[ 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: 117 KB, 1600x899, 08BASELCVR.jpg [View same] [iqdb] [saucenao] [google]
6909341 No.6909341 [Reply] [Original]

When AI lets anyone create art in seconds, how do people stand out?

What sets great art apart?

>> No.6909345

Good marketing

>> No.6909349

>>6909341
trad
/thread

>> No.6909354

>>6909341
There are no "stand out" AI artworks.
I don't think most people need to do much different to stand out.

One way you can stand out is drawing a bit more unpredictably, don't tighten everything, let your human mistakes show.

>> No.6909356

>>6909341
when every whore has an onlyfans, how do you stand out?

>> No.6909360

>>6909356
Duct-tape a banana to my butt

>> No.6909380

>>6909349
but what sets great traditional art apart? most of it happens to be beg trash, and the rest is pretentious masturbatory shit

>> No.6909388

>>6909341
Do what it can't do.

>> No.6909410

>>6909388
>It can't do anything but blobs
>It can't draw bodies
>It can't draw faces
>It can't draw hands
You are here
>It can't draw yotsuba

>> No.6909413
File: 138 KB, 600x1029, badinkstudios_screenshot.jpg [View same] [iqdb] [saucenao] [google]
6909413

>>6909380
>but what sets great traditional art apart? most of it happens to be beg trash
Well of course it can't be beg trash (and have you seen most digital art?).
With (good) trad there's a level of craft and attention to detail rarely seen in digital. It can be beautiful but also 'raw', lacking the polished rendering found in typical photoshop work (that veneer so easily replicated by AI). It often more readily reflects the hand and personality of the individual artist.
On a practical level trad produces an original piece of art that can be sold. Also there's the 'wow' factor, just go to any YT video with a good trad artist and read the oohs & aahs in the comments sections. I've seen these reactions far too often in real life to assume it's just politeness. Normies often don't see digital as 'real art' (and I've even seen them conflate AI with digital, thinking it was always done that way), saving their respect for traditional (not saying it's right, but it is what it is).

>> No.6909414
File: 807 KB, 1269x1016, 1.jpg [View same] [iqdb] [saucenao] [google]
6909414

>>6909341
prompt what you’re passionate about

>> No.6909415

>>6909341
All the things that also make regular art stand out, there's no Nvidia card or technical skills that can give you a sense of taste or aesthetic or mood (sure, you can study them, but it takes very long to not seem like the artistic equivalent of babys first fanfiction). There's a reason a lot people with "amazing" skills cannot get a following or at the very least make some friends

Sure, AI gives you more shit to throw at the wall, but if you don't know what's good in the first place how do you pick things? Think of the elf anime big tiddy painting guy. You can say normies eat whatever slop you give them, but not everybody is interested in lowest common denominator limitations (or has enough resouces to not just be mediocre shit)

>> No.6909425

>>6909414
I've been so zen on AI since most people are using it for low level uncanny coom and memes. Art be democratized, lol.

>> No.6909463

>>6909341
You have to pay for the latest version. Those with newer and bigger AIs can now win the art speedrun.

>> No.6909495

>>6909410
There's a ton of shit it hasn't been able to do since day 1 and likely will never be able to do but you keep coping about it over here. Just go back to prompting bro.

>> No.6909513

>>6909495
>will never be able to do
literally everybody agrees that machines will become smarter than any human. what's precisely that unique shit they will never be able to do?

>> No.6909538

>>6909341
You don't. Earth is basically an AI retirement home. Let's hope our overlords supply free tendies and wi-fi.

>> No.6909545

>>6909495
>There's a ton of shit it hasn't been able to do since day 1
I think focusing on the technical abilities of AI is the wrong way to go as a critique. It will obviously improve, as it has in the last year. What it will be unable to do is replicate the particular perspective and vision of an artist, the unique symbolism and idiosyncratic worldbuilding the greatest artists do (think William Blake, Junji Ito, Robert Crumb).
>>6909513
>literally everybody agrees that machines will become smarter than any human.
No doubt, but mere IQ and technical knowledge will not replace the more holistic perspective of humans (where wisdom & dare I say, spirituality resides).
>what's precisely that unique shit they will never be able to do?
Precisely the unique shit, where the biographical and experiential exists. AI could never replace Robert Crumb, William Faulkner, James Joyce, Stanley Kubrick, William Blake, Junji Ito, Charles Bukowski, Henry Darger, Inio Asano, Kentaro Miura, Alejandro Jodorowsky, David Lynch ... the list goes on and on.
Sure, AI could IMITATE them stylistically, but art is more than just 'style,' it is a phenomenology.

>> No.6909565

>>6909545
>(where wisdom & dare I say, spirituality resides).
can you be any more bluepilled.

>> No.6909568

>>6909545
Yeah you have absolutely no idea what you're talking about. Back to proompting.

>> No.6909571

>>6909513
Who said anything about intelligence? We're talking about art you tard

>> No.6909573

>>6909565
>can you be any more bluepilled.
Wisdom & spirituality bluepilled? Bugman take.
>>6909568
>Yeah you have absolutely no idea what you're talking about.
Convincing.

>> No.6909588

>>6909341
AI art usually has the same look every time. At least the animu ones I've seen. It can't be compared to an artstyle with a very strong visual identity.

>> No.6909598

>>6909545
Abstract stuff like gesture and cartooning are actually impossible for diffusion just by the parameters of how it actually works.

>>6909568
Retard

>> No.6909644

>>6909598
>impossible
You mean, in the sense that it is not "moving a plotter" around to create rhythmic lines, or that it's impossible to mimic the look of gesture drawing?

>> No.6909652

>>6909341
>When AI lets anyone create art
It's not art, it's just imagery.

>> No.6909682
File: 73 KB, 1188x523, this kills the ai bro.png [View same] [iqdb] [saucenao] [google]
6909682

>>6909341
https://www.youtube.com/watch?v=mTBCzH1UyNY
https://www.youtube.com/watch?v=5sNvD8ePFHs

>> No.6909726
File: 653 KB, 1710x1938, AI burnout.jpg [View same] [iqdb] [saucenao] [google]
6909726

>>6909682
>picrel
really all those arguments I remember from the first few months. They were throwing anything against the wall: both right and left coded 'arguments' ("based" gloating, cringe victimhood, and the dumbass democratization argument).
Now they're entering the burnout phase, and the more honest among them admit they've overdosed on their own supply. Seems an endless flood of vacuous imagery is ... boring ... and not really art at all.
Makes some okay memes, though.

>> No.6909762

>>6909380
>what sets great traditional art apart?

The fact that it can and does exist. Great AI art doesn't exist and never can.

>> No.6909763

>>6909545
stfu tripfag and neck yourself. take your bait thread with you

>> No.6909773

>>6909341
>When anyone can take a photo, how will photographers stand out?
Hasn't been a problem for them, why should it be a problem for us?

>> No.6909811

>>6909341
Name one standout AI artist. Or a standout piece of AI art.
Stop opening Google in a new tab.
Name one yourself right now, off the top of your head.

>> No.6909846
File: 1.49 MB, 1024x1024, hU3c2Cxz8DqKaAKqOHIVu.png [View same] [iqdb] [saucenao] [google]
6909846

>>6909811
me

>> No.6910019

>>6909811
>
>>6909846
HIM.

>> No.6910029

>>6909644
It's not something you can code, dude.

>> No.6910053

>>6909341
okay, but what's the deal with the taped banana?

>> No.6910095
File: 376 KB, 1080x1377, Transart.jpg [View same] [iqdb] [saucenao] [google]
6910095

>>6909726
"Tranny regrets transition" phrase

>> No.6910152

>>6909341
Having control of what they draw. If you even want an "AI" (it's just an interpolation program that mutates we had this shit since the turn of the century) to make something semi-consistently, you need to have a longass convoluted script for it to read from and multiple trials. God forbid you want it to make a video. Meanwhile, if YOU want to make something, you just fucking do it. Sure, it takes a long time, a lot of patience, and hard work, but it is YOU who are in control, not a program which doesn't know what it's doing.

>> No.6910173
File: 218 KB, 1024x1024, 1698523197561245.jpg [View same] [iqdb] [saucenao] [google]
6910173

>>6909811
anon

>> No.6910177

>>6910053
made realize everyone that professional art and artists are just a money laundering scheme
https://en.wikipedia.org/wiki/Comedian_(artwork)

>> No.6910178

>>6910173
>anon
That's me! But I didn't make pic real, nor do I think it is a standout, it's just an anime girl with a pepsi jacket?

