Can someone give me the rundown on the AI art debate?

>because they're fucking retarded
No, they're simply not sufficiently trained. They are getting better at training every day.

true

it's good enough now to destroy artgays, apparently

we're reaching levels of cope not previously thought possible

nobody is saying it will keep improving forever
and that doesn't need to be the case for it to be good
some of the output right now could reasonably be considered good, so saying it will never be good doesn't even make sense

>learning by looking at art is bad
Why do idiots talk in circles?

see

https://i.imgur.com/UPXjExu.gif

>let's explain latent space to art majors who can't count

>racist
sheesh, really?
does your reddit friends know?

Typical chud dogwhistle.
I thought the ai-cucks were the racists, very sad to see they aren't the only ones.

it does, because it needs to readjust its weights to match the input.
Why are you lying like a retard?

stop lying, how can it refer back to the original input to correct the output.
You don't know

stop lying
Training data[edit]
Stable Diffusion was trained on pairs of images and captions taken from LAION-5B, a publicly available dataset derived from Common Crawl data scraped from the web, where 5 billion image-text pairs were classified based on language, filtered into separate datasets by resolution, a predicted likelihood of containing a watermark, and predicted "aesthetic" score (e.g. subjective visual quality).[15] The dataset was created by LAION, a German non-profit which receives funding from Stability AI.[15][16] The Stable Diffusion model was trained on three subsets of LAION-5B: laion2B-en, laion-high-resolution, and laion-aesthetics v2 5+.[15] A third-party analysis of the model's training data identified that out of a smaller subset of 12 million images taken from the original wider dataset used, approximately 47% of the sample size of images came from 100 different domains, with Pinterest taking up 8.5% of the subset, followed by websites such as WordPress, Blogspot, Flickr, DeviantArt and Wikimedia Commons.[17][15]

End-user fine tuning[edit]
To address the limitations of the model's initial training, end-users may opt to implement additional training for the purpose of fine-tuning generation outputs to match more specific use-cases. There are three methods in which user-accessible fine tuning can be applied to a Stable Diffusion model checkpoint:

An "embedding" can be trained from a collection of user-provided images, and allows the model to generate visually similar images whenever the name of the embedding is used within a generation prompt.[34] Embeddings are based on the "textual inversion" concept developed by researchers from Tel Aviv University in 2022 with support from Nvidia, where vector representations for specific tokens used by the model's text encoder are linked to new pseudo-words. Embeddings can be used to reduce biases within the original model, or mimic visual styles.[35]
A "Hypernetwork" is a small pre-trained neural network that is applied to various points within a larger neural network, and refers to the technique created by NovelAI developer Kurumuz in 2021, originally intended for text-generation transformer models. Hypernetworks steer results towards a particular direction, allowing Stable Diffusion-based models to imitate the art style of specific artists, even if the artist is not recognised by the original model; they process the image by finding key areas of importance such as hair and eyes, and then patch these areas in secondary latent space.[36]
DreamBooth is a deep learning generation model developed by researchers from Google Research and Boston University in 2022 which can fine-tune the model to generate precise, personalised outputs that depict a specific subject, following training via a set of images which depict the subject.[37]

This is why I said "when they use it to generate images".

>looking at images is bad
You are a retard. You will always be a retard. You are also a lying piece of shit. Your suffering is your own making. You choose to be stupid.

you are retarded. The ai process is circular. Its a feed back loop thats how it learns.
Ffs you people are morons

You know how in Excel, you can fit a line to a series of points on a graph? Machine learning boils down to that, but the line is wibbly, and in billions of dimensions instead of 2.

There are a bunch or weights associated with each color and shape and they 'turn on" when you give it the right input.

https://youtu.be/1CIpzeNxIhU

Alright, thanks. I think I got the gist of it. The video is great.

its only 3

how does it make any sense a non-human can claim inspiration? retard

your computer cannot be inspired you idiot. You ai shills are applying human emotions to a mindless machine.

a computer does not have emotions you moron

you are mentally disabled

Anon, you think a computer has emotions. You are either a bot or intellectually disabled.

you are mentally retarded, this isn't debatable. You think your computer is a person

that is what you are arguing. I am arguing a computer is not human. Everything else is an irrelevant strawman. You simpleton.

A computer cannot claim inspiration you retard.

yes it is. being inspired is uniquely human. you are an idiot. Again you are trying to apply human characteristics to a machine like a moron

intelligence and selfawareness what you lack poojeet

>he thinks AI is actually intelligent

you are actually dumber than an AI which is not intelligent at all.
Protip: use a chat ai to argue for you.

you think your computer can talk to you, you are a schizo tard

you genuinely are an idiot aren't you? you have no self awareness at all.

my god you are subhumanly stupid. AI isn't true AI

So you admit to lying.
Good to know.

be honest you don't know how it works. Don't claim you do.
Anyone that has looked at it knows its copying artists.

you are retarded

you are.

its truth you are certifiably retarded

no it isn't, you people are clinically retarded. Did Einstein steal when he thought up his theories? he looked at a clock.

Kek there it is the retard is a joo ranting poltard. If everyone stole and there us no inspiration why didn't have cars, computers and planes 10k years ago?

