#comics for sale | Explore Tumblr Posts and Blogs

davidmariottecomics · 7 months

Text

Copyrights and Wrongs, Part 4 (A.I.)

Greetings!

We come, at last, to the end of our discussion on copyright (for now)! And we're talking about it specifically in relation to "Generative A.I.", a buzzword for copyright theft tools! And, hopefully, if you weren't feeling secure with your understanding of copyright before now, you'll be feeling pretty groovy after this and will be able to submit an informed and thoughtful comment to the Copyright Office's A.I. study, where they are requesting public comment--particularly from those most affected by A.I.--to help determine their future policy on it.

Where We Are Right Now

Honestly, like most things in the world, it's kind of a mess with a ring of hope.

First and foremost, remember that all of this is new and at the beginning of legislation, so from a legal standpoint, they're playing catch-up to the technology, which isn't uncommon. It's also happening at a tumultuous time in the U.S. as the country is dealing with the coupling of increased facism (hey, fuck KOSA) and unyielding capitalism. It's almost like those things are all related in some way...

The Copyright Office has started fielding submissions of A.I. generated material, and so far their general policy seems to revolve around a matter of substantial contribution. So, obviously, there was that really terrible comic that made the rounds a while back, and that I alluded to last week, that was one of the first big test cases. It was fully "illustrated" (stolen) by A.I. with original text written by the "creator" and submitted to the copyright office. Initially, it was registered, then modified as the A.I. component became clear, and ultimately the copyright was registered for the original text, but not granted for the visual component.

But, my understanding is that there have been works that use a similar enough model or technology or "A.I." that have been easily registered with the copyright office because the amount of technological contribution was minor. For example, if a similar tool was used to intuitively do color flatting, but the final color rendering--all the final tones, effects, shading, etc--was all done by the artist, that wouldn't pop. Or if an artist created 95% of a work and fed their own piece into an engine that then finalized that last 5%, that might not pop. And I think the essential difference, that they're currently recognizing, is the distinction between a human being putting in their own work to create something that is copyrightable and using "A.I." as a tool in the process of creation and a tool being left to fully create the "new" work.

Meanwhile, a number of lawsuits have been filed and are in various stages of processes, which is where the frontlines of the battle are being fought. A lot of the claims are being brought specifically on the basis of copyright violation (and a few cases of trademark violation). For probably pretty clear reasons, creatives of all sorts are saying "hey, fuck this noise, you took my work without permission and your robot is allowing people to inherently copy it for their own benefit, without payment or attribution."

The complication--and the reason why we need clearer directives put forth from the Copyright Office, as well as some legal wins to better define case law around this--is that the A.I. developers are basically claiming that:

1. They don't actually have the data. Either the way their system functions, they don't actually keep it on hand, they just let the program go wild online or through whatever they at one point used to train it and they don't know how exactly it chooses what it does, but it just goes and makes something and there's so much information in these systems that you wouldn't actually be able to prove whether it used your thing or not.

2. Along those same lines, since they have so much data, if you can't definitively prove that your work was taken, that you suffered damages, your lawsuit should just be dismissed. Doesn't matter if your work turns up in Book3 and was probably torrented to get there (which... if we could prove that would actually get most of this stuff thrown out as violation because the FBI says you wouldn't steal a car! Nah, but, really, if we could prove that the way the works were obtained to fuel the datasets was illegal, that'd be a huge boon to these cases), if you can't prove that your work has specifically been violated because, well, they have 183000 other books it could've stolen from, the case doesn't have clear standing and should be dismissed.

3. That they are not inherently responsible for how their technology is used by consumers. Basically, your work being in their dataset does not cause harm by itself and any damages that may result from your work being in the set are actually the responsiblity of the person who, say, used "A.I." to write like a Sparknotes-esque summary version of your book. The company didn't ask the program to write that, the user did, and if the user then sold it, the responsiblity rests with them.

4. That it's fair use. Basically, as long as they're using publicly available data, or data submitted by the users with the *understanding* that they are knowingly submitting their own material, the resulting generation is a transformative work. We're going to put a pin in this one and come back to it.

