[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/ic/ - Artwork/Critique


View post   

File: 204 KB, 1078x1705, 169859764230472263.jpg [View same] [iqdb] [saucenao] [google]
6911296 No.6911296 [Reply] [Original]

I think this is the best place to ask thjs:

Is there realistically any way to curb or prevent art style "theft"?

https://twitter.com/Nekojira/status/1633882318787198978?t=LsUvA7hJad-Kq7QdxsaYrg&s=19

>> No.6911298
File: 238 KB, 1073x1348, 169784803302180651.jpg [View same] [iqdb] [saucenao] [google]
6911298

>>6911296
https://twitter.com/Abyou17/status/1633971596825358336?t=yjjSHAJqXSfiEPRhFovo7g&s=19


https://civitai.com/models/5126/torino-aqua-style-lora


It's joever

>> No.6911304

neck yourself faggot.

>> No.6911321

>"theft"
theft

>> No.6911331

>>6911296
>Laws against stealing style
>Disney salivating at claiming every possible style with their warehouse of art

>> No.6911336

>>6911331
>faggot OP
>bumps his bait thread
stop trying to frame it as a style thing

>> No.6911424

not about "style". never was. only sub 80iq retards think it is.

>> No.6911442

>>6911296
right click -> save as =/= theft.

>> No.6911452
File: 1.70 MB, 960x1280, 1686551097969177.png [View same] [iqdb] [saucenao] [google]
6911452

>>6911296
>I feel so painful just by watching this
how am i the only one who'd feel honored / ecstatic to find out someone trained a model with my work
grow up

>> No.6911459

>>6911424
What's it about then?

>> No.6911462

>>6911459
le precious OC

>> No.6911470

>>6911459
work

>> No.6911475

>>6911452
Grown ups protect their intellectual property, they don't let tech companies profit from them by stealing their artwork.

Sign a damn license agreement if you want to use someone's artwork in an image data set to train you image generator, it's not that hard to have ethical business practices.

>> No.6911480

>>6911336
Not OP. Did you sage with your reply?

>> No.6911481

>>6911470
look at this faggot
>>6911475
grown ups dont care about ip
>Sign a damn license agreement
why? its not like im selling any of this. im creating for the sake of creation

>> No.6911490

>>6911481
You are not creating, and neither is the diffusion model.

https://www.youtube.com/watch?v=-MUEXGaxFDA&ab_channel=JimmyMcGee

>> No.6911491

>>6911490
pyw

>> No.6911503

>>6911481
no arguments huh. I accept your concession.

>> No.6911508
File: 227 KB, 776x1119, 1469823061768.jpg [View same] [iqdb] [saucenao] [google]
6911508

>>6911491

>> No.6911509

>>6911503
its not even fun to dunk on you guys anymore, its just sad now

>> No.6911512

>>6911491
>oh no i've been called out, better ask him to pyw
fuck off back to /g/ AItourist

>> No.6911516

>>6911509
you never managed to dunk tripfag-kun

>> No.6911521

>work

>> No.6911529

>>6911508

How did this picture turn into a meme? It's not bad.

>> No.6911530
File: 158 KB, 400x371, 1381659850250.png [View same] [iqdb] [saucenao] [google]
6911530

>>6911296
put the word NIGGER at like 50% opacity into the middle of your artwork

>> No.6911531

>twerk

>> No.6911537

>>6911529
it's extremely generic I guarantee you'd scroll right past it if you saw it on artstation or wherever that's why

>> No.6911538

>>6911470
Elaborate

>> No.6911541

>>6911538
nb4 nebulous statement about computers not having feelings

>> No.6911544

>>6911490
I actually watched that video in its entirety around a week ago and part of the video is full of blatant misinformation. Has no clue what he's talking about and still pedals the "it just copies it can't make anything unique " meme. I genuinely wanted to take him seriously but he's not going to do basic fucking research that he's no less of a lazy gullible nigger than the average degenerate here.

>> No.6911549

>>6911296
What is being stolen? What monetary loss does this cause for the original artist?

>> No.6911550

>>6911490
>>6911544
you know youre doing something right when bro feels the need to make an hour long video "debunking" a piece of tech and he doesnt even understand

>> No.6911551

>>6911538
A work is a creative expression that is fixed in a tangible medium, such as a painting, a photograph, a song, a book, or a computer program. A work can be protected by copyright, which gives the author or owner the exclusive right to control how the work is used.

>> No.6911552

>>6911544
But the video is right, a computer cannot create anything from scratch, it scans through multiple images in its database to compose a frankenstein essentially. Do you even Computers?

