn't just translated into other languages, but has each of the characters localized as well. Neither Harry Potter being British nor being male is fundamental to the story. There is no reason the French, Aria Potter couldn't save the world through the power of her mother's love, and with the help of her friends. Well except the fact that it would likely make JKR lose her mind, since she doesn't even tolerate fan fiction.

Is it possible to make these changes now, sure? It just isn't really practical for even really big name authors, much someone who only sells a few thousand copies of each book.

[–] [email protected] 2 points 1 year ago

I understand that authors and artists in a more general sense are very, very concerned. The problem is that this has nothing to do with AI. If I were to generate a story based on their work, I simply cannot distribute it without legal repurcusions. Doesn’t matter if I used AI or not. The problem lies, once again, at the publishers. They can churn out copy after copy using AI and abuse the artist in that regard. Something similar is happening with DC Fables and it’s creator(who in a gigachad move just threw the entire IP into the public domain).So what we need is copyright reform. Artists deserve to be paid adequately for their work and should be protected from being ‘impersonated’ by publishers using whatever means, not strictly AI. All these ban AI discussions miss the underlying point completely, being copyright reform. AI just sped up the proces 100 times.

[–] [email protected] 5 points 1 year ago

Well if Georgy-boy would finish a book, maybe his job wouldn't be on the line to a robot?

[–] [email protected] 1 points 1 year ago

*Former author, George R.R Martin.

That mofo just been watching TV and playing videogames the past decade+.