As you have probably seen, there are AI generated images on this site, in some of my videos, and in my poetry book. I'm well aware of the contentious nature of these tools, my decision to use them is not out of a lack of consideration for the perspectives of others. If you've reached a different conclusion for yourself, that's fine. If your perspective doesn't fit neatly into the categories of "full support" or "total opposition", you're in good company.

This is long, but my thoughts are complicated.

My AI History

I started using neural network "AI" generators in 2019, with a particular bit of rendering tech called "Style Transfer". It was a precursor to modern generators, where I could provide a fractal base image, and a stylistic source image, and render a stylized fractal. Or in reverse, I could inject fractal designs into a subjective image as a style to get trippy interpretations.

At the time, this was met with wonder, amazement, and encouragement from all directions. My fans liked it, my friends liked it, people at the markets I attended like it. One piece I did this way was selected to go to the literal moon in a time capsule, and has become my most successful design of all time. I never used these tools to deceive, I was always discussing the workflow and the fancy tools that are involved.

Then Style Transfer grew up into Dall-E, and for just a brief moment the doodles Dall-E spat out were still quirky and messy enough not to cause too much alarm. I took advantage of this cool moment in tech history to artificially illustrate a book of my poems, fulfilling a lifelong goal. Between the time I sent that book to print, and the time the first copies arrived in the mail, Stable Diffusion and Midjourney hit the scene, the drama dial turned way up, and the internet demanded that everyone quickly develop a strong opinion right away.

I took a step back, no longer feeling welcome in many of the art spaces I used to enjoy socializing in. Online, there is very little room for nuance in these conversations, and passionate debaters on both sides are quick to toss out dehumanizing accusations. Whether it's a "soulless leach without a speck of creativity at all" or a "reactionary Luddite who can't keep up with progress", there are some nice ugly strawmen to beat on while talking about you.

Then I Talked to Real People

Another local artist asked me what happened to the poetry book I was working on. I explained that I didn't want to make people hate me by putting AI art up for sale. We had a conversation about the ethical complaints, how the tech works, and the way I used it, and she told me she couldn't imagine anyone actually being angry about it. I wasn't so sure, the anger is HOT online, so I started talking to other artists at the markets.

Every single one that I talked to said the same thing, they saw AI as a new tool that can be used in good ways and in bad ways. None of them suggested that hiding my poetry book away was a good move, and several pointed out that I don't know how many fans of AI art there are out there.

Still uneasy, I talked to some of my closest friends. People who are engaged with internet culture, some of whom work in creative industries. People I could really trust to call me on my bullshit. They'd come to many of the same conclusions as me already, and told me they wouldn't look down on me continuing to explore the tools as they evolve.

So I took a deep breath, and introduced the poetry book. Customers loved it, it was neat and trendy, and I sold enough to order a second print run. I started mixing some of my AI Art generations in my prints bins next to my fractal art, and had the same positive response. Not only are my AI prints selling, my fractal art sells more often when the AI art is part of the display. The anger that is so present online does not manifest at the markets I attend, or if it does it stays quiet and keeps to itself.

Wait, I've seen this before.

Over a decade ago, the Gamergate controversy burst out into the internet, and established propaganda techniques that continue to be used to this day. I won't bore you with a breakdown of Gamergate, but it's fair to say that it was an astroturfed campaign that dressed up a questionable ethical complaint and turned it into a majorly divisive cultural issue, which effectively encouraged reactionary conservative young men to isolate away from broader gaming culture with extreme hostility, and encouraged everyone else to exclude those young men for the safety of their communities.

Gamergate was very clearly an early test run of what later became the MAGA cult. Both were orchestrated by Paul Manafort, who you may remember was involved in negotiating Russia's support of Trump, did illegal deals with foreign nationals in order to change the GOP's policy platform, and is considered by some to be an architect of the current Ukraine war. Yeah, that guy proved himself by turning the gaming community into a social battlefield.

