This article was first published on LinkedIn in June 2023. It has been lightly edited/updated and republished here in January 2024.
I have wanted to write an article about generative AI art for a while now. This is not the first draft I have written.
The trouble I ran into, that stopped previous attempts in their tracks, was this: as a subject it is both surprisingly simple and, at the same time, horribly complicated. Finding a middle ground in which to talk about generative AI and its impacts has been a challenge.
But I think I have found a way in, and it was all thanks to one of those annoying YouTube get-rich-quick videos.
What is the problem and why should we care?
Artists (and creatives in general) have been sounding the alarm about generative AI for months now. We have concerns about how these programmes work, how they learnt to do what they do, and how they are being used to replace us across the creative industries. I don’t believe it is hyperbolic to say that it has caused something of an existential crisis across large parts of the creative industry. As I write this in early summer 2023, Hollywood writers are on strike, in part because of the threat they are facing from generative AI.
I care, and lots of my peers care, about the rise of generative AI art because we are at risk of losing work. I cannot pretend there is not a selfish element to this. I love doing the job I do and I wish to continue doing it for many years to come.
But we also care because there is a real risk to the quality of the art (the images, stories, music, and videos) we all enjoy. A race not to the bottom, but to a middling mediocrity, where art = content; increasingly bland and an endless parody of itself. This affects all of us. Everyone who enjoys art. We will all suffer the consequences if the human element is removed from the art we look at, read, or listen to.
Artists also care because these programmes have been built by exploiting our work, without consent or compensation. If you paint, draw, write, sing, narrate, or make music, then your work is forming the backbone of these generative AI art programmes. They could not exist without our work but we had no say in how, or even if, our work was used in the creation of these programmes.
And we care because there are young artists, just starting out, who need opportunities to find a place in this industry. As generative AI gobbles up a greater and greater share of the jobs in the marketplace then the routes into the industry for young artists become fewer and fewer. With fewer new viewpoints and experiences and world views entering the industry, the more staid and pedestrian the art world will become.
But why should anyone else care? That is the question.
If we are going to reign in the galloping advance of generative AI and set appropriate, reasonable, human-focused controls on the use of these programmes then we need to have wider support. Support from people who enjoy the art we make and who believe in the work we do.
This means we need to share our stories and explain some of what has happened, is happening now, and is likely to happen in the future. To highlight the problems with these programmes.
And in the process, I hope, we earn that support.
What is generative AI art?
The simple answer goes something like this…
Generative AI art includes any image created using a generative AI programme (popular ones include DALL-E, Midjourney, Stable Diffusion, and Adobe’s Firefly). It is the image generated by one of these programmes in response to a written prompt entered by a user.
If you type “painting of a spaceman riding a unicycle” into a generative AI programme, then the image it spits out in response to your prompt is what we are talking about.
The person typing the words into the programme is the “prompter” and the image generated in response, by the programme, is the “generative AI art”.
Of course it is more complicated than that, but we are striving here for a middle ground between simplicity and complexity. For now, that will do as a definition.
Not all AI is created equally
I have used the term “generative AI” to draw a distinction between what these types of programmes do (ie generate an image in response to a prompt) and what other Artificial Intelligence programmes can do.
Lumping an image-generating programme into the same category as a programme created to analyse complicated medical data, for example, does not make much sense. Other than the inclusion of the phrase “AI” in the name, there is little similarity between them, and the same can be said for the benefits they can bring, risks they pose, and the harm they can do.
When an artist rails against “generative AI art” they are not attacking Artificial Intelligence in general. There are undoubtedly many excellent uses for AI. Generative AI art is not, in my opinion, one of them.
“Jobs won’t be lost, you can become expert prompters”
When artists began waking up to the ramifications of generative AI programmes late last year, one of the first concerns to be raised was the impact on jobs. If a company can get one of their employees to type a prompt into one of these programmes and generate images themselves, then why would they need to hire an artist or a photographer to make illustrations or take photos for them?
Fear not, we were told in a seemingly endless parade of condescending videos and essays, jobs will not be lost. You can become prompters instead.
The theory goes something like this: A casual person on the street can type a prompt into the programme and get it to generate an image. But think how much better the image will be if the person typing the prompt is an artist!
I will spend only a short amount of time on this because the argument has become largely irrelevant, as we shall shortly see. But to address a couple of things before we move on…
Being an artist does not automatically make you better at writing words into a programme. Rather, the suggestion being made is that a prompter with an artistic background can more easily spot when an image spat out of the programme has artistic merit. An artistically-minded prompter can then take those “better” images and plug them back into the programme, asking it to develop that image (or a section of that image) further. Or, an artistically-minded prompter could add words or phrases into their prompts to guide the programme towards a certain art style that they know would be a good match for the subject. Or they could sketch a rough outline of the image they want and feed that into the programme alongside their written prompt.
