Why It's Morally Okay To Steal A.I. Artwork (The Jimquisition)
Added 2024-03-11 13:26:19 +0000 UTCA.I. advocates love to pretend they're real artists, and claims of plagiarism are little more than appropriative attempts to steal a sense of validity. Let's look at how these hacks justify themselves and see how little techbros understand the concept of hard work.
Comments
While the person you talked to is correct in what they said, the example is (unintentionally) beside the point. It addresses a common but *completely false* idea about how generative AI works. It's the naïve imagination of the computer acting as a "collage machine", which stores copies (either full or in parts) of the training images and then reassembles bits and pieces of them into the output images. This is completely wrong and bears no resemblance whatsoever to what actually happens inside an AI model. No part of any of the training images is stored anywhere inside the resulting neural network. The AI won't — and, in fact, CAN'T — reproduce a copy of any of those images from memory, not even when deliberately prompted to do so. The entire point of training the network on hundreds of thousands of publicly posted images is to teach it *abstract ideas* like "This is what a pencil line looks like", "This is the look-and-feel of an oil painting", "This is a common pose in portraits", and so on. That's the reason why OpenAI and Stability AI are deliberately not seeking dismissal of the central allegations of copyright infringement in their respective court cases. All the other points of the suit got dismissed, but *this* part they want to take to court, in order to set a precedent. Once the dismissal phase is over, and discovery begins, they'll just back a dump truck up to the court house and enter a *literal metric ton* of scientific papers into evidence that proves beyond even a shadow of reasonable doubt that the simplistic imagination of a "collage machine" is 100% false. The real villains in those cases are the sleazy lawyers taking advantage of the aggrieved artists. Instead of telling them, "That's not how copyright works. That's not how machine learning works. That's not how *anything* works. Attempting this lawsuit has no chance of success and will be a waste of our time and your money," these guys told their clients, "Suuuure. We'll go with the 'throw any old shit against the wall and see what sticks' approach; no problem! Here's our bill for writing up your list of ludicrous, unsupported, wild accusations and submitting them to the court. Ooops, they got dismissed. Here's out bill for amending them and submitting them again. Oh dear, the amended complaints were dismissed, too. Who could have foreseen that (besides any competent lawyer)? Expect our bill for amending them *again* real soon."
Sebastian Weinberg
2024-03-13 21:24:38 +0000 UTCI just had a conversation about this topic with someone who works in a theater and also has quite an extensive understanding of art in general and it was interesting to hear their explanation, when I came with the arguments of this video. So idk, I thought I'd share this, bc I really love Stefanie, watched Jimquisition and other videos for a long time (was actually introduced to it by that person I talked to about this topic). They told me that an essential problem is in the way we think about art and copyright and ownership. That transformative action can be very different, and that we have a very "old" understanding of what art typically is or was. But he also said that for example if i went into an art museum, took pictures of every painting, put those together in a collage of sorts, even without really thinking about it, that could stil be called art, so why exactly shouldn't AI-generated art be called that? I guess it would be great if artists who's art got used to train AI could get something out of it, and not only that, but in general that ppl get something out if AI uses their data in any way. But it is still a kind of transformative art form and art doesn't have to inherently have any real meaning, reason and it certainly doesn't have to involve a lot of effort or work. In theory a lot of things can be called art, and we just typically use excuses like that to discern "real art" from non- or bad art... It was certainly interesting to hear that perspective, especially from someone in that industry. So my question would be (and I hope it's ok to ask) to Stefanie: What do you think makes AI-Art so problematic? Like, is it "only" bc it's not "real art"? And then I would ask, does that mean if I bake a cake, but use a machine to mix the dough that would be no real cake in the end, bc I took an easier way? Can I only make real art if I have studied something for a long time or have a degree or deep knowledge of something? Wouldn't that mean a lot of what we today consider art wouldn't actually be art? That was actually a lot of questions... Idk, I plan to work in IT, so I do think this is a very important topic, I think there are things to be discussed and of course a lot of things to be weary about when it comes to AI in general. I just wanna kind of understand what the actual problem is, instead of talking about something completely different and missing the point, you know? Ok, sry for the tangent, I just really felt like I wanted to engage with this more for once ^^* Have a great day anyone
Burgerism
2024-03-13 15:30:36 +0000 UTCNormally I rush to watch your videos every Monday, as soon as they are posted. But when I saw the title, I stopped, because I was afraid of being sadly disappointed. Was this going to be a cogent exploration of the legal reasons why fully AI-generated material should not be copyrightable, in order to prevent SEO-like gaming of the broken copyright system, by the same people who are now patent trolls and domain name squatters? Or was it going to be yet another ill-informed hate-screed, based on nothing more than reflexive, fearful luddism, jealous gatekeeping, and mis-directed anger? It took me two days to finally watch it, and needless to say, OF COURSE I was disappointed by a video that offered no new insights, no interesting perspective, and very, VERY clearly not an ounce of research into the real dangers posed by the unchecked proliferation of generative AI. You were content to preach to the choir, getting plaudits for simply echoing the conventional wisdom of "That's not REAL art!" Well, no shit, Sherlock! Way to get that low-hanging fruit! Congratulations on a take so lazy and aggressively _mid_ it could have come from a large language model. The fight against the corporate mis-use of generative AI is going to be one of the most important battles of our generation. Wouldn't you rather contribute to THAT fight instead of playing gatekeeper against some poor schmucks who pay $10 a month to Midjourney, so they can see their ideas turned into pretty pictures, and are unreasonably proud of their "creative" efforts? Is THAT the appropriate target for your hatred? Doesn't it bother you at all that your rhetoric is entirely identical to the Real Gamers™ railing against easy modes and "filthy casuals"? You had no problem mocking them for their elitist attitude, but now that there's an "easy mode" for something YOU take pride in, suddenly that same attitude is A-OK! What's worse, your arguments regarding generative AI "stealing art" or "violating copyright" reveal that you haven't made even the slightest effort to learn how AI training actually works. Right now, you are the equivalent of a creationist shouting "Evolution is false, because I never saw a cat spontaneously turn into a dog!", making it obvious they have no clue what evolution is or how it works, yet expecting to have their argument taken serious, and be debated by people who actually know what they're talking about. How are you expecting to successfully argue against generative AI, if you won't even make a token effort to learn how it works?
Sebastian Weinberg
2024-03-13 08:41:11 +0000 UTC
