Stop Rushing To Copyright As A Tool To ‘Solve’ The Problems Of AI
from the embrace-the-creativity dept
A few months back, I got added to a group chat of Hollywood writers/actors/directors, all seeking to understand what the fuck was going on with AI. And one topic that came up consistently early on was “will copyright protect us” and, if not, “how can copyright be changed to better protect us?” I’ve already made it clear that I’m skeptical of the various copyright lawsuits against AI companies, claiming that the training of their LLMs violated copyright law. While there are some arguments against it, it seems to me that training is the equivalent of learning from, and we’d never say that reading a book and learning from it violates copyright law. Similarly, various lawsuits about search engines and book scanning suggest a (correct) recognition that scanning copyright covered works to create new (even commercial) products is fair use.
However, I’m somewhat perplexed at the focus on copyright law as the tool that many people want to use to “fix” what they insist is a problem.
I fear it’s a symptom of Hollywood spending decades falsely convincing people that copyright was the only tool out there for “protecting artists.” Of course, this was always a myth. Copyright was created from the beginning as a tool to protect the middlemen and gatekeepers, not the artists and creators themselves. But, one of Hollywood’s greatest tricks has been convincing the creators, whom Hollywood itself is exploiting, that the copyright tool they’re using for said exploitation is in the interest of artists.
So, it’s no wonder that many artists instinctively reach towards copyright as the tool they are focused on in order to “deal with” questions around AI. And that leads to confusing nonsense hearings, such as the one held last week by the Senate Judiciary Committee, focused on copyright and AI. That hearing (as too many congressional hearings are these days) was full of nonsense, like Senator Marsha Blackburn claiming that “fair use” really means a “fairly useful way to steal.” No, Senator, as the Supreme Court has made clear, fair use is the only way in which copyright law can coexist with the 1st Amendment and is a fundamental speech right. Which is something a senator should understand.
But the focus on copyright, again, seems misplaced. As I explained in that group chat I was brought into, creators expecting copyright to protect them are in for a world of hurt. Because even if Congress goes and changes copyright law, it will be changed (as always) to favor the interests of the studios, the labels, and the publishers, who have always made sure that copyright works to their advantage over those of the creators and (especially) over the public (despite the Constitution requiring any copyright law to benefit the public first and foremost).
If content creators are concerned about the impact of AI on their own livelihoods, copyright is likely a poor choice overall. Similarly, if they expect that there is some legal tool that will magically protect their jobs from new technologies, I fear that they are not going to end up being particularly happy either. Historical attempts to ban technology all fail, replaced by society eventually figuring out how to use the technology in a more reasonable way (often leading to more products, and even more revenue for those who were scared of the technology).
As we’ve pointed out over and over again, the Hollywood studios spent years fighting back against the VCR. The MPAA’s Jack Valenti claimed in a congressional hearing about the VCR that: “I say to you that the VCR is to the American film producer and the American public as the Boston strangler is to the woman home alone.” Literally four years after he made that claim, home video revenue provided the Hollywood studios more revenue than box office receipts.
And we’re starting to see creative folks figuring out ways to embrace, rather than shun, AI. In April, the artist Grimes didn’t rely on copyright when she told everyone to go ahead and use her AI-powered voice in any songs they wanted, in exchange for sharing 50% of the revenue back with her. She just asked people to register the songs on her website. In fact, she explicitly does not claim ownership of any of the songs others make.
Within a month there were hundreds of songs, some generating real revenue for Grimes (and whoever created them). And Grimes seems really thrilled with all this. As she told the NY Times in response to a question about whether or not the AI Grimes “negates the need for… real Grimes.”
No, I don’t think so. Maybe for me, but I kind of want that. Feeling really amazing from making beautiful art is something that has typically been behind a gate for a lot of people — extreme amounts of time and energy, years of technical training. I think it’s valuable that there’s a tool with which, if you have a beautiful idea, you can make a beautiful thing and access that.
In a different interview, with NPR’s Planet Money she made a similar, but slightly different point:
I run into absurdly creative humans all the time, but not a lot of people get to be artists. A lot of luck is involved in that. It’s hard to build a fan base, and it’s hard to get your work in front of the public. So if there’s ways to reduce these algorithmic barriers by letting people inhabit my being, then I think we’re moving in a direction I really like.
