How does it work then? I see lot’s pf people claiming to know how it works… only to not actually know how the training works exactly, only a superficial understanding.
How is access limited and at the same time you are bullying everyday Joes who are actually using it?
Ah yes, because people in 3rd world countries earning $1 an hour or less to label that data for the image gen can 100% afford the $10/month for a subscription or a pc to run locally.
Delete all software and turn off your computer or be a hypocrite.
How so?
The stuff they use for training is free for any artist to train on.
The fact that you think AI training and humans looking at thinks are the same thing tells me you don’t know how humans art nor how machines train.
You don’t own the definition of art and nobody you will encounter in a post of any sort is even doing it for major profit.
True. However, this argument should not be about semantics;
I got news for ya.
You don’t own the definition of art.
This is not about definitions, I won’t spend time arguing semantics with you. Also, why re-state yourself?
AI is for everyone, but is made for the rich to get richer, like literally everything else you see or do online
Without social development, all forms of technological development will do nothing but allow for greater forms of torment.
The fact that you think humans don’t use neural networks trained by experience to generate art (or anything else we do) tells me you don’t know how humans art nor how machines train.
Simply, Diffusion models somewhat (and this is extremely hand wavy) work as the “reverse” of an image recognizer.
It is taught the concepts of images through image learning (build neural circuits) to detect image features. Then, you do the reverse, iterate on noise to generate features.
I don’t care how you perceive the term art. This ain’t art. The Meta lawsuit comes to mind. The one where they were caught illegally training their LLM on authors’ works without their permission, using a pirated source, while still trying to argue that it was perfectly fair.
If I use computer software to type up a letter instead of writing it, I’m not benefiting off the backs of everyday joes
If I use a computer to calculate math, I’m not stealing a working Joe’s job
If I use a computer to type a prompt for an image generator and it spits out an image, I’m benefiting off the backs and the works of the unknowing artists the AI vacuumed up
If I use AI to write a book, I’m benefitting off of the authors’ works that Meta never paid any money to, while shadily downloading all of their books from torrent websites
Your comparison here falls flat because AI image generation is a unique scenario. Computers aren’t the issue; corporate AI is.
The AI community had an opportunity to be conflict-free and fair. A public utility that wasn’t created via theft and exploitation. Companies had an opportunity to ask the art community to willingly contribute to it and have something everyone can equally benefit from. Capitalists took that opportunity away and fucked up the entire thing.
Capitalism is the core of the problem and AI art ain’t art.
And I’m sure all the AI everyone gets to use, are “collateral” products, that were realized, while they keep the goal of creating the AI that will ultimately replace all the employees and make the rich independent of the very annoying human workforce in all areas.
Edit: lemmy kept converting the 4 and 7 to numbered bullet points, converting them to 1. And 2.
That’s why the formatting on the numbers is strange, using only blank spaces to separate.
I wasn’t wrong but i should have qualified it. There are instances where companies have pirated art, but the majority is stuff you can freely access online. I agree that they shouldn’t have the pirated art that was behind pay walls. What they do with it isn’t the problem there, it is that they have it. I should have said that the pirating of art isn’t fundamental to the process and really probably was due to overzealous people tasked with finding data to train on and who, like most of us, grew up in the Naptster/limewire era.
![[Rusty Creates] 'Artists'](https://lemmy.world/pictrs/image/ee6ccbcc-d001-4b28-a88f-d39c2b75f860.png){kind=link}
How does it work ?
Paywalls limit access, cost of hardware to run locally limits access.
Can some people access it, yes, is access limited, also yes.
Strawman? maybe?, it’s unclear how it’s related and as a singular statement is mostly nonsensical.
It absolutely is not, there are several ongoing lawsuits and repeated strikes about this exact thing.
This i agree with.
I agree with this also.
AI is for profit, not for everyone.
The major difference here is the scale but you’ll have to look in to that yourself.
How does it work then? I see lot’s pf people claiming to know how it works… only to not actually know how the training works exactly, only a superficial understanding.
Ah yes, because people in 3rd world countries earning $1 an hour or less to label that data for the image gen can 100% afford the $10/month for a subscription or a pc to run locally.
How so?
The fact that you think AI training and humans looking at thinks are the same thing tells me you don’t know how humans art nor how machines train.
This is not about definitions, I won’t spend time arguing semantics with you. Also, why re-state yourself?
Without social development, all forms of technological development will do nothing but allow for greater forms of torment.
The fact that you think humans don’t use neural networks trained by experience to generate art (or anything else we do) tells me you don’t know how humans art nor how machines train.
Simply, Diffusion models somewhat (and this is extremely hand wavy) work as the “reverse” of an image recognizer.
It is taught the concepts of images through image learning (build neural circuits) to detect image features. Then, you do the reverse, iterate on noise to generate features.
I don’t care how you perceive the term art. This ain’t art. The Meta lawsuit comes to mind. The one where they were caught illegally training their LLM on authors’ works without their permission, using a pirated source, while still trying to argue that it was perfectly fair.
If I use computer software to type up a letter instead of writing it, I’m not benefiting off the backs of everyday joes
If I use a computer to calculate math, I’m not stealing a working Joe’s job
If I use a computer to type a prompt for an image generator and it spits out an image, I’m benefiting off the backs and the works of the unknowing artists the AI vacuumed up
If I use AI to write a book, I’m benefitting off of the authors’ works that Meta never paid any money to, while shadily downloading all of their books from torrent websites
Your comparison here falls flat because AI image generation is a unique scenario. Computers aren’t the issue; corporate AI is.
The AI community had an opportunity to be conflict-free and fair. A public utility that wasn’t created via theft and exploitation. Companies had an opportunity to ask the art community to willingly contribute to it and have something everyone can equally benefit from. Capitalists took that opportunity away and fucked up the entire thing.
Capitalism is the core of the problem and AI art ain’t art.
Edit: a wild article appears.
hallelujah.
kill the capitalists, not the technology.
4 Is definitely wrong
7 Yes
And I’m sure all the AI everyone gets to use, are “collateral” products, that were realized, while they keep the goal of creating the AI that will ultimately replace all the employees and make the rich independent of the very annoying human workforce in all areas.
Edit: lemmy kept converting the 4 and 7 to numbered bullet points, converting them to 1. And 2.
That’s why the formatting on the numbers is strange, using only blank spaces to separate.