>> No.6910179

>>6909341
AI (in so far as you describe it) is not a tool for artists, it's a tool for commissioners to replace artists.
Always remember that.

>> No.6910209

>>6909846
What the actual fuck am I looking at? Some fucking knee nipple shit two right arms and some flesh tumor on her right? How the fuck do you not notice this shit it's like a mangled pile of limbs with no clear outline?
>>6910173
Generic woman in pepsi outfit in a somewhat generic artstyle. I don't get what's so memorable.
>Didn't even post 'caust even doe it'd be a free win just by shock value itself
If this is all AIfags need to be impressed, then the whole fucking movement is stuck in the mud. It won't be until a leading follower with a large enough community are incredibly critical of their program's artwork even when it's "GoOd EnOuGh!!!" for it to get better.

>> No.6910220

>>6910209
ok now post your art

>> No.6910222
File: 103 KB, 1024x1001, maurizio_cattelan_La_Nona_Ora.jpg [View same] [iqdb] [saucenao] [google]
6910222

>>6910053
I always found it hilarious that Maurizio Cattelan finally got mainstream attention from the banana taped to the wall rather than any of his masterful sculptures, which I guess proved the exact point that he was making with the piece.

>> No.6910224

>>6910209
/beg/ detected

>> No.6910241

>>6910220
>>6910224
>Huh, you call my taste shit? WELL I better see you post something that rivals Michelangelo's masterpieces for your opinion to even mater! Even then I'd probably call it shit!
The point I made was that more and more people are getting complacent because it's "good enough" so AI is stagnating. Not that pro artists should be the ones leading AI art, but people genuinely passionate about both AI and art no matter the skill level instead of lazy coomers who are impressed with generic waifu #13270981. It doesn't take a biologist to know what shit smells like.

>> No.6910248

>>6910222
damn. maybe he should duct-taped the pope to a wall.

>> No.6910249

>>6910241
^^didn’t read lol

>> No.6910250

^Of course you didn't read you illiterate retard

>> No.6910253
File: 30 KB, 411x550, 23_001.jpg [View same] [iqdb] [saucenao] [google]
6910253

>>6910248
haha wtf he's way ahead of me

>> No.6910504
File: 528 KB, 2047x1996, F9iBqslXsAAY9D4.jpg [View same] [iqdb] [saucenao] [google]
6910504

>>6909341
Nobody knows what “great” art is. We’ll have opinions now, change our minds in a decade, and on and on.

>> No.6910642

>>6909811
this guy >>6910504

>> No.6910645

>>6910222
>masterful sculptures
Honestly? Bland as hell, even if it has skill in the medium, it lacks any sort of impact, even as just an image of someone being crushed under a rock.
...I guess it true what they say, it's better to get a bad reaction, rather than no reaction at all.

>> No.6910648

>>6909811
All the replies to you so far have proven your point more or less (though several of them are clearly jokes). Not one of them list an actual artist or name.

Suppose I'll do it (sorta) since you asked, not that I like AI;
Who's that guy who makes porn images who's constantly accused of using AI and tracing, but swears that he doesn't (but absolutely does)? He's pretty famous, though it could be argued not entirely for his art, and more so for his reputation and the fact that it's porn.
See, I actually listed one, and I didn't even look it up in google, just like you asked.

>> No.6910692
File: 258 KB, 1024x1024, so.jpg [View same] [iqdb] [saucenao] [google]
6910692

AI resolution is too small and the results are full of errors. It's nor ready for prime time yet.

>> No.6910695

>>6910648
So he's famous because of people calling him out on blatantly lying and scamming them? That's supposed to be a good example?

>> No.6910829

>>6910692
this, for now
>>6910179
ai will need a human to fix shit for at least a decade.
after that, yeah: pretty much the same people that use to hire or supervise drawing monkeys now will do everything by themselves

>> No.6910831

>>6909341
Process, backstory and actual product that can be touched physically. (Digital "art" btfo)

>> No.6910834

>>6910692
>AI resolution is too small
You can generate 8k now.

>> No.6910842

>>6910831
it's called a "printer". should try one some day

>> No.6910869

>>6909811
>>6910648
The actual joke and point of the "ai art" is ze nwo (basically supegomunissm). Suddenly phrase% You will own nothing and be happy " hits completely different when it's finally starting to crawl in to ze reality. It's all part of the ze plan to make human cattle as miserable as possible. Read on Agenda 2030. It's happening. Not even doomposting or shitposting my dear Anons. Skynet is here to stay. >>6909341

>> No.6910887

>>6910695
>blatantly lying and scamming them? That's supposed to be a good example?
Seems like the average AI artist, so yes.

>> No.6910894

>>6910834
no you can’t, upscaling isn’t generating and the fewer pixels there are at gen time, the more artifacts there are

>> No.6911060

>>6910648
I think you're thinking of Shexyo? Dude is basically a joke lol

>> No.6911262
File: 175 KB, 1920x2025, notoaiart.jpg [View same] [iqdb] [saucenao] [google]
6911262

>>6909341
the death of the boring, tasteless human drawing machines is a good thing
real ART always has been about creativity and ideas, not investing over 9000 hours in autistic rendering

>> No.6911268
File: 49 KB, 489x576, 1697236172695520.png [View same] [iqdb] [saucenao] [google]
6911268

>>6911262
>muh idea fags in action

>> No.6911327

>>6911262
>ART always has been about creativity and ideas
But is it? Most /ic/ seems to agree that art is just about the composition, or even the render craftsmanship, regardless of the lack of intent, passion, or human emotions whatsoever

>> No.6911329

>>6911268
who is this four chan?

>> No.6911382

>>6911262
A beautiful effortless looking drawing of a hand by a human is not a creative idea, but is real art and worthy of praise.

>> No.6911389

>>6911382
is it worthy of praise just because of the effort, like a participation award? or is there something concrete that makes it more valuable as an object of beauty?

>> No.6911390

>>6909341
>What sets great art apart?
the gameplay

>> No.6911391

>>6911389
yes

>> No.6911579

>>6911262
Your grammar is slipping sirs

>> No.6911723
File: 311 KB, 2048x1640, 85src4rg0gu41.jpg [View same] [iqdb] [saucenao] [google]
6911723

>>6909341
AI is cute, but writing a prompt is not a proper way of creating visual art.
try again when AI is capable of following visual instructions.

>> No.6911727
File: 1.58 MB, 1165x1205, teaser.png [View same] [iqdb] [saucenao] [google]
6911727

>>6911723
I mean, we are not far off from that...

>> No.6911747

>>6911727
whenever photoshop adds this feature, drawing monkeys are toast

>> No.6911798

>>6911747
>2 more features

>> No.6911805
File: 2.73 MB, 1536x1840, 02135_.png [View same] [iqdb] [saucenao] [google]
6911805

>>6909341
idunno sorry

>> No.6911813

>>6909410
>>6911798
/ic/ will never accept defeat, the goalpost moving is eternal.

>> No.6911832

>>6911813
It's a drawing board you stupid nigger.

>> No.6911836

>>6911813
what does it even mean to accept defeat?

>> No.6911866

>waiting for anon to post pics of the work my ai art is "copying"

>> No.6911872

>>6911866
You were provided some sources. it quite literally used those images in it's process.

>> No.6911875

>>6911872
and the final image looked nothing like the images the model was trained on. how would you even know if the prompt i used grabbed tokens from those pieces anyway?

>> No.6911878

>>6911875
>grabbed tokens from those pieces
it's not that.
The whole process to make an AI image includes the training. you can't separate that.

>> No.6911882

>>6911878
>it's not that.
what is it then?
are you saying "based on" or "inspired by" = "copy"? i thought copy implies... them looking the same

>> No.6911894

>>6911060
Yeah, that's the one. Not sure if I could have found his name with just my description anyway, even if I could have usde google.

>>6911262
>not investing over 9000 hours in autistic rendering
The reason everyone thinks AI is so good is because it imitates those autistically rendered image, you autist. If the AI couldn't render the light shimmering on the goop dripping out of the gaping holes you guys are all proompting, you idiots wouldn't nearly be so keen on it.

