|
https://www.youtube.com/watch?v=3OzEPDW3Fdc Now Tayne I can get into.
|
# ? Feb 2, 2023 19:48 |
|
|
# ? Jun 10, 2024 20:05 |
|
Futanari Damacy posted:💯 No argument is being put forth to effectively protect artist's livelihood. All the arguments are to protect their intellectual property from "theft". Either because they have dumb principles about ownership, or more charitably because they have some naive hope that an AI trained on public domain stuff won't take people's jobs. The only thing you could ever prevent an AI from depicting is poo poo like Mickey Mouse because that is an actual copyright. Beksinskis style isn't copyrighted so if Disney wanted to take his job without "stealing" anything they could just hire a bunch of artists for minimum wage to ape his style for their AI to learn from. Or they could just use his work directly because if done properly there is no way to tell what an AI was trained on. Even in your completely un-anarchist communist utopia all the law will ever be able to do is to make AI capable of reproducing Mickey Mouse illegal.
|
# ? Feb 4, 2023 07:33 |
|
https://twitter.com/copyrightlately/status/1621667685485641729
|
# ? Feb 4, 2023 07:38 |
|
i dont like getty i hope they eat poo poo
|
# ? Feb 4, 2023 07:52 |
|
how do you file a lawsuit without actually making sure that you're suing the right company
|
# ? Feb 4, 2023 07:53 |
|
|
# ? Feb 4, 2023 08:06 |
|
my favorite thing about ai art is the sensation of the computer just sort of shuffling a pile of poo poo together on a table and molding it into shapes the way you might try to shape a pile of garbage and then doing one of those "there you have it" hand motions
|
# ? Feb 4, 2023 08:13 |
|
my favorite is when the algorithms just keep drawing more teeth and fingers
|
# ? Feb 4, 2023 09:57 |
|
Mercrom posted:No argument is being put forth to effectively protect artist's livelihood. Pro-AI is "destroy artists' livelihood"; anti-AI is "don't destroy artists' livelihood", both sides are quite matter-of-fact about this
|
# ? Feb 4, 2023 18:13 |
|
Futanari Damacy posted:Pro-AI is "destroy artists' livelihood"; anti-AI is "don't destroy artists' livelihood", both sides are quite matter-of-fact about this is your argument about copyright or livelihood because "require ip licensing" is not going to actually stop ai generation from being used it's just gonna get Getty et al. more money from the next generation model if you believe ai is gonna replace artists then there's no policy short of banning it wholesale that will protect them but also banning stuff to protect jobs is extremely lame and is just a weird artifact of a hosed up culture where it's believed people dont deserve to eat if they don't work. if artists were provided for thru ubi or space communism or whatever they could just make whatever art they wanted without needing to also receive a paycheck from bigco inc. for designing lovely billboard ads that the artist and end user both hate
|
# ? Feb 4, 2023 18:59 |
|
"How are you going to protect artists from what we are going to do to them" is itself a strange way to demand an argument as opposed to just not doing the harmful thing to begin with
|
# ? Feb 4, 2023 19:19 |
|
the question was not "how are you going to protect artists" it was "why are you supporting strengthening ip law for the benefit of corporate interests in the name of 'protecting artists' despite the fact that it won't meaningfully protect them from the thing you actually care about (job replacement)" RPATDO_LAMD has issued a correction as of 19:26 on Feb 4, 2023 |
# ? Feb 4, 2023 19:23 |
|
In all seriousness it really is just as simple as don't pay for legacy media and the money you would have wasted on that, pay actual human artists to make things you would like This satisfies the apparent need for the solution to be somehow socialist and worker-oriented, as well as the anarchist compulsion to stick it to the man, and not least the consumerist demand for content itself IP is a weird diversion to get onto because we are literally all in agreement that following it only benefits big companies and its purpose is not to help artists protect their work. The disagreement appears to be that to me anyway, that doesn't mean you should just steal artists' poo poo willy-nilly with the sole justification being "this is the future, get with it" e: ^ how do I respond to thing I didn't do? "learn to read better" I guess
|
# ? Feb 4, 2023 19:26 |
|
Futanari Damacy posted:https://twitter.com/Dexerto/status/1620091517259087874 Haha, wow, how ridiculous. How do you "accidentally" watch a deep fake porn video of Pokimane??? Like what site would you even go to to do that? Haha. Does anyone know?