>> No.6911556

>>6911551
>creative expression that is fixed in a tangible medium
glad we agree. so which part doesnt apply to generated images?

>> No.6911558

>>6911538
NTA but essentially when a piece of art is made it takes in a lot of the author in mind, how hes feeling, how he sees it, how he strokes it (the pen dumbass), how he shades it, how he styles it, a piece of art is inherently dependent on the author and its intention, the AI does not understand that, hence why most of the pieces it produces are inherently soulless products of a Frankenstein that he put together without understanding, because its mass produced and bland.
In terms that an artlet like >>6911556 this retard can understand, its handmade food by a chef vs mass produced food by a food factory.

>> No.6911561

Artists and art culture are entirely dominated by commie left wing cucks.
What do you think the means of redistribution from the haves to the have nots look like?

>> No.6911563
File: 109 KB, 1032x512, 2085786821.jpg [View same] [iqdb] [saucenao] [google]
6911563

>>6911558
>>6911541
whoah doggy art is in the eye of the beholder amirite? one which only sees muck is himself muck
really early gens do have more soul desu

>> No.6911568

>>6911563
>Paralysis demon

>> No.6911570

>>6911556
depends how the generated images came about

>> No.6911571

why are we doing AI debate again? draw MOAI

>> No.6911572

>>6911571
*Cough* I was forced to do 15 Halloween pieces this month already no more...

>> No.6911573
File: 51 KB, 512x512, 1757441649.jpg [View same] [iqdb] [saucenao] [google]
6911573

>>6911568
>execution of an angel

>> No.6911581

>>6911490
Half this video is just a completely unhinged schizo rant about evil corporations that has jack shit to do with AI. The few actual sensible points he does touch on (ethics of obtaining the training data, algorithmic bias involved in letting an AI make decisions, etc.) are blatantly obvious to anyone who actually works in the field. It's taught at my uni's intro to ML class for instance, and dozens of papers talk about those issues and how to resolve them. He doesn't even really have a call to action or thesis for anything- it's just fatalistic doomershit that doesn't really leave any hope for the future beyond living in an endless corporate dystopia, which is really unlikely to happen given that AI will only improve it's utility to society as it gets regulated more.

>> No.6911587

>>6911581
>>6911550
>>6911544
>>6911491
Hit a nerve huh?

>it's easy to use rhetorical tricks to make "AI" seem intelligent
>like all good cults "AI" promotors use mysteries to sneak their ideas in the door
>we might not know exactly what the hidden layers are doing
>but we know how machine learning works and we know what it produces
>"AI" can pick up on some complicated patterns but it's a far cry from creativity
>by feeding it a lot of images a model can trick you into thinking it's creative
>but it's really just fitting whatever prompt you give it to the most likely output
>researchers have been able to prompt these models to spit out their own training data
>so it's clear the models are not synthesizing something new the way a person could

>> No.6911595

>>6911587
>ai is able to create 1:1 copies of the images it was trained on
>therefore it cannot create anything new
you realize how idiotic this sounds? its barely worth responding at this point

>> No.6911599

>>6911595
and yet, you replied to him. because you're a nodraw tourist fag desperate to bump his fucking bait thread. the subject has been discussed at length and there's no point to make yet another thread.

>> No.6911600

>>6911459
Total AI kike death

>> No.6911606

>>6911296
Being a noname beg whose art nobody cares about

>> No.6911608
File: 565 KB, 672x1440, 1689064089790411.png [View same] [iqdb] [saucenao] [google]
6911608

>> No.6911616

>>6911587
>it's easy to use rhetorical tricks to make "AI" seem intelligent
>we might not know exactly what the hidden layers are doing
>but we know how machine learning works and we know what it produces
Who cares?
>"AI" can pick up on some complicated patterns but it's a far cry from creativity
>by feeding it a lot of images a model can trick you into thinking it's creative
>but it's really just fitting whatever prompt you give it to the most likely output
Isn't this how creativity works? The only difference between this and a human studying human anatomy from thousands of poses is the fact the machine is automating it. The guy in the video goes on this whole segment jerking off Earthbound like it came out of the aether when that game wears its obvious 80's americana influences on it's sleeve.
>researchers have been able to prompt these models to spit out their own training data
>so it's clear the models are not synthesizing something new the way a person could
The paper he cites induces the replication behavior by prompting with captions from the LAION dataset directly, which is unlikely to happen in most real-world applications as end-users don't have access to the training data. This is a well-documented problem and there are various papers addressing solutions to this (remove duplicate image-caption pairs, add a lot of duplicate images with different captions, tweak training duration and dataset size hyperparameters, etc.) There's more in this paper https://arxiv.org/pdf/2305.20086.pdf

>> No.6911618

>>6911595
>>therefore it cannot create anything new
this is true though. Its essentially doing a picture merge.