Current Caselaw

Naruto v. Slater et al is a particularly bizarre case that has a lot of weight in this fight at the moment. It's a really wild little story about a nature photographer who went out to take photos. He stepped away from one of his cameras for a minute and a crested macaque named Naruto picked it up and took some selfies! The photographer printed them in a book of his and said "look at these monkey selfies!" PETA said "Hold on! That monkey deserves restitution! He took those selfies and holds that copyright, not you, hoo-man!" And so they sued on behalf of Naruto. And the judge said "Well, no. I'm not giving a monkey copyright. He's a monkey and copyright is something that goes to works with human authorship." Following this case, the Copyright Office issued directives claiming "only works created by a human can be copyrighted under United States law, which excludes photographs and artwork created by animals or by machines without human intervention" and specifically lists "a photography taken by a monkey" as something that doesn't count. Now, whether Slater, the guy who owned the camera, actually owns the copyright on Naruto's selfie is a little... ambiguous. The court's dismissal of the case definitively says Naruto doesn't own it, but it doesn't specifically say Slater actually gets to maintain it or if the image just doesn't have copyright.

But I bring all of this up because, again, this is the standard that the Copyright Office currently follows. They're looking for human intervention and they're looking for substantial contribution.

Author's Guild v. Google, Inc also comes up a lot, particularly in the fair use defense. And it's one that's... tricky. Basically, you know how you can look up a book on Google and often there's a little thing on the side of the screen that lets you read the first like 20-ish pages (I looked up House of Leaves for an example, and they'll let you preview 77/738 pages). Well, authors said "they're copying our books without permission, that's a violation of copyright." And Google and the courts said "no, it's transformative. They aren't doing it in competition with the books, they're creating a new way of access that refers back to the original work in market purposes, not a supplanting version of it." And this is the exact sort of precident that makes these current proceedings really legally rough. Like, it is understood and appreciated as a function of Google (well... old Google, not post-"A.I." Google...) that it is a search engine and that for it to work, it needs to be able to conduct searches based on the information it's given and be able to post results for ya. And that means taking what's publically available to it to search, which often includes copyrighted material. Now, the exact nature of how they "transform" copyrighted material into more easily searchable material is something that gives me pause--I do feel like there should be more of an opting-in from the copyright holders--but I can see how the court came to the conclusion it did.

The bigger issue is, of course, that this is one of the big cases that "A.I." developers are using as justification of fair use. The data sets they're using are essentially acting like Google--searching through existing material like a search engine--but with a new result at the end which is transformational of the original information. There are a couple of differences here that I think disclude that premise, but again, pin in that for a second.

Feist Publications, Inc., v. Rural Telephone Service Co. is a really interesting one in this whole discussion too. This case was about a telephone book. Short version: both Feist and Rural made telephone books. Feist covered a larger area than Rural. Feist ldecided to print a phonebook with a dozen areas all in one. They contracted with 11 of them, but Rural held out, so Feist just copied the names and numbers and printed them anyway. Rural said "that's infringement" and the courts said "no, they're just printing facts. There's no significant creative authorship to the phonebook, and a set of factual data can't be copyrighted." This one's important on two fronts. First, it's one of the major cases that begins the definition of threshold of originality. Basically, there needs to be some actual creative thinking involved in making a copyrightable work, but it requires actually trying to inject thought into the result and make it distinct from existing works, not just having done the effort of work. That is to say, sure, it took time and work to originally compile the phonebook, but it is organized alphabetically and just a set of facts and was not arranged in an artistic manner. Second, and something that I think will be interesting going forward, it set that the U.S. doesn't recognize database rights. I will be very interested to see if, say, once competition starts stealing the datasets from some of these bigger A.I. developers, they'll suddenly worry about their own copyright--which actually doesn't exist for those datasets.

