Blog Archives

My (Current) Thoughts on AI Images

There are numerous AI (“artificially intelligent”) programs designed to allow someone to use text prompts and maybe a few simple other buttons to generate images that have never existed before. People have been playing with them for years, and the fact they were improving has been clear for a long time, but in my opinion they have taken a major leap forward suddenly and recently. As with any disruptive technology, this opens a number of cans of worms, and some of those worms seem likely to crawl into the game industry sooner, rather than later.

This broke open for me last month, when I used prompts to produce the following image using Midjourney — an AI image creation program, that allows you to enter text prompts and style notes, ask it to create variants of options it presents, and eventually upscale a thumbnail to a higher-res image. I pay for access to Midjourney, and for the rights to use the images it creates with my prompts in commercial products.

(Prompt and variation choices by me, using Midjourney, and who to credit here is one of the fundamental questions of AI art)

I was blown away that my effort to use prompts to have Midjourney return an image of a dungeon entrance, such as was of the right quality and style I could use it in a professional ttRPG adventure, was successful. I had been sharing images I created on social media that were much less successful, so I shared this one as well, with the comment “So, this is the first MidJourney AI image I’ve prompted that I believe I can use as-is for a ttRPG product. It’d make a great cover, even.”

It would have been more accurate to say it was the first image I thought I *could* use, since I had no immediate plans to do so, but I said what I said. While the response has evolved as more people replied, the early pushback referred to my suggesting I would put that image in a product as “anti-art,” “unethical,” and “gross.”

I wanted to dig into those responses, so I tried to ask leading and clarifying questions. I likely should have waiting longer before replying, because my knee-jerk responses to being called gross can be more confrontational than I prefer to make my professional communications. having taken more time, and having had a number of conversations with different people in a number of different places, I thought compiling my current thoughts and positions in one place would be a good move.

As a starting point, I recommend anyone interested in the question of the history, legality, and morality of AI art read the recent Engadget article by Daniel Cooper ” https://www.engadget.com/dall-e-generative-ai-tracking-data-privacy-160034656.html. The article doesn’t try to definitely settle any issues, but it’s a good rundown of what some of the big questions are.

Speaking of big questions, I was asked a lot of them when I publicly declared I had AI art I can put in a product. I don’t want to call out anyone specifically in a venue where they can’t reply (and I absolutely do not want to encourage anyone to engage in people I disagree with online in anything other than a polite and professional manner, so please don’t), so I have tried to summarize my position on big questions below. Given there are public links to this article, it’s fair game for anyone who wants to debate or disagree with any of my thoughts or statement.

We should all acknowledge there are unsettled legal and ethical questions about AI-generated images. For example, AIs are trained by looking at images online, almost always without compensating the owner of those images, or asking permission. However, the AI do not just remix existing visual elements, nor copy them into a database and go back to grab pieces of them. They look at existing art as references, to learn from them. I do not see an ethical difference between that and a living artist doing the same thing, which is commonplace and well-accepted.

Another common concern is that art generated by an AI may take jobs from existing traditional artists. Such concerns have been raised by new technologies before, including photography and Photoshop. Pragmatically, I note that as of time of writing this, my art budget has not been reduced at all by my use of Midjourney (and this article is the only “professional” product I have such images in atm, and only because it is crucial to understanding why I am even discussing this). I pay several artists, to the tune of several thousand dollars a year, and do not foresee that going down at all. Midjourney can do many amazing things, but at least in my hands it can’t produce something like a recognizable band of adventurers, especially not reproducing them in multiple different illustrations.

However, that pragmatic note aside, I do not accept the argument that my using an AI to create my own art is an ethical or moral failing on my part. The system requires input and decisions from me to generate art I can use, and I do consider the images in question to be “art,” even if a machine was substantially used in its creation. All the mechanical aspects of fixing an image with a camera are machine labor, and it is accepted as art. If I created an image with a spin art machine, or a spiralgraph, that is accepted as art. Nor is the claim that I am unethically avoiding paying artists convincing to me. I can and have used public domain images in commercial products, and that neither puts money in any artist’s hands, nor inspires anyone to claim I am being immoral.
I absolutely understand the concerns of professional artists that this new technology may make their careers more difficult, or even impossible. I have seen the same AI-driven changes begin to influence how professional text is generated and sold. The program Grammerly is used by at least one company I am aware of to replace one human editing pass, and I have been told other places use it in place of any paid editor. Things like resumes and ad copy have AIs dedicated to producing them, and that trend is only going to grow.

In my opinion the answer to those challenges is to work for a world where creators don’t have to depend on companies paying them to create what those companies want, and constantly working to do so as cheaply as possible (as opposed to claiming that finding a cheaper way to obtain images is immoral, which would also apply to using stock art for example, which I do all the time). I suspect this problem is going to spread, farther and faster than expected, and no effort to convince the world to not use it is going to be enough to save creators that can be replaced by it. Those answers may include seeking direct patronage from fans (such as through Patreon and Ko-Fi), a Universal Basic Income, or some other answer no one thinks of using every truck driver, warehouse worker, delivery and transportation job, and basic manufacturing employment begins to be eliminated due to Ais doing it cheaper.

But this technology is not going away, and I suspect it will always have things it does not do as well as living artists. At the moment, I am exploring what it can and can’t do, and I am continuing to research, consider, and come to my own conclusions. There are risks involved in adopting any technology early, and I am weighing them. For example, without paying a much higher subscription fee, my Midjourney art can be seen and used by others with the license, so I have no exclusivity. Further, if the courts decide these images are being created by a machine, rather than being created by humans using machines, there’s a good chance that legally they will be impossible to copyright, adding another layer of complication for using it commercially.
On a personal level, I am enjoying many of the images I created as things to look at. I enjoy creating art this way, and am growing to appreciate the skill it takes to do so. Like silkscreen, quilting, coil pottery, or tie-dye it’s a very different kind of skill, but that’s not on its own enough to say it isn’t art.

So, on AI art I know what camp I’m in but, obviously, not everyone is going to agree. This tech is happening, so it’s worth learning and thinking about. I don’t currently have any plans to actually put AI art into commercial products, but I’m, absolutely using it for my own entertainment and in social media posts. I’m still considering how to proceed in cases where AI art looks like it may be a good match for a product’s needs, and may use it as the basis for human artist efforts, or as underpainting for more traditional art, or as-is, or as-is but paying an artist extra money for each AI piece I use in a form of patronage just to keep the human artist industry vibrant, both out of love of art and (pragmatically) because I believe a healthy visual creative field is necessary to make the best possible game products, even with access to AI-generated images.

Speaking of Being a Patron to Maintain a Healthy Industry of Creatives…
I have a Patreon. It helps me carve out the time needed to create these blog posts, and is a great way to let me know what kind of content you enjoy. If you’d like to see more essays on industry issues, or Pathfinder 1st or 2nd edition, 5e, or Starfinder content (or more rules for other game systems, fiction, storytime posts, game design articles, worldbuilding tips, whatever!), try joining for just a few bucks and month and letting me know!

If you prefer, you can drop a cup of support in my Ko-Fi. It’s like buying me a cup of coffee, but more convenient!