A Way Forward on the Generative AI Problem

Generative AI—that kind that is used to produce images and text at a prompt (or series of prompts)—is, shall we say, a major controversy in the indie-writing community. This particular topic has hit the hot-button issue of class, and since it appears as if the indie-writer community is heavily left-leaning (and specifically the social-justice-progressive faction of the left-of-center), this has made for a truly volcanic situation.

I think it’s more of an issue among self-publishers than traditionally published writers because of the fact that we indies have to do everything ourselves that a traditional publishing contract would otherwise handle. This includes cover art, illustrations (where applicable), and marketing images/videos. Traditional publishers make arrangements for these things, and with the oodles of money that they usually have at their disposal, they have no issue commissioning expensive, professional-quality art for published works. Independent writers must either do all such things themselves—if they have artistic ability—or commission others to do so. And they must foot the cost of this out of their own pocket if they elect to do the latter.

Because of this fact of indie publishing, detractors and defenders (albeit fewer of these) both argue against or for GenAI on a class basis. Those against it say that its training algorithm steals the work of small artists and presents images based on their copyrighted work. When authors then use GenAI-generated images for illustrations and cover instead of commissioning a human artist, this further deprives independent artists of income. Those in favor of GenAI say that indie authors have fewer financial resources than traditionally published authors and that GenAI is a way for them to have professional-looking images without the expense.

(Full disclosure: I favor the “against” viewpoint on a class basis. I see the point of the pro-AI people, but I think in this case, the harms outweigh the benefits. In addition, if it’s okay for a writer to use AI-generated images instead of human-created art, why isn’t it okay for a reader to generate an AI book according to their personal preferences too? And I am not saying this from a viewpoint of extreme class privilege, in terms of being able to commission thousand-dollar art at will. I have a comfortable middle-class income, but I am not independently wealthy. I have created my own cover image for my novel The Inheritors, but that’s because it’s an image of a weather phenomenon rather than a human. It wasn’t that hard for me, an atmospheric scientist, to do, in other words.)

There is also a fault line for AI usage on the basis of disability, with some disability advocates arguing that it is a way to help people with certain kinds of disabilities participate in the writing community. It seems that this argument is more in favor of AI that produces written output rather than graphical, however, and there are non-generative forms of AI, which are more useful to disabled people than generative AI.

Additionally, there are objections to AI on an environmental basis, such as the amount of computing cycles it uses and therefore the electricity needed to power these algorithms. My position on electricity consumption becomes apparent in my novel The Inheritors, but in short, it’s this: Make the grid zero-emissions and it’s not a climate issue how much power someone consumes. The problem with high power usage is how the power is generated. If it were all generated by clean power, high power usage wouldn’t matter. I am an “abundance liberal” and something of a futurist, so I don’t support undoing the conveniences and benefits of modern society. I just support greening them and going to zero.

Although there are concerns about deforestation to build immense data centers for blockchain algorithms, and water usage for cooling the supercomputers in data centers, AI algorithms do not become progressively less efficient with more data added in, as blockchain does by design. Processors continue to get faster and smaller, and cooling technology continues to advance apace. Water cooling is efficient for removing heat, but it’s not the only option for cooling—and electrical cooling approaches will be “lifted” by the “rising tide” of a green grid. Some people on the far left may oppose pretty much any industrial or computing construction for any purpose, but I’m not interested in arguing with anti-modernity advocates—which is what they are, no matter how much they choose to hide their social-engineering desires behind “marginalized people” (most of whom are not at all inclined to give up technology). My position on that sort of posturing is clear in my novel, but I’ll lay it out very explicitly here to be sure. Large, costly, hot data centers are necessary for scientific computing, e-commerce, and the Internet writ large. AI/ML, used properly, is a tool that can advance all of these things and improve life. Weather models run out of data centers. So do biomedical algorithms. So does all the e-commerce that we indies are utilizing. So does social media, which we also use. If there are tradeoffs to be made, I side with modern life, because it has benefited more people than any “back to nature” movement, whether left-wing or right-wing. If you’re anti-abundance, we’re probably not going to be able to talk about this subject. And you also should get off the Internet to be consistent with your stated beliefs.

