Bit of a late arrival here–both to the discussion and the whole…AI shenanigens…I am bit of an old school out of touch. So-AI or companies of AI’s use human made arts to train theri AI and tha’s the theft bit or? How does it all work?
2 Likes
With mobile games, alot of them use third party advertisers. And both them and the actual game companies often have their offices in countries where copyright laws are either really lax or non-existent.
In its most summarised form, yes. It also does this without the consent or knowledge of the artists.
So I’m not going to pretend to be an expert on the matter, nor are most here but the way I personally understand it is like this:
You are a programmer, and you want your program - your machine learning program, your AI, however you want to call it - to create anime art. In order for it to do so, it needs to know what humans look like, and what anime looks like. So you feed it images of humans drawn in anime. Through that it knows the shape of humans, where the arms are meant to go, how many eyes a human is meant to have and how big they’re meant to be, and the more images it’s fed, the more accurate it’ll be because it’ll have more references under different poses, light, etc.
The ethical side of this is that the images fed into it can be anything from big companies and their animated films or series, to recently-deceased artists who’s images and style is being approximated by machines in order to make the person using said machine in order to make a quick buck. And without the images put into the AI, it simply wouldn’t work. Some people compare it to the IRL practice of artists using references to ensure that what they are drawing is accurate, but it’s really most akin to recycling. It’s taking existing things and twisting them into something new, compacting thousands (sometimes hundreds of thousands, more recently billions) of images into a crude approximation with basically zero oversight from another human being.
Using AI to create something “good” is incredibly easy, but it’s never special. You can’t express intent or ideas through the press of a button. All you have to do is type “big boobed supergirl blue sky flying” and apparently that’s enough for some people to call themselves “artists”, when in reality if one were to draw - or even photoshop - such a thing into being, it would take far more work.
Tbh claimingit to be real art or claiming to be an artist through ai stuff sounds petty…
Though my understandnig on the topic is…well…basically on the level of 2+2=4 and I do not mean to be rude in any way, shape or form
But as you said isn’t it the same as a human learning by studying other art and media etc? At least to a degree? I draw myself, admittedly not on a professional level, but I tend to use a lot of…arts as reference? or a way to hepl me learn how to draw an arm, where they go, proportions etc.
I too can claim an art is mine after copying someone–isn’t it the act of lying and claiming it to be self made tat is theft or?
The thing is that they don’t really ‘train’ the AI as much as they massively data scrape artists without any regard for their consent or copyright; and the most worrying part, they haven’t just done that with art, but with over one billion images of real, unsuspecting people in Facebook, some of which minors or people long-since deceased.
They recognize aspects that make an image (sort of like Google Reverse Image does) and they sort of just try to mash them together as you use tag lines when you demand the AI engine to process an image. They’re not really trained to ‘create’ anything as much as they are trained to get better to take leftovers from already existing images, putting them together with duct tape and smoothing out the edges.
In academia, quoting any information from another source without at least citing said source and that said information wasn’t something you made up on your own research (and in some cases even asking for permission from the author) is considered plagiarism, and can lead to some very serious things; and that is without entering into the minefield that is copyright for things like art, music, and many other things.
1 Like
Are you familiar with youtube and how videos get taken down if they have certain music used without permission?
The difference really is that a human puts in the effort to learn these things and has a reason to. Maybe someone wants to draw because they saw something pretty and want to do the same, maybe they want to draw because it allows them to express themselves, maybe they want to make it a career, maybe they just want to do something fun. But when a human chooses to learn any art skill, they’re not just copying it - they’re learning to adapt it, they’re learning what they like about it, what they don’t, how they can adjust what they already know to draw it more to their liking.
With AI, there can never be intent. Sure, you can put an expression like “happy”, or “sad”, or “angry” into the prompt and it will make the character(s) it draws appear any of those emotions, but intent can run so much deeper. You can express emotion through drawn style, colours, rendering choices. The ability to express emotion through art - or music, or video, or any creative medium - is absolutely stripped away through AI because it’s not a thinking human with consideration for these things.