Answer the question esl, Why didn't we have the modern world 10k years ago. If everything was solved in your view.
You are a retard that lacks genius.

sound pretty dumb, but of artists are not so tech savy, tell them computers are just machines

>poltards
who are they?
we are trans-humanists here

Ai art is real art!
Transwomen are real women!

yes
/pol/morons not welcomed

Its not soulless. You are the one without soul. Of course you could never comprehend the nature of soul when you are spiteful wretch. A sad and pathetic create, not deserving of pity.

Why can't people appreciate all kinds of art? Who made you the gatekeeper? No one. You are fucking nobody.

Ok, you can go be interested in "REAL art" in your little corner and stop bothering us.

>pictures of taylor swift with bits of dog in them
SOVL

I never said they were, you're missing the point. YOU claimed that banning AI art now would be a retroactive punishment and therefore all acts being committed today are legal and will never be punishable. I'm saying that if a court rules that the use of AI derived from copyrighted laws constitutes copyright infringement under the laws which ALREADY EXIST, you're wrong.

https://i.imgur.com/i6OAN6A.png

Have you ever profited of fanart, anon?

tell that to all the small time artists who sell porn of characters from existing anime

[...]
literally who

Obviously there are limits a company is willing to go to enforce its copyright. But one major difference is that the most litigious IP lawyers of all are the stock image owners and similar, who even managed to successfully sue Google into gimping Google Images.

https://archive dot ph/UFqY3

>No, you can't retroactively punish someone who did something that was LEGAL when they did it.
But its illegality is unknown. Let me put it another way: if I deliberately dump a massive amount of a new chemical into a water supply, not knowing the long term effects, and it ends up causing mass death, I can definitely still be charged for murder, even if the illegality was unknown at the time. The question isn't the illegality of copyright infringement, it's whether AI art constitutes copyright infringement. And if it is ruled that it does under older laws, it will therefore be illegal even prior to the ruling.

>Copyright is FIRMLY ESTABLISHED
kek copyright law wouldn't be massive if it was
>and DOES NOT COVER PROTECTION OF STYLES OR COMPUTER TRAINING
Only a couple of state level courts have made any rulings on it, and the rulings haven't even been consistent. If I train a model on a single image and it outputs something extremely close to that image, I've obviously still committed copyright violation.

>kek copyright law wouldn't be massive if it was
That's what makes it so firmly established, every fucking edge-case has been argued in court

>If I train a model on a single image and it outputs something extremely close to that image, I've obviously still committed copyright violation.
No one is arguing against that, do you have any other strawmen arguments ?

>Then you've acknowledged that AI art can be copyright infringement
Yes, if you use an AI and you create a 1:1 or close to a 1:1 version of existing copyrighted art, you are in copyright violation, just like with every other method.

Nobody has ever argued otherwise.

Nobody is interested in generating or training a model to do that, as it is pointless.

>look at art is stealing
Artists cannot comprehend how wrong they are. They are on such a low level of existence, its a wonder they can function in society.

Theres no point in explaining how something works that you can never comprehend, so I wont bother wasting either of our time.

fatter!

I SAID FATTER!

That isn't AI made, but than you for raiding his Patreon.

>r34/ehentai
Thank you for telling me.

I know you don't know how it works. You're like a five year trying to lie to an adult. It doesn't work.

low effort trolling my man

kek
>fanart is the same as a program that rips the code from images, weights it, farts out a duplicate then repeats until it has a knock off.

>farts out a duplicate
You are becoming increasingly desperate with your bullshit

yes, you can over-train it to get that effect
but it's not the intended effect

you don't understand it
you are taking an extreme example to support your ignorant biased view

you are lying through your teeth and hiding behind an authority fallacy like a fraud.
I've researched how it works and it is stealing from artists. It doesn't matter how it does it.
Its being fed their art and out putting art like theirs. Full stop.

liar.

white people lie, it's known

>he thinks its about ethics
you fucking tard, the black box issue is when they neural networks get so large no one can understand them.
Its humanly impossible to understand it.

when we simplify the sample size we see that it copies the image exactly.
People have likewise created understandable models showing that for AI comps

yes? give it two images it blends them together

The fact you can't understand it has nothing to do with what its doing.
You cant understand it because its doing a million crossword puzzles faster than a human can by trial an error to reach its answer
The cross word puzzle is that its scrapping the code for colour, roundness, contrast everything

>artists looked at my art and now they make art that looks like mine!
>NOOOOOOOOOOOOOOOOOO!!!!!
Please shit yourself to death.

do you? you keep going back to this fallacy. Its not an argument.

I can also Control + C and then Control + V. Don't need an AI for that. Photoshop already does this.

>models that are a few gigabytes have 5 billion images in them
lol lmao

>rips the code from images
This doesn't even mean anything lol

There is no "code" in digital images anon. They are just numbers.

There is no "code" in either digital image or sound data. A digital image is constructed of a header specification that tells you its format and dimensions etc, and then arrays of integers representing the pixel data in each colour channel. Its just numbers.

How are you defining the word code here?

except prompt engineers

except prompt engineers

Prompoofters

A Promteur. It sounds more fancy.

prompt autist*

I am BUTHOQN TOV
that's my sig 😉