The analogy sucks because it implies AI is in any way beneficial.
He’s a modern day John Henry.
this is clear proof that AI art is soulless and real artists will always outperform AI
As long as progress continues and humanity survives, computer generated art will eventually outperform humans. It’s pretty obvious, as far as science knows you could just simulate a full human consciousness and pull images out of that somehow, but able to run that in parallel, never deteriorating, never tiring. It’s not a matter of if “AI” can outperform humans, it’s a matter of if humanity will survive to see that and how long it might take.
deleted by creator
Real regrettable take, come back in 5 years for a nice snack
To the many, many, downvoters…you’re completely insane if you think AI art which has been a thing for like 18 months won’t improve to the point that it’s better than flesh bag artists ever.
You clearly don’t understand how these things work. AI gen is entirely dependent on human artists to create stuff for it to generate from. It can only ever try to be as good as the data sets that it uses to create its algorithm. It’s not creating art. It’s outputting a statistical array based on your keywords. This is also why ChatGPT can get math questions wrong. Because it’s not doing calculations, which computers are really good at. It’s generating a statistical array and averaging out from what its data set says should come next. And it’s why training AI on AI art creates a cascading failure that corrupts the LLM. Because errors from the input become ingrained into the data set, and future errors compound on those previous errors.
Just like with video game graphics attempting to be realistic, there’s effectively an upper limit on what these things can generate. As you approach a 1:1 approximation of the source material, hardware requirements to improve will increase exponentially and improvements will decrease exponentially. The jump between PS1 and PS2 graphics was gigantic, while the jump between PS4 and PS5 was nowhere near as big, but the differences in hardware between the PS1 and PS2 look tiny today. We used to marvel at the concept that anybody would ever need more than 256MB of RAM. Today I have 16GB and I just saw a game that had 32GB in its recommended hardware.
To be “better” than people at creating art, it would have to be based on an entirely different technology that doesn’t exist yet. Besides, art isn’t a product that can be defined in terms of quality. You can’t be better at anime than everybody else. There’s always going to be someone who likes shit-tier anime, and there’s always going to be parents who like their 4 year old’s drawing better than anything done by Picasso. That’s why it’s on the fridge.
So your argument, if I’m understanding it correctly, is that:
- You believe the model of polygon-based rendering in video games has diminishing returns. No argument. Not sure what this has to do with the generated art which doesn’t have similar constraints and doesn’t work the same way.
- Art is subjective, so calling something better or worse is pointless. Also no argument, this is why it’s absolutely ridiculous for people to be saying all AI generated art is universally bad. It has its purpose in the same way caricature “artists” in European historical districts have a purpose…in theory.
It sounds like we’re on the same page, but you have a reason (which you’ve been unable to coherently represent) you think AI generated art will never improve to the point of being good.
AI art isn’t bad because of its inherent quality (though tons of it is poor quality), it’s bad because it both lacks the essential qualities that people appreciate about art, and because of the ethics around the companies and the models that they’re making (as well as the attitude of some of the people who use it).
AI has no concept of the technical concepts behind art, which is a skill people appreciate in terms of “quality,” and it lacks “intent.” Art is made for the fun of it, but also with an intrinsic purpose that AI can’t replicate. AI is just a fancy version of a meme template. To quote Bennett Foddy:
For years now, people have been predicting that games would soon be made out of prefabricated objects, bought in a store and assembled into a world. And for the most part that hasn’t happened, because the objects in the store are trash. I don’t mean that they look bad or that they’re badly made, although a lot of them are - I mean that they’re trash in the way that food becomes trash as soon as you put it in a sink. Things are made to be consumed and used in a certain context, and once the moment is gone, they transform into garbage.
Adam Savage had a good comment on AI in one of his videos where he said something like “I have no interest in AI art because when I look at a piece of art, I care about the creator’s intent, the effort that they put into the piece, and what they wanted to say. And when I look at AI, I see none of that. I’m sure that one day, some college film student will make something amazing with AI, and Hollywood will regurgitate it until it’s trash.”