>>6911805
Is this also AI? I will admit, I kinda dig it, though it is a bit too nonsensical and abstract, it'd be nice if everything made sense, and felt in place, like the two figure statues do,

>> No.6911895

>>6911882
>I copied but
is still using copying in the process

>> No.6911899
File: 372 KB, 1024x1024, screaming.jpg [View same] [iqdb] [saucenao] [google]
6911899

>this one is very interpretable in the same way human art is
lel they're finally learning

>> No.6911904
File: 170 KB, 1004x1169, FUuooRaX0AM183D.jpg [View same] [iqdb] [saucenao] [google]
6911904

Good morning, sirs! ai slop is gonna replace artists any day now

>> No.6911909

>>6911899
"d u7c5 ebo76r80fred747856sdxs680chandbrsfa"
interpret my prompt. it's art.

>> No.6911911
File: 126 KB, 1024x1024, 7e829e7768414ab9a0d8e82eea6a5255.jpg [View same] [iqdb] [saucenao] [google]
6911911

AI sux lol

>> No.6911913

>>6911909
better than a duct taped banana any day

>> No.6911917
File: 109 KB, 1032x512, 2085786821.jpg [View same] [iqdb] [saucenao] [google]
6911917

>>6911904
>>6911911
early models have unironic sovl now

>> No.6911918

>>6909341
fuck off, pajeet

>> No.6911920

>>6911913
disagree. that banana was nutritious.

>> No.6911921
File: 549 KB, 448x704, 20220825104116_3516957410.png [View same] [iqdb] [saucenao] [google]
6911921

>>6911917

>> No.6911938

>>6911917
i can see pretentious faggot hacks using those ironically as an statement

>> No.6911948

>>6911917
>>6911938
As AI enters into a physical dialogue with the corporeal and unseen, myth is manifest as sensually tangible experience. Expressed through the violent metamorphosis of mutating, ever fateful identities and thoughts, the poetry and mythology of Classical antiquity - its sense of tragedy and transformation – emerge invigorated and renewed. With an incomparable surface and cascading sense of destabilized forms, the surface of 2085786821.jpg pulsates with a frenzied sensuality that reaches beyond allegory to the absolute itself. Sumptuous in opposites and allusion, 2085786821.jpg offers an allegorical appeal to form and its ongoing transformation.

>> No.6911958

>>6909341
Ironically, the only thing AI is really succeeding at is highlighting how much better actual artists are.

>> No.6911963

>>6911958
>shake the tree, let the dead leaves fall to make for anew
this is true, but far from the only thing its succeeding at

>> No.6911965

>>6911963
wanting to genocide indians is nothing new

>> No.6911976

>>6909495
>There's a ton of shit it hasn't been able to do since day 1 and likely will never be able to do
Like? Give examples

>> No.6911988

>>6911389
>like a participation award
Ai is the ultimate participation trophy endeavor

>> No.6912030

>>6911976
Draw a stylish complex graffiti tag(you know the type) that says niggerstonguemyanus

>> No.6912357

>>6911813
Stop projecting >>6909682

>> No.6912602
File: 105 KB, 1024x1024, 1695982617887113.jpg [View same] [iqdb] [saucenao] [google]
6912602

>>6911878
you can separate that.
for example, if you as a human study anatomy, you are literally copying the image, but what you take away from it at the end is not the image, but your understanding of it (the anatomy, or whatever you studied).
it's that kind of relationship.

did you know that the training process is very similar to this?
>AI makes literal random shit
>it then compares that random shit with the target image
>it makes small adjustments to its "weights" by figuring out the parts that contributed the most to the result being wrong.
(weights are essentially the connections between neurons inside a network. all of them. this is important to remember)
>now AI makes random shit again, but it's a little closer to the target image now.
>it repeats the process over and over until it can somewhat recreate the image using its own means.

now if you only did it on one image, it would result in copying, since that's all the network knows.
but what happens when you train on a large amount of images?

remember what i said about weights earlier? if you train on multiple images, there will be some connections that will be used over and over. strengthening that connection
(for example a face from the front always has its eyes and other features in certain areas inside the face. if the AI keeps making faces, it will realize that some of the paths it takes to generate a face it is always the same, or similar at least.)
and it's through this process the AI can "generalize" features. not by learning from individual images, but by seeing patterns that span across all the different images.

>> No.6912604

>>6912602
Explain your understanding of LLMs in your own words or you can't be taken seriously

>> No.6912611

>>6912602
>if you as a human study anatomy,
Stopped reading here. Stop comparing a the development of a machine to humans learning to draw. It's silly.

>> No.6912612
File: 45 KB, 919x256, 1696695117280239.jpg [View same] [iqdb] [saucenao] [google]
6912612

pic related is a good visualization of what the AI "understands". (this is not a completely accurate showing though. this is the result of researchers trying to reverse engineerhow SD interprets things)
a token like "bald" will have specific weights tied to it.
and remember what i said about weights? they're essentially a snapshot of how a network is configured in order to do a task (like creating a certain image).
you could say they are like a recipe for how to create an image.

>>6912611
i explained to you exactly how they compare.

>>6912604
lmao it is in my own words.
i'm just trying to tone down on the berating and needless emphasis. i don't care anymore. now i'll only write for the people who do care to understand this shit.

>> No.6912613

>>6912612
>i explained to you exactly how they compare.
Ok I read it. Sounds like machine learning and human learning are incomparable.

>> No.6912614

>>6912613
*That is to say, the human understands something about what it is learning from. Code doesn't, it's just clusters of pixels.

>> No.6912617

>>6912613
ok. then we'll have to agree to dissagree.

>> No.6912623
File: 924 KB, 814x610, 1695813762047678.png [View same] [iqdb] [saucenao] [google]
6912623

>>6912614
but that is wrong.
AI does not have human level understanding, but it has understanding of all the tokens in its system.

when you tell it to make a bird, it makes a bird because it knows what a bird looks like and what features it has.

just like with the "bald" example earlier. it understands the features. it's not just making an entire image or an arrangement of pixels.

mirrors are still my go to example for this. pic related. these were prompted using "mirror", among other things.
it clearly understands something about mirrors. but it lack the spatial reasoning ability and the accuracy to fully understand what mirrors are.
but limited understanding is not no understanding.
this is not just pixel crunching, otherwise you would never be able to generate something new together with a mirror.

>> No.6912642

>>6912623
Mimic!=understand. It may very well mimic perfectly one day, still not understanding.

>> No.6912649

>>6912612
>lmao it is in my own words.
I asked you to explain LLMs not image prompting

>> No.6912661

talking of understanding in such systems simply a category error, mistaking associations between language tokens for associations between tokens and physical, social, or mental experience.

>> No.6912705

>>6912642
>>6912649
>>6912613
>>6912604
stop fucking replying to attention whores. especially the ones creating such bait threads

>> No.6912706

>>6912649
i'm explaining image diffusion models though?
how are LLMs relevant? they work differently and i'm also not as knowledgeable on their specifics.

i know that they are next token predictors. afaik that also requires a degree of reasoning capabilities in order to do it well.
but i don't know much about their training process and how they use their training data.

>>6912642
>>6912661
you're probably conflating "understanding" with "human level understanding".
it obviously does not understand what a face, a dog, a turtle truly is in their entirety, but that's also not within its capabilities to begin with.
what it does understand is how these things are represented visually.

>> No.6913277

>>6909345
/thread

>> No.6913413
File: 969 KB, 2000x1225, brushwork_lightwork.jpg [View same] [iqdb] [saucenao] [google]
6913413

>>6909349
>>6909588
>>6909652
>>6909762
>>6911382
Keep huffing that hopium. You are only going against the smartest people in tech.
>>6909811
Me

>> No.6913415
File: 115 KB, 1024x1024, OIG (16).jpg [View same] [iqdb] [saucenao] [google]
6913415

>literally the same as every other AI "artist"
>it's not even bait, AIfags are this delusional

>> No.6913428

anything you can shove a sketch into, yet?
i've only fucked around in photoshop ai

>> No.6913472

im about to prompt so fucking hard right now

>> No.6913530

>>6912623
It's impressive how poorly you understand what understanding is. It's not recall - this is a mistake I see autists in particular make a lot.

The way all of these systems work has a very convenient analogy:

Imagine a program that is trained on the set of all possible board states in tic-tac-toe. It responds to a move with the move that would generate the boardstate that it weights most favorably for winning the game. Its performance, because tic-tac-toe is a simple and solved game, is the same as any discretely programmed tic-tac-toe program - i.e., perfect.

