Art and Random Thoughts - June 24, 2023
The Stairs in the Forest:
This week I’m in the middle of a few incomplete projects, so I’m sharing a piece here I made earlier in the year. In the parlance of classical artists of yore, this piece is an example of me “winging it”. I basically sat down in front of my sketchbook and said “I think I’ll draw some shit with trees”. I browsed Pinterest for an hour to see what it could stir up in the ol’ imagination, and found a photo of some old stone steps in the woods. Surely, I thought, these steps should just keep going up and up and up… What could be up there? Well that was a mystery for another day. I finally had some shit to draw with trees.
This piece was relatively quick to do and I finished it in the course of one night. The big trees were just big smears of graphite powder with a few touches from the old electric eraser to give them some bark. The thing about drawing leaves, is that sometimes when you draw lots of really small things really half-assedly, when you zoom out (or look at it on a phone), the mistakes disappear and it looks much cooler than it does at full size…
Anyway, if you look carefully you might find all sorts of hidden shapes in there. Shapes that might have been old things built long ago by people long gone. Is there treasure there? Probably not. The real treasure, after all, is the health insurance plan you’ll need if you risk life and limb climbing all the way up those stairs.
One thing I learned is that sometimes a simple concept (and a pinch of mystery) will make a great piece. Not everything needs to be a grandiose flying city…
(*but many things should definitely be a grandiose flying city)
Weekly thoughts: where AI art goes from here
One thing I keep wondering is if the next generation will think about AI art the same way we think about innovations from the last couple centuries— from cameras to computers and digital art. Will it become normalized while we become the old curmudgeons dismissing modern technology? As much as I want to shake my fist at it and tell it to get off my porch, I know it’ll be hard to avoid. Especially when we look back at some of the old innovations that have become normal for us…
Back in the 1800’s the invention of photography basically rendered realistic painting obsolete. What you’d once pay an artist to painstakingly capture with a lifetime of skill could be done by a person with a small machine in a much shorter amount of time. But while many painters were at first opposed to this new invention, later on it spurred whole new styles of painting, and ignited the art of photography. (Will AI drive artists into new creative paths too? It’s hard to say. They’d need jobs to do that)
Back in the late 90’s (ahem, 1990’s) when digital art software was just starting to get good, there were some brief debates about whether digital art was “real art”. After all,“real art” at the time didn’t come with undo buttons or perfect color matching, or copy and paste, or magic wands, etc. Some things that used to take a lot of manual labor could now be done with the click of a button. In the fine art world, it took a while for digital art to gain acceptance, and even today it hasn’t really eaten into the fine art market very much. But in the end, digital art did replace traditional art almost completely in some industries— concept art for film and gaming development, art for marketing and advertising — basically anything commercial where the gains in productivity from digital tools could be the most profitable. Ironically these are the domains where AI is threatening to take over the most.
20 years ago cheap stock photography and illustration started becoming a booming online industry. Cheap illustrations, clip art, and photos of handshakes and awkward smiling models could be bought for fractions of a dollar. Projects that used to pay high prices to photographers for custom shoots, (or to Getty images for high priced licenses) were replaced with vast selections of cheap stock photos. Today these microstock companies are a thriving industry with hundreds of thousands of corporate customers.
Today all of these things— cameras, digital art, stock images, have become normalized. With each new “innovation” a set of skills that artists earned a living for was de-valued. Over the years, the groups fighting against these innovations dwindled, and people entering the industry today see all these high-tech replacements as commonplace.
Is “AI art” going to become just as normalized too someday?
Probably. But just as digital art didn’t fully supplant traditional art, AI art still has a number of hurdles that may keep it from fully supplanting human-made art.
Copyright: We’ve yet to see the legal system fully digest this pickle. On the AI-development side, we have companies that sucked up vast troves of copyrighted images from all over the internet to build their for-profit image generation tools. (There’s a lot of businesses that certainly wouldn’t be happy with their media libraries getting harvested for this purpose— we’ve yet to see them move on it legally). On the other end of the pipeline, images generated by AI may not be copyrightable, since they’re not produced by a human, making their use by potential corporate customers rather limited.
An insatiable need for more data. In order to improve their models, AI companies will need ever more images to train on. While AI aims to replace artists, they’d need an ever-increasing supply of human-made art to continue improving the technology. That is, the humans whose jobs they plan to take with this tech. In a way, it’ll be like a snake eating its own tail. Soon, AI may be choking on art that it generated itself, and it may already be doing so…
There will be a limit to how good it can get. Even CGI in films today — while it’s good, it’s still not perfect. In software development, its often “easy” to get a new technology from 70% perfect to 80% perfect, but its often insurmountably expensive go from 98% perfect to 99%, and effectively economically impossible to reach 100%. AI will always have its weak points— niche subjects with little training data, or combinations of things it just can’t quite get right.
We’re still in a lot of unknown territory with AI generated images. The only sure trend I see is that AI art will have no real “value” very soon, if it ever did. That is, “art” that is cheap and can be made instantly in great volume will have no worth whatsoever— either to pay for or even to take the time to glance at. Once the internet is flooded with this stuff, it’ll be the human-made art— art made with care, created with inspiration, where the details matter - that’s the art that’ll be valued.