True, sadly most of those copyright are probably owned by other megacorp. So they either collude to surppess the entire industry or eat each other alive in legal clashes. The latter is happening as we speak (the writers for NYT are probably long retired, but NYT still owns the words) so I guess we'll see how that goes.
>OpenAI found a way to benefit by learning from these images. The same way that every artist on the planet benefits by learning from the images of their fellow artists.
If we treat AI like humans, art historically has an equally thin line between inspiration and plagiarism. There are simply more objective metrics to measure now because we can indeed go inside an AI's proverbial brain. So the metaphor is pretty apt, except with more scrutiny able to be applied.
When discussing our work, we can name them.
When one of our influences comes out with a new body of work, we can gush about it to our own fans.
When we find ourselves in a position of authority, we can offer work to our influences. No animation studio is really complete without someone old enough to be a grandfather hanging out helping to teach the new kids the ropes in between doing an amazing job on their own scenes, and maybe putting together a few pitches, for instance.
We can draw fan art and send it to them.
None of these are mandatory, but artists tend to do this, because we are humans, and we recognize that we exist in a community of other artists, and these all just feel like normal human things to do for your community.
And if an artist suddenly starts wholesale swiping another artist's style without crediting them, their peers get angry. [1]
1: https://en.wikipedia.org/wiki/Keith_Giffen#Controversy
OpenAI isn't gonna tell you that it was going for a Cat & Girl kind of feel in this drawing. OpenAI isn't gonna offer Dorothy Gambrell a job. OpenAI isn't going to tell you that she just came out with a new collection and she's still at the top of her game, and that you should buy it. OpenAI's not going to send her a painting of Cat & Girl that it did for fun. OpenAI isn't going to do anything for her unless the courts force it to, because OpenAI is a corporation who has found a way to make money by strip-mining the stuff people post publicly on the Internet because they want other humans to be able to see it.
How many of those words do you remember learning? How many can you confidently say you remember the person or the book that taught you the word? 5? 10? Maybe 100?
That's how brains work. We ingest vast amounts of information that other people put out into the world. We consume and it incorporate it and start using it on our own work. And we forget where we even got it. My brain works this way. Your brain works this way. Artists' brains work this way. GPT-4 works this way.
The idea that a visual artist can somehow recall where they first saw many of the billions of images stored in their brain -- the photos, movies, architecture, paintings, and real-life scenes that play out every second of every day -- is laughable. Almost all of that goes uncredited, and always will.
This is what it is to learn.
There is a strong correlation between how copyrightable a concept is and how well you can point to where you learned it.
To take it to an extreme, you obviously can't just use one of the available neural net lossless compression algorithms to circumvent copyright law or citation rules (e.g., distributing a local LLM that helpfully displays the entirety of some particular book when you ask it to), you can't just tweak it to make it a little lossy by changing one letter, or a little more lossy than that, etc., while on the other hand, any LLM that performs exactly the same as a markov model would presumably be fine, so there is a line somewhere.
A company builds an AI tool. That AI tool is trained on a ton of artists' work over the years. The company opens up the AI tool and asks it to draw, "X but in the style of Y," where Y is come copyrighted artwork. The company then prints the result and puts it on their packaging.
What's the difference?
I'd argue there isn't one. The copyright infringement isn't the ability of the artist or the AI tool to make a copy. It's the act of actually using it to make a copy, and then putting that out into the world.
Again, take it back to basics: do you believe it is permissible to share a model itself (not the model output, the model), either directly or via API, that can trivially reproduce entire copyrighted works?
So yeah, of course imo it should be permissible to share a model that can reproduce copyrighted works. Being "capable of being used" to violate a law is not the same thing as violating a law.
A ton of software on my computer can copy-paste others' work, both images and words. It can trivially break copyright. Hell, there are even programs out there than can auto-generate code for me, code that various companies have patent claims for. Do I think distributing any of this software should be illegal? No. But I think using that software to infringe on someone's copyright should be.
(Note: This is different than if the program distributed came with a folder that included bunch of copyrighted works. To me, sharing something like that would be a copyright violation.)
> It is clearly not the same as your mind, because your mind can and, assuming you want to stay in good standing, will provide credit for influence
I forgot to respond to this, but it's not true. Your mind is incapable of providing credit for 99.9% of its influence and inspiration, even when you want it to. You simply don't remember where you've learned most of the things you've learned. And when you have a seemingly novel idea, you can't always be aware of every single influential example of another person's work/art that combined to generate that new idea.
Compression and the output from LLMs are cousins. The model tries to predict what continuations are likely, given context. Indeed, it takes a lot of effort to make LLMs less willing to just output training data verbatim. And conversely, you can get compression algorithms to do things similar to what LLMs do (poorly).
Whether this also describes most of human cognitive process, is subject to debate.
Ultimately, only high courts in each jurisdiction can decide. I can imagine a case where some highly advanced nations decide different interpretations that cause conflict. Then, we need an amendment to the widely accepted international copyright rules, the Berne Convention. Ref: https://en.wikipedia.org/wiki/Berne_Convention
Nobody working on a new cancer drug actually did any work for me. They did labour for other reasons, and were happy with it.
There it is okay for me to steal their recipe and sell their cancer drug.
They were happy until their copyright got stolen, I guess. Then got unhappy.
LLMs are collections of GPUs crunching numbers. "Inspiration" doesn't really apply to them.
A better analogy is sampling, and musicians remixing music are very much required to pay for the samples they use.
If I sample a track and play it in my home I don't properly compensate anyone.
If I ask GPT to create a cool new comic based on the article and then delete or use it privately it, same applies.
To me this is a strong point in favor of the idea that OpenAI has no business using their work. How can you even think it's ok for OpenAI to use work that was not done for them without paying some kind of license? They aren't entitled to the free labor of everyone on the internet!
At the risk of answering a rhetorical question: because copyright covers four rights: copying, distribution, creation of derivative works, and public performance, and LLM training doesn't fit cleanly into any of these, which is why many think copying-for-the-purpose-of-training might be fair use (courts have yet to rule here).
I think the most sane outcome would be to find that:
- Training is fair use
- Direct, automated output of AI models cannot be copyrighted (I think this has already been ruled on[0] in the US).
- Use of an genAI to create works that would otherwise be considered a "derivative work" under copyright law can still be challenged under copyright.
The end result here would be that AI can continue to be a useful tool, but artists still have legal teeth to come after folks using the tool to create infringing works.
Of course, determining whether a work is similar enough to be considered infringing remains a horribly difficult challenge, but that's nothing new[1], and will continue to hinge on how courts assess the four factors that govern fair use[2].
[0]: https://www.reuters.com/legal/ai-generated-art-cannot-receiv...
[1]: https://www.npr.org/2023/05/18/1176881182/supreme-court-side...
[2]: https://fairuse.stanford.edu/overview/fair-use/four-factors/
> But if we are treating proper compensation as stifling technology, I'm not surprised people are against it.
That's just it, nobody looking to get paid by OpenAI actually did any labor for OpenAI. They did labor for other reasons, and were happy with it.
OpenAI found a way to benefit by learning from these images. The same way that every artist on the planet benefits by learning from the images of their fellow artists. OpenAI just uses technology to do it much more efficiently.
This has never been considered labor in the past. We've never asked artists to "properly compensate" each other for learning/inspiration in the past. I don't know why it should be considered labor or proper compensation now.
But we shall see what the courts decide!