Art and Random thoughts - June 10, 2023
This was a week of riffing on some old ideas in new and maybe interesting ways (we’ll see next week…). I could probably take the last 10 years worth of art that I’ve done and trace each piece to a lineage of riffing; one piece inspiring the next, either from seeing another direction it could have taken, or simply fixating on all the mistakes my eyes latched onto. On that note, this week I’ve reached back to the past to pick a featured piece…
This week’s artwork
The above piece was the result of 2 years of on and off work that I started in 2007. I made this one in a period where I wanted to get very technical, and wanted to make something big and complicated. So I looked at the general panoply of fantasy art at the time to find the biggest castle I could, and set out to make something bigger (because who doesn’t like a big castle? Nobody, that’s who).
The above version was colorized in photoshop; the original piece was done in graphite; and the original outline for it was a Process (with a capital P there) that managed to bog me down for most of the two aforementioned years. I plotted this whole thing out with architectural side views and top views and wrangled that all into some sort of 2-point perspective system that involved T-squares, triangles, and even a parabola (don’t ask).
At the time, my only presence on the internet was an art website that nobody visited and a DeviantArt account that also nobody visited. This piece, however, somehow made it onto the front page as a “Daily Deviation” which gave it the largest burst of viewers that any of my pieces had gotten until that point.
Although that reception was great, a small thing I took away from this experience was that I never wanted to spend 2 years working a piece ever again. I’ve found that the longer I take to work on a piece, the harder it can be to stay motivated to see it through. (If I spend months on a piece, it’ll really burn my muffins if it doesn’t turn out really good). Last year I set out to challenge myself to make one piece a week, often just winging it, experimenting, and seeing what works. If it doesn’t, then I’ll just make a new piece a week later. I’d say I’ve learned more in a year of doing that than I did in any year prior (but more on that another day…)
Weekly thoughts: Inspiration vs AI
It seems these days that AI-generated images have been sucking a lot of air out of the room (a.k.a Art Twitter). One argument I keep hearing from AI promoters is how AI is just “learning” to create art the same way a human artist learns. They suggest that a computer digesting artwork and generating images is the same as an artist viewing another artists work and then making art of their own. The implication they suggest is that it’s therefore not “theft” when a for-profit corporation harvests millions of pieces of art online to train its AI, because it’s just doing what artists can do.
Of course, this shows no understanding of how an artist works and it also shows little understanding of how AI works. AI can only do what computers have always been good at: consuming data, and producing data. To AI, artwork is just data. It reads pixels of an image, it finds common patterns among them, and if it sees enough of them, say 10000 pictures of a dog, then we can say it “knows” what a dog is. And with that data it can find the common bits between them and generate an image that might statistically be likely to look like an average dog.
I don’t know about you, but I don’t know any artist whose process remotely resembles that. Artists aren’t photocopiers, nor do they simply produce some sort of averaged-out version of all the pictures they’ve ever seen. Nobody looks at a thousand still-lifes then makes a still-life that’s just an uncanny mix of them all. The human mind is so much more advanced than that. A person could meet a dog once and make art based on that one encounter. They could make art inspired from many things they’ve only had fleeting glimpses of. A person could string together concepts totally unrelated and make a piece of art from it. An artist might see a single image of something and then make something wildly different in response. To put it bluntly, artists can come up with crazy shit, and that’s what makes artists great.
Another major difference is that AI has no conceptual sense of what’s in the images it’s consumed. To AI it’s just flat, two-dimensional data. It doesn’t know that dogs walk and bark; that they have muscles and skeletons and stomachs that want to eat literally everything; it doesn’t know that they’re alive, that they feel happiness and sadness, and may obey or misbehave; it doesn’t know what it’s like to have a dog for 10 years of your life, and those are things that an artist and not an AI could comprehend and use to inspire a piece of art. To AI, a dog is just a bunch of pixels.
(For the record, I’m a cat person. Never had a dog; don’t know why I started using a dog example, but once I started I had to run with it. Sorry, cats, didn’t mean to offend)
Anyway, anytime someone suggests that image-generating AI is “just doing what artists do”, remember that these programs are just spitting back averaged-out pixel data, and they definitely are not doing any of the actual weird shit that artists really do.