However, this program doesn't understand what an X is. Or an O. It doesn't even understand the goal is to get 3 in a row. It is trained to copy the conclusion - the boardstate - and it functions as if it had understanding of the game, but the reality is it's just a complexly weighted die. Note: this tic-tac-toe model is still without understanding even with PERFECT KNOWLEDGE IN ITS TARGET FIELD (every possible boardstate) - it's not about the size of the dataset.

Today's LLMs are only different from 20th century LMs in the first L - larger training datasets. That's literally it. Scaling up the same memoryless Markov processes to insane sizes doesn't grant them memory OR understanding. It's still the same fundamental architecture and thus inherits the same deficiencies - you can't scale up a 3 bedroom floorplan with no bathrooms to 3 billion bedrooms and expect indoor plumbing to arise magically from "emergence".

To imply it has understanding is to grant it literally metaphysical properties that do not exist within its code, its training data, or even the trained model. It's retarded.

>> No.6913564
File: 880 KB, 941x527, 明日晴れますように.png [View same] [iqdb] [saucenao] [google]
6913564

>>6909425
most =/= all
you can't deny the fact that creativity is not limited to someone who can draw

>> No.6913569
File: 2.20 MB, 2048x1152, 1691010001188914.png [View same] [iqdb] [saucenao] [google]
6913569

>>6913564
nature is creative and motherfuckers out here saying a computer cant be - let alone the fact that the creativity is shared (compounded?) between it and the user

>> No.6913582
File: 287 KB, 1522x1334, 20231031_040221.jpg [View same] [iqdb] [saucenao] [google]
6913582

Judge has dismissed the first lawsuit against OpenAI.

>> No.6913595

>>6913582
Stability, not (closed)OpenAI.

>> No.6913599

>>6913582
the only aspect that was dismissed with prejudice are the copyright act claims by the two artists that didn't have their works registered, Everything else has been dismissed with leave to amend.

>> No.6913606

>>6913599
good fucking luck to them in round two. they barely have any idea how this shit works. they have no case.

>> No.6913612

>>6913606
lucky for them, neither does the judge. And I'd argue stability doesn't understand their own blackbox.

>> No.6913619

>>6913612
with even a modicum of knowledge one can deduce that there was no copyright infringement.

>> No.6913627
File: 588 KB, 1292x840, Screenshot-41-transformed.jpg [View same] [iqdb] [saucenao] [google]
6913627

>>6913619
Uh huh
>reeeeee that's a bug!
doesn't matter, still copyright infringement.

>> No.6913638
File: 799 KB, 960x1280, 1698632508176668.png [View same] [iqdb] [saucenao] [google]
6913638

>>6913627
>>reeeeee that's a bug!
what are you even trying to say?

>> No.6913643
File: 570 KB, 1080x1278, Fakedismiss.jpg [View same] [iqdb] [saucenao] [google]
6913643

>>6913582
Fake shit.
However I doubt (((they))) will let artists win. Israel need AI to fund the war and to control public opinion, they invested heavily on it.

>> No.6913647

>>6913638
The AI could churn out an image that was overfit without the user intending. for example, a young user in Afghanistan might prompt and use "Afghan girl", while unaware of the copyright photo. That's on Stability. Who knows what other innocent prompts would lead to such situations?

>> No.6913652

>>6913643
whats even more gay than people not reading the full order are people going "AKCUALLY DAS NOT HOW IT WORK"

>> No.6913745
File: 62 KB, 1694x606, 20231031_055723.png [View same] [iqdb] [saucenao] [google]
6913745

>>6913652
They have read it. It's basically the judge saying "Your case is fucking non-existent, do you have anything else to add before I laugh you out of my courtroom?"

>> No.6913811

>>6913745
Stability Ai's days are numbered either way, they aren't doing good financially and open ai will have a monopoly

>> No.6913812

>>6913582
Yes aisisters, Jewish billionaires buying off judges means they are in the right

>> No.6913813

>>6913811
open source wins every time. altman is a black hole.
>>6913812
someone didn't read the entire order

>> No.6913826

>>6913813
Open source makes no money, stay delusional emad

>> No.6913830

>>6913826
I'm not delusional, I graduated from Oxford with a master's degree.

>> No.6913832
File: 366 KB, 1346x843, Screenshot_20231031_072320_X.png [View same] [iqdb] [saucenao] [google]
6913832

This is you.

>> No.6913861

>implying i have a partner
art is the only partner I need

>> No.6913921

>>6913413
> You are only going against the smartest people in tech.
Same people who think cooking is a waste of time and try to push their disgusting beige gloop as food. Why the fuck should I care what they think?

>> No.6914027

>>6913921
because they are smarter and more successful in every metric than you.

>> No.6914032

>>6914027
If they are so smart and successful, why are so many of them in jail?
https://www.youtube.com/watch?v=V36kSqwjaaw

>> No.6914115

>>6913832
>Autistic tranny #1 upset autistic tranny #2 is proompting more than talking/gash licking "her" partner.
kek

>> No.6914161

>>6913530
i'd say your post highlights the flaws in your own understanding.

>Today's LLMs are only different from 20th century LMs in the first L - larger training datasets. That's literally it.
...that's not true afaik. transformers only came about in 2017.

>However, this program doesn't understand what an X is. Or an O.
it doesn't need to. it could a an A and B or red and blue, it doesn't matter. again you conflate broader human understanding with a more specific form of understanding:
because why would a game NN need to understand the broader meaning of what a "X" is? just because humans understand?

>tictactoe
it is a solved and incredibly simple game. you don't need a NN for this.
here, you actually can solve the entire game just by memorizing all game states - because there aren't that many.
but take a game like go instead.

>It doesn't even understand the goal is to get 3 in a row. It is trained to copy the conclusion - the boardstate - and it functions as if it had understanding of the game, but the reality is it's just a complexly weighted die. Note: this tic-tac-toe model is still without understanding even with PERFECT KNOWLEDGE IN ITS TARGET FIELD (every possible boardstate) - it's not about the size of the dataset.
it's the opposite anon. because it has perfect knowledge, because it's a solved game, you don't need any understanding to beat this game.
but if we ramp up the game complexity, your argument will quickly fall apart.

for example: a game like go cannot be played like this at all.
you cannot copy or remember all the board states in go, there's an astronomically large amount of them. so it cannot predict all the outcomes. then how do go engines function? how does it play at an above human level?
does it not need the ability to evaluate the board and the moves being made? is that not an understanding of the game?

cont

>> No.6914190

cont.

>>6913530
>It is trained to copy the conclusion
this is fundamentally wrong. and this only can work at all with easily solved games like ttt.
the actual true goal is pattern detection and generalization, the ability to apply what is being learned to new situations.

and ask yourself what generalization really means.

>> No.6914242 [DELETED] 

please reply to me, I'm waiting.

>> No.6914530
File: 439 KB, 778x800, f765924f-1841-4c02-9e0c-1407c09e2da0.png [View same] [iqdb] [saucenao] [google]
6914530

>>6909341
Character

>> No.6914599
File: 252 KB, 1254x1633, 169827593555626258.jpg [View same] [iqdb] [saucenao] [google]
6914599

>>6909341
While we're on the topic of AI:

>"In a class action lawsuit filed against AI-generated image service providers Stability AI, Midjourney, and DevianArt, a US district judge ruled that determining whether generated images may be in direct violation of copyright laws is “not plausible” at the moment"

x.com/franklingraves/status/1719053140220137846?t=r0-PifXz93qq4aN9vAK0xg&s=19

https://www.computerworld.com/article/3709691/artists-lose-first-copyright-battle-in-the-fight-against-ai-generated-images.html

>> No.6914601
File: 133 KB, 1245x1626, 169793679142263718.jpg [View same] [iqdb] [saucenao] [google]
6914601

>>6914599
>" 'The other problem for plaintiffs is that it is simply not plausible that every Training Image used to train Stable Diffusion was copyrighted (as opposed to copyrightable), or that all DeviantArt users’ Output Images rely upon (theoretically) copyrighted Training Images, and therefore all Output images are derivative images' " - Judge Orrik

Note: Sarah Anderson, aka: Sarah's Scribbles. I'm sure some of you are familiar with her work:

https://x.com/SarahCAndersen?t=bsgT2dGZlyhLePfFxEJ1gg&s=09

>> No.6914609

>>6913627
Nigger that's just an img2img at a really low filter. In other words they put it through a filter and are then pretending that this was generated by the model alone. It's not even worth explaining it to you anyway since like the Sarah Anderson shit, it will be dismissed because they will be unable to replicate any of their copyrighted shit with literally any model. You are worthless. You are untalented.