And the last case for right now that I'll mention is Twentieth Century Music Corp v. Aiken. This was a particularly wonky case. A restaurant owner was sued for having the radio on for his customers, under the basis that the radio station had the right to broacast songs and that was fine and good, but the restauranteur did not have the same broadcasting rights within his establishment. And this was a major copyright case. Ultimately, the court decided that, no, having a radio (or TV or whatever) on in an establishment did not inherently violate copyright because it is not a performance of that work. But the really important thing in this case is it definied again the goals of copyright. Copyright exists at the intersection of the authors, the disseminators, and the users. The rulings describe copyright's purpose “to secure a fair return for an ‘author’s’ creative labor'” and “to stimulate artistic creativity for the general public good." And I'd contend, as so many other people are doing right now, that generative A.I. is not securing a fair return for author's creative labor, nor stimulating artistic creativity for the general pubic good. It is engaging with existing, copyrighted material outside of fair use and without return for the original author, and supplanting creativity--both in discouraging existing authors and in flooding the markets with creatively bankrupt material. As Gallup reports, 3 out of 4 Americans believe that A.I. is going to reduce jobs and 80% of Americans do not trust that companies will use A.I. responsibly.

Transformative Use

Okay, here we are. I've said so much about all of this, but here's the crux of my arguement, at least. The standard for copyright is that to qualify for copyright, materials must meet certain standards of originality, creativity, and fixation. The products of A.I. (and "products" is used particularly pointedly here) certainly qualify for fixation. They can be seen and in some form exist, much as I may personally hate it. But I don't believe they meet the standards of originality and creativity.

A comparison I have frequently made is that of a blender. You can write a grocery list and by most standards, that's going to fall under the same idea as a dataset. Now, in one version, you can use your expertise to buy your own groceries and combine them to make a meal--taking the time, effort, skill, and your existing or gained in the moment knowledge to create something nourishing. Maybe it goes wrong and you burn something, or maybe you forget a key ingredient, or maybe you decide to experiment and the result is less than appetizing, but it's all a result of your thought and efforts. Now, if that's how you're approaching any sort of creative work--art, writing, music, film, etc--that should qualify for copyright. What "generative A.I." is doing is being fed a grocery list, *somehow* obtaining the items on the list, but in a vague way where they may or may not be the correct items on the list, they may or may not include more or fewer items than on the list, and the legality of how they were brought to you is unclear, and putting them all in a blender. Yes, you get a sludgey smoothie at the end of it, and the smoothies may be distinct in what they put out, but they lack the skill, refinement, and knowledge of even an amateur chef working with the ingredients. They just regurgitate what they're fed, and if you just add a pinch of salt at the end, that's not really doing anything.

But this only applies to why works created by A.I. shouldn't be copyrightable. The other crucial part is whether they're violating copyright. So much of the defense of fair use relies on the transformative standard. Is the original material being infringed upon? In the first place, let's just say that, as long as we're talking about "A.I. works" as commercial enterprises, most of them are not going to fall under fair use under the standards of parody, satire, or commentary. When we see something like that bad Batman comic I mentioned last time, or the bad "A.I. illustrated" comic I mentioned earlier, those are clearly meant to be taken seriously as new, original works. They aren't commenting on or engaging with existing copyrighted work as a way of making a point. They're seeking to create a new work with existing elements, but particularly in the case of that terrible Batman comic, without a real understanding of the work they're engaging with. The artistic styles are all over the place specifically because the "creator" couldn't be bothered to know whose work he was taking from--and, I'd contend, if he was specifically attempting to copy the works of specific artists in an attempt to replace their work, that also would not fall under fair use standards.

Additionally, to the argument that the developers are making, they say that their databases are so full of information that you can't *really* tell what the original works were, and in that way, they are not a repurposement for the identification of the original work. Which, again, to that bad Batman comic, I think is a really damning defense. I, and many other people in my industry, were able to clearly and recognizably point out what authors and works were being taken from--no, maybe not every work that was being used, but enough that it would generally be considered plagiarism. The products created are meant to be in competition to, not additive to, the original works without acknowledgement of the works that they are taking from.

Judge Pierre Leval has described it as "Transformative uses may include criticizing the quoted work, exposing the character of the original author, proving a fact, or summarizing an idea argued in the original to defend or rebut it. They also may include parody, symbolism, aesthetic declarations, and innumerable other uses." These aren't critiques. They don't expose the character of the original author, provide facts, summarize anything other than the exact work itself, or parody. And they don't make aesthetic declarations, they take and recycle others aesthetic accomplishments. The whole issue is that they claim they're making something new, something different, something worthwhile from whatever data they may have within the system, but the system is a tool. It doesn't have the human recognition to actually synthesize the material and make a transformative work. It just regurgitates what it's fed.