I’m not saying that everyone needs to do as Grimes does and release their voice or likeness for anyone to use, but with every new technological breakthrough we see the same pattern. We see some folks freak out about it, and quickly rush to copyright law to try to “protect” the old ways of doing things. And that never works. But then we see some more creative folks embrace the new technology, not worry about the copyright implications, and figure out what’s best for their fans and what’s best for creativity as a whole.
This conversation would be a whole lot more productive if we all focused on figuring out the creative ways to embrace the technology and do more with it, then falling back on trying to use the wrong tools (copyright) to try to hold back the tide.
Comments on “Stop Rushing To Copyright As A Tool To ‘Solve’ The Problems Of AI”
Marsha Marsha Marsha
Anytime there’s something nonsensical in the Senate, I can count on “my” esteemed Senator to be right up in there, a black hole of truth, intelligence, and good sense.
Wait...
Once you bought the book. You cannot just place it in a library and lend it out to different people/entities to learn from.
We are not talking about one person buying a book and learning from it but about one Borg buying a book (if even buying it) and the collective learning from it.
That does not scale. The easiest way to make copyright scale would be to nullify it. But then one would need to come up with a replacement for its original purpose (the current excesses and overhead are not really worth preserving, and planing without them would provide a lot of breathing room for new ideas).
Re:
um. What? That is literally the opposite of true.
Re: Re:
Public libraries need special licenses and operate under special conditions. Really.
Re: Re: Re:
That depends on the details. In the USA, anyone can lend out physical books, or other media such as DVDs, without any special license. (Remember how the film companies got quite annoyed at Redbox, about 15 years ago?) It’s when they make copies, such as for digital “lending”, that they need a license, so of course that’s where the copyright cartels are screwing them over. And outside the USA, even lending or selling physical material may require licenses or royalties.
Anyway, while “we” might never say that learning violates copyright law, Poe’s law is always very much in effect when it comes to copyright maximalists. There’s probably someone saying that.
Re: Re: Re:2
A case in point is Major League Baseball’s claim that “Any … account of this game, without the express written consent of Major League Baseball, is prohibited”. Someone did try to get permission to tell a friend what happened, but was unsuccessful. That is exactly “learning”, plus telling a friend what was learned.
Re: Re: Re:
They most certainly do not.
https://www.techdirt.com/2020/05/20/libraries-have-never-needed-permission-to-lend-books-move-to-change-that-is-big-problem/
Re: Re: Re:
Um. No. Not really. That’s wrong.
https://libguides.ala.org/copyright/firstsale
Re: Re: Re:
“The first sale doctrine, codified at 17 U.S.C. § 109, provides that an individual who knowingly purchases a copy of a copyrighted work from the copyright holder receives the right to sell, display or otherwise dispose of that particular copy, notwithstanding the interests of the copyright owner.”
source
Re:
Once you have bought a real book, you can let your friends read it, give it to someone else, or sell it. For the last two, the new owner can do the same. Most physical books are read by more than two people.
The pattern being that an image of your head is used to create an AI television host, who then comments on investigative reporting about the misdeeds of Network 23?
Grimes is not the entire world of creatives though. You go “I’m not saying that everyone needs to do as Grimes does” but then talk about letting people like Grimes and Trent figure out “what’s best for creativity as a whole”.
Letting megastars decide how technology progresses, top-down rather than unions and more advocating from the bottom-up? It doesn’t seem a lot different from letting something as powerful as Hollywood dictate the way that technology goes.
Re:
Where did I say anything along those lines?
Re: Re:
When you talk about Grimes and Trent Reznor and point to them as the kinds of people that will steer people toward “what’s best for their fans and what’s best for creativity as a whole.”
Re: Re: Re:
The second to last paragraph of what Mike wrote starts with:
It seems you skipped that paragraph in your rush to get a place in the peanut gallery.
Re: Re: Re:2
And you missed my first comment where I said:
Using someone with money and connections like Trent Reznor when talking about seeing “some more creative folks embrace the new technology”, it gives me pause as to whether he actually means it when he says “I’m not saying that everyone needs to do as Grimes does” because he’s pointing to two examples of megastars embracing technology.