>> No.6914611

>>6909341

>AI
>LETS
>ANYONE
>CREATE
>ART

>> No.6914617

>>6909413

you are mentally ill and need to seek professional help immediately

>> No.6914620

>>6913647
Okay but the outputs are derivative. You're allowed to make derivative art. Whether it be yourself or a machine

>> No.6914622

>>6913826
>Open source makes no money,
And? That's typically never the main goal

>> No.6914623

I like what the pro-AI redditfags are calling it. Generative AI, or gAI for short.

I don't feel economically threatened by gAI at all, but it's ugly and I don't like seeing it spammed everywhere. Once the boomers get a hold of it, it won't be trendy anymore but we'll have to put up with boomer gAI's instead.

>> No.6914633
File: 251 KB, 1672x516, duplicate_images_1.jpg [View same] [iqdb] [saucenao] [google]
6914633

>>6914609
Mate, it's from a paper on retrieving training images.
https://arxiv.org/abs/2301.13188

>> No.6914635

>>6914623
>Once the boomers get a hold of it, it won't be trendy anymore
It hasn't been trendy in months. The hype died off within a month back in December of last year. Only Turbo autists (me) and perma begs (most of (you) ) have it rent free in their heads. Most normal bags are still getting filtered by chat GPT despite the fact that it literally does whatever the fuck you tell it to do most of the time. It will never become mainstream because you actually have to have a functioning brain to set up and run local SD. Copro SD is constantly getting sensor to the point where even tame shit gets blocked so that will forever be a meme. It makes me question why they even bother implementing it in the first place when it doesn't even make them (much) money and they feel the need to babysit people all the time

>> No.6914638

>>6914633
Mate, those are Cherry Picked results from a shitty filter. Any retard can fake those results even without using sd. They take a photo, run it through a filter and then pretend that that's copywriting fringement. I don't know why you're even still arguing when the mother fucking judge that doesn't even really know how it works even agrees that it is not inherently copyright infringement because anything and everything it generates is derivative. The only way anything AI generated could be copyright infringement is if you took an existing art piece, put it through the img2img tab at a denoise strength of like 0.1, and then claimed it as your own.

THEN you can claim that that PARTICULAR output is C.I.

>> No.6914639

>>6914620
Please read the fair use laws. You haven't. Pop quiz: what is the fifth factor in determining judgement in a fairuse court case?

>> No.6914642
File: 72 KB, 680x502, F9vNq6kacAAtqj1.jpg [View same] [iqdb] [saucenao] [google]
6914642

>>6914599
>>6914601

kek. ok pajeet.

>> No.6914647

>>6914638
Got stable diffusion 1.4 on your PC? You can try the prompts yourself and get those images, honest to god. Hell, just prompt bing chat with "Bloodborne box art"

>> No.6914652
File: 68 KB, 481x680, 6573476867980567.jpg [View same] [iqdb] [saucenao] [google]
6914652

>>6914599
>>6914601

fake news bitch

>> No.6914673

>>6914642
He's essentially saying " actually provide some fucking evidence next time and maybe I will take you seriously you useless monkey"

Like.. who goes through all the trouble of filing a lawsuit against a company only to not provide sufficient evidence? Any other appeals submitted should be denied for that reason alone if you're that lazy and incompetent

>>6914647
Will post results later on.

>> No.6914681

>>6914652
So only one out of the three were denied......why is that?

>> No.6914694

>>6914161
>>6914190
>transformers
Basically just streamlining the training process with parallel computing; doesn't change the models themselves, just makes larger training data more feasible. It's almost like you don't know what you're fucking talking about.

>it doesn't need to.
It needs to have some understanding. It doesn't. It's emulating the end result of understanding - it's basically cargo cult cognition. It's almost like you're ascribing an ability to it that it doesn't possess and just coping with "hurr durr it's just not HUMAN understanding" because you can't actually prove your claims and are banking on a type of understanding no one has been able to prove even exists. This is "the AI works- I mean understands in mysterious ways" levels of retarded.

>you don't need a NN for this
...Ah, you ARE autistic. Let me make this clear to you: ITS SIMPLICITY IS WHERE ITS ILLUSTRATIVE POWER OF WHAT THE MODELS ARE ACTUALLY DOING ARISES.
IT
IS
A
FUCKING
ANALOGY.

If you can't understand what an analogy is, you sure as hell have no business preaching about understanding or lack thereof - these are concepts beyond your cognitive capacity to engage with.

>but if we ramp up the game complexity, your argument will quickly fall apart.
Actually it continues to hold, because this is essentially how they had ChatGPT playing chess - it has no internal representation of the concepts or rules in the game, just weighted boardstates. It's almost like you don't know what the fuck you're talking about.

>this is fundamentally wrong
Incorrect. It doesn't emulate the process of constructing a response to a prompt, it emulates exclusively a fully constructed response. It doesn't emulate the process of image construction, it emulates the end result, and where it appears to have some process it is only emulating the details present in the end results within its training data that arise from a given real process.

It's almost like you don't know what the fuck you're talking about.

>> No.6914727

>>6914639
Stable diffusion is inherently transformative you moron.

>> No.6914735

>>6914727
Failed the quiz. Read fair use laws. Go on. Fucking go on! Git! Get outta here! Skedaddle! Vamoose! Don't come back without the correct answer to the popquiz question.

>> No.6914738
File: 371 KB, 1080x2267, 169815166555876255.jpg [View same] [iqdb] [saucenao] [google]
6914738

>>6914735
You asked what the fifth Factor was you talentless monkey and I gave it to you.

https://libraryguides.salisbury.edu/copyright/fairuse#:~:text=Fair%20use%20%22fifth%22%20factor%20%2D,traditionally%20weigh%20against%20fair%20use.

>> No.6914752
File: 26 KB, 500x286, F17211B5-3C94-41BB-A2F1-96FD323AAF42.jpg [View same] [iqdb] [saucenao] [google]
6914752

>>6914611
Anyone can read write now and look at the damage that did. Fuck Guten(((berg))).

>> No.6914753

>>6914738
While not the 5th I meant(I was referring to how garbage pail kids was judged), What is transformative about a painting being used in the same context? That's not transformative.

>> No.6914758

>>6914753
If I hand draw a painting of Mickey Mouse in the style of Van gogh, that is transformative because it's not a carbon copy of one of his existing works. SD worth the same way. It can copy Styles pretty well but it is literally impossible to make a carbon copy of anyone's existing work with sd. You have not even shown concrete examples and Cherry Picked pictures do not count

>> No.6914764

>>6914758
>transformative
Not the way the law works. You have it right there in that pic. It has to be a new use. An entirely new context.
Using ai images made of data from copyright images to compete with those very images you trained in, not transformative. It doesn't simply means changed. Yes the English definition might, but that's not what is being discussed.

>> No.6914766

>>6914764
>It has to be a new use
And sd is in essence, a new use

>not transformative.
It is, because the outputs are new ans unique.

>> No.6914768

>>6914766
>Want to use copyright images
>Write some software
>Feed images in
>Get new images out
In that process, it went back to images for use as images. Not transformative.

>> No.6914772

>>6914766
Dumb jeet

>> No.6914773

The Effect of the Use Upon the Potential Market

Another important fair use factor is whether your use deprives the copyright owner of income or undermines a new or potential market for the copyrighted work. Depriving a copyright owner of income is very likely to trigger a lawsuit. This is true even if you are not competing directly with the original work.

For example, in one case an artist used a copyrighted photograph without permission as the basis for wood sculptures, copying all elements of the photo. The artist earned several hundred thousand dollars selling the sculptures. When the photographer sued, the artist claimed his sculptures were a fair use because the photographer would never have considered making sculptures. The court disagreed, stating that it did not matter whether the photographer had considered making sculptures; what mattered was that a potential market for sculptures of the photograph existed. (Rogers v. Koons, 960 F.2d 301 (2d Cir. 1992).)

>> No.6914781
File: 90 KB, 1286x776, 169856418806828878.jpg [View same] [iqdb] [saucenao] [google]
6914781

>>6914647
Nta but I have sd spun up right here now. What should I test to prove this shit ain't copyright infringement?

