Playing with AI art

Despite the aches and pains of being in my 70s, this really is an incredible time to be alive. We have so many new toys to play with! I've just spent a few hours playing with AI (artificial intelligence) art, making fanciful photos of my family.

AI art generation is the process of using artificial intelligence to create original and realistic images and art from a text description or a prompt. AI art generators use machine learning algorithms and deep neural networks to learn from large sets of existing art and find patterns, styles, and features that can be used to make new art.

But, wait! you say. Doesn't that mean that all it's doing is taking bits and pieces of pre-existing art and photos that real people created without compensating them for it? (In the above piece of generated art, if you look carefully in the lower right corner, there's an artifact, a bit of some visual that doesn't really fit but obviously came from somewhere.)

Well…yeah, sort of… But that's looking at it from an irrelevant perspective. Consider this: If this were 14th century Florence, and you were an artist painting, say, a fountain. Did you design the fountain? Did you pay to have it built? Do you intend to remimburse the builders after you've sold your painting?

Of course not. We have understood for centuries that art consists of observing reality, de-constructing it, and then putting it together in new and novel ways. Sometimes the reconstruction works; often it doesn't. But, sometimes, it awakens a new pathway in the brain, and the soul quickens just a bit. But we've never thought the artist should pay for the design on the model's dress, or the biotech company that made the genetically modified apple in a still life, or the the owner of horses in a pastoral setting. And just as a human artist puts a picture together out of everything they have learned, seen and experienced up to that point, so does AI.

Successfully getting an AI to produce the picture in your mind takes a new kind of skill. And, in my experience, one AI isn't enough for me to get what I want; I often use several, as well as traditional graphics editing apps. Just as AI takes bits and pieces to create an image, so do I. See, AI isn't an artist and can never be, as it lacks the emotion that is at the heart of all art. So it's a tool, just like a brush or a can of paint or a camera. To actually get art out of it requires an artist—me, in this case.

Here's what should have been a simple case; but it serves as an example. I wanted to create a graphic of an artist in 14th century Florence, Italy, painting a fountain. I decided to start out with Fotor, which despite an often confusing user interface, usually gives me the best results. (Do not waste money on Topaz, which is very expensive and not nearly as good.)

Anyway, here's my text prompt and the result:

When this came out, I laughed out loud. The painter in it is literally painting the fountain! So I decided to try again with more detail:

To be honest, I suppose the man in the picture could be making a small painting. But the picture in my mind had the artist at an easel. So I tried again:

Uh…so, where's the fountain? I have found that when my descriptions become too verbose, the AI often leaves out one or more components. So, I tried again (and remember, each one of these costs like a penny at Fotor, which is why I often start with that):

So this time, I got the fountain, but not the easel. It's at this point that I considered asking for separate pictures (a 14th century artist at an easel, and a 14th century fountain in Florence) and just putting them together myself. But I decided to try another of my favored AIs, HotPot, and it gave me the illustration I actually used above, on the second try. (But it costs closer to a dime for each generation on HotPot.)

Anyway, today I decided for fun that I would do fanciful artworks with the faces of my family members. This would combine crreating environments and characters using AI generation; then using AI face-swapping to put the desired faces on the target characters.

Dorothy

Karen

Jenny

John

Michael

Have I come up with a Mona Lisa or Rembrandt? Of course not. But did I have fun? Of course! and that's what makes this such an exciting time to be alive!