|
SCheeseman posted:I was paraphrasing a dumbass response I got when described human memory of visual data and AI art generation tools as conceptually similar. Apparently this means I think the brain runs 7zip. Except the "unless it does idk" I added that because I was afraid it was gonna be taken seriously, didn't work rip all good friend, I'll direct my condescending ire toward stux
|
# ? Sep 20, 2022 01:48 |
|
|
# ? May 29, 2024 23:10 |
|
|
# ? Sep 20, 2022 02:06 |
|
What was the prompt for this one
|
# ? Sep 20, 2022 02:09 |
|
New York Times Best Seller "Mal Molt Dosta Dood Doolth", or "storybook about a baby seal disguised as a loaf of bread that only eats Doritos and drinks Mountain Dew in a magical forest of smiling happy rat families" Except for the hilarious titles, Stable Diffusion does a better job IMO: Mercury_Storm fucked around with this message at 03:38 on Sep 20, 2022 |
# ? Sep 20, 2022 02:31 |
|
SCheeseman posted:I was paraphrasing a dumbass response I got when I described human memory of visual data and AI art generation tools as conceptually similar. Apparently this means I think the brain runs 7zip. Except the "unless it does idk" I added that because I was afraid it was gonna be taken seriously, didn't work rip This is exactly how your brain works. you don't store an mpeg, when you remember things you are imagining the past
|
# ? Sep 20, 2022 04:42 |
|
Hey is it art now?
|
# ? Sep 20, 2022 05:54 |
|
I'm tempted to start making art installation pieces based on prompts.
|
# ? Sep 20, 2022 06:05 |
|
BARONS CYBER SKULL posted:Hey is it art now? WHERE, I ask, is the S O U L ???
|
# ? Sep 20, 2022 06:34 |
|
sometimes you just get 5 perfect images: "the death of modern art, pencil sketch by R. Crumb"
|
# ? Sep 20, 2022 06:35 |
|
BARONS CYBER SKULL posted:Hey is it art now? nothing is art
|
# ? Sep 20, 2022 06:47 |
|
Posting is art.
|
# ? Sep 20, 2022 07:04 |
|
Hadlock posted:What was the prompt for this one I got a bunch of fun results with this, generated for a forum game, and though only the last one was actually used, some of the others are too cool not to share. "retro sci-fi hazmat suit, high fashion. retro futurism. vogue magazine coverpage." "hazmat suit. avant gaurd ffashion designed by Zdislaw beksinski and giger" (typo included) Retro sci fi Hazmat diver suit. Underwater. High-end fashion magazine. Highly detailed sharp focus face. [GEN]. cover art by anderson ,arnold armitage, loish, thomas kinkade (final, edited very with photoshop to ad a logo and fix the hair)
|
# ? Sep 20, 2022 07:35 |
|
Pay me for reading my post. Its art.
|
# ? Sep 20, 2022 07:47 |
|
I mean, I wouldn't mind if Open AI/Midjourney/etc had to pay out like 10% of income to (in-copyright) artists used in prompts proportionally to how often they're invoked. Imo, Rutkowski should be compensated for inadvertently turning out the secret ingredient of fantasy AI art.
|
# ? Sep 20, 2022 07:51 |
|
Burning Rain posted:I mean, I wouldn't mind if Open AI/Midjourney/etc had to pay out like 10% of income to (in-copyright) artists used in prompts proportionally to how often they're invoked. Imo, Rutkowski should be compensated for inadvertently turning out the secret ingredient of fantasy AI art. ok but you have to pay me for my prompts then
|
# ? Sep 20, 2022 07:55 |
|
Yeah absolutely, they should respect the copyright to Greg Rutkowski's style. One sec, I'm getting a phone call from a lawyer...
|
# ? Sep 20, 2022 08:12 |
|
yeah, i'm not saying any of that
|
# ? Sep 20, 2022 08:15 |
|
I’m gonna steal Greg’s hands
|
# ? Sep 20, 2022 08:26 |
|
What does Greg Rutkowski own the copyright too? Because the concept "You can not learn from my images without paying me," is definitely a wild one.
