Revisiting AI

I posted a while back about one of the AI art generators because I was excited to see what that could do. As someone who does a lot of my own covers and has the ideas but not the skills to do really complex covers and gets frustrated with cover designers who flake or take a month longer than they said, I thought it was a potentially interesting new option.

But I never really pursued it. When I was able to get into the beta on that one I did about five prompts and thought, “well, this is crap.” It was not able to take my simple description and do anything the least bit attractive with it.

I figured that was that. Creepy, not-quite-right images that just don’t work for anything outside of horror.

That’s changed pretty fast though. A few friends of mine who are authors have really dug in on using Midjourney and some of the stuff they’ve created is really, really good. To the point that one set up a shop for selling notebooks with covers they created from Midjourney and another is or is planning to do a Kickstarter with images they’ve created there.

On the flip side I’ve seen some pretty prominent authors publicly state they’re not going to support the use of AI and most I’ve seen who were playing around with it early on have said they are done even posting personal use images with it because of the harm it does artists.

(Which, man, that cover that was done for Ilona Andrews is amazing…)

I’ve decided I’m personally sitting this one out for the time being. Because it’s become a hotbed of controversy. And there are some clear ethical violations going on out there with building models.

For example, I’ve heard of people building models using one current, active artist’s work. That’s just some serious bullshit right there.

And there’s another AI that’s so sloppy about stealing other current artists’ work that the generated images still show the original artist signature.

Tech seems to have this thing these days where it’s like, “can we do it, let’s go!” and then “oh fuck, that has problems, let’s fix them on the fly.” It really harms a lot of people along the way in the name of “progress” (which is otherwise known as making a bunch of dudes with no social skills or empathy rich).

I don’t think AI art or writing is going to go away so I expect at some point it will become something I use indirectly even if not intentionally. (One of the SFF publishers was just put on blast for using an AI-generated image someone had put up on a stock photo site. How we’re supposed to know that’s what that is, I don’t know.)

I do think AI art and writing need some ethical parameters. Because it’s one thing for one artist to take years to learn another’s style and then to reproduce that style over the course of a couple of weeks for each piece and it’s something else entirely to have a computer learn that style in the course of a week and then churn out replicas one a minute. Unless we want to either subsidize new artists so they can create and still live or we want to just stop having public displays of new creative works, something will have to change about this.

Training an AI on public domain content I think is probably fair. Training it on copyrighted content (like this blog is and any current blog is) without permission, not cool.

And using it to basically replicate one specific artist’s work when that artist is still alive and active–that’s asshole-level behavior right there. (Especially the example I saw with some dude who wouldn’t admit he was doing it replicating a woman-artist’s work.) Not surprising that it happens, but also not something we have to as a society reward or accept. Shunning is very effective for a reason…

But, yeah, I think this particular cat is out of the bag and it’s going to be ugly for the next few years and change the industry substantially over the next ten.