I am not saying that the controversy around AI generators is a scheme by some shady right-wing manipulator. Revolutionary new technology isn't a piece of propaganda that you suddenly inject into the conversation like an accusation about "ethical standards in journalism". But I can't shake the feeling that the incredibly divisive nature of some of the conversations has the same structure as Gamergate, MAGA, and Meme-stock clickbait. I assume that means they have a similar purpose.

Self- Righteous Anger Sells the Best

The trickiest part about this kind of divisive rhetoric is that it is carefully constructed to make you sure of two things. The first is that you are right and should feel good about being right, and the second is that anyone who disagrees is wrong and needs to be made to feel bad about it until they change their mind. This mental framing is used all the time for manipulation, because it requires the least amount of effort for most people. It's the bread and butter of a media bubble, and it's why you have that feeling that the people who disagree must have never really thought things through.

Before I get into the specific complaints about AI, several of which I agree with, I want to point out how unhelpful anger and hostility is in this situation. The only people who benefit when we spend a bunch of time bickering with each other are the people who run the sites we keep refreshing, or the people who run ads on those sites. 100% of the bullying that happens on both sides of the issue towards other artists is wasted effort, because it has no practical impact on the development of the technology.

Even for the things you feel very convinced you are correct about, the anger that those feelings bring needs to be focused towards the people with influence and control over how this tech is getting used and developed, the managers at the large corporations that are adopting the tech, and the money keepers who are deciding which creative projects get funded. The Writer's and Actor's strikes in Hollywood are a fantastic example of effective ways to push for better policies and standards around AI. And for what it's worth, I really like the model of use that they've come up with for the tools in those negotiations.

This Tech is Inevitable, and Inevitably it Will Improve

Adobe has already put AI into Photoshop (and really, other versions of machine learning have powered tools in PS before). Microsoft is building it into their operating system, Google is building it into search. Every major social media platform is investing in it. The fundamental tech isn't patented, so development in this space is wide open, with tons of money for the first big productivity boosters to establish themselves. This isn't like NFTs (which I called bullshit on from the beginning) where the hype will die down and the tech will fade away without a use case. This is like email, making a big impact even in a rudimentary form.

Depending on who you ask, an AI art generator model that is trained entirely on rights-cleared images is either already here, almost here, or right around the corner. No one who's following the tech is predicting that the controversial training data it used at first is intrinsic to the success of the tool. Adobe says Firefly is ethically clear, but doesn't prove it. Open Diffusion is a community project to accomplish this, but it still has a few kinks to work out to be fully free of the taint of "stolen" data. When these new models are proven and adopted, the most odious complaint will evaporate, and all of the complaints that really matter will still remain.

This is why I don't put so much weight on the ethical angle of complaint regarding AI. I don't think that scraping public internet files is as close to "theft" as some people say but I understand why they say it. I think that a lot of the content that is allegedly "stolen without compensation" was actually signed away in exchange for access to social media platforms (in the form of EULAs and ToS that we all clicked "I agree" to without reading), but I don't blame anyone for not predicting that. Mostly, I don't think any strong opinions here are going to last more than a year or so.

Bigger Problems Are Looming

No, I don't mean climate change and the unnerving rise of fascism. Even once the ethical complaints about AI generators are gone, there are valid and important complaints about how we are going to cope with a world that includes them. Those big companies I mentioned before aren't putting billions of dollars into building these tools because they think they won't have an impact on the job market. It is absolutely, obviously going to be the case that workers will be expected to increase their productivity to be on par with people who use AI in order to remain competitive. That sucks, and as someone who's had three different jobs end due to automation I'm well aware that companies don't give a shit what happens when they don't need someone anymore.

