AI is stealing writers’ words and jobs…

[Jfdlsjfd]

At this point, I would say that the general consensus is that the answer is no, you cannot copyright things created without human authorship. Of course, there are three important things to remember-

1. There is no bright-line test about how much human involvement is needed. A human artist can use "AI tools" and create an image and copyright it. So the only real dividing line right now is that you can't just have the AI be the author of the work.

Which is the general interpretation on copyright generally, not only in the US. It is also quite logical. Imagine you get an AI generator to create a specific image that is protected by copyright, like a very good depiction of Elsa in Frozen. You could say "Hey, I commited no copyright infringement since the AI created the artwork, not me." If you use AI to create something, you're a human, and you exert control over what you must be the author.

2. This is (to date) a statutory issue based upon how the law is written; it's always possible to change the law. (Maybe- it's also possible for courts to reinterpret the Constitution's grant within Art. I).

We're in the infancy of a life-changing technology (not hyperbole, I think it's on par with the Internet) and it's very much possible that laws will be adapted over time to reflect the interest of the citizens over time, when there will be enough examples to assess the overall effects.

 

[Sacrosanct]

I don't know. It's the US legal system. If they convince the judge that it is fair use, then apparently it will be. I am not privy into the legal argument they will develop to show that use by Books3 is fair use. But it's not the same as saying "we'll pirate it anyway", but I am certain they can pay lawyers to write a convincing argument to defend their idea. Any idea when the outcome is expected?

It seems clear to me. Books3 used all of those images without consent or compensation. They flat out state "unauthorized". We have a definition and a precedence of what Fair Use actually is, and it's not what Meta is claiming it to be. If someone violates someone's copyright, you don't get to also use it by claiming the first person falls under Fair Use when it doesn't. It's still under copyright, regardless if someone else gave it to you if they don't have the rights to do so.

 

[Jfdlsjfd]

I don't click through videos, but it appears that this was more of a "thought experiment" than a real legal attempt.

Here is a quick (contemporaneous) legal analysis.

Given what we know of machine-generated works since that point, I would say, quite strongly at this point, that this didn't do anything.

You might be interested in an AI tool that reads video for you an print a quick summary of it .

https://www.summarize.tech/

I own no part of it and just submit it for the QoL of everyone.

 

[Sacrosanct]

This is an example of how AI should be used. It's Priyanjali Gupta, who used AI to recognize sign language (and I'm sure inspired that episode of Echo).

 

[Jfdlsjfd]

It seems clear to me. Books3 used all of those images without consent or compensation. They flat out state "unauthorized". We have a definition and a precedence of what Fair Use actually is, and it's not what Meta is claiming it to be. If someone violates someone's copyright, you don't get to also use it by claiming the first person falls under Fair Use when it doesn't. It's still under copyright, regardless if someone else gave it to you if they don't have the rights to do so.

They might develop the idea that they didn't copy any part of the Books3 database, just ran an AI training tool on it that didn't rely on copying any part of it, for example, therefore they don't have to compensate since they didn't copy or redistribute the litigious database during training. They woud then claim that the result is a derivative work, containing no part of the original (like a commentary) and derivative work enjoy autonomous copyright under US law (not that I am sure it's necessary, since they don't distribute their AI model either). I really don't know what their legal argument will be, but it's just an example. I am pretty sure Snarf could come up with something much better if given the legal budget of Meta.

 

[Sacrosanct]

I really don't know what their legal argument will be,

They state it up front in that quote. That any material used by Books3 should be considered Fair Use. What they don't say is why it should be Fair use, because it doesn't seem to qualify. Books3 just took all these copyrighted images without consent.

 

[Jfdlsjfd]

Yes, but also ... no.

Okay, one way to think about this is to view this in terms of extracting economic damage on the "AI Firms" (the companies making AI products). Viewed that way, a person might think that a class action is a great idea, because you can get a bunch of plaintiffs and leverage that.

But there's a flipside to this as well. Imagine you are the AI Firms. Negotiating all of those individual licensing agreements? That would be untenable. Instead, you actually want a class action! Why? Because it gives you a single entity to negotiate with. Sure, you have to do the payout upfront, and that sucks, but then ... that's it. And the class action will cover (and have a preclusive effect upon) EVERYONE ELSE.