>>6914768
By your logic any fan art of any IP ever is infringement because the artist used official copyrighted material as a reference. Why is it ok when you or does it but not when a machine uses math to do it? Your argument makes zero sense
>>6914772
>No argument
.

>> No.6914782

>>6914781
>art of any IP ever is infringement because the artist used official copyrighted material as a reference.
It is. I don't disagree.(if they are selling it)

>> No.6914789

>>6914781
Check that paper linked further up. Bloodborne is an easy one to try tho. Guarantee if you gen a handful one or more will be the PS4 boxart/promo image with the back facing hunter with his arms to his sides holding the weapons, in that foggy yharham street.

>> No.6914791

>>6914789
If they are not carbon copies, they aren't copyright infringement. That's assuming it actually does what you say will do. They have to be perfect copies of original things like the original PlayStation logo in order for it to be copyright. Except even then that wouldn't be copyright because merely drawing the PS4 logo it's not copyright infringement. It's only copyright infringement if I make a brand, slap their logo onto my brand, and then claim it as my own. Me drawing the logo is not illegal so I'm a machine trying the logo is not illegal either.

>> No.6914799

>>6914791
No, I'm saying the fact the machine produces these, means it hold that data, and they are selling access to it. Copyright infringement. They released the open source stuff to cover their ass, but rake in the dough via investors. Google was given special rights to allow for hosting images in the search irrc there was a court case about it.
Obviously the image will be a lossy version, not sure why you want to shift the goal posts...(ok I know why)

>> No.6914801

>>6914791
Post your University of Dehli law degree, faggot.

>> No.6914817

>>6914789

https://files.catbox.moe/axrcou.png

Tried bloodborne. It tried and failed miserably at generating thr official box art. It looks like a child with Parkinsons tried to copy it purely from memory. No where near a carbon copy. Not inherently infringement because you couldn't even pretend this is official bloodborne art. Therefore neither I or the software used to make this UNIQUE and TRANSFORMATIVE piece of shit image violated anyone's copyright.

No I will reiterate, as I keep telling you, it is only copyrighting fringement on MY end if I make a carbon copy of box art (some fucking how) and then claim it as my own. Using copyrighted artwork to make money off of new artwork is not copyright infringement or else every artist that uses patreon should be sued into Oblivion. Make it make sense and use common sense.

By the way just to prove I actually used SD to generate this, run this image through a Metadata reading site like:

https://www.metadata2go.com/
>>6914799
>means it hold that data,

If I hold the data you mean has to physical copies of the original training data then you were incorrect. It really has a rough idea of how bloodborne box art should look. As I just demonstrated it cannot even accurately copy it. The data set images are not stored in the model I'm using or else the model itself would be several hundred if not several thousand terabytes in size ( or openai used special black magic to invent an entirely new method of image compression that could fit all those images from the LION datasets into a roughly 4gb bin file). Basically a very beefed up TI-84 calculator. You know how if you knew what you were doing you could draw pictures on the wrapping section of those calculators? That's essentially what this is except the art actually looks kind of good....kind of.....


https://youtu.be/q1OEXc_Gio4?si=R5-qUUhLmqXeqqQg
>>6914801

>no argument

Come on dude at least the other dude sewms to have a brain......
Next.

>> No.6914820
File: 1.49 MB, 997x962, Screenshot 2023-10-25 100054.png [View same] [iqdb] [saucenao] [google]
6914820

>>6914817
regarding that google hosting images thing. It only applies to the U.S. jurisdiction, and other countries may have different laws and interpretations of fair use or fair dealing. For example, in Germany, the federal court of justice ruled that Google’s image search engine does not violate copyright by showing thumbnails of copyright-infringing pictures. But in France, the court of appeal of Paris ordered Google to pay €1.6 million to a group of photographers and photo agencies for displaying their works without authorization.

>> No.6914822
File: 136 KB, 1024x1024, OIG.5SAV.jpg [View same] [iqdb] [saucenao] [google]
6914822

is better

>> No.6914825

>>6914817
>If I hold the data you mean has to physical copies of the original training data then you were incorrect.
I don't mean that, no. just that the training/funding is sus, and may be subject to laws relating to fair use.

>> No.6914828

>>6914820
>regarding that google hosting images thing. It only applies to the U.S. jurisdiction, and other countries may have different laws and interpretations of fair use or fair dealing.
Well it's a good thing for them that Open Ai/stability ai is a US company and therefore how they trained the models is 100% legal....right? As far as I'm aware of most if not all of their operations are on us soil so I don't even know why you're bringing up foreign countries.

>But in France, the court of appeal of Paris ordered Google to pay €1.6 million to a group of photographers and photo agencies for displaying their works without authorization.
And they are well within the right to do that. Why? Because Google uploaded and displayed EXACT....COPIES....OF...THEIR...WORK...

Now tell me, is the image I linked a carbon copy of any work anywhere?

>> No.6914832

>>6914828
>stability ai
I thought they were london?

>> No.6914836

>>6914828
>Because Google uploaded and displayed EXACT....COPIES....OF...THEIR...WORK...
even if google ran them through img2img at 5 percent de-noising, I'd say the photographers would have a case.

(ok even 50 percent... 100 percent... if the case is that it violates copyright to use the images in the first place)

>> No.6914843

In any case, it is important to note that fair use is not a right, but a defense. This means that Stability or OpenAI cannot claim fair use as a justification for using any images without permission, but only as a possible argument against infringement claims in court. They should be aware of the legal risks involved and take steps to mitigate them, like ensuring they have proper licenses for the images they use, obtaining consent from the rights holders when possible, and respecting the moral rights of the original creators.

>> No.6914844

>>6914836
>I'd say the photographers would have a case.

Why would they even do that? Of course they would still have a case because you're still using the original image. That's entirely different than telling someone or something that knows how something looks to draw something new. As I keep telling you but apparently you keep ignoring, the original images and the data set used to train the models are not stored directly into the model.

>>6914843
>This means that Stability or OpenAI cannot claim fair use as a justification for using any images without permission, but only as a possible argument against infringement claims in court.
So do I need to ask nickelodeon for permission whenever I draw sandy cheeks fan art?

>> No.6914847
File: 664 KB, 888x574, Screenshot_2023-10-31-22-28-30-376_com.google.android.documentsui.png [View same] [iqdb] [saucenao] [google]
6914847

Modern art = I could have done that but you didn't

:)

>> No.6914850

>>6914694
first thing: calm down. you really should not speak this confidently or lash out about things you have barely any understanding of. i'm a relative novice as well but at least i try to keep myself somewhat educated on this stuff.

>Basically just streamlining the training process with parallel computing
transformers introduced self-attention and it a significant architectural change.
it's like the difference of
>having a very narrow tunnel vision and not being able to see the things that are in the distance
vs
>having a birds eye view over the entire field (context).
imageAI gained something from this as well in the form of cross-attention, which is a part of SD. it would be impossible without it.

being able to have a "bigger view" ultimatively means it has an easier time seeing patterns (!).
this is why it is a significant qualitative change and not just a change in efficiency, although it's that too.


>It's emulating the end result of understanding
i already explained the in depth process in here >>6912602
(the last paragraph)
i'm describing the process where understanding happens. i can't go any more in depth than this. or do you want me to explain exactly what neurons do what task during the generation of an image?

1/3

>> No.6914851

>>6914694
> ITS SIMPLICITY IS WHERE ITS ILLUSTRATIVE POWER OF WHAT THE MODELS ARE ACTUALLY DOING ARISES.
your analogy doesn't even show the need for a NN at all. (nor does it apply, but more on that later)
did you know that even chess isn't complicated enough to need a NN? the strongest chess engine, stockfish, is still mostly non-NN based.

what does it tell you about your analogy when it only works for simple games, whereas you NEED a NN for games where things are more complex and open? (even just a videogame like starcraft or minecraft)

the problem here is your analogy clearly does not apply to go/starcraft/minecraft.
in fact, it only applies to manually created programs.

it really doesn't apply even with chess or tic tac toe. i'll go over your original post again:
>It doesn't even understand the goal is to get 3 in a row.
it depends on how the AI is made. if you train an AI recursively through its own games. (without any other inputs other than playing with it) it would eventually "understand" everything about the game of tic tac toe. it would not understand how the field actually "looks" like or what pieces it is playing with, but it would understand

>that the game is played on a 3x3 board
>you can't play over existing pieces
>3 in a row means victory or defeat.
>the best strategies, through trial and error.
now that would be a very narrow type of understanding, but an understanding nonetheless.