Even looking past the corporate job market, there's another huge issue that AI art poses, and one that is so much of a hassle that I completely support art communities that prohibit AI art. Without trying hard, I can make 50 awesome art images with a generator in an hour. If I have a silly idea, my friends might all generate their own takes on it, and suddenly there's a handful of people who all want attention for the thing they just created. Content feeds, especially popularity driven feeds, are vulnerable to content flooding. Even if you tell every AI artist to limit their contributions to 1 a day or 1 a week, a community can still be overwhelmed with AI art to a point that it becomes hard to discuss anything else.

That's going to get even worse as it becomes corporatized. The future of marketing is AI-tailored to your interests at the content level (not just the curation level), and advertising is going to morph to match them to try to compete for the same attention you usually give artists or creators. The world is going to be overflowing with AI generated content in any avenue that doesn't actively seek to reject it. This is going to devalue the market for creativity if we do not manage to make big changes to how we find and engage with media, the same way that Spotify devalued the market for musicians and Youtube devalued cable TV.

The Only Way Out is Through

If the tech is inevitable, those problems are too. It's OK to spend a little time dealing with the shock of their arrival, but at some point we have to move past denial and rejection and start having conversations about how to build tolerable realities that include this technology. There are no easy answers, and a lot of old precedents don't apply anymore.

What can we do? We can teach each other the difference between what good use and bad use looks like. Nothing can stop someone from drawing a very awful thing in Photoshop, but culturally, we understand that someone who uses Photoshop in that way is creating a problem themselves, and we don't blame Photoshop for that awfulness. We can do the same with AI, if we talk with each other about it.

We can teach each other how to value traditional artisanship differently from generative artisanship, the same way that we learned to distinguish digital art from fine art. It's true that someone typing a prompt into a generator is not doing the same thing that a painter or a sketch artist does. It's a problem for painters when customers are confused into thinking an AI image took the same kind of effort, but the solution is to teach customers the difference.

At the same time, just because the actions are very different, that doesn't mean someone who practices writing effective prompts and editing the results is without value. That's still a kind of creative effort, and we don't do any good by trying to convince someone otherwise. We do not have to take value away from a painter to give it to an AI artist, for the same reasons we don't take value away from a painter to give to to a sculptor. We build artistic spaces and communities that specialize or more inclusive ones that encourage different mediums, and it works out. My presence at local art markets has not prevented the people around me from being successful with their art, nor has it prevented me from being successful with my non-AI art. I charge what works for me, they charge what works for them, and we both find customers (sometimes the same customers even).

What Does Bad Look Like?

Training data is not the only potentially gross part of AI art. Lying about your workflow is scummy. Automated imitation is shitty, especially when it's intentionally done to steal someone's thunder. Filling the internet with overly sexualized and unrealistically idealized renders of anime girls (or worse, real children) is pretty gross, and I don't have to like that just because I use the same tools as the people who make that content. Using AI generated images to harass someone, or to deceive someone about the real world? That's very disgusting, and I'd reconsider a friendship if I caught a friend using the tools that way.

And God help me, I don't want to see a Calvin and Hobbes comic from anyone who isn't Bill Watterson. If there's a line to be drawn, it's that.

What Does Better Look Like?

So what steps do I take to hold myself accountable?