|
# ? Feb 4, 2023 19:26 |
|
turn off the TV posted:how do you file a lawsuit without actually making sure that you're suing the right company stable diffusion's outputs literally reproduce the getty watermark. I'm a fan of "stable diffusion is a kind of lossy compression" argument. RPATDO_LAMD posted:is your argument about copyright or livelihood because "require ip licensing" is not going to actually stop ai generation from being used it's just gonna get Getty et al. more money from the next generation model "Professional Visual Artists" don't neatly fall into two categories of "megacorp slop generator" and "starving vulnerable indie artist" no sane person believes that "AI" will replace every single last artist, but it perfectly reasonable to assume that "AI" could replace some of them running large-scale text-to-image models takes a huge amount of capital, plus the input data of many, many real people who were not asked for consent or offered compensation. as these models exist right now, it is hypercharged surplus value extraction and obviously exploitative. exploitation may not be novel to "AI" VC projects, but isn't something to celebrate
|
# ? Feb 4, 2023 19:27 |
|
to be clear the litigation is lose-lose: either Getty/Disney & Co get their way or Peter Thiel does
|
# ? Feb 4, 2023 19:32 |
|
fyi running these models does not take significant capital, they are 2-8gb in size and run on consumer hardware. any random person can just download an existing model (for free) and make jpegs without involving the wallets of any corporations. creating them requires capital for sure though, and big companies (google/openai) are not going to share the models they create/train internally
|
# ? Feb 4, 2023 19:40 |
|
webcams for christ posted:to be clear the litigation is lose-lose: Killer AI robots with Rainbow Flags or Killer AI Robots with MAGA Flags.
|
# ? Feb 4, 2023 19:41 |
|
webcams for christ posted:stable diffusion's outputs literally reproduce the getty watermark. The actual complaint in the lawsuit seems to be quote:This case arises from Stability AI’s brazen infringement of Getty Images’ intellectual property on a staggering scale. Upon information and belief, Stability AI has copied more than 12 million photographs from Getty Images’ collection, along with the associated captions and metadata, without permission from or compensation to Getty Images, as part of its efforts to build a competing business. As part of its unlawful scheme, Stability AI has removed or altered Getty Images’ copyright management information, provided false copyright management information, and infringed Getty Images’ famous trademarks. A German nonprofit named LAION is the actual organization that scraped Getty's servers and collected the captions and metadata, converted them, and then redistributed them as part of datasets called things like "LAION-5B" or "LAION high resolution" along with URLS pointing to the locations that the images are hosted at. https://laion.ai/blog/laion-5b/ You can do searches with it here: https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2Fknn.laion.ai&index=laion5B-H-14&useMclip=false If the lawsuit was just "they downloaded a bunch of our pictures" that'd be one thing, but AFAIK they don't have any way of actually proving how many of getty's images were actually used since probably nobody at any point tracked that and SAI didn't do any meta scraping.
|
# ? Feb 4, 2023 19:51 |
|
Honestly, part of the issue seems to be we're litigating "is bots scraping websites/images legal/moral/ethical" all over again. The lawyers seem to be believe the answer is an unequivocal "no, it's copyright infringement as gently caress, sic em judge", but if that's the case then literally every search engine and aggregator service in existence is also illegal as gently caress and also need to be shut down as well. Which, I wouldn't cry if the internet as we know it was reduced to a pile of slag, but people seem to have a weirdly myopic view of the implications of the claims being made here.