>> No.6911620 [DELETED] 
File: 1.36 MB, 864x1208, 01064_.png [View same] [iqdb] [saucenao] [google]
6911620

>>6911568
thanks for reminding me about the time i did a whole sleep paralysis demon series

>> No.6911624

>>6911620
you did this anon? how much time did this take?

>> No.6911639 [DELETED] 
File: 1.52 MB, 864x1208, 01047_.png [View same] [iqdb] [saucenao] [google]
6911639

>>6911624
it took about three hours to start getting something interesting but i had been working for a day and a half straight at that point

>> No.6911641

>>6911624
10 seconds to generate. about 1 hour looking for prompts to steal.

>> No.6911654 [DELETED] 
File: 1.44 MB, 864x1208, 01057_.png [View same] [iqdb] [saucenao] [google]
6911654

there are only ~200 images from that session before i found these, i should honestly revisit i can do better now

>> No.6911697

>>6911561
All it takes to make one of them a capitalist is telling them you copied/sold their art without giving them squat.

>> No.6911709

>>6911620
What does this have to do with sleep paralysis? Even with the tools to generate you fail to make actual art

>> No.6911724

>>6911296
You're cute.

>> No.6911733 [DELETED] 
File: 734 KB, 512x744, 00097_.png [View same] [iqdb] [saucenao] [google]
6911733

>>6911709
those are sleep paralysis demons

>> No.6911735
File: 139 KB, 793x884, 135871923.jpg [View same] [iqdb] [saucenao] [google]
6911735

>>6911618
It's not true though, 99.9% of what is made with AI is new.

>> No.6911743

>>6911733
You're just a schizo

>> No.6911748

>>6911735
This is correct, it generates new images, I'm not for ai and the only reason it's as good as it is, is by stealing large datasets of actual artists but it does create new things unless there's overfitting where theres repeats of the same image in a dataset and then it ends up taking more traits copying basically 1to1, but pretending that what it does is "photobashing" is either misinformed or a complete cope.

https://youtu.be/nIRbN52PA0o?si=EOZbEEKdL-VgrOIm&t=482

>> No.6911749

>>6911748
>cope
>says the coping machine learning pajeet

>> No.6911753 [DELETED] 
File: 1.99 MB, 1024x1280, 00003_.png [View same] [iqdb] [saucenao] [google]
6911753

>>6911743
thanks i hope you enjoyed

>> No.6911757

Can AI create artworks that are beyond the quality of its data?

>> No.6911765

>>6911748
advanced probabilistic photobashing is still stealing, nigger thief

>> No.6911767

>>6911757
No. It can only mix and match, never create something by itself.

>> No.6911773 [DELETED] 
File: 2.73 MB, 1536x1840, 02135_.png [View same] [iqdb] [saucenao] [google]
6911773

can anon point me in the direction where it copied this from? id love to find more pls thx

>> No.6911782
File: 1.04 MB, 876x930, 564654.png [View same] [iqdb] [saucenao] [google]
6911782

>>6911773

>> No.6911787 [DELETED] 
File: 2.30 MB, 1024x1768, 01978_.png [View same] [iqdb] [saucenao] [google]
6911787

>>6911782
center and center left are cool! not quite the same unfortunately, hardly a copy

>> No.6911811

>>6911452
>why yes, I enjoy watching other men engage in sexual acts with my wife. It's an honor that other men find her as attractive as I do!

>> No.6911818

>>6911618
Not only if I'm not how it works, you cannot even prove that's not how it works. Meanwhile I can prove to you that it in fact makes you unique pictures.

>> No.6911820

>>6911552
You are correct that it cannot create something holding new because it has to learn Concepts from human created material. You are whoever is still an idiot for suggesting that it merges shit together by constantly scanning a database. You can run stapled accusing without an internet connection and without the original data sets. The safe tensor binary files do not contain pictures. They are just a bunch of mathematical weighs that the computer can understand. You can almost think of it as incredibly advanced Computer code. You know how computers cannot understand human speech bottom cells see we have to have interpreters and programming languages? It's the same way with this kind of. The models do not have pictures saved in them. If it does then prove it.