Final Bits, I Swear

Apparently you can opt-out of Dall-E now, but the process is arduous. So much of what these folks are relying on is making their processes so difficult, people don't engage with them when there is an option to opt-out, and such misinformation and obfuscation on how things are created that it's difficult to identify when your work may be used, and for the public to understand how the work is being harvested.

I'm supremely frustrated with companies like Disney who keep using "generated A.I." in their promotional materials. I'm also disappointed in shows like Last Week Tonight where John Oliver and crew shared a bunch of A.I. generated Johns that had been posted to Reddit. Like, Disney's a company that has literally changed copyright law in the U.S. to suit their purposes. John Oliver's only now back because the writer's strike is over, where the writer's won protections of their jobs against A.I. Like... the hypocrisy is astounding and disappointing that they cannot square their own desire for copyright protection with protecting others.

You may've seen the recent A.I. manga lettering debacle (hey guess what, it was bad!). Here's an interesting little conversation with some actual letterers about it.

Finally... yeah, I do actually think there are uses of A.I. as a tool that are fine. And most of that comes down to being a person who inputs your own data. I think about how Pixar has a "character generator" that works on these same principles, where they drew a bunch of base features and fed them into a system and the system helps them animate by populating backgrounds with characters who have random combinations of those features. But, that works off of stuff that the Pixar animators put in. Or, again, fairly similarly, I know meteorlogists have talked about how this sort of system can be really helpful in taking the data they've recorded and helping them predict upcoming weather patterns. But these are so vastly different from how bullshit like ChatGPT or Dall-E or any of these other scam machines are being operated and I just wish people could understand that.

I think that's it for this week! Don't forget to submit your own comments on the Copyright Office's study and also to say fuck KOSA to your senator!

What I enjoyed this week: Blank Check (Podcast), Craig of the Creek (Cartoon), One Piece (Manga), Pokemon Violet (Video game), The Archive Undying by Emma Mieko Candon (Book), Sex Education (TV show), Only Murders in the Building (TV show), Yu-Gi-Oh: Duel Links (Video game), Wipeout (TV show), Whisper of the Heart (Movie), Josee, the Tiger and the Fish (Movie)

New Releases this week (10/4/2023): No new books from me this week.

Final Order Cut-Off next week (10/9/2023): Godzilla Rivals: vs. Mechagodzilla (Editor) Sonic the Hedgehog: The IDW Comic Art Collection (Editor)

New Releases next week (10/11/2023): Sonic the Hedgehog #65 (Editor) Sonic the Hedgehog Halloween Special (Editor) Sonic the Hedgehog: Seasons of Chaos TPB (Editor)

Announcements: I'm still doing a member drive over on my Patreon! You can read about it in a public post there! If you join, renew, or updated to the Feature Fan ($10) tier or above, you're going to get a Mystery Comic Grab bag! And as a patron, you're going to have a bit more choice on what all it is! All the info is on there, so if you're curious, please do give it a look! And it'll be going through all October!

If the Patreon isn't your cup of tea at this time, or you wanna do more of a one-time donation, from Oct 3rd to the 18th, so right now, I have 2-3 comic Mystery Bundles in my Ko-Fi store! Same premise as the Patreon--there'll be a bundle for grown ups and one for kids. They're pay what you want with a $15 minimum. If you send $25 or over, I'll ship you a trade paperback too!

It will be US only on both the Patreon and Ko-Fi just because shipping internationally's very expensive right now. But, for international folks, I will put together a nice little digital goods bundle for ya!

Wanna support me otherwise? I'm on Ebay, there's my webstore, or you can support Becca through their channels! Good news: We don't need a new laptop! But your support is still appreciated because we, like so many other folks right now and I get it, are kinda struggling.

Pic of the Week: OOO, been a while since we've done one of these! The pic of the week is Becca's Rouge the Bat, over on Bsky! And big ups to Michelle Perez for encouraging everyone to do more Rouges. But the one I'm posting is the promo for my membership drive/Ko-Fi sale! Y'know how I'm always talking about advantages of the Patreon? They've got a different pic of the week!

#copyright explainer #comics and copyright #comics #ai is bad #fuck kosa #rouge the bat #comics for sale #sonic #transformers #godzilla

10 notes · View notes