What the AI copyright fights are truly about: Human labor versus endless machines
- Reference: 1704312129
- News link: https://www.theregister.co.uk/2024/01/03/kettle_ai_copyright/
- Source link:
The New York Times [1]opened fire on OpenAI and its champion Microsoft just last week. The newspaper was upset that "millions" of its articles were allegedly used without permission to build bots including ChatGPT that output "memorized" copies of those stories or ones that are "substantially similar."
Your humble hacks got together today to discuss it all, which you can replay below – or listen via your favorite podcast distributor: [2]RSS and MP3 , [3]Apple , [4]Amazon , [5]Spotify , and [6]Google .
On the surface, these cases are about alleged copyright infringement by Big Tech at a time when it's still up in the air as to how that law intersects generative models. Do the plaintiffs have a solid legal footing, or not; can they be fairly compensated if necessary, and how would that work; and does the law need to change, why should it change, and how?
As we wait for those questions to be answered, it appears to us that below the accusations of infringement, a moral argument has been brought before the courts against the displacement of people by hyperscale generators of material that were built on human work. What is the future of human labor if people can be mimicked by an endless army of machines with no sign of recompense? Will that work in court, and should it?
[7]
Settlements, payments, and agreements may be reached to straighten all this out before long.
[8]
Meanwhile, a Beijing court has [9]ruled AI-generated content can be copyright protected. America has [10]taken a similar stance: if a human was the driving force of the creative work, and they happened to use AI to produce it, it can be protected by copyright.
On today's 18-minute show we have, clockwise from top left in the thumbnail, vultures Chris Williams, Brandon Vigliarolo, Thomas Claburn, and host Iain Thomson.
[11]
This episode was produced and edited by Nicole Hemsoth Prickett. You can find previous Kettles [12]right here . ®
Get our [13]Tech Resources
[1] https://www.theregister.com/2023/12/27/the_new_york_times_files/
[2] https://feeds.simplecast.com/Vnvf8Fkd
[3] https://podcasts.apple.com/us/podcast/the-register-kettle/id1713589041
[4] https://music.amazon.com/podcasts/fe62a659-3451-496b-b064-f30744c2ad65/the-register-kettle
[5] https://open.spotify.com/show/3SkH00VOX02KYNaAbYzajj
[6] https://podcasts.google.com/feed/aHR0cHM6Ly9mZWVkcy5zaW1wbGVjYXN0LmNvbS9WbnZmOEZrZA?ep=14
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZZXnHCsy6rWQvqHIi9q4XQAAAZU&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZZXnHCsy6rWQvqHIi9q4XQAAAZU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://www.iam-media.com/article/beijing-internet-court-rules-ai-generated-content-eligible-copyright-protection
[10] https://www.theregister.com/2023/03/16/ai_art_copyright_usco/
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZZXnHCsy6rWQvqHIi9q4XQAAAZU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[12] https://www.theregister.com/Tag/kettle
[13] https://whitepapers.theregister.com/
They're not
These lawsuits are NOT about human / machine labor. At least not in the specifically mentioned cases. Most of the authors suing AI companies would be likely more than happy allowing their works to be used in AI models - as long as they get compensated for that, similarly to how they get royalties when their works are used or built upon in the traditional way. After all it's just fair, that if someone is making money by building on your works, you also get to make some money off that.
The real problem for artists and authors is not that AI could produce works just as good as they can (because AI can't, and with the help of AI they could produce even better works also themselves, at least for now), but that AI companies simply took everyone's copyrighted works without permission and built something on top of that, that now displaces the original works and authors. And not only that, but they will keep doing this also in the future, if they're not stopped and not forced to licence works and pay the authors.
It's like if some company employed millions of workers to build a huge building, and then won't pay said workforce, by arguing that the completed work does not resemble to any of the single bricks that any of the workers put into it. Obviously that's not how thing work, should work, or could work in the long term. Not in the logical and not in the economical sense.
The good news is that most likely AI companies will be forced to actually start licencing works from authors anyway, not only because of legal, but also just because of pure technical reasons. That's because that will be the only way they will be able to ensure in the future that they're actually using human generated high quality content instead of just feeding back AI generated low-quality content to their models which will lead to the degradation and implosion of those in the long term.