That's it. You're done.

(FWIW, this is a similar tactic that was used by, inter alia, Spotify, so it's not unheard of.)

In other societies, the result would be a tax on GPUs used to inference artwork as a compensation. Not sure it would satisfy the "ethical" argument, even if it would make it legal.

 

[Jfdlsjfd]

They state it up front in that quote.

I am guessing that their legal argument will be a little longer than that quote, especially since it's a quote from a magazine and not their legal department.

That any material used by Books3 should be considered Fair Use. What they don't say is why it should be Fair use, because it doesn't seem to qualify.

That's not what's in the quote. It says "Meta has acknowledged using parts of the Books3 dataset but argued that its use of copyrighted works to train LLMs did not require "consent, credit, or compensation." ". The only quoted part being "consent, credit and compensation". And then that the plaintiff had an "alleged" copyright. They never mentionned Fair Use and the part where the article says "contending that any unauthorized copies of copyrighted works in Books3 should be considered fair use." isn't a quote, it's the original work of the author of the article. It could be an oversimplification of their legal claim.

Books3 just took all these copyrighted images without consent.

Yup. But if we rely on the line where it says that Meta "argued that its use of copyrighted works to train LLMs did not require "consent, credit, or compensation."", it is coherent with the argument that copyright grants exclusive rights to the copyright holder, none of them preventing AI-training in the specific way they used it. It might be a stretch, it might be convincing, I don't know what the final legal outcome will be, but I am pretty sure it will a more elaborated argument that "a third party infringed copyright, therefore it's fair use to copy the work forever now", which is quite unconvincing but might not reflect their analysis.

 

[Snarf Zagyg]

They state it up front in that quote. That any material used by Books3 should be considered Fair Use. What they don't say is why it should be Fair use, because it doesn't seem to qualify. Books3 just took all these copyrighted images without consent.

Well, the thing about Fair Use is ... anyone can tell you the four factors, but telling you how they would apply in a novel situation is notoriously tricky. Because it's not like there's any way ex ante to know how a court is going to apply the factors, and none of them is dispositive. In addition, many courts will use additional other factors (either from case law or on an ad hoc basis) that can play a role.

Anyway-
1. Purpose and character of the use, including whether the use is of a commercial nature or is for nonprofit educational purposes (this is also where courts will add in a "transformative" component, especially if it's commercial work).

My take- this cuts heavily against AI Firms, given that it's a commercial use. Mitigated by the argument that it's transformative (they are not simply cutting and pasting, but using it to train).


2. Nature of the copyrighted work.

This also cuts against AI Firms, given that the corpus of Books contains many creatively expressive works.


3. Amount and substantiality of the portion used in relation to the copyrighted work as a whole.

This is arguably the worst factor for AI Firms, because, they are using the "whole" work. That said, there are two things to consider; while difficult, it is not impossible to find fair use of an entire work. In addition, I think that AI Firms would argue that they aren't "using" the whole work, given that the outputs are constrained to not regurgitate the whole thing.


4. Effect of the use upon the potential market for or value of the copyrighted work.

Here, we are looking at the damage to the market for the original work done by the alleged unlicensed product. I think that the AI Firms have a decent case in terms Books3 on this factor ... I am far less certain about how this would work for some of the image generating AIs, which I believe can mimic the styles and outputs of certain artists, and arguable would affect the existing or future market for the copyrighted works. Honestly, you could probably make the same argument for Books3 at a certain point, to the extent that the corpus contains enough of a writer's output to mimic the writer and damage the market for their books. So ...


But yes, I do have some difficulty seeing how they could prevail under a "fair use" argument, given that I've seen (IMO) much better fair use claims shot down by the courts before. But you never know!

 

[trappedslider]

This is an example of how AI should be used. It's Priyanjali Gupta, who used AI to recognize sign language (and I'm sure inspired that episode of Echo).

I should note that this isn't GENERATIVE AI but instead is facial (hand?) recognition