I guess to put this as a TL:DR because I am admittedly rambling here (whoops)
A person making a piece of art may create something inspired by a certain style or film or other artist, but an AI will make something derivative of it.
Mhm! Even if you make a remix you need to ask for permission still or it is counted as derivative work. Though making an exact copy of a single song is one thing.
In this case isn’t it more similar to combining a variety of random different songs etc and making a new one out of a really big pool?
A remix of a song implies that you are indeed saying that this is your cover or remix of a song that already exists, while attributing the credit of the original song to said artist.
AI is more like taking bits and pieces from Michael Jackson’s discography, pasting them together with a machine that does it for you, and saying that actually it was all you, and nobody else did anything about it, you just ‘trained’ the machine with some ‘data’ you refuse to disclose in an attempt to not have to pay royalties.
So bascially putting your own touch–not a native speaker–like…finess? a piece of your personal take?
1 Like
Yeah! No two people draw in the exact same way, because no two people have the exact same life experiences or artistic inspiration. But if everyone’s using AI, it’s always going to look the same.
To compare it to roleplay: Any one of us could make a female blood elf hunter dressed in red and gold, but only you would make Tyedréa.
1 Like
I got an image that I can share later. But it has a good quote on it that highlights the problem with AI.
An artist may take 7-15 images to use as inspiration for their next project. It might be a man holding a red apple, a cliff and a rainbow with a thunderstorm in the background.
An artist will create a piece using these images as inspiration to create their own style.
An AI uses them as samples. You will get THAT man holding the apple, THAT cliff and THAT rainbow with the thunderstorm in the background.
Nothing is created. It is taken
1 Like
I get that though isn’t in this case sample size is much bigger? Going by earlier comments–hundreds of thousnds…etc? Wouldn’t it make it A man rather, an amalgamation of all the collected data etc?
Yes. But it is still taking. Not creating.
An artist without internet access can still draw.
An AI without internet is a brick program
Well yes they are not the same, for AI internet is what our brain, limbs, eyes etc are.
I…Likely can’t draw well with my foot if I were to lose my hands.
The most often cited dataset is billions of images, but obviously quite a small part of it is f.e. fantasy digital art - it’s still pretty easy to coax generators into giving you a pretty clear facsimile of a particular artist’s style, or with a prompt specific enough steer it into giving you nearly the same image over and over.
That is deffo bad. Using it to make an exact replica or copy of someone else’s work, my questions were more so nudged towards pieces that came off different due to the larger data set more…generalized, less personal. As that bit fascinates me the most.
But yes, I agree that is…quite–well can’t type in a profane stuff here so you get what I mean.
1 Like
And? Ignoring that people have been shown to be able to draw with their feet with practise… they can still create.
AI can’t. Not yet. It just takes. Could be 10 images or 10,000 images. But it still only takes. Never creates.
This has been something that’s irritated me in the past. I was told “You’re just discriminating against disabled artists” to which I simply linked this:
https://www.youtube.com/watch?v=TpfcFBhclyI
Human willpower can lend itself to endless creativity. I’m inclined to believe anybody who says “I can’t draw, so I’ll just use AI” honestly just has a skill issue.
Or even if somebody truly cannot draw - they have zero motion in any limb or have some sort of physical condition that genuinely prevents them from doing so - there are so many other means of creation. Art is but one of many.
Writing - stories, scripts, poetry, songs, documentaries, opinion pieces, essays, blogging, reviews, even something as mundane as a Twitter hot take is a form of expression.
Music - Picking up an instrument, honing your vocal abilities, writting through sheet music, composing through digital audio workstations, sound design, foley work for films.
I’m kind of blanking on other means of expression RN but what I’ve listed is certainly not the limit on how people CAN express themselves.
…or instead you can just get a machine to crap out an image that looks like thousands of other crapped out images by other bozos willing to take the shortcut.
1 Like