Re: Re: Re:3
The whole point of the article is that big artists (and studios) almost always fall back on copyright when new technology emerges that they see as negative which makes it very much relevant to use examples of big artists that don’t.
If you don’t agree with that assessment, tell us then who Mike should use an as example of artists embracing new technology when a majority of all the other big artists think it’s bad? Especially for making the point that copyright isn’t always the solution.
Re: Re: Re:
Not this nonsense again. When I used to talk about Reznor people would say exactly what you said. So then I’d talk about smaller up and coming artists doing creative shit, and people would complain “that only works for up and coming artists because they have nothing to lose.”
You’ll make excuses for everything.
I even wrote a whole article where I went through artists starting big, then medium successful, down and down, to up and coming, to show that creativity can work across the board:
https://www.techdirt.com/2010/01/25/future-music-business-models-those-who-are-already-there/
It’s been years since I’ve had to respond to such nonsense, so thanks for reminding me of this old article.
50%? 500%?
So if 10 artists allow me to make music using their voices/instruments like Grimes, and I make a wonderful song using all of them, where exactly am I to come up with the 500% necessary to pay them all?
Re:
That’s entirely your problem since you choose to use those voices with the attached cost.
Re: Re:
It’s more a problem of the public because it means that such a song will never be made, and the justification of copyright is to incentivize creations, not disincentivize them.
Re: Re: Re:
If you promise to share multiples 50% of a possible income, regardless of what you are doing, paying those multiples of 50% is still a you problem – copyright or not.
Re: Re: Re:
Which has rarely, if ever, happened before the Statute of Anne, and in practice, even after.
Copyright issues before the printing press was legitimately an issue of who got to copy what, simply because the ability to write was time-consuming to teach and, even before the Middle Ages, was limited to the rich or the religious orders (who were coincidentally also rich).
They used to actually go tobwar over this, mind.
Fastforward to the Stationer’s Company bullshit. The damn greedy jerks never wanted to, ahem, encourage creativity, but to maintain andbossify a monopoly on publishing, and, as a very nice aside, censor what they felt would harm the Monarchy.
History lesson sink in yet?
It might be the other way around. After decades of having the same lesson repeated that copyright infringement is theft, the Internet now turns to amoral computers with no sense of right and wrong to create works of art inspired by others. There’s no theft without knowledge and intent, so these computers who can only be acted upon are a godsend. The idea that artists need to find a constant source of “seconds” to break the law is a pretty good indicator that we need copyright reform
So how about the inverse of the premise...
Can AI make copyright better? Some sample scenarios:
1) un-copyrightable AI content overwhelms copyrightable content on social media and spreads from there.
2) The debate about whether AI-generated material deserves copyright protection (and who should hold such a copyright) produces meaningful reforms for copyright as a whole.
3) automated copyright filters or blockers on social media are significantly improved by AI to the point where they’re functional. Or conversely, that AI filters are successful at detecting fraudulent DMCA claims.
Re:
For #1 the way this makes copyright “better” is by devaluing copyright to the extent that it drives IP trolls out of the market. Or going even further, (making copyright better but destroying many art forms) producing enough quality content that human creators stop monetizing their own work.
Not that I think either of these scenarios is likely. If AI gets that good, we’re more likely to start looking at legislative changes granting copyright to AI-generated works (see #2).
Re: Re:
But then if an AI never dies, then their work never ever enters the public domain. Considering that the word “robot” itself comes from the Czech word for a forced laborer, let’s treat it like the machine it is and have it make this public domain art for us all to enjoy.
Re: Re:
On top of what Crafty Coyote has said, there’s no real way to regulate what happens when AI gets granted copyright ownership. In the same way that Naruto the monkey was not allowed to have ownership of his selfie, despite PETA’s desperate attempts to say otherwise.
An AI cannot fight for its right to hold onto IP in the same way that a human can, regardless of whether the human has the legal and monetary resources to do so. Granting AI permission to hold copyright would simply have it devolve into a fight over corporations and other monied individuals competing for the right to vacuum up as much AI generated content as possible for licensing.
Is copyright law, in its current smegstorm state, the right tool to solve anything?
Re:
If your problem is “how to permanently entrench the current class divide and make people revisit Marxism” then yes.