2/3

>> No.6914852

>>6914694
>Actually it continues to hold, because this is essentially how they had ChatGPT playing chess - it has no internal representation of the concepts or rules in the game, just weighted boardstates. It's almost like you don't know what the fuck you're talking about.
chatgpt playing chess is not the type of machine learning we're talking about....
chatgpt hasn't trained directly on chess, so its capabilities would be limited.
remember: AI is about pattern generalization. this is why large high quality datasets are so benefitial for it. and chatgpt is a language model in the first place, so it learning chess is double the challenge.

and it is arguable if it has no internal representation of the concepts or rules of the game or if there are simply other limitations, like context size, that prevent it from playing chess accurately according to the rules.

>Incorrect. It doesn't emulate the process of constructing a response to a prompt, it emulates exclusively a fully constructed response.
yes. and in order to do that, it needs understanding. this is why these factors emerge from AI. otherwise it can just say random unrelated shit all the time. after all that is a "fully constructed response".

a famous example someone gave to illustrate this is the following sentence at the end of a mystery story:
>"and the culprit was ____"
in order to grab the right token, you need an overall understanding of the entire story.
a LLM is not capable of that much atm, but ANY RIDDLE shows this idea in action.

and if you understand what this means, you also understand that this does not only apply to riddles and mysteries, but to every token it makes. it needs to have a degree of understanding. even if it is narrow and different from a human's understanding.

3/3

>> No.6914853

tripfag at it with the mental gymnastics again kek
i genuinely hope you're getting paid for this

>> No.6914861

>>6914844
>As I keep telling you but apparently you keep ignoring, the original images and the data set used to train the models are not stored directly into the model.
I know this. latent space is spooky.
>So do I need to ask nickelodeon for permission whenever I draw sandy cheeks fan art?
using them as a business. god..

>> No.6914864

>>6914861
No...or else patreon artists would be constantly ass fucked

>> No.6914867

>>6914864
>or else patreon artists would be constantly ass fucked
It sincerely is copyright infringement. if disney wanted to, they could fuck the artists in the ass. Nintendo does it to fan games all the time.

>> No.6914888

>>6914867
Their art is transformative. That's why they CANNOT do that or they would do it already. Remember when they tried to sue the family of a toddler because the toddler song one of their copyrighted songs on youtube? It's because they did not understand their use or copyright laws at the time. Now they do and that's why they don't try to pull that nonsense anymore.

Unfortunately however, current copyright law DOES prevent people from, let's say, having a Spiderman themed grave Stone for a deceased child because the father made the mistake of using actual official art on the tombstone.

It seems you've conveniently forgotten how greedy this corporations are

>> No.6914899

>>6914888
No, wrong. you don't understand transformative again. Selling fan art is quite literally illegal, but suing people costs money. Fan artists are unlikely to have lots of money. And fans get mad when you treat their peers in a way that looks like "mean-spirited big company sues little starving artist", so big companies hold off. (aside from nintendo, they don't want fan games to exist, so will sue if you try to make money off a fan game)

>> No.6914904

"Intellectual Property Rights
You need to either wholly own the content you submit to Patreon, or have the rights to use, display, and resell it. Any content you submit must comply with trademark and copyright laws.

Content submitted for merchandise creation containing unauthorized copyright-protected or trademarked content is absolutely prohibited. "

See look. even patreon acknowledges it. huge companies just turn a blind eye.

>> No.6914938

>>6914899
Making fan art is not illegal because it is transformative.... it is new. Is unique. You're making dumbass Arguments for the sake of dying on a hill

>And fans get mad when you treat their peers in a way that looks like "mean-spirited big company sues little starving artist"
>Retard thinks corps actually care about the internal drama of pink haired unemployed artists.

I'll give you the benefit of the doubt. Explain why fan art is somehow not transformative

> (aside from nintendo, they don't want fan games to exist, so will sue if you try to make money off a fan game)
So then why the flying fuck doesn't Nintendo sue artists for making fan art? They have a US Branch so they could do that if they wanted to try right? They routinely prevent event organizers from hosting Smash tournament events because that's (somehow) legal due to their extremely tight control of their ip. They show open hatred and contempt towards their fan base so if anyone would sue fan artists if they could, it would be them. You contradicted your own point and you didn't even realize it

>> No.6914944

>>6914904
>You need to either wholly own the content you submit to Patreon
The content and question is fan art. Making fan art does not mean you are declaring ownership of a particular ip. Me join the crew figure of Mickey Mouse in Ms paint does not mean I'm declaring that I own Mickey Mouse. You are misinterpreting what they're saying. What you just quoted prevents me from taking ribs of Rick and Morty episodes, uploading them to patreon, and then claiming that those are MY Rick and Morty episodes that I created.

Fanart.
Is
Not
Copyright
Infringement
You
Retard
Glue
Eater

You would think someone from /ic/ would know this shit.

>Content submitted for merchandise creation containing unauthorized copyright-protected or trademarked content is absolutely prohibited. "

This applies to official artwork that you read from Cartoon Networks website. In other words I cannot take official art from the website, print out some T-shirts with Steven Universe on the front, and then sell t-shirts.

If what you were saying is true then fan art on patreon would not even be fucking possible. Patreon is always under constant fire from the feds and credit card companies because occasionally some loli artist will try to sell realistic loli porn ( which is actually illegal by the way) in the credit card companies will threaten to drop support on patreon unless they tighten up control over that shit.

So, halfwit, if fan art was truly illegal then patreon would not allow it because otherwise companies like Nintendo or Nickelodeon or Cartoon Network or Warner Brothers or whoever the fuck would be constantly threatening to sue them

>> No.6914946

>>6914938
Dying on the hill that selling fan art is illegal? I will. it is. nothing you say changes that. consider looking up the law.

>So then why the flying fuck doesn't Nintendo sue artists for making fan art?
I'm sure they would fly some fucks if there was an organized nintendo only fan art selling event(comparable to a smash tourney).

It's hilarious how wrong you are.

>> No.6914952

>>6914946
I'm not talking about physical events. What you see is correct, they would sue artists on Twitter for uploading fan art. If they are cleaning up to shut down physical events they are greedy enough to attack people online. They don't even like certain gameplay footage of their games uploaded to YouTube because apparently that counts as copyright infringement. Meaning if someone does like a 10-hour let's play of a new Mario game and uploads of the youtube, they are within their rights to strike that video down. Other gaming companies could do that shit with their IP to other game footage uploads YouTube but a lot of them choose not to in good faith.

Your idiotic. Simple as that. Why would they choose to shut down events but not Target fan art? Because they fucking can't

>> No.6914953

Selling fan art is not illegal? huh, as the owner of Hot-Topic, I no longer have to get licenses to print... I'm gonna save me some money!

>> No.6914954

>>6914938
>Making fan art is not illegal because it is transformative.... it is new. Is unique.
None of this is relevant, and as long as one can reasonably be confused as to whether the fan art is officially licensed or off-brand, you're infringing copyright laws in most jurisdictions. Whether you like it or not, companies own certain IPs (like Nintendo does Mario or Capcom does Street Fighter), and they have the right to sue people who make use of such IPs without license, unless the use falls within the definition of fair use. Much if not most of fan art falls outside of this category, since it just replicates the designs verbatim.
>So then why the flying fuck doesn't Nintendo sue artists for making fan art? They have a US Branch so they could do that if they wanted to try right?
1. They don't need to, a cease and desist letter is enough in almost all cases.
2. Are you aware that corporate lawyers don't work pro bono? If yes, then can you put the two and two together as to why it's not worth their trouble if the C&D does the job?
(anything above is not legal advice)

>> No.6914960

>>6914952
>Your idiotic
your an dummy

>> No.6914970

>>6914954
>None of this is relevant, and as long as one can reasonably be confused as to whether the fan art is officially licensed or off-brand, you're infringing copyright laws
>draw stick figure of.....actually no one in particular. Just a fucking stick figure
>nickelodeon: "ummm aucktually that looks like doodle Bob so I'm gonna sue you now k thanks"

>Whether you like it or not, companies own certain IPs (like Nintendo does Mario or Capcom does Street Fighter), and they have the right to sue people who make use of such IPs without license,
Yes....if it's copyright infringement. Bernard does not copyright infringement. I'm not repeating the explanation again because at this point you're being dumb on purpose. You still haven't explained why fan art isn't fair use or isn't transformative.