• I don't use artist or art community names in my prompts. I'm not trying to communicate someone else's art through mine, and I don't want the generations I create to make noise that detracts from other creators.
• I don't use AI to imitate. This is sometimes a tricky distinction, but I am expressing myself when I use AI generators. I'm not trying to mimic someone else, I'm exploring and finding the images that fit the aesthetic and vibe that I designed for Complicated Reality. (That's how my fractals work too, even though the algorithm powering them is very different.) It is no coincidence that the colors in my AI generations match the colors in my fractals. It is no surprise that the things I write about are the same things I generate art about. I am the binding thread that pulls this all together, and that's what I'm investing creatively when I use AI to express myself.
• I Don't Present AI Work Where It Isn't Welcome. When an art community says "no AI art", I don't put my AI art there, I don't start conversations there about AI, and I don't complain that I'm being "left out". I have other mediums to establish my authenticity with, and do not need to disrupt communities with drama. I only put my AI art in places that explicitly welcome it, or in places I've personally built to accommodate it (such as this website and the art markets I help host).
• I Disclose My Workflows. I do not have any need to pretend that my generated work is anything other than what it is. At my art table and online, I make sure people understand what they are looking at. Not just to cover my ass, but because people talking about the cool tools they use was how I found my way to using cool tools, and I want to pass that torch on. If my art inspires someone to try making their own, I'm more than happy to point them in the right direction.
• My Prices are Based on My Effort. Working in fractal or text or AI all feels like the same kind of work to me. It's still time I spend thinking creatively, applying creative skills, and chasing creative results. When I'm setting a price for something that includes AI work, I'm not picking some random value for that based on other art that looks like it. I'm counting up the time I'm putting into creating and polishing the work, the cost of materials or services, and setting a price that works for me and lets me grow my art.
• AI Isn't the Only Art I Do. I make fractals. I write poems. I tell stories. I design puzzles. I host art markets. Even though AI art is fun and easy, I don't allow it to become my primary focus, because maintaining and growing my other art skills is still important to me. I don't have any doubt that I'm an artist, so it's not hard for me to prove it in multiple ways at once.

How I Use AI Generators

Principals are nice, but how do they translate in practice? For starters, everything I write is done with fingers on a keyboard. Other than spell check and grammar correction, I do not have any desire to put my name on machine generated text. I don't have any qualms with people who do (and think some people would greatly benefit from doing so), but writing is simply too much fun for me to hand it off to a robot.

When it comes to image generators, a lot of my AI designs use my fractal images as a base input. One of my favorite workflows is to make a fractal, then describe what I see in the fractal the way that people describe what they see in a cloud, and use the fractal and the description as a prompt to start generating. Now that I'm running AI models locally, I can take much more control in this process and my designs closely reflect the fractals they are based on.

For a while, I explored the use of AI generators as an aid for aphantasia. I don't see images in my mind at all, my thoughts are clusters of words and concepts in a dark space. Turning those thoughts into prompts and letting an AI take over for visual imagination was a thrilling experience, and a few of the renders in the AI art section here come from that period. In the end, I didn't feel like this process improved my ability to imagine visually. I still use AI to visualize a thought from time to time, but only casually for personal use, not as artwork I publish.

Finally, while this is less artistic in nature, I also have a freelance position where I get paid to write training content for various AI models. You may have seen headlines that these tools are trained by underpaid workers in terrible working conditions for pennies, and it is true, that does happen. But the people who pay me, pay me $20+ per hour to write stories and poetry for their robots, which is way more than any other job I could find in my city in 2 years. Plus I get to work from home and set my own hours, which makes it a perfect companion job for my art, puzzle, and market positions. (My opinions about using AI developed long before this job appeared, in case you're wondering about bias.)

An Optimistic Future

We've covered a lot of the negatives of AI use. That's important, but I also want to include a small dose of excitement about this tech. A terrible consequence of the very spicy discourse online is that a lot of people close their eyes to the cool parts about it. You do not have to support everything about AI to appreciate some aspects of it, nor do you have to fully reject it just because some aspects aren't to your liking. We can't find a balance if we can't also talk about the good.

A world where people feel empowered to express themselves artistically is a more beautiful world, even if it is a noisier one. But perhaps more importantly, there's a very powerful onboarding ramp that can take someone from generating AI images to learning traditional digital art skills. We could be motivating a whole new audience of people who want to be creative to learn how much depth there is to digital art, and in the process help them respect the joy of craftsmanship and artistry. We can equip less-verbally inclined people to express their feelings by letting them collaborate with a judgement-free machine to help find a suitable expression. We can laugh at corporate execs who say silly things like "art doesn't make sense to me" because art can be something that everyone does instead of something that requires a career's worth of dedication to.