|
# ? Sep 20, 2022 08:32 |
|
Also most copyright isn’t even owned by the artist at this point. In practice the freelancer on deviantart would get 3 cents over their lifetime and you’d open yet another money tap for Disney, Sony and a bunch of obscure Luxemburgish holding companies Also https://copyrightalliance.org/education/copyright-law-explained/limitations-on-a-copyright-owners-rights/fair-use-exceptions-copyright/ quote:The fair use exception permits a party to use a work without the copyright owner’s permission and without compensating the copyright owner for such use in certain circumstances. The copyright law identifies certain types of uses, including criticism, comment, news reporting, teaching, scholarship, and research as examples of activities that may qualify as a fair use. “Ah but training an AI is not actually teaching, scholarship or research”. Idk, we’ll see. And if so, ok then we should change the law to allow it. It’s not hard, but I know a few multinational corporations that will lobby against it every inch of the way
|
# ? Sep 20, 2022 08:49 |
|
i'm not saying there is or should be any enforcable copyright in using the images to train up the software. just saying it would be swell if corps would share part of the income with the artists who's work are evidently giving their tools their appeal with customers.
|
# ? Sep 20, 2022 08:52 |
|
Deltasquid posted:Also most copyright isn’t even owned by the artist at this point. In practice the freelancer on deviantart would get 3 cents over their lifetime and you’d open yet another money tap for Disney, Sony and a bunch of obscure Luxemburgish holding companies Training an AI is already fair use according to US court precedent. See google vs authors guild, or the turnitin.com lawsuit. That said, Japan has already added an explicit carveout for training machine learning models.
|
# ? Sep 20, 2022 08:55 |
|
In theory yes, but in practice there’s no real way of knowing which style the consumer wants to emulate, and styles weren’t protected before and I doubt they will be in the future. That’s kind of like wishing that a music-generated AI would share, for example, income with Sabaton because people make music “in the style of Sabaton” because the consumer is insufficiently educated to know “Swedish ballad power metal” will yield similar results. Edit: re compensating Greg Rutkowski
|
# ? Sep 20, 2022 08:57 |
|
I can see a situation where AI models become like cell lines used in biology research, there's a bunch of named cell lines that seem to replicate forever, they have cell lines for testing liver drugs, cell lines for testing heart medications etc etc. Sometimes lines are created as branches from another well known line There's obviously the dall-e/midjourney/stable diffusion lines now There is now the "waifu" line from the stable diffusion 1.2; I'm sure someone will take the stable diffusion 1.4 and apply a much larger "waifu" data set to it Somebody else is going to take the future 1.4 waifu model and they'll append a celeberty photography database to it. Then a couple months later someone will append to the wifu-celeberty model the national geographic photography database to it, and share it on bit torrent. And then someone will train 10,000 images of horses and re-share that, etc etc. Someone will finally train the updated horse ai on a couple thousand images of bugs bunny and share that via bit torrent. ad nauseum. An old copy of the getty images photography database (With tags) will get leaked and we'll train against that. Doesn't really matter whose images it's been trained on at this point, it's free and you can train new images to the existing model, probably indefinitely. Unless you give the exact prompt with the image, it'll be (I presume) impossible for someone like disney or fox to go after you for using an AI trained on xyz copyrighted materials to generate your work
|
# ? Sep 20, 2022 09:01 |
|
Deltasquid posted:In theory yes, but in practice there’s no real way of knowing which style the consumer wants to emulate, and styles weren’t protected before and I doubt they will be in the future. That’s kind of like wishing that a music-generated AI would share, for example, income with Sabaton because people make music “in the style of Sabaton” because the consumer is insufficiently educated to know “Swedish ballad power metal” will yield similar results. yeah, I know it's extremely unlikely - it's not like Spotify where you play specific songs by specific artist even when playing "Sabaton radio". But precisely because it would benefit Disney/etc more than random artists, they might get interested in a potential stream of revenue - but more likely by building their own branch of heavily restricted model.
|
# ? Sep 20, 2022 09:11 |
|
Personally I’m just happy something like stablediffusion exists. Much like how gimp and paint.net allow digital artists to use open source alternatives to things like Photoshop, this will open the door to adoption by all artists instead of reserving AI art to an elite class. And training those datasets without fear for injunctions because a single frame of steamboat willy was used in their training, and subsequently keeping those datasets open source, uncopyrighted/uncopyrighteable, is an important aspect of that. All the hand-wringing about the medium feels a bit like how photography was supposedly going to kill the visual arts. “Real” Art will adapt (probably dive headlong into the money laundering business) and “not real” art (i.e. pop art, including, yes, a bunch of deviantart sanic OC’s and people making collages of their favourite anime characters kissing) will flourish. And the vast majority of consumers will give AI art a whirl and crap out the equivalent of bathroom stall selfies while artists who dedicate time to master the medium will be more distinguishable fine art photography, by comparison.