I concede that all of the above is arguably true.
But it conveniently forgets that none of our artistic skills – the things that mark us out as artists, the things we have practiced and honed – are being deployed here. It does not utilise our brush work or our rendering skills. Ultimately, we are still typing words into a prompt box, just like the next person. It is our artistic “eye” that gives us the edge over the person on the street.
But we weren’t born with an artistic eye. It is something learnt. In fact, there are thousands of videos and tutorials available on how to develop artistic skills, how to better use colour and perspective and shades of light and dark to improve your art. These things are not closely-guarded secrets; they are openly shared.
How long will it take for the non-artistic prompter to develop an “eye”? A few months? A couple of years? And once they do, then our edge has disappeared. An ex-artist prompter is no different from a prompter who has learnt about art. Both are equally as capable of typing words into a programme and then assessing the images it spits out.
At best, even if I was to buy into the idea that I could out-compete other generative AI prompters due to my artistic background, it is only a temporary thing.
It’s even more temporary than that!
I said above that the argument had become largely irrelevant. This is because generative AI art programmes are not the only AI tools to have been released into the wild in 2022 and 2023.
And this is where we need to introduce that YouTube video I mentioned at the top of this article.
The thesis of this video (and many more like it, this is not intended as a specific attack on this one YouTuber) is that you don’t need to be a good prompter, or have an artistic eye, because you can use ChatGPT to write your prompts for you.
ChatGPT is a large language model. If you are not familiar with it, then there are lots of very good explanations out there; I would recommend part three of Henry Neilsen’s series of articles about AI. (While you’re there, check out part two, which features an interview with me.)
In this YouTube video, they are using ChatGPT in place of an “expert” prompter:
- You, the human user, ask ChatGPT to write you a prompt based on a simple premise
- ChatGPT spits out a prompt for you to use, complete with all the artistic verbiage needed to guide the generative AI art programme towards a particular style
- You copy and paste the prompt into your chosen generative AI art programme
- The generative AI art programme spits out an image
All of those “don’t panic, become a prompter” videos are looking very 2022 all of a sudden.
We are automating the fun part
What are the AI programmes doing here? They are automating the best parts of the process. The fun parts. The rewarding parts.
What are you, a human user, doing? Typing a few words and then doing a spot of copy-and-paste.
We have made machines to automate the creative process. This is not a good thing.
“Show me the harm”
When artists and other opponents of generative AI art raise issues, it is easy to counter with an accusation of doom-mongering. Where are the real world harms? Who is really being affected by all this?
The second half of that YouTube video helpfully illustrates one such harm.
What business does this YouTuber suggest as an outlet for their new ChatGPT/generative AI workflow? Creating travel posters of specific locations and selling them on Etsy.
Hang on a minute! That’s what I do. Or, at least, it was.
One half of my business sees me working on book covers and character art, and the other half was taken up with creating and selling prints of specific places to sell on Etsy. Generative AI evangelists are encouraging you to use generative AI automation to compete with the products that I create and sell.
But surely I can out-compete a generative AI art programme. Right?
How long does it take them to create a new design? A few minutes.
How long does it take me? Days.
How do they decide what designs to use on their prints? By waiting for their generative AI programme to spit out something they think looks nice. (Does it matter if it doesn’t actually look like the real place? Of course not!)
How do I? By visiting places, taking photos, soaking up the atmosphere of a place, and trying to impart some of those feelings and emotions through the choices I make when composing and painting a new design.
How do they produce their prints? By outsourcing the printing and posting to a drop shipping company.
How do I produce mine? With a local printer, using environmentally-conscious materials and inks, before packing it with a hand written note and taking it to the post office myself.
You may decide that this level of automation is a good thing. That the ability to produce more, quicker and cheaper, is a positive for the world. I disagree.
I believe in my process but I know I cannot compete with the automation these new businesses are employing. I believe my prints are a better product than the prints they produce, invested with more feeling and personality and produced in a more sustainable way, but I know I cannot charge a premium for that.
I am not holding myself up as particularly virtuous or special, but I do believe that there is greater value in something created by a human, with intention.
Can artists like me compete against automated art? No. Not without help.
Who can help?
The public can. Not by paying a premium for human-made art. I don’t believe that is a fair solution. All it would achieve is to bar those with less money from having access to non-AI art.