Looking beyond art, there are some really, really awesome things that AI can help us with. From engineering a sustainable living environment, to revolutionizing food and medicine production, to equipping school children with a conversational archive of everything they might be curious about, to a radical shift away from artificial scarcity as an effective economic model. It may feel like artists are the ones getting hit the hardest right now, but I can promise you that the generation that grows up with AI will be open to AI managers and CEOs pretty soon.

There's a LOT of HARD WORK to do before these systems can be effective solutions for government, but I am positive that modern campaigns are already highly dependent on automated analysis and that's only going to become more prevalent. When every person can ask their phone to generate a reasonable response to, say, the active climate crisis, we're not going to tolerate leaders who say it's too hard to figure out. The ability to create fake news is rising, but so is the ability to learn how to spot it and resist it.

Automatic Bias

We do need to be careful and conscious of the many different ways biases in the data we train on produce biased results, but part of that is going to require us to come to terms with AI agents observing the real world. I want the healthcare assistance AI to be able to look at my face and tell me if there's a sneaky sign of a health problem, and that requires training the AI on a bunch of real faces.

A world of AI trained only on carefully licensed data archives is going to reflect whichever biases those archives record, so the ability to observe reality will be critical for building a well-rounded intelligent agent. If the public already lost its privacy to cell phone cameras, that is only going to be more true when adaptive AI agents are moving through the world (in robotic or software forms).

Be Wary of Corporate Allies

While this is feeling like a less likely outcome now, one of the early developments in the "Anti-AI Art" movement set off large alarm bells. An association was formed called the Copyright Alliance, which put itself out as an invitation to get "real artists" a "seat at the table" while discussing the future of AI. What wasn't given as much focus was who else had seats at that table, and when I found out Disney was the company coordinating the effort, I suspected something shadier was really happening.

Disney has a wonderful treasure trove of high quality art in its archives that it could use for its own AI models for a long time. They also have billions of dollars, and don't need to skip paying real artists just because they could use AI, but they want to. The biggest threat that AI poses to Disney is the threat YouTube poses to Hollywood: Smaller, cheaper studios can come closer to competing on creative quality. It would be in their best interests if AI models were only economically feasible for giant studios like them.

So when Disney is stoking the fires of people arguing with each other, and using sharp accusations to draw deep lines of division, I have to advise that we all look at their actions through a skeptical lens. When Disney and a handful of other large business representatives sit down at that table we have a seat at, they still hold the winning share of votes, but they've funneled your attention towards yelling at artists instead of yelling at CEOs.

Shame Doesn't Work, Try Empathy

Obviously, I've thought a lot about all of this. That doesn't mean I'm right, and it doesn't mean there's something wrong with either of us if you disagree with some or all of this. But if you've made it this far, I hope you can see that the best paths forward don't require hostility towards each other. We don't have to accept the overly simplified binary representation of the issue that drives click-revenue. We can talk about the good and the bad, and we can work together to build a better future.

But we can't do that when the primary currency of the conversation is shame and guilt. The solutions that those emotions offer are temporary and flawed at best, but more often result in the exact opposite of their intent. If you make it clear that your contribution to a conversation will hurt, you only convince people not to have that conversation with you. Don't let an overabundance of those feelings online convince you they are the only valid ones to have. There's hope and passion to be found, if you're willing to look for it.

Is AI Art Good, though?

One thing I have avoided touching on here is the actual artistic quality of AI generators today, and what I think skilled AI use actually looks like. I could probably write about as much about those topics as what I've already written here. But in short, I think AI art has some really cool aesthetic properties, some really characteristic noise that I don't like, and there's nothing inauthentic about the reactions it can provoke. I think getting better is a matter of taking finer control over the output, not just creating impressive images, and that the amount of control a text prompt gives you is inherently, intentionally limited. Finer control will require additional or more mature tools. Over time, I'll write focused articles about these ideas, and you'll be able to find them in the Guides section.