If anybody tells you that AI generated art is the bee's knees ... don't you believe it. Dream Wombo handed me a lot of rubbish today, but I saw the potential in some of it and spent about seven hours painting the images you see here. AI is a very mixed bag: as a concept, it's not fair to "proper" artists, and at the same time, 90% of the results you get out of it are ... weird. My experiments are almost finished ... horses with five legs? People with seven fingers on one hand? People with two heads? A horse with two torsos? Say, what?!!
So, I think I'm just about done experimenting. I've learned a lot ... learned what I don't like about AI generated images, as well as what I do like. Some engines are better than others; none is as hot as they're reputed to be, but the best can certainly produce the raw materials to make excellent pictures.
The one thing I dislike the most about AI (and this is the thing that would/will stop me using it very much) is that it's totally random. You know, in your mind's eye, the image you want. The engine hands you something that's so idiotically different, you wonder what it was thinking. Example: You ask for a teacup with teapot ... it gives you (I kid you not!) a woman with thirteen fingers and flowers spilling out of the left side of her skull, holding a teacup the size of a bucket, while a teapot the size of an oil drum grows out of her left elbow.
What?!!!!
So, experiments complete. I might play with AI occasionally when I'm totally bored, but I'd prefer to imagine my own images and create them from scratch. Like this one:
That was me. All me. Painted in photoshop, after being composited from a couple of my own photos, plus a royalty free image of a deer that was sourced from one of those free wallpaper sites where they make you look at 200 commercials before you get to save a picture. You could also go to Wikimedia for sources, or go to any one of dozens of free image sites. These provide the elements from which a picture is composited, and... I'll be totally candid --
This is something I'll play with when I'm bored, but for all useful purposes, the only thing I'd use it for is to generate the occasional picture element, when I can't afford to pay a bucketful of money to buy a prop.
Fair to say, though, that AI generated art is in its infancy. Come back in five years and see what's happening in this field. The way computers and software advance, I have an intuition that five years from now you won't be able to tell something that was produced by a machine from something a human sweated blood over, for a week. And that's what worries me the most. In fact, it's worrying the entire creative community -- musicians, writers, artists, the lot of us. Many people can see their livelihoods going away. I'm lucky, I don't have to worry about that. I'm retired, and this is just a hobby...
I like to paint. Like...
The "raw" image showed enough promise for me to paint on it for about ninety minutes. But check out the original raw image ... those eyes are doll's eyes; then, as if that wasn't enough, the computer set his hair on fire!!! And the shoulder of his tunic is going right into his skin. There's no depth to the image, and nothing is "happening" in it. It's good raw material, but as a finished image... nope. Good thing I like to paint, right?!
Same deal with the Chinese princess. The original offering from Dream Wombo was filled with flotsam and jetsam, the face had weird buttons and bits added in for no reason, and the hair went "wrong" at one side. Took me an hour and a half to rescue that one, too!
Mmmm. Experiments complete. I shall now quit bellyaching about AI, and go back to painting, which is what I like best in any case! I'll try to post more often, too ... I've been neglecting this blog terribly -- 2022 did a number on me (don't even ask), but this is 2023, and almost through to the end of Q1. Time to drive on. So -- here we go.