>> No.6911825
File: 1.63 MB, 960x930, 564655.png [View same] [iqdb] [saucenao] [google]
6911825

>>6911787

>> No.6911834
File: 3.28 MB, 2621x3276, 20220223_193658.jpg [View same] [iqdb] [saucenao] [google]
6911834

>>6911825
i like those a lot, i wish i knew who the artists were - reminds me of zach lieberman

still
not a copy

>> No.6911841

I'm an artist, and I'm on the artist's side on the AI discussion
That being said, if you can only make one fucking face and one haircut at the same angle and the same lighting and the same pose with little variation, and that's "your thing" you should shut the fuck up

>> No.6911842

.

>> No.6911845

>>6911765
It's not photobashing though

>> No.6911846

>>6911841
I fear not the man who has practiced 10,000 kicks once but the man who has practiced one kick 10,000 times

>> No.6911849

>>6911296
don't draw and no one can steal your art.

>> No.6911850 [DELETED] 
File: 174 KB, 1280x720, maxresdefault.jpg [View same] [iqdb] [saucenao] [google]
6911850

>>6911296
destroy this fail country of pajeets

>> No.6911860

>>6911834
Its all in the laion data set. https://rom1504.github.io/clip-retrieval

>> No.6911861

>>6911820
The "mathematical weigth" is a mirror of the original picture and requires the original picture and it's destiled from the original picture, just the same as painting the mona lisa with crayons instead of oleum, or plain taking the painting scan and inverting the colors with a filter.

>> No.6911868

..

>> No.6911871
File: 1021 KB, 2869x1446, workflow.png [View same] [iqdb] [saucenao] [google]
6911871

>Personally drawn works are to be submitted for critiquing. Do not claim authorship of work that you have not personally drawn or contributed to
you should take off the "contributed to" part because thats obviously false

>> No.6911874

>>6911871
if artists were in control of hiroshimoot. yeah, shit would change. read the room though, AI art does not belong.

>> No.6911877
File: 27 KB, 611x611, 169825005705433314.jpg [View same] [iqdb] [saucenao] [google]
6911877

>>6911861
>The "mathematical weigth" is a mirror
They aren't mirrors you can almost think of them as rough estimates of what a certain concept should look like. As people keep telling you, it is incapable of making one-to-one identical copies of images in the data set. It never does that it sucked in extremely rare cases. Give me an example of it doing that since you claim it does that regularly.

https://civitai.com/models/180511/sashakhmel-or-sashakhmelsfw-style

The artists that this lora is trained on has all of their art publicly available on multiple sites (their Twitter page, danbooru, some of it on r34, etc).

Direct me to any images of theirs that look identical to the Showcase images on the civitai website. I'm giving you the opportunity to prove me wrong by something as simple as comparing two images Side by side.

Oh wait you can't because you are full of shit. You love to hear yourself talk. You love to THINK you are an intelligent person when you aren't.

>> No.6911879

>>6911296
No such thing as art style theft

>> No.6911889

>>6911874
>read the room though, AI art does not belong.
sad state of affairs

>> No.6911930

>>6911874
are you speaking on behalf of /ic/?

>> No.6911946

.

>> No.6911954

..

>> No.6911955

...

>> No.6911964

>>6911955
Four dots

>> No.6912114

>>6911964
Five dots

>> No.6912199

Styles do not have copyrights. Quit your bitching.

>> No.6912201
File: 112 KB, 1400x700, 169794437966882969.jpg [View same] [iqdb] [saucenao] [google]
6912201

>>6912114

>> No.6912412

>>6911877
>you ask me to prove you wrong, but also you set the terms of what the probatory evidence should be, and said evidence doesn't contradict my claim
Why would I do that
>and you claim it does that regularly or something
Never made such a claim
I don't know if I'm as smart as I think I am, but I can safely say I'm smarter than you because you are arguing with somebody else

>> No.6912417

>>6911846
That's fine , but if I can replicate your 10k only kick in one try, you don't get to cry about how that is "your kick"

>> No.6912488

>>6912412
You implied it makes varbon copies which it does not, so I accept your concession

>> No.6912751

>>6912488
I didn't implied it's a carbon copy and you'll knew that, if you understand what the "mirror" and the later explanatory "destiled" words means in the english language
The only thing I concede is you being more stupid than me, now with evidence

>> No.6913625
File: 232 KB, 500x553, carl faggot police.png [View same] [iqdb] [saucenao] [google]
6913625

>>6911296
It's called Glaze. They are also looking to incorporate a new program called Nightshade to sabotage the image generates.
I know a lot of the ai fags are shills bc I don't see nearly as meany of them on alt chans. That includes one made by /pol/ fags and one about technology.
OP is a faggot that needs to die.

>> No.6914345

>>6912751
What do you think "mirror" means?

>> No.6914347

>>6913625
>snake_oil_shill.txt