But that’s outside the context of your original post, in which you said that AI art would someday be better than what humans can make. And this is where my point about video game graphics comes in. AI is replicating the art in its training set, much like computer graphics seeking realism are attempting to replicate the real world. There’s no way to surpass this limit with the technology that powers these LLMs, and the closer they get to perfectly mimicking their data and removing the errors that are so common to AI (like the six fingers, strange melty lines, lack of clear light sources, 60% accuracy rate with AI like ChatGPT, etc.), the more their power requirements will increase and the more incremental the advancements will become. We’re in the early days of AI, and the advancements are rapid and large, but that will slow down and the hardware requirements and data requirements are already on a massive scale to the tune of the entirety of the internet for ChatGPT and its competitors.
improve to the point of being good.
So… first you say that art is subjective, then you say that a given piece can be classified as “good” or “bad”. What is it?
Your whole shebang is that it [GenAI] will become better. But, if you believe art to be subjective, how could you say the output of a GenAI is improving? How could you objectively determine if the function is getting better? The function’s definition of success is it’s loss function, which all but a measure of how mismatched the input of a given description is to it’s corresponding image. So, how well it copies the database.
Also, an image is “good” by what standards?
Why are you so obsessed with the image looking “good”. There is a whole lot more to an image than just “does it look good”. Why are you so afraid of making something “bad”? Why can you not look at an image any deeper than “I like it.”/“I do not like it.”, “It looks professional”/“It looks amateurish”? These aren’t meaningful critiques of the piece, they’re just reports of your own feelings. To critique a piece, one must try to look at what one believes the piece is trying to accomplish, then evaluate whether or not the piece is succeeding at it. If it is, why? If it isn’t, why not?
Also, these number networks suffer from diminishing returns.
Also:
In the context of Machine Learning “Neuron” means “Number from 0 to 1” and “Learning” means “Minimize the value of the Loss Function”.
Youve gone off the rails here, I don’t know what argument you’re trying to make.
Looks like op changed their phrasing to “outperform” but that has the same definition problems.
In any case the argument I’m making is simple
For a given claim “computers will never ‘outperform’ humans at X” I need you to prove to me that there is a fundamental physical limitation that silicon computing machines have that human computing machines dont. You can make ‘outperform’ mean whatever, same fundamental issue.
Their explanation was wasted on useless people like you.
flesh bag artists ever
Dehumanization. Great. What did the artists do for you to have them this much?
Also, do you have any idea of how back propagation works? Probably never heard of it, right?
We’re all flesh bags, what are you talking about? Explain to me how you are not a bag full of flesh (technically a flesh donut, if you consider the sphincters).
I’ve heard of the neural net back propagation, but I’ve just now learned that it’s called that based on flesh bag neural nets. What about it?
We’re all flesh bags, what are you talking about?
So, in your eyes, all humans are but flesh with no greater properties beyond the flesh that makes up part of them? In your eyes, people are just flesh?
based on flesh bag neural nets
That is false. Back propagation is not based on how brains work, it is simply a method to minimize the value of a loss function. That is what “Learning” means in AI. It does not mean learn in the traditional sense, it means minimize the value of the loss function. But what is the loss function? For Image Gen, it is, quite literally, how different the output is from the database.
The whole “It’s works like brains do” is nothing more than a loose analogy taken too far by people who know nothing about neurology. The source of that analogy is the phrase “Neurons that fire together wire together”, which comes with a million asterisks attached. Of course, those who know nothing about neurology don’t care.
The machine is provided with billions of images with accompanying text descriptions (Written by who?). You the input the description of one of the images and then figure out a way to change the network so that when the description is inputed, it’s output will match, as closely as possible, the accompanying image. Repeat the process for every image and you have a GenAI function. The closer the output is to the provided data, the lower the loss function’s value.
You probably don’t know what any of that is. Perhaps you should educate yourself on what it is you are advocating for. 3Blue1Brown made a great playlist explaining it all. Link here.
Correct, humans are flesh bags. Prove me wrong?
I’m not sure what the rest of the message has to do with the fundamental assertion that ai will never, for the entire future of the human race, “outperform” a human artist. It seems like it’s mostly geared towards telling me I’m dumb.
It’s proof of nothing really. Just because a drawn picture won once means squat. Also, AI can be used alongside drawing - for references for instance. It’s a tool like any other. Once you start using it in shit ways, it results in shit art. Not to say it doesn"t have room to improve tho
Also, imagine if the situation were reversed and an AI drawing was entered instead to a drawing contest. People would be livid, instead of celebrating breaking the rules.
