It’s becoming pretty clear that some people are much better at harnessing these emerging tools than others. Some people are already billing themselves as “AI artists,”
This is evil bullshit. Whatever creativity or value rather an artist creates with these "tools" will be sucked out by the AI directly from the photoshop canvas and recycled the next day except the artist wont be necessary this time.
and I predict that many will find success in this new art form.
Until the pupil has surpassed its many masters obviously.
At which point Parent will be on Mars.
The artist is necessary for creating anything legitimately new that doesn’t suck. The critic is necessary to filter out all the crap the AI would be happy to create. Look at all these examples where the text prompts literally say which artist’s visual style to copy.
Making derivative artwork is in the process of becoming completely commoditized. That’s good in the sense that it allows novices to create visual content that they previously could only imagine (presuming they can adequately describe it).
But this also means the half-life for visual imitation will be drastically shortened, because even viral new styles will be done to death by artful copies in a matter of weeks.
Is XKCD art? Visually unchallenging (?), but certainly creative and thought provoking. What we’re mostly talking about at this point is giving kids (and some adults) who can’t draw well yet a chance to express what’s in their heads onto the page.
Creating compelling results which connect with people requires practice, patience, the willingness and ability to explore, a whole lot of cherry picking. Iteration.
There will have to be protections for content creators, which will have to evolve to accommodate AI’s increasing capabilities to mimic, some might say “steal,” the style and character of an artist’s work. It’s simpler in the case where the AI clearly imports entire assets from a single source, much harder when it almagamates tiny fragments of a large number of different artists’ works. Some kind of tax may be necessary to ensure that human creators of source materials are fairly compensated. Nothing happens without their work as the basis of creation.
You perfectly know fine art isnt the issue here because its a small industry anyway and you guys arent investing dozens of billions to take over a niche industry.
For those who dont know what we are really talking about here: it's the illustration and entertainment industry, which combined are probably well over a trillion dollar industry.
Creating compelling results which connect with people requires practice, patience, the willingness and ability to explore, a whole lot of cherry picking. Iteration.
It doesnt. And you know it. It only takes taste. Which is very easy for an AI to acquire with intelligent supervision. It will be way better than the 99 percentile of artists very quickly. The rest is monkey work and is worth nothing.
There will have to be protections for content creators, which will have to evolve to accommodate AI’s increasing capabilities to mimic, some might say “steal,” the style and character of an artist’s work. It’s simpler in the case where the AI clearly imports entire assets from a single source, much harder when it almagamates tiny fragments of a large number of different artists’ works. Some kind of tax may be necessary to ensure that human creators of source materials are fairly compensated. Nothing happens without their work as the basis of creation.
One of the first things I noticed on the SD website is the cleverness of the bullshit. With "ethic" and stuff everywhere. Did you wrote it? Are you CEO?
There wont be taxes mate, there will be questions :)
Seriously, I don’t know that. According to Statistica, the total value of fine art transactions was over $65 billion in 2021. But fine, let’s take it as stipulated: commercial art is a far bigger market.
If there really is that much at stake, change is inevitable.
“It doesnt. And you know it. It only takes taste. Which is very easy for an AI to acquire with intelligent supervision. It will be way better than the 99 percentile of artists very quickly. The rest is monkey work and is worth nothing.”
You are (aggressively) psychic! As an AI practitioner, if I could so easily bottle up taste and sell it, I surely would. Supervised learning requires a source, and current methods require vast (and thus expensive) corpora of training data. The better it understands what is “good” by the standards of that data set, the less ability it has to generate novel work and redefine what is, in fact, good.
Visual style, in particular, is subject to fashions whim, what ML practitioners call “concept drift.” What is exalted one day, is scorned and ridiculed in another.
If taste were so easily captured in a model, Hollywood would never have to worry about making another bomb. Game studio executives could sleep blissfully. A vast amount of money is lost due to bad design choices, across countless industries. Frankly, I think a Taste Oracle is worth more than an AI Snuffaluffagus.
“One of the first things I noticed on the SD website is the cleverness of the bullshit. With "ethic" and stuff everywhere. Did you wrote it? Are you CEO?”
I am just not that clever. I see all that as CYA. The fact that they released everything, code and models, undermines the notion that they expect to corner a multi-trillion dollar market.
Paranoia, anger, aggression, maybe this is appropriate in the face of unknowable change. But people are divided at the outset of this revolution, and remain conquered by the future Mars dwellers.
To clarify, I work (Earthbound) in Computer Vision and Perception; my preference is for concrete problems.
As an AI practitioner, if I could so easily bottle up taste and sell it, I surely would.
And would you lie to protect it?
Snuffaluffagus
You're the aggressive one by my book :)
To me taste is subjective and ill-posed. I accept that people have the right to their own personal tastes.
What would you lie to protect?
If people can make art out of literal garbage, I think they can make art by manipulating a Turing-complete class of machines (recall that LLMs can perform program synthesis).
I agree that naive and even not-so-naive usage of AI tools will mostly be artistically superficial, but I believe that (some) people will break through the creative limitations inherent in the copycat nature of this machinery.
Now maybe thats okay. Maybe there is no need to lie about it. If there is paranoia, it doesnt come from me I believe.
And I'm not talking about emotions...
It’s becoming pretty clear that some people are much better at harnessing these emerging tools than others. Some people are already billing themselves as “AI artists,” and I predict that many will find success in this new art form.