Re:
It is if you’re a wealthy investor/owner of “intellectual property” and your perceived problem is not being wealthy enough.
It sure is curious to see Hollywood advocating for a model where you buy something, say, an actor’s likeness and voice, once – and then for that single one-off payment you own it permanently, can remix and re-engineer it as you please, and use it to whatever personal or commercial end you see fit.
Re:
Woah there.
Just because Hatsune Miku is a runaway hit and proof positive that it can work (her voice bank provided for by a Japanese voice actress) doesn’t mean Hollywood will take the hint.
Re:
Sure is curious for the same actor to expect to be paid for doing nothing after they have sold their likeness.
Re: Re:
OP’s got a point though. Hollywood has never been the sort to say that you should be allowed to pay for something once then own it forever and do whatever you want with it.
Re: Re: Re:
True, but Hollywood accounting means that they often only pay once.
To be fair, the main reason why copyright law worked out wasn’t due to any particular effort from Hollywood. They lucked out because judges didn’t have an incentive to doubt anything they said, or take a closer look at the evidence brought in those cases.
The successes reaped by the lawyers were then attributed to copyright law, which is where they got vital talking points like “$150k per infringement!” Realistically, even their biggest successes like Jammie Thomas-Rasset and Joel Tenenbaum were, at best, victories that were primarily Pyrrhic in nature. Artists were not enriched by cases going to court involving small downloaders.
So the outrage by artists is somewhat understandable, when they get told that copyright doesn’t protect what they think it should. They’ve never known copyright to fail. The problem is that right now, they’re dealing with a post-Prenda era, where copyright law is no longer the “I win!” button that they’ve been historically used to. Judges and civilians alike are more willing than ever to question their morally-positioned arguments.
At the very least, they’ve got most of the civilized world backing up their writer-actor strike, because fuck Hollywood.
That's an incredible leap
Pretending people learning from a book is equivalent to an LLM harvesting billions of works (without permission, attribution, or compensation) is quite the leap. The scale alone makes the comparison ridiculous, let alone the ethical implications.
I agree though that we can’t “copyright” ourselves out of this. Copyright only applies if little people download stuff, not if a big company does it.
I find the conversation around generative AI such a depressing indictment of our industry. Maybe it’s my mistake holding us to a higher standard.
Re:
Do you have any idea of how many copyrighted works you see each day? You’d be surprised at the scale of copyrighted information you ingest and use in your daily life.
Re: Re:
Human brains and machines don’t ingest information in the same way. Please stop pretending they do.
Re: Re: Re:
Nobody said they did, it’s about how it’s processed and stored, and later used to produce an output. Ever heard of the expression “but on a computer”? It aptly demonstrates the fallacy that when something is done with a computer it’s somehow different.
Re: Re: Re:2
And humans and machines do this differently too. The machines doing this are capable of doing so on a speed and scale that a human can’t match, alongside this.
Humans and machines are different things and what a human does vs the scale and speed at which a machine does something, they need to be regulated and treated differently in many circumstances for us to have a functional society. I’m sorry that you can’t see that and have to depend on petty false gotchas like an NFT bro.
Re: Re: Re:3
Perhaps, but if you’re thinking of copyright as the tool to do it, you might very well not get the result you’re looking for.
Re: Re: Re:3
That argument has been trotted out in some form every time some new disruptive technology come along that upsets the current status quo.
On what false gotchas am I depending on? Be specific.
Re: Re: Re:4
Stuff like this. A fake gotcha where you equate humans looking at stuff with computers scanning and scraping up stuff and treat both like they’re the same thing for the sake of your argument:
Re: Re: Re:5
You still haven’t explained why it’s a fake gotcha, you just say I’m wrong. If a computer mimics what a human does but faster, what’s the difference?
Re: Re: Re:6
You’re still trying to say that what humans and machines do are the same, and acting like a person going about their day and seeing or hearing or playing copyrighted stuff is the same as a machine trawling the web faster than any human can as its main task. That’s where the fake gotcha is. I think I’m done here, to be honest.
Re: Re: Re:7
You should have no problems explaining the difference then instead of just saying it’s a gotcha, right?
Ie, you actually can’t explain the difference.
Re: Re: Re:7
To be honest, that’s what pro-copyright entities have been saying for years. Which is why they justified shit like claiming that someone using a popular song as a ringtone committed grand copyright infringement at the same level as BitTorrent when their phone goes off in the subway.