Therefore, the main fault line over GenAI is class. As I said, I think the “nos” have the morally correct answer. Surprised after that rant? Don’t be. I’ve thought this out, and I try to think through issues on their own merits rather than on whether my political tribe is making them. The rest of this blog post is based on that conclusion.

To be very clear on this, I do not support an implicit “we are entitled to your money because we are professional artists (and also marginalized/queer/poor/BIPOC/fill-in-the-blank social justice buzzword) and you are not” attitude that I see on social media quite frequently, because an indie writer does have other ethical options. They could create their art themselves. They could use public-domain art, like Standard Ebooks does. They could go with extremely simple covers (bright colors, geometric shapes, and huge bold text only), or use covers that services like Draft2Digital can generate based on existing templates (rather than GenAI). These are all perfectly legitimate options, and no small artist is being “stolen from” if a writer chooses not to use their services. Some of us actually have artistic talent as well as writing talent. And those who don’t… well, either they get very lucky, or they pay the price for their questionable cover choice in reduced sales of their book. My point is that nobody is entitled to someone else’s business, and “you’re morally obligated to commission artists instead of doing anything else” is not a winning argument, either morally or pragmatically.

But there is another argument that is a winning one, both on a practical and a moral basis. It is a demonstrable fact that GenAI is trained on copyrighted material, and in the case of art, this is disproportionately the work of small artists who have posted it to the Internet to share it with the world. They are 100% correct that GenAI rips them off. This is unconscionable, and this is why I come down on the side of “against GenAI.”

So what can we do about it?

First, I really think that the social accounts that are constantly harping on GenAI usage are not helping. They’re preaching to the choir. The people who are still using GenAI for art either do not agree with the copyright concern or don’t care. They are not going to be swayed to stop doing it because some account on social media is constantly going on about it. If the goal here is to stop GenAI usage, this is a failing approach (which makes me suspect that that’s not the point, but rather, that the point is in fact to preach to the choir).

“Okay, but what if a large, anti-GenAI account calls out writers by name for using it?” Nope. “Callouts” don’t work. This isn’t 2018 when an angry online faction could (whether rightly or wrongly) drive someone offline or tank their career with a public pillorying campaign. What’s more likely to happen now is that the target will simply become angry and defensive. And then MAGA-type influencers will swoop in: “You’re a victim of cancellation? The woke mob has come for you? Don’t let them get you down; we support you!” And yes, they enthusiastically, defiantly support GenAI. They spam their horrible playground, X, with AI-generated videos of Donald Trump. It’s a way to “troll the libs” and “vice-signal,” in contrast with virtue-signaling. And they’ll welcome their “cancellation victim” into the “club” if that person becomes MAGA. This has happened time and again, and it has radicalized far too many people who did not need to be lost to right-wing authoritarianism.

Don’t do public callouts. It will not work for the overarching goal, stopping GenAI theft, and will have primarily bad effects.

To actually achieve this goal, I think we need a three-pronged approach.

1. Convince individuals not to use it because it looks like crap. We can tell the difference. It’s either too sloppy (six-fingered people), too shiny, or too perfect. Human-created art always has flaws in it, and they can be charming. The idea that GenAI images are “soulless” may be a bit of a cliche now, but it’s a true one. But, you might say, what if GenAI stops looking like crap? What if it really does start to look nice? Frankly, we should already have Parts 2 and 3 in place if and when that day comes….
2. Convince book retailers not to sell content that was generated with AI. Some have already taken this step and banned book sales where the cover or text used GenAI. Let’s see if we can get them all to do it. And in the meantime….
3. LOBBY. ORGANIZE. ADVOCATE. Get it clearly defined in law as illegal copyright infringement to sell, as a commercial product, materials that contain GenAI images or text. I think that focusing on commercial sales has to be the way to do this, and here’s why. If it became illegal to produce any image or text that was based on existing copyrighted intellectual property, that would ban all human-created fan art and fan fiction too. I doubt anyone wants this outcome. But it is illegal copyright infringement to sell fan fiction or fan art for profit without the permission of the copyright holder, so I think this is what we have to do for GenAI too. And yes, this means accepting its presence in a noncommercial context. But putting a stop to commercial use of it is a win I think we can achieve.