Fair Use Benefits Everyone

This reminds me of a billboard I saw about 15 years ago when the state of Delaware was considering changing its laws to permit new casinos in addition to the three that had been grandfathered in when the restrictions went into place.

Delaware Park, one of those three casinos, put up a billboard opposing the bill, and it said, "No more casinos! THREE IS ENOUGH!"

It just goes to show that corporations will take any position -- even a casino taking an anti-casino position -- as long as it's self-serving.

From that article:

> The New York Times’ lawsuit, meanwhile, asks everyone to ignore a half century of legal history and progress, and toss aside the longstanding principle of fair use in their own narrow pursuit of an undeserved payday.

The New York Times lawsuit isn't purely about training on content - it's about memorization. They included examples of OpenAI's models being able to reproduce NYT articles entirely from the weights.

There's also stuff about RAG in there - about OpenAI (and Microsoft) features that run web searches and then summarize the results. I've not been following developments closely, I'd be interested to know how that part of the lawsuit is shaking out so far.

I agree on this point!

And also, training an open weights AI on the outputs of an AI trained using "fair use", is itself "fair use" that benefits everyone.

OK. Obviously the use-case has changed.

Consider a random example of fair use: print an excerpt of this three panel comic you would like to discuss.

Does building a business around the automation of fair use constitute fair use? I would say no.

Translation from corporate speak: when we steal from you that's good; when you steal from us that's bad

The irony of Sam Altman's company pretending to care about what's "fair" is so very rich.

He literally made billions exploiting the intellectual work of millions of his fellow humans, yet has the audacity to whine about how the system isn't fair.

Generational level of hypocrisy from OpenAI. What a nerve putting "authors" in the list first. Fair Use benefits OpenAI and the labs much more than anyone else.

Just be honest and don't try to blend it into "Everyone".

Fruitfulness of discussion might be aided by bearing in mind that US and European copyright law traditionally come from very different perspectives. Basically, societal utilitarianism vs work-is-an-inalienable-part-of-the-artist author's natural rights. "Copyright should last 5 years because starving keeps artists productive. And copyright owners have no say at all in their works being used for LLM training because it's good for society" is consistent with US foundations, at least pre-Berne. "Authors are not permitted to give up their right to personally refuse a use of their already-sold work, nor to give anticipatory permission for use in future technology" is IIUC unremarkable in a European context. An AI suggests[1] as philosophical background. People are coming at this from vastly different places. Bearing that in mind might help avoid talking past one another. IIUC.

I also note, when the internet was just catching on, and email was becoming less niche, but copyright law had not yet established sending email as an implicit grant of permission to make copies, most all use of email violated copyright. Network->disk->memory - all copies without permission. I'd get email with footers vaguely like "You are hereby granted permission to <do email> for the purposes of XYZ Corp - all other use constitutes infringement". There was a multi-year lag between accepted practice and legality. IIUC, forwarding email is still problematic today. Having lived through that, "Arg! <New tech> violates copyright!" gets less traction with me. And with copyright term extensions, and Berne on the US side, having a taint of unprincipled corruption, that also makes it less compelling.

[1] https://cyber.harvard.edu/people/tfisher/iptheory.pdf

Surely OpenAI would thus not be opposed to others distilling their models.

Is distillation fair use then?

But let me guess, the trained model is a novel work and is protected, right?

> Authors, journalists, creators, academics, accessibility advocates, legal experts, librarians, economists and business leaders are voicing their support for the longstanding principle of fair use and its application to AI training.

Have you heard of the landmark $1.5 billion class-action copyright settlement in Bartz v. Anthropic?

Absolute cinema!

ironic that they can talk out of both sides of their mouth, fair use is important for their training but if you distill it's definitely not fair use anymore

Are they just trying to rage bait everyone?

They must understand that this can literally only make people hate them more.

There is no open-ended "fair use" doctrine in the EU. Not everyone after all.

Bullshit.

Copyright originally maxed out at 28 years. Corporations rewrote the laws over and over, so that now copyright can last over 100 years, turning intellectual property into yet another rent extraction scheme.

Now, in this case, OpenAI argues that “fair use” is for the good of all. OpenAI doesn’t give a shit about what’s good for all. They want the ability to define intellectual property rules so that they can extract rent from someone else’s property.

If they want to benefit everyone, why aren’t their models open? For that matter, why is their name OpenAI, when they are not open? They’re publishing tons of work in mathematics, supposedly to benefit everyone. Instead, why don’t they forgo all research themselves and dedicate their work to supporting existing mathematicians in using AI effectively, so that the people that built the base of knowledge get credit for the new work?

Because OpenAI wants to extract the maximum shareholder value out of other people’s work.

Does anyone actually believe PR like this anymore?

I'm not sure what I find more morally repugnant to be honest. The largest wholesale theft of human labour ever seen, or the company that has made billions from this heinous act disingenuously and cynically trying to claim it as in any way "fair" (whilst simultaneously crying foul about distillation)

This is a wild re-imagining of what "Fair use" means in the US courts, which have a relatively straightforward four point test, and of which OpenAI, a commercial venture that produces exact copies of an article and thus acts as a market-replacement for the original work, fails handily.

What OpenAI wants is "free use".

LLMs trained on fair use are themselves fair use. Weights pls

What AI is doing is socializing the costs but privatizing the profits. This has been the model for every social media platform going back to forums. Everything after USENET, really.

If you're using the collective output of humanity to train your LLM then humanity should own the output. Problem solved.

One can easily understand this as the worker's relationship to the means of production.

> The New York Times’ lawsuit, meanwhile, asks everyone to ignore a half century of legal history and progress, and toss aside the longstanding principle of fair use in their own narrow pursuit of an undeserved payday.

There's absolutely no "fair use" precedent that says that you can program a computer with copyrighted material. They're just asking for an exception essentially based on their compression being lossy, which of course the law takes the absolute opposite opinion on e.g. distributing a 1080p of my bluray disc has been determined to definitely not be fair use.

Their essential case is that they use the word "training," and people training themselves from books is fair use, so since they use the word "training" then what they're doing is fair use. It's stupid, and judges going for it are either stupid or corrupt. I'm not actually copying your book, I'm training my printing press.

Silly billionaire, fair use is for people, not megacorps!

But seriously, fair use is for students, teachers, comedians, and all the people who can't pay or couldn't possibly pay enough for permission from copyright holders. A corporation training a closed model for profit can, and should, pay. If American copyright law is too onerous for megacorps to contend with, maybe it should be reformed.

Sorry, but no. If that was a statement was coming from an actual Open AI I would welcome it, but instead it comes from a pre-IPO "Open"AI, after it siphoned as much information as they possibly could.

Whatever fair use is supposed to have been, generating never-before-seen amounts of revenue by hoovering up every word ever written is not it.