AI Art Thoughts

It's fair to say that AI art such as Stable Diffusion, DALL-E 2, and MidJourney are really popular right now. Just yesterday, the Cortex Podcast released an episode about the ethics of AI art.

The State Fair Art Contest Affair

A notable source of outrage against the ideas of AI artwork was an AI-generated art piece winning the Colorado State Fair art contest. Here's the cliff notes of the piece based on the creator's comment: - He used a "special prompt" - He generated "100s" of images with it, tuning it over time - He selected his top three images, upscaled them, and had them printed to submit in the digital art category The image itself is available for viewing on Wikimedia and is rather pretty. It lacks a lot of detail, but you can get a semblance of what's going on in it.

I don't think that AI art should be put into the contest as "traditional" digital art and especially not with traditional-traditional art. I think this because I believe that the skills being judged for those types of art are not being put to use in AI art.

Jason Allen (the creator of that winning AI piece) said that he had to generate hundreds of images before settling on just three. Even accounting for time taken to tweak prompts and parameters, the biggest hurdle to AI art seems to be processing power and time spent generating the images. If that's the case, deeper pockets will immediately mean better art because your "skill" will scale with how much you want to spend on cloud GPUs.

Contrast this with "traditional" digital art (which I'm just going to call "digital art" from now on): someone with Photoshop can compete with someone using Krita and they largely have the same tools available. There will be small differences at higher skill levels for niche tools exclusive to one or the other, but fundamentally both rely on the talents of the artist more than how good their computer is.

Likewise with traditional art, someone using pro-level paints versus student grade paints will still produce similar levels of quality. The best video example I found after a short search was this one. You can tell a difference between the two pieces, but they're both impressive and have greater dependence on the artist's skill than their tools.

To claim that the "skills" of the AI artist won the digital art competition is comparing apples to oranges. It's like claiming that I'm a math wizard who cracked the NTLM password hashing because when I happened to be good at prediciting a user's password based on the company they work for. It's a totally different skill set.

I feel kinda awkward talking about the AI art. I'm no expert on machine learning. I'm also no expert on art. I've been learning a lot about the latter, though.

I find it ironic that I decided to start learning to do art just as AI art started to take off.

I wrote the above a few days ago. Today I learned that furry-community art website FurAffinity has formalized a policy on AI art.

2.8 Content Lacking Artistic Merit
Content lacking artistic merit is not permitted on FA, and includes items such as:
-- Submissions created through the use of artificial intelligence (AI) or similar image generators.
...
Content created by artificial intelligence is not allowed on Fur Affinity.
AI and machine learning applications (DALL-E, Craiyon) sample other artists' work to create content. That content generated can reference hundreds, even thousands of pieces of work from other artists to create derivative images.
Our goal is to support artists and their content. We don’t believe it’s in our community’s best interests to allow AI generated content on the site.

Good policy, good reasoning.