|
# ? Feb 4, 2023 20:11 |
|
LAION may be a german nonprofit but it's a german nonprofit specifically created by StabilityAI to do all the scraping instead of them doing it directly as some kinda legal shield. albeit with the side benefit that "other AI researchers can also take the LAION data if they want I guess" but also, even if LAION did the original scraping, those scraped images are still copyrighted by getty and Stability can still be sued for copyright violation for the act of copying them from LAION's servers to Stability's. the legal question at issue here is not whether SAI copied the images (they definitely did), it's whether it counts as "fair use" under US copyright law and then in additoin to that Getty is also suing for trademark violation over the reproduced watermarks
|
# ? Feb 4, 2023 20:23 |
|
LAION Flux
|
# ? Feb 4, 2023 20:25 |
|
Tree Reformat posted:Honestly, part of the issue seems to be we're litigating "is bots scraping websites/images legal/moral/ethical" all over again. The lawyers seem to be believe the answer is an unequivocal "no, it's copyright infringement as gently caress, sic em judge", but if that's the case then literally every search engine and aggregator service in existence is also illegal as gently caress and also need to be shut down as well. there have already been lawsuits about this vs google, the verdict was "yes this is copyright violation (duh) but goog has a fair use defense in part because their use of the search results is transformative and does not compete with the original content etc etc" LegalEagle's video goes over that google stuff and some of the AI lawsuits from a lawyer's perspective: https://www.youtube.com/watch?v=G08hY8dSrUY
|
# ? Feb 4, 2023 20:32 |
|
RPATDO_LAMD posted:LAION may be a german nonprofit but it's a german nonprofit specifically created by StabilityAI to do all the scraping instead of them doing it directly as some kinda legal shield. Where did you hear/read that LAION was created by Stability AI? From what I understand they're separate entities, but Stability AI has covered their rental fees.
|
# ? Feb 4, 2023 20:35 |
|
turn off the TV posted:Where did you hear/read that LAION was created by Stability AI? From what I understand they're separate entities, but Stability AI has covered their rental fees. How kind of them to help charity
|
# ? Feb 4, 2023 20:35 |
|
RPATDO_LAMD posted:there have already been lawsuits about this vs google, the verdict was "yes this is copyright violation (duh) but goog has a fair use defense in part because their use of the search results is transformative and does not compete with the original content etc etc" So the copyright side of all this really boils down to if the artists/Getty are able to convince the judge "right, so copyright infringement like this was okay, but copyright infringement like that is not", huh. The whole "what is transformative vs derivative" just seems like a completely arbitrary distinction based entirely on what judge you get.
|
# ? Feb 4, 2023 20:38 |
|
idk I just think that from the perspective of suing over copyright infringement and data scraping that seems like something that you'd want to take into account for your lawsuit
|
# ? Feb 4, 2023 20:40 |
|
turn off the TV posted:Where did you hear/read that LAION was created by Stability AI? From what I understand they're separate entities, but Stability AI has covered their rental fees. Seems I was wrong. LAION already existed but Stability funded the creation of the 5B dataset specifically: https://techcrunch.com/2022/08/12/a-startup-wants-to-democratize-the-tech-behind-dall-e-2-consequences-be-damned/ posted:Mostaque says that Stability AI funded the creation of LAION 5B, an open source, 250-terabyte dataset containing 5.6 billion images scraped from the internet. (“LAION” stands for Large-scale Artificial Intelligence Open Network, a nonprofit organization with the goal of making AI, datasets and code available to the public.) The company also worked with the LAION group to create a subset of LAION 5B called LAION-Aesthetics, which contains 2 billion AI-filtered images ranked as particularly “beautiful” by testers of Stable Diffusion. Tree Reformat posted:So the copyright side of all this really boils down to if the artists/Getty are able to convince the judge "right, so copyright infringement like this was okay, but copyright infringement like that is not", huh. that's literally the definition of fair use yes (but it's more nuanced than just "transformative y/n". watch the video). that's also why it's legal to sample a song, or quote part of a book in a review, or parody a film, without getting your dog shot by the copyright cops despite that being literal exact copying. laws being subjective and up to human interpretation by a judge/jury is also very cool and normal. if everything was just "beep boop follow these exact mechanical rules", everything would be run by an army of bloodsucking lawyers exploiting every technical loophoole they could find. i mean, even more than it already is
|
# ? Feb 4, 2023 20:45 |
|
RPATDO_LAMD posted:that's literally the definition of fair use yes (but it's more nuanced than just "transformative y/n". watch the video). I did, it was just as mushy there as everywhere else I've read about it. quote:that's also why it's legal to sample a song, or quote part of a book in a review, or parody a film, without getting your dog shot by the copyright cops despite that being literal exact copying. I mean, I had it drilled into my head in school "even a few words being the same as somewhere else is plagiarism, don't ever do it or you instantly fail the assignment, we have a computer program that will know if you did", so every book report was an exercise in describing the book without copying anything from it. quote:laws being subjective and up to human interpretation by a judge/jury is also very cool and normal. if everything was just "beep boop follow these exact mechanical rules", everything would be run by an army of bloodsucking lawyers exploiting every technical loophoole they could find. Eehhh, I'd prefer laws to be set in stone when they're made, and when they inevitably prove inadequate, they get changed, replaced, or ideally repealed by the issuing body of elected representatives, rather than this constant ad-hoc patching of vague laws by capricious unelected bureaucrats.
