AI :: Lower courts have misconstrued fair use for too long

[rich in distress]

https://www.digitalmusicnews.com/2023/05...t-opinion/

Hope it gets traction.

[Will Collier]

Give the headline writer a raise:

[jonny]

Nice try, but fair use was a bit of a stretch. Sure, one has been colored in, but they are both pieces of fine art. Hardly transformative.

[chopper]

RIAA. I think we can all agree they are the guys in white hats doing the Lord's work.

Not.

[$tevie]

rich in distress wrote:
https://www.digitalmusicnews.com/2023/05...t-opinion/

Hope it gets traction.

It's incredibly specific to this particular matter, which in essence is that Vanity Fair used the Warhol print instead of the original image for an article and should have paid the photographer who shot the original. It doesn't apply to works sold in galleries or auctions, or such. If you read the court document, it kind of lays the ground for it to NOT be a precedent.

[$tevie]

Will Collier wrote:
Give the headline writer a raise:

I saw this on Twitter! So good. Can't believe it was the Wall Street Journal of all publications.

[rich in distress]

I think the gist of the issue is that it was instructed to consider Andy’s work derivative, and as such, the original CR holder had the prerogative of usage of Andy’s, granting it for a single event only
Then again, I didn’t read the court document, so your point may be stronger.

Anyway, I find amusing the notion provided by the music exec, that training AI with copyrighted material renders its output a derivative work.
Good times…

[Tiangou]

rich in distress wrote:
Anyway, I find amusing the notion provided by the music exec, that training AI with copyrighted material renders its output a derivative work.

An algorithmic transformation of a digital image or digital music is a derivative work.

There is literally nothing "original" about AI output in the legal sense.

In this use-case, AI is a utilitarian device that performs sequential operations upon pre-existing materials to make mash-ups of them.

US copyright does not protect “any idea, procedure, process, system, method of operation, concept, principle, or discovery.” (It also does not protect works prepared by an officer or employee of the US Government as part of that person's official duties.)

And AI output can't be independently copyrighted because there the "author" is neither independently adding something creative/new, nor is it a human. Both are required for Copyright protection.

However, when a human takes AI output and further transforms it, that work may be copyrightable. Details here:
https://www.ropesgray.com/en/newsroom/al...l-guidance

[rich in distress]

Even then, training AI with copyrighted material would deter the fair-use defense… that’s what I find amusing. And yet, it’s debatable.

https://www.copyright.gov/fair-use/summa...ir1992.pdf

[Tiangou]

rich in distress wrote:
Even then, training AI with copyrighted material would deter the fair-use defense… that’s what I find amusing. And yet, it’s debatable.

https://www.copyright.gov/fair-use/summa...ir1992.pdf

Not sure why you use that case as an example. There's a back and forth about whether APIs, and portions of embedded firmware are so purely utilitarian as to be uncopyrightable. One of the compromises in IP law has been to call decompiling of functional code to provide access to computing hardware a fair use. (Tho recent cases have eaten away at that exercise of fair use.)

This doesn't seem to provide legal precedent for training AI.

Training AI to create images and music is NOT equivalent to copying source code without which a computer won't function.

The argument for opening source code is that it's utilitarian. But AI is trained on art, which is not utilitarian.