The Ghibli thing is a great example; who's still actually doing that? It was a passing fad.

But let's not pretend that that passing fad has changed the fact that Ghibli's films are absolute masterpieces that will continue being enjoyed by generations to come. Because they are and still are.

We throw plenty of smart people at plenty of hard ideas. If a company really wanted to, they would find a way to make this feasible.

If it all was non-profit - then no one would raise the ethical issue.

No one is stealing anything. It's not theft. There has been no crime. None of this is anywhere near criminal law.

I could make a more nuanced argument on copyright infringement. But to make that steelman, I'd need to accept a too large overton window shift, so I'll decline to do so here.

Music conglomerates have money and their lawsuits will probably settle the issue.(unless they settle) That will be applied for all copyrighted works, regardless of the medium.

I believe going against the big guys is the reason why the big ones don't yet have music generation LLMs.

Just because you obfuscate what's happening by calling it "learning" and pretending your model is actually just looking at pictures the same as a human, doesn't make it true.

I'd prefer looking at what (potential) consumers actually do rather than what they say. "Saying" is a really weak signal.

Where did you get that idea. Global economy is ~200T/year PPP. 0.1% of that split across every artist you want the training data from would be insanely difficult for the vast majority of them to turn down. Which makes sense as art isn’t that big a percentage of the global economy compared to say housing, food, medical care, infrastructure, military spending etc.

Obviously the incentive to take without compensation is far more appealing, but that doesn’t mean it was impossible to make a reasonable offer.

That's kind of an interesting concept: "since the scale of my transgression was so big, I should get away with it scot-free."

Didn't that exact social experiment took place in the US last year? I thought the result of that was disastrous if media reports are to be believed.

OTOH I remember creator of Wordle closed the "low few mil" deal instantly, so I do believe it unlikely that people turn down few _hundred_ months worth of salary. But those artists are not from regions with 50-100x less median income and/or wider income distribution relative to US - I think they're concentrated in relatively high-income-low-disparity regions - so I don't think there's backwater wherever that lifetime income there is equivalent to no more than 6 months worth in US that has abundant supply of artists.

And IMO those artists are basically engaged in a geo-scale dumping of media contents. It's the same phenomenon as how moving consumer electronics manufacturing to US instantly multiply costs by small integers instead of just incurring premiums in percentages. If that phenomenon were to be quenched and those effects were integrated into economy anyhow, that will change the global balances of power to some statistically significant degrees, like, we'd be seeing flying rocket amphibian McBoatfaces everywhere. That might be interesting, but I'm not sure if that's an interesting kind of an interesting thing to see.

The starving artists I know would be extremely happy to get even one cookie to lick. I know an artist prodogy that paints on canvas and has work in a sizable gallery, at least one institution patron, and is constantly hosting paid workshop events. He architected and built his own custom 40ft ceiling pine art house covered in beautful stained wood and arches, with large metal cuttings and engravings of wild horses on the railings. This artistic prodigy is still starving and works as a handyman/construction worker part-time. He is strongly opposed to AI, by the way.

Most artists are "starving artists"; there are extremely few artists that can support themselves by their creations alone. Many artists make no money at all, and many artists seem to work or create alone as individuals, meaning that they almost always lack the funds or community resources to protect their creative work.

No, far better that we have four rent-seekers who gobble up everything that anyone is naive enough to share with the world, then turn around to demand profit in order to keep up with the new pace of the world that they’ve created.

There is a reason why we call it styles, because it’s a recognizable pattern someone came up with maybe after decades of work.

So if scan a book you are making a copy. In some copyright jurisdictions this is allowed for individuals under a private copying exception - a copyright opt out, if you like - but the important thing is private use. In some jurisdictions there is also a fair use exception, which allows you to exploit the rights protected by copyright under certain circumstances, but fair use is quite specific and one big issue with fair use is that the rights you are exploiting cannot result in something that competes with the original work.

Other acts restricted by copyright include distribution, adaptation, performance, communication and rental.

So if you copy a book, digitize it, and write a program to analyze the word frequencies it contains you may, in some jurisdictions but not all, be allowed to do this.

If you’re doing it locally on your own machine you are simply copying it. If you do it in the cloud you are copying it and communicating the copy. If you copy it, analyze it and train an AI model on it that could be considered fair use in certain jurisdictions. Whether the outputs are adaptations of the training data is a matter of debate in the copyright community.

But importantly if you commercialise that model and the resulting outputs compete with the copyright protected material you used to train, your fair use argument may fail.

So when you buy a book you are actually party to what is effectively a licence granted by the copyright holder, albeit it to the publisher. But as the end user of the book you are still restricted in what you can do with that copyright protected work, through a universal end user licence encoded in law.

The four factors of fair use in the US:

> the purpose and character of your use

Commercial, for-profit. Not scholarship, not research, not commentary, not parody, etc.

> the nature of the copyrighted work

Absolutely everything. Artistic, creative, not purely factual.

> the amount and substantiality of the portion taken, and

All of it, from everyone.

> the effect of the use upon the potential market.

Directly competing with those whose data was copied.

It shouldn't be!

The model works by training on what features humans can make sense out of the image they're presented with, if the image and the observations of the image's feature were clear/observable enough. Then the generation makes use of those observations. I'm just using 10% as an arbitrary number to describe proportions. If the generation were 100% of the observations from the same image, the model would be overfitting, and many would have deemed it to have produced a copy.

> Just because a race car uses kinetic energy, gravity, and friction to propel itself, the same way a human does, doesn't mean it's doing the same thing as a human.

WTF does this even mean? A race car uses concepts from Newton, just as how a human uses gravity to train it's muscles to move be it knowingly or unknowingly. But you don't see them (car makers/humans) paying rent to Newton after he discovered gravity. Come on!

If I buy a book and use it to prop up a table, the author likewise does not own the table, or any works I undertake on that table.

If I buy a book and rip out the pages to make a collage, the US is the only legal jurisdiction where I run even slight risk of civil penalties.

An LLM is downstream of a book. Using a book to make an LLM does not confer any rights or privilges towards the LLM on the original author, just as using a hammer or nails dont permit the hammer or nail manufacturers any royalties on what I make, even if I build a hammer making machine with them. Theres no right to the works of people who build on your work without reproducing your work, at least outside of strict copyleft.

Its like demanding a cut from people who learned how to use photoshop by watching your photoshop tutorial youtube videos.

This is why the most successful cases against LLMs have been on the "Did they purchase the book" side of the fence, and not on the "What did they do with it" outside of the one case, where the legal company tried to use the LLM to 1:1 reproduce the content they had a limited license to, but thats obviously a no go and they should have known better.

> Is it transformative if I take all the pages in Hanya Yanagiharas A Little Life and use a thesaurus to change every second word?

If you meant it literally.. I'd think that such a version would be a sort of parody. It'd be up to lawyers doing their cross-examinations to prove the work was intended for such a purpose though..

> Or a more realistic scenario: what if I translate it to Spanish without license from the author? That's not allowed, and yet I have "transformed" the work in the same way that an LLM does.

Probably a lawyer would answer this better than me, but the 'content' is the same and would violate copyright. There's also other factors, like if it was translated/distributed for free.

Besides that I regard that LLMs to hold mathematical observations in contrast to a translated work. So long as the user ensures the output isn't close to what's already available imo it fits the transformative criteria.