>since it just replicates the designs verbatim.
Most fan art DOESN'T DO THIS at all. Many artists have their own drawing Style. Their own unique we have drawing eyes, faces, limbs, anatomy, shading, etc. Drawing characters in their own art style is by definition transformative

https://twitter.com/Cheekie_/status/1715408787644064187?t=bjYdej5HtEZG4HgWODloxQ&s=19

Why hasn't this person been sued yet by Warner brothers? Don't pull that " they don't want to piss off fans" shit because this artist is basically a nobody, as are most online artists. The backlash they would receive from taking down this artist page would be a drop of water in an ocean.
>They don't need to, a cease and desist letter is enough in almost all cases.


And yet curiously you cannot pull up any examples of anyone receiving a cease and desist letter for drawing smut of princess peach. I'm sure they're well aware of rule 34 and gelbooru and other image hosting sites that are dedicated to hosting the supposed copywriting infringing art rights? Weird how they NEVER get those lettwrs cuz of fan art. Hmmmm...... I guess they must REALLY like those sits and give them favoritism

>> No.6914977

>>6914970
now take this on >>6914953 oh, just gonna ignore that? you know it's illegal, no matter the style. if hot-topic printed fan art from properties they had no rights to, they would get sued, simple as.

>> No.6914982

>>6914970
>anyone receiving a cease and desist letter
https://www.reddit.com/r/legaladvice/comments/9vyis2/nintendo_sent_a_cease_and_desist_for_painting/
not smut, but it's something. just first result from googling. I imagine the cease and desist would have been the same had they painted her with fat tits in a bikini. maybe it happens more than you think.

>> No.6914984
File: 47 KB, 747x377, Screenshot 2023-11-01 115613.png [View same] [iqdb] [saucenao] [google]
6914984

>>6914982
here's another
https://twitter.com/Cptn_Alex/status/1310750779737694209

>> No.6914987
File: 420 KB, 2172x2325, 627d121bd4fd200d73814c11_60bcd0b7b750bae1a953d61d_autoencoder.png [View same] [iqdb] [saucenao] [google]
6914987

>>6909341
Autists ITT seem to ignore that literally every picture exists in the latent space. You can take a picture of your room right now, and there's a very close representation of that very same image in SDXL.
Copyright law applies just like in any image you can create in any other tool ever.

>> No.6914991

>tool

>> No.6915004

>>6914987
>literally every picture exists in the latent space
not quite. the data added to the process "creates" the image, (the prompt, parameters, LoRAs etc)

copyright can apply differently for different situations, the creation of that latent space may be an act of infringement, since they built it on copyright work and are selling it. no human hands interfered, it's data in data out. stealing. theft. bad boys. bad boys II. Emad was literally punished by getting bad karma(he fell down 2 flights of stairs). that's proof enough for me.

>> No.6915012

>>6914982
That figure or painting looks dead in your identical to the official peach character so yes.... even though it's extremely scummy there within the right to do that because it looks TOO similar to the original art/character model. ( I'm saying this assuming pic rel is what they painted. The imgr link they posted is dead so I'm not 100% certain that's it)
>>6914977

If you make a carbon copy of the Hot Topic logo then yes that is copyright infringement. Fan art on the other hand in most cases are not carbon copies. How many times have I said that?

You have still failed to prove that fan art by itself is inherently copyright infringement. It's only copyright infringement if the fan art looks way too similar to the original official art. Which, again, isn't the case most of the time because a lot of artists have their own unique style that looks nothing like how Nintendo draws their characters. Even /ic/ vets will tell you you cannot copyright an art style in regards to drawings.

>>6914984
Do you have any links to what he drew that triggered the lawyers?

>> No.6915014

>>6915012
Forgot the picture

https://files.catbox.moe/xv4caz.png

>> No.6915016

>>6915012
>Do you have any links to what he drew that triggered the lawyers?
no, all I gathered was he was painting nintendo characters on controllers.

You keep dodging, fucking annoying. Can you hire an artists with a distinct style and sell t-shirts of nintendo characters at hot-topic with no license? Fuck no. I rest my case.

>> No.6915018

>>6915014
his style is unique in a sense, it's not a carbon copy. it is the same pose/obviously reffed, but so is a lot of fan art people sell.

>> No.6915041

>>6915018
We must have different definitions of unique because had you not told me that's what he drew I would have assumed it was some official Nintendo amibo. You should steer clear from creating anything Nintendo related if that's actually what his style looks like

Regardless, my point still stands. He only got the letter because his style looked TOO similar to Nintendos. In this particular case I kinda side with Nintendo even though he probably didn't mean to commit copyright infringement.

>>6915016
>no, all I gathered was he was painting nintendo characters on controllers.

Oh shit that's even worse. You're effectively doing it twice because you're not only drawing official characters to similarly to the originals but you are essentially modifying their own Hardware without their permission. And then selling it without their permission.

>Can you hire an artists with a distinct style and sell t-shirts of nintendo characters at hot-topic with no license? Fuck no. I rest my case.
Yes. People literally do that all the time..... not specifically at Hot Topic but in general. I live in a very tourism heavy town and plenty of Street vendors make caricature art. The kind where they intentionally make you look like an abomination. They also sell LEGAL FAN ART t-shirts of characters like Ben 10 and Pikachu. Can you take a wild guess why that's allowed?

>"but but then why don't I see unofficial fan art t-shirts at Wal-Mart? "

Because people don't want unofficial art t-shirts. They want the real deal. If a 10-year-old wants above a picture he's going to want the one that looks mostly like the original art style. A freelance artist will have a much harder time getting their unique LEGAL merchandise into a big bucks retailer like Walmart then a mega corporation like disney. Most normies would consider but deviantART tear fan art versions of the shirt terrible anyway so Walmart wouldn't even consider them for that reason alone

>> No.6915088

>>6915041
>Street vendors
Come on stop playing dumb. you lost the argument. get over it.

>> No.6915150

>>6909341
shit like performance art, live drawing with freakish skills

99% of artists get the rope though, not saying I want that but I think it's what's going to happen

>> No.6915163

>>6915150
>99% of artists get the rope
drawing monkeys were always overrated
the best thing about ai is that the only people left in the craft will be those that sincerely love the craftsmanship by itself, regardless of the economic incentives

>> No.6915164

>>6915150
>live drawing with freakish skills
artists will be the competitive weightlifters in the AI image age. AI artists might show up with a forklift and get laughed out of the building.

>> No.6915166

>>6915164
the people laughing will be people the general public will laugh at, like people who glue themselves on roads to protest climate change

>> No.6915168

>>6915163
pretentious people like you produce nothing of value

>> No.6915169

>>6915166
I'm going to glue myself to a computer to protest AI art.
>as if I'm not already glued for all intents...

>> No.6915201

>>6915088
Clearly you did because you cannot refute anything I said. You make idiotic statements like "dude you lost dude get over it" because for whatever the fuck reason you were incapable of using common sense. You are incorrect. You are utterly retarded. Good night.

>> No.6915218

>>6915201
I'm amazed you can be so utterly wrong and keep at it. It's SO commonly known that selling fan art is illegal and companies can shut it down if they so wished(lucky for fan artists, they don't wish). Run away faggot.

>> No.6915222

>>6915218
why did nintendo have to buy the rights for that super mario bros porn parody?

>> No.6915231

>>6915222
parody is protected under fair use.

>> No.6915234

>>6915222
Nintendo didn't have to buy anything. They chose to buy the rights to the parody to halt its distribution.

>> No.6915245

https://www.owe.com/is-fan-art-legal-fair-use-what-about-mash-ups-copyright-myths-and-best-practices/
Here's an in depth article by Linda Joy* covering laws around fan art. again, if you think selling fan art is totally legal, you are wrong.

*an attorney, painter and former graphic artist/illustrator. She practices intellectual property law, arts law and mediation for artists

>> No.6915434
File: 186 KB, 610x369, aitrannies.png [View same] [iqdb] [saucenao] [google]
6915434

>> No.6915946

>>6915245
NTA
>site covers most shit I knew already
One that's missing is using the copyrighted assets. There was an incident where an artist alley had to take down prints that contained official logos which if you remember sounds a lot like the Inktober incident.