Re: Re: Re: neither do people
Literally nobody ingests information in the same way as everyone else. False dichotomy.
Re:
Also, a LLM doesn’t harvest copyrighted works like a combine harvester harvests crops.
That’ll be the “research arm” of the “AI generator” company.
And those are two different conversations. A LLM merely uses the copyrighted work to “learn”, albeit with a ton of user vetting, because apparently, the algorithms were never designed to mimic the entire learning process, complete with memory retention and all the funny stuff humans do to process said information.
Also, if copyright was at the heart of the issue, do explain why Disney hasn’t already either sued the damn grifters or worked oit a deal with them. Unless I’m mistaken, that shit IS newsworthy enough for every country on the planet to take notice.
Re:
Just how many of the books have you read, music tracks you have listened to and movies you have watched, have influenced how you say what you say?
I feel it’s telling that grimes released her voice and appearance for 50% revenue cut. At the very least, that’s how much she feels she ‘should’ get for A.I. voice, which to me implies that there should be some monetary compensation for using someone’s likeness.
It’s also a substantially different debate (as far as I can tell) from the deep learning art/writing, where the authenticity or validity of the libraries used for deep learning voices is not really questioned..
It only takes a short clip to make an AI voice. What should actually happen in the case that someone’s likeness is stolen? Should they sue for 50%? 100%? I don’t see an answer here, and I don’t necessarily expect one to come around anytime soon, but I think that’s the underlying concerns many actors have, and we can see this as a flashpoint in the current SAG-AFRA strike.
Re:
How much money should someone get if someone else makes a portrait of them, using publicly available materials and sources? It’s likely that only existing celebrities will have the resources and clout to argue these terms and conditions, as well as drag other people through the judicial system if it comes to that.
This is the point where I think you’re going to have to look at intent, and that’s not something even the courts can reliably get right sometimes, seeing that it’s basically left up to lawyers trying to outwordplay each other. But we can look at examples such as using Photoshopped images of celebrities to scam people with dodgy drugs or crappy investment ideas. As far as I can tell, the lawbreaking comes from the fraud and deception, not the usage of a public figure to portray something inaccurate – although this could change.
The bigger issue is that once people realize the value of their data, they won’t feed it to OpenAI or others so fast. Also Meta is now releasing open-source LLMs which will incentivize proprietary data models. We’ll have next-level garbage-in-garbage-out because those who have the best info will have that amplified by their version of a bot, with no incentive to share their data.
AI cannibalism and feedback loops are already dumbing down GPT, which will go the way of rented bowling shoes or house bowling balls versus custom-drilled.
Re:
How will the prevent the owners of an AI from feeding it material that the owners have legally obtained?
Re: Re:
It won’t, which is why rightsholders and authors are desperately thumbing their way through copyright law to find something that supports their position.
Copyright is the mechanism how authors are supposed to get compensation for the work that they do. It’s government’s promise that copyright laws will be enforced to the level required for authors to get suitable compensation for the products available for consumption. So far, this has not produced riches to the authors, and instead of getting proper compensation, authors are struggling to gain any money reward for the endless work.
While rejecting compensation for selected groups of people might be useful way to keep money budgets in control, in the long term, it eats the legimate activity in the society and gives power to scams and illegal activity rather than genuine attempts to improve lives of your fellow citizens.
When book authors, song writers, software developers and artists cannot get their money situation fixed by the copyright payments, but pirate sites are gathering millions of illegal money from the marketplace by ripping off artists works, authors lose confidence to the government’s promises about compensation level.
This loss of trust to the government is dangerous and we’ve already seen it in usa and latin america, and the recovery back to the path where police, fbi, nsa, copyright owners are enjoying trust that they deserve for their hard work of enforcing copyright laws.
Re:
If there was “millions of illegal money” made by pirates, the IRS would have found it a long time ago.
As it stands, even the most dedicated of agents couldn’t find any assets from The Pirate Bay to pull beyond their servers and operating costs.
The truth is that the “profits” made from piracy have always been exaggerated to make it sound like nerds robbing a bank. If giving away content was that profitable, content creators would have used that strategy to claim that money instead.