Instead, the public can help us by rejecting art made with generative AI automation. If given a choice between two things, one made by a human and one automated by a programme, then choose the human-made one. That doesn’t have to be a vote with your wallet either. Something as simple as sharing human-made art on your facebook or twitter or when talking to your friends and family is fantastic support. (Just make sure you let everyone know who the artist is, of course. Crediting is caring.)
At the same time, choosing not to share art made with generative AI programmes also goes a long way to supporting human artists. Let’s avoid normalising automated art.
Who else can help? The companies whose existence is predicated on the work of artists, that’s who.
In this example, it would be great if their get-rich-quick scheme fell at the first hurdle when Etsy refused to let them open a shop on their marketplace. But, depressingly, that is not the case.
After watching that video I went straight to Etsy and searched for AI art. Oh my! There is a lot of it. The place is awash with the stuff. Narrowing my search to those locations I have painted, it didn’t take long to find AI art prints in the search results. While some were openly crediting their generative AI roots, not all of them were so transparent. It is not always going to be easy for the customer to know.
After some further, ever-more-furious Googling, I came to the conclusion that Etsy doesn’t care. What was once a home for hand-crafted products made by real artists is now a marketplace like any other and a welcome home for automated art.
I am no longer on Etsy
I can no longer support a marketplace that doesn’t support human artists. My shop is now empty. The virtual shelves are bare. All that is left is a note letting people know why I have decided to shut up shop.
I transferred my print shop to this website but the lack of reach, exacerbated by the death spiral of twitter and instagram, platforms that had previously been good at driving traffic to my shop, resulted in extremely low sales in the second half of 2023. As such, from the start of 2024, my shop has now closed.
I am fortunate that print sales are only one part of my business and I can continue to focus on book art and character commissions. However, both of these areas are also under threat from automated art.
This is just one example
It is important to note here that this is by no means the greatest harm being perpetuated by these programmes. My relatively privileged position as a white, western, not-totally-brassic man allows me to weather these sorts of problems when others may not be able to. There is a lot more that can, should, and is being said about the wider harms that these programmes are causing.
My aim here is not to raise this one issue above other, more harmful problems. Rather, it is to give just one example of a real-world problem these programmes are causing.
If you need more examples you only need to speak to artists, and writers, and musicians, and voice-over artists. This is not even the only example I can give from my own personal experience. It is affecting all aspects of the work I do, from print sales to book covers to magazine work to character art commissions.
These programmes could not exist without us
What is even more infuriating is that none of this would be possible without the work of human artists and photographers. To dive into the tricky technicalities of exactly how these programmes have automated the creative process would take more space than I have here. I will save that for another day.
However, the short version is this: Billions of images were taken from the internet and used to train the programmes. This was done without the consent, or even the awareness, of the people who own those images.
My work was taken. They went to my website and found images of my prints and they used those images to train their programmes. And now those programmes are being used by people who are positioning themselves in direct competition with me.
Now imagine that scenario repeated millions and millions of times over, as every artist whose work was taken finds themselves having to compete with the bastardised ghosts of their own work.
But wait, there’s more
This article has gone on long enough and I feel like I have barely scratched the surface of the topic. Please remember that this is not intended as a complete overview of generative AI art, its issues, or the work going on to minimise the harm it causes.
Rather, it is a personal rumination on just one of the ways in which this new, disruptive technology is affecting me and my work.
If you are looking for more information about generative AI automation then I have listed some further resources below.
I will return to this topic, I’m sure. Probably right after watching another triggering YouTube video. For now, all I would ask is that you support human artists and the work they create. I believe our world will be a richer place for having humans at the heart of the art we enjoy.
Glaze: A University of Chicago project working to protect internet users from invasive uses of machine learning.
Protecting Artists from AI Technologies: A GoFundMe by the Concept Art Association, who are raising funds to push for greater legislation and to repair some of the harms already suffered. Their fundraising page explains more about the ethical issues of generative AI automation.
Center for Artistic Inquiry and Reporting Open Letter: An open letter calling for restrictions on AI illustration in publishing.
AI Doesn’t Care – That Means We Have To: Henry Neilsen’s set of articles exploring different types of generative AI programme.
Getty Images sues Stable Diffusion: An article from The Verge covering the UK legal action being taken by the stock image site Getty Images.
Class Action lawsuit against generative AI companies: An article from Bloomberg on the class action lawsuit that has been brought in the US by a group of artists.
The WGA strike and AI: A Wired article about the current writers strike and how AI affects screenwriters.