(Now obviously the fact that they will licence content won't technically limit anyone giving them AI generated content, but once the authors actually enter a contract and will be paid for their works, they will be also held liable for those - and that's why AI companies will pay them, even if pennies.)
"allowing their works to be used in AI models - as long as they get compensated for that"
That sounds like humans versus machines to me.
"AI companies simply took everyone's copyrighted works without permission and built something on top of that, that now displaces the original works and authors"
Again, labor versus machines IMHO.
I think you're over thinking this. Yes, the cases are about copyright allegations. The NYT is rather specific. But a lot of the cases have an undercurrent of something along the lines of: it's not fair that these models learned how to imitate us and are now pushing us out of the market.
This isn't just purely over copyright, but copyright is how the plaintiffs hope to solve it. That's at least my impression of it all.
C.
Re: "allowing their works to be used in AI models - as long as they get compensated for that"
"Again, labor versus machines. "
No, because it would be just as wrongs if it was humans who did the same thing that computers do. The point is that there's no real effort involved on part of the AI companies to create the results, or it's negligible compared to what went in creating the original works, whose authors do not get compensated for their works by any means, and it's the AI companies that reap all the revenue.
The situation is very similar to what's search engines and social media sites were and are still in, and which is also utterly unjust, because all they do is take everybody else's content, in most cases without even explicit permission to do so, slap ads next to them, and call this a product, not compensating the original authors by any means, despite those being the ones who supply the actual value in the service.
It's evident when you ask yourself: what would still have value without the other? Would search engines have still any value if they had no articles to show? Obviously not. Would articles still have value if there were no search engines? Definitely. Similarly: could AI companies provide a useful service to the general public if they couldn't take everyone else's content to train their models? Definitely not. Would everybody else's artwork, books, etc. still useful to people is AI companies didn't exist? Definitely.
Again, the problem with AI is not that AI "imitates" other authors. It wouldn't even be a problem if it would just plain out copy them. Most artists, authors don't mind if others are using their songs, photos, etc. in their derivative creative works, even if it's verbatim. Actually, most professional artists create their artworks with the explicit intention and assumption that somebody else or a lot of other parties will reproduce these works. But they only don't mind or even want that if this happens with their permission and when they get royalties for every reproduction or use. The real problem is that the AI companies 1. took all everyone's works without permission, 2. don't even want to pay royalties on the use of said works, not even after the fact, and not even pennies.
The "argument" the AI companies have that they're not using a single authors' works, and that the works their systems produce incorporate all influences from all authors in the world, and therefore they don't copy and don't even resemlbe in most cases the works of a single one of those authors or a specific work of them. But if anything, that should be an argument against the AI companies, because it means they violate the rights of not only a single author, but the rights of countless or all authors in the world, every single time their systems produce something.
Again, it's irrelevant whether the AI companies are producing what they do through algorithms or humans - and that's why it's not a humans vs robots issue. It would be just as wrong and unjust if they'd produce the same results using humans, by for ex. the humans clipping together pieces of text from others' works. The problem is not that, but the lack of permission and compensation.
a moral issue: the future of human labor versus an endless army of machines
I disagree that this is a man vs machine issue. When bookkeeping was done by rows of clerks at desks adding up figures, they could be and were replaced by computers that were far faster and more accurate at doing that work. It may be tempting to compare that example, by noting how much quicker a computer is at generating a page of text than a human journalist, but the journalist also had to find the story and research it, before forming what they found into a finished article. The AI simply copied that and regurgitated it. Without the published work of the journalist, you could run the AI for a million years and it would never come up with the story itself.
There's the underlying concern
Don't get me wrong, yes, there is the copyright issue that the courts have been asked to consider.
But I think it's more than just simple copyright infringement at play here. It's questioning the very nature and validity of LLMs: their ability to do a passable impression of a human, and how that displaces humans when used at scale, and how those people are compensated.
Some artists may not even want an AI to replicate them, money or not.
Which I think is at the heart of some of the complaints.
C.
nope
Reset copyright back to 14 years, renewable once, retroactively. Then we can talk.