|
# ? Feb 4, 2023 20:54 |
|
RPATDO_LAMD posted:that's also why it's legal to sample a song the song analogy is useful because it is not legal to sample a song for a without properly licensing the track if you want to monetize your new recording. it's very cut and dry and algorithmically enforced when you try to host your new music on a streaming service songwriting is where it's a lot more interesting, messy, and much more arbitrary
|
# ? Feb 4, 2023 20:57 |
|
Futanari Damacy posted:Pro-AI is "destroy artists' livelihood"; anti-AI is "don't destroy artists' livelihood", both sides are quite matter-of-fact about this Futanari Damacy posted:"How are you going to protect artists from what we are going to do to them" is itself a strange way to demand an argument as opposed to just not doing the harmful thing to begin with There's a paid service you can use right now to make almost perfect text-to-speech using any voice you train it on. There's actual dangerous AI out there and no one is stopping that. This discussion is only about how to mitigate the damage. I'm gonna make a radical anarchist suggestion and say that maybe less IP laws would help. If you want to reduce the potential power of large corporations who build AI, maybe make AI itself unprotected by IP law. That's very unrealistic but so are most anti-capitalist ideas. It also probably won't help much considering capital has the resources to protect their property anyway. But hey at least my idea won't help bring about a future where Disney can sue regular people for making unlicensed art that looks too much like what their proprietary AI generates.
|
# ? Feb 4, 2023 20:58 |
|
your high school's plagiarism rules are not the united states copyright code
|
# ? Feb 4, 2023 20:58 |
|
It is possible to provide attribution to each subpixel and each audio sample. I hope they're forced to.
|
# ? Feb 4, 2023 21:54 |
|
Pixel (0,0) is derived from all 5 billion images in the dataset: [link]. Pixel (0,1) is derived from all 5 billion images in the dataset: [link]. Pixel (0,2) is derived from all 5 billion images in the dataset: [link]. Pixel (0,3) is derived from all 5 billion images in the dataset: [link].
|
# ? Feb 4, 2023 22:04 |
|
Fuschia tude posted:Pixel (0,0) is derived from all 5 billion images in the dataset: [link]. Pixel (0,1) is derived from all 5 billion images in the dataset: [link]. Pixel (0,2) is derived from all 5 billion images in the dataset: [link]. Pixel (0,3) is derived from all 5 billion images in the dataset: [link]. If that's what it takes then go. or threshold contribution.
|
# ? Feb 4, 2023 22:04 |
|
mawarannahr posted:If that's what it takes then go. or threshold contribution. I think it would be easier to just destroy the Internet's capability to transmit images and video. Let's do that instead, unironically. In fact, stop reading my posts, you're stealing my words with your eyes and hippocampus.
|
# ? Feb 4, 2023 22:08 |
|
human beings are granted different rights and enjoy a different legal status than products or commodities hth
|
# ? Feb 4, 2023 22:11 |
|
webcams for christ posted:human beings are granted different rights and enjoy a different legal status than products or commodities hth the ai is a tool and not legally responsible for anything. the human using it is
|
# ? Feb 4, 2023 22:14 |
|
webcams for christ posted:human beings are granted different rights and enjoy a different legal status than products or commodities hth stfu and get a lobotomy already, or you'll be hearing from my lawyers
|
# ? Feb 4, 2023 22:14 |
|
|
# ? Jun 10, 2024 20:05 |
|
Mercrom posted:the ai is a tool and not legally responsible for anything. the human using it is
|
# ? Feb 4, 2023 22:23 |