Just because a drawn picture won once means squat
True, a sample of one means nothing, statistically speaking.
AI can be used alongside drawing
Why would I want a function drawing for me if I’m trying to draw myself? In what step of the process would it make sense to use?
for references for instance
AI is notorious for not giving the details someone would pick a reference image for. Linkie
It’s a tool like any other
No they are not “a tool like any other”. I do not understand how you could see going from drawing on a piece of paper to drawing much the same way on a screen as equivalent as to an auto complete function operated by typing words on one or two prompt boxes and adjusting a bunch of knobs.
Also, just out of curiosity, do you know how “back propagation” is, in the context of Machine Learning? And “Neuron” and “Learning”?
Also, imagine if the situation were reversed and an AI drawing was entered instead to a drawing contest. People would be livid, instead of celebrating breaking the rules.
Except that already happened, and people were livid. Your correct assessment of such a scenario says a lot more than your half hearted defences for AI art.
Wat?
It’s the opposite of the OP’s headline.
Aimbot works because being good at games is essentially bending your skills to match a simulation, aimbot can have the simulation parameters written into it.
LLMs are blenders for human-made content with zero understanding of why some art resonates and other art doesn’t. A human with decent skill will always outperform a LLM because the human knows what the ineffable qualities are that make a piece of art resonate.
100% yes but just because I really hate how everyone conflates AI with LLMs these days I have to say this: The LLM isn’t generating the image, it’s at most generating a prompt for an image generating AI (which you could also write yourself)
What a stupid fucking idea for a contest. “Press a button until something interesting pops out. Best button pusher wins.” Glad it got subverted like that.
You seem like you’re fun at parties
Haha, humans taking over AI jobs!
An AI generated contest… thats the most low effort contest ever. Glad this person did what they did and used real skill.
Like a book recommendation contest
even a book recommendation contest could only be this dumb if the only way to participate was to shout out random words and hope it would coincide with a good book title
Concur. I would actually like a book rec event
I highly recommend Schachnovelle by Stefan Zweig
You cheated. Of course if you’re an artist, you’re going to make better art than AI.
Somehow they invented a contest that’s more boring than watching paint dry.
I’m shocked i haven’t seen captchas of the form “choose which image is AI generated”
People who use Lemmy would be able to tell the difference most of the time, but the average person would have zero idea.
Just look at any of the YouTube videos with obviously AI generated clickbait thumbnails that get 10s of millions of views. Or all of the shitty obvious Photoshop thumbnails that existed before AI.
Yes everyone on lemmy is both mentally and physically superior to the average “normie”, who is completely incapable of rational thought.
It’s more like people on Lemmy know what to look for and generally have a hatred of AI “art,” so they’re looking to spot it.
Exactly they are just smarter, their brains are more attuned to being right
finally someone who gets it
Im pretty sure that’s called cheating.
If you listen to the AI preachers at their pulpits, AI gen works the same way that people making art does, so how could it be cheating if they’re the same thing?
on stupid contests like these it’s allowed
it’s like… yeah you can tweak every single parameter and build your own checkpoints and stack hundreds of extra networks on top of one another and that is certainly a skill, but creating art with intent is an entirely different skill. and the first one won’t give you shit if the contest is about creating art with intent.
But is the outcome important part, or which skill was required to make it
Prompt engineering is not easy. Getting something as good as a real artist’s work is very hard, especially if you’re not an artist. Of course actual art is going to win every time.
Yeah, not only promting, but adjusting temperature, combining different checkpoints and loras, addding additional steps to upscale, redraw certain parts etc.
In the end its very much just a numbers game though, see if you got lucky after 100+ images
Prompt “engineering” goes to show that you can stick the word engineer on the end of anything to make it sound official and important. Talking to a chatbot is not engineering of any sort
The trick is the artist invested time
Yeah, a real picture engineer worked hard on it
“Prompt engineering” now that’s a new one, what next? Prompt researching?
Vibe coding
i don’t want to shit on someone’s honbies but i will take dick from a Kamen Rider fanboy over a AI Prompt Engineer any day
deleted by creator
