Welcome to Trippytown, population you!

Thoughts on AI Generated Art

I have started to use AI art generators for characters and covers. I’m going to talk about my take on the legal, ethical, and use of the generators.

First, the legal issues. There aren’t any. Remixes of various kinds – collage artists and hip hop musicians leap to mind – are allowed to use other work without pay or even recognition of use so long as the work produced does not infringe on a copyright. It has been elaborately legally tested that an artist does not hold an IP on “style,” either. It has to infringe on a specific intellectual property. If I write something that could be confused stylistically as belonging to, say, Stephen King but does not draw on any characters, situations, plots, dialog, etc., from any of his works, I’m good. It is not enough to say “Stephen King could have written this.” It must have actually been written by him.

Secondly, the moral issues. Again, I find none. Every artist, except for a few first artists lost in the dim recesses of pre-history, has found inspiration in other artists. If you’re complaining that your art was “used” to generate other people’s art… take a good look at your stuff, buddy. I’m willing to bet that others have inspired you, and traces of that inspiration can be readily found in YOUR work.

Also, this technology is coming for me, too. For thirty bucks a month, I can join an AI server that will dish me out AI-generated stories. Oddly – and this does surprise me – AI-generated writing is not as good as AI-generated visual arts. Particularly in longer works, the AI stuff gets WEIRD. The technical term for the strange tangents AIs take in writing are called “hallucinations.” I suspect that in any sufficiently long text work, the odds of a statistical outlier coming to dominate a section of the narrative – and thus determining the rest of the narrative – approaches one-hundred percent. Apparently, stories are hard.

(And on a personal level, I have resisted using those tools, which I now realize is absurd. Technology DOES march on, and I shall march with it. Jasper, here I come.)

Additionally, I’m sorry visual artists, but the world changes. Technology marches. A fair bit, if not most of the visual artists complaining about AI-generated art come from the digital end. How quickly we forget the cries of artists who had to draw and paint with pens, pencils, and brushes when art creation became more about skill with computer peripherals and programs than with physical objects!

I have sympathy for their position. The AI-generated art is already so good that it is clear that their professional days are numbered. In another decade, we might be at the point where no one can compete with computers for art in the visual, musical, or writing spheres. And I include cinema in this analysis. You better believe someone in the big production studios is paying attention to this AI stuff.

Lastly, for the moment, at least, to get the best results still requires effort. Creating AI-generated images, right now at least, is like having a conversation with a very skilled artist who doesn’t know how to obey instructions. So, you have to try to find the exact right words, in the right phrases, in the right order, with the right emphasis – sometimes including examples from other places – to get specific, consistent results.

And I will say this next part with tremendous respect for visual artists, but the AI isn’t harder to talk to than a human artist. Getting beautiful, consistent, true-to-your-vision results are next to impossible with human artists. This might be a fault with me, but I feel I’m pretty good at describing what I want, and the energy expended talking to human artists to corral their creativity to align with my vision is at least as hard – if not harder – than with AI-generated art.

However, it is also very limited. It is great at stuff like landscapes, cityscapes, anything static. It is good at human portraits… so long as they don’t show hands, anyway. Anything with action, especially involving more than one person, will be a mess. Trying to get it to generate, say, two people wrestling will result in surreal results, if not body horror. I’ll make an imagine right now with the prompt “two people wrestling.”

See? And when you try to make it better, you get this

As you might notice, that is not an improvement. It’s borderline body horror nightmare fuel, really.

On the other hand, as a game master for tabletop RPGs, this is pretty amazing, because the randomness of the images generated can be used as immediate inspiration for the more loose flowing narrative style of RPGs. It is also easy to present players with striking images to further the feel of the game. So, that’s pretty cool.

I suspect ere long I will have a lot of new art for this site, too.

Those are my thoughts about using AI generated art for the moment!

Leave a Reply