AI art and human obsolescence
Part of Musings on philosophy; related to The human knowledge horizon.
Musings dated 2023-05-06. I wonder what I’ll think of this article in the future.
AI art is solid cocktail conversation topic these days because anyone can weigh in. Those sympathetic to artists muse sourly over the ethics of training AIs without permission. Engineers and economists discuss compute costs together. In the corner, zoomers and philosophers take shots.
Part of what makes the topic fun is that largely, everyone already agrees. Stealing is bad, AIs who are more competent than humans are scary, and capitalism is probably going to produce a situation nobody is happy with. In general, most people seem to regard the future of AI art negatively.
My opinion
I want to document a contrarian take, that AI art is actually a force for net good. It will (1) allow the world to produce a greater quantity of art, (2) spur the average person to train a more critical eye when interpreting art, and (3) have a net zero or net positive effect at lowering the barrier for individuals to produce meaningful art.
1) AI artists will mean a greater quantity of art
This is for certain. It’s already tough to distinguish some AI art from real art (see Stable diffusion fraud for my experience with that). Sure, quality is subjective, but that AI art shares the same believability ballpark with human art convinces me it’s already on par with most subjective experiences of art.
2) Which means people will train a more critical eye
With one’s senses flooded by generated (and retouched) art, the only way to cope will be finding a way to filter away bad art. But what is “bad art?” People will need to decide on their own. And in the process of doing that, I suspect the average person will develop a more robust opinion on aesthetics.
Or, maybe not, and everyone will continue to overvalue art with hype surrounding it. But IMO this is no worse than today’s status quo. I sense many people defer their opinion on aesthetics to someone else, usually a museum curator or gallery owner, unconsciously allowing that person’s opinion’s on art to color their own. And so, under the right circumstances (such as visiting a museum), people can convince themselves they like the aesthetics of virtually any kind of art.
So, I see AI art is sort of a meritocratizing force to this. Pretty soon, stable diffusion will be accessible to anyone with a good idea. So, anyone with an idea can create art, thereby making art less exclusive. The average person will need a more critical eye to filter through the noise, and dig deep to understand why they prefer certain art above others.
I like to think this will breed more thoughtfulness. Or maybe it will breed disdain for anything not human-created. I’m optimistic we as netizens will find a middle ground.
3) And for those seeking to create great art, the barrier to entry will either lower or remain unchanged
With mainstream AI art, I predict one of the following futures:
(A) more art is produced, in general
(B) more art is produced, in general — and it is of a higher quality.
There’s probably some variance in opinion on whether (A) or (B) is more likely. It’ll depend on what you consider “good” art (a whole article topic on its own). I’m torn on this topic because I suspect AI artists will result in a ton of generic art flooding the internet. Is generic == bad? Right now I feel the answer is “no,” but maybe I’ll change my mind.
This also assumes demand for art is roughly inelastic. Right now I feel the answer to that is “yes.”
Anyways, with the barrier lowered, I think more art will be created, thereby increasing the odds that some aesthetically great art shows up.
A strong argument against this section is that “created” != “discovered.” That is, even if a great artist is out there, they have a harder time distinguishing themselves and becoming findable. I don’t have a great answer to this. I hope it doesn’t turn out this way, though.
Appendix
I glossed over some huge topics that usually come up in cocktail conversation so I’ll try to cover them here since they’re interesting.
The churn of art culture may accelerate
From a conversation with Christie:
Hot take but I think in terms of art created it will be close to a net good
True creativity these days already comes from people purposefully breaking the mold and doing something funky w it
Ai art will become the new mold, everyone’s gonna use it as a blueprint for their drawings
Then the drawings will all look samey
Then someone will break the mold
Then the mold will change
Repeat
Maybe, like, the cycle of art culture will accelerate
But that’s all I think
Capitalism’s reaction to AI art might be scary
The idea that a team of environment artists on a cartoon might shrink to 20% of its size is spooky. Netflix already tried this and credited an AI artist in their credits to serious backlash, but I doubt the negative sentiment will last.
I’m fortunate and grateful that my income isn’t from art so I’m not threatened. This lack of threat surely influences my view on AI art being good, too.
The problem of operator deficiency
I saw a meme post about how artists shouldn’t be worried about being replaced because people can’t communicate what they want to the AI and I LOL’d
This is pretty true imo. The job of a stable diffusion model is to extrapolate words (a frickin tiny dimensional space) into a whole image (frickin huge dimensional space). There’s so much room for randomness that the best models are differentiated by their method of effectively injecting randomness.
Like the whole thing is so chaotic. Surely there will exist a tool in the future that’s more effective than words in the future, an Adobe Photoshop of Al art, with dials and knobs like “lock alpha layer” that will do something funky with the image.
The future of the robo-
Not very well collected thoughts but just a musing:
Visionary artists will probably inpaint or retouch their work using AI. In that sense, stable diffusion is basically a hyper-advanced paintbrush. You can let your robo-paintbrush totally loose and allow it to handle (basically) the creation of your whole art piece, or you can use it with intent.
I think AI art powered by human intent will start winning out over generic AI art pretty soon.
More thoughts
2023-06-06: https://reddit.com/r/StableDiffusion/comments/141hg9x/controlnet_for_qr_code feels like an evolution of art. This is a new thing altogether that relies on human creativity yet uses the new tools.
2023-09-05: More on AI art as a meritocratic force. I think anyone being able to create art means we’ll see a greater diversity of perspectives.
“The old masters” were masters of their craft because virtually nobody had the time or economic privilege to spend time learning art. So while that period produced beautiful art, it was an echo chamber.
Individual artists came next, pushing forward weird movements like impressionism that was resisted at the time but eventually appreciated.
Between then and now, art has become increasingly accessible. AI will make it even more so.
2023-12-01: Thoughts on formalism.
2024-10-15: Scott Alexander posted an AI art turing test: https://docs.google.com/forms/d/e/1FAIpQLSdqpfY0OXLQoO_UNkhKTAtQbmh8EX_xpAAaGV6mxlBDms9CzQ/viewform
5- but a reluctant 5, because there’s plenty unethical about where these models come from. That being said: any answer other than 5 feels a little disingenuous, unless the respondent rejects artistic formalism entirely (i.e. they believe art must always be judged as a product of its historical or author’s context).
To criticize AI art as bad while confusing it for human art is basically admitting that human art can be bad just as AI art can be good. Which seems like a perfectly valid take. But — if that’s your take — I think you’d have to vote 5 here.
2024-10-26 Here it is happening in real time! https://www.reddit.com/r/ChatGPT/s/rX3lZza8SN someone knows it’s AI art and appreciates it all the same. The comments are a minefield