|
# ? Sep 20, 2022 09:26 |
|
Burning Rain posted:i'm not saying there is or should be any enforcable copyright in using the images to train up the software. just saying it would be swell if corps would share part of the income with the artists who's work are evidently giving their tools their appeal with customers. we should make sure everyone in society can live with dignity. but that doesn't really have anything to do with AI art imho
|
# ? Sep 20, 2022 09:41 |
|
Hadlock posted:I can see a situation where AI models become like cell lines used in biology research, there's a bunch of named cell lines that seem to replicate forever, they have cell lines for testing liver drugs, cell lines for testing heart medications etc etc. Sometimes lines are created as branches from another well known line ever read that short story about the first man to upload his brain? he was useful for simulated slavery becuase his default state was ignorant of the possibilities of what could be inflicted on simulated brains. his brain pattern got copied and used for e-slavery by billions of people over thousands of years. whenever he become non compliant they just reset him to believing he was uploaded yesterday
|
# ? Sep 20, 2022 09:45 |
|
Rutibex posted:we should make sure everyone in society can live with dignity. but that doesn't really have anything to do with AI art imho Show me an artist with dignity, I know I don’t have any
|
# ? Sep 20, 2022 09:50 |
|
Rutibex posted:ever read that short story about the first man to upload his brain? he was useful for simulated slavery becuase his default state was ignorant of the possibilities of what could be inflicted on simulated brains. his brain pattern got copied and used for e-slavery by billions of people over thousands of years. whenever he become non compliant they just reset him to believing he was uploaded yesterday I have not, that sounds very interesting I was thinking earlier, the european version of NASA announced their new space shuttle and I wanted to use stable diffisuion to do a montage with the american space shuttle, and I realized because of the datum the model doesn't contain anything newer than ~July 30, 2022 so I can never do that since the model doesn't know what future (real) technology looks like, unless you start off as a base image. It also doesn't know what, for example, the 2024 ford mustang looks like, and I'm sure there will be other problems over time as it becomes increasingly outdated
|
# ? Sep 20, 2022 10:16 |
|
It looks like a 2022 Mustang. But I think the models will get updated and retrained every so often. The $500k or whatever it cost to train SD isn't that much in the grand scheme of things.
|
# ? Sep 20, 2022 10:18 |
|
BARONS CYBER SKULL posted:With a good enough prompt structure, you can turn it into anything… these have come out good. what kind of hardware are you running it on?
|
# ? Sep 20, 2022 12:35 |
|
Mad props to anyone who can guess the prompt for these: hint: Same prompt in Craiyon axolotl farmer fucked around with this message at 13:23 on Sep 20, 2022 |
# ? Sep 20, 2022 13:06 |
|
echinopsis posted:these have come out good. what kind of hardware are you running it on? just a 2080 super, hardware doesn't really change much other than: how you set it up, and how long it takes to render Moongrave fucked around with this message at 13:46 on Sep 20, 2022 |
# ? Sep 20, 2022 13:43 |
|
BARONS CYBER SKULL posted:just a 2080 super, hardware doesn't really change much other than: how you set it up, and how long it takes to render right.. I'm running out of vram going above 512x512, on a 8gb video card. but tbh I haven't really tried to find a solution yet. maybe need to gently caress around a bit more
|
# ? Sep 20, 2022 13:54 |
|
echinopsis posted:right.. I'm running out of vram going above 512x512, on a 8gb video card. but tbh I haven't really tried to find a solution yet. maybe need to gently caress around a bit more
|
# ? Sep 20, 2022 13:58 |
Imagine four balls on the edge to a cliff...
|
|
# ? Sep 20, 2022 14:26 |
|
Sedgr posted:Imagine four balls on the edge to a cliff... How many dimensions is the latent space containing the cliff?
|
# ? Sep 20, 2022 14:29 |
|
TheGoonspiracist posted:I'm tempted to start making art installation pieces based on prompts. these legit look like modern art installations. artists are about to get turbo lazy when they start using AI to guide their projects.
|
# ? Sep 20, 2022 15:25 |
|
|
# ? May 29, 2024 23:10 |
|
Sedgr posted:Imagine four balls on the edge to a cliff... Ding ding ding!
|
# ? Sep 20, 2022 15:42 |