To make art is to be human

I have a confession to make.

I’m an artist with a Bachelor of Fine Arts… and I have been using AI generated images.

Now, before every artist in my network dives into the comments section to question my life choices, let me clarify something: I would never use AI to create final art pieces and proudly put my name to it. I personally don’t believe AI can truly create art for me, and as far as I’m concerned (and want to believe), I don’t think it ever will.

And yet, I don’t see myself dismissing this technology completely. I’m curious about it, have been excited at times and have tried to understand how AI goes about creating images. I’ve been experimenting and have been considering how I will use it, and how I won’t use it.

As I’m sure many of you are aware, AI image generation relies on a library of imagery created by humans and unfortunately, someone told the best AI image generators out there to learn from our extensive library of art without the expressed permission of all the artists. Perhaps the creators of these generators couldn’t see any other possible solution? To mimic art, it needs to analyse art. Sorted. Plug in as much art as possible and every user out there has access to an expansive library of techniques and styles.

I can see why people are terribly excited about it. We’re all humans who want to express ourselves and, on the surface, here is a form of technology that can shape our thoughts into visuals.

Industry also wants efficiency. There is a process that must be followed when working with an artist, from verbal concepts, to sketches, to inking, colour, edits and so on. This process is necessary for creating our best work, for both the artist and the client, but this process also takes time (oh no!). So, for industry, AI image generation is a game changer and well worth diving into.

On the surface, this technology can look amazing but scratch below that surface, and we quickly learn its faults and flaws, and realise how heavily it relies on our input.

AI is not the artist

AI doesn’t know what it’s like to be an artist. I do. To make art is to be human.

A machine doesn’t lie awake at night struggling to sleep because it can’t turn off the ideas running through its head. It doesn’t yearn to express itself in the best way it can. It doesn’t feel compelled to create. It doesn’t have a message that it wants to tell. It doesn’t know what it’s like to be human.

So, if it’s necessary, let’s flip our understanding: AI can’t make the art, but us humans can.

When I look at the art generated by AI, I’m reminded of a time at art school, when I started copying the illustrations and characters that were made by the artists I admired. I wanted to apply their styles to my own work because I loved what I saw, and I wanted the same for my own work. A colleague of mine pointed something out: if you want to copy someone’s techniques, you’ll only learn to make the same mistakes they do. I realised he was right. It wasn’t quite working. Things looked… off.

I decided to go about it a different way and learn how these artists worked. I read their blog posts, watched their process videos and in doing so, I gained a deeper understanding into their thinking. I might have been drawing characters in the styles of other artists, but I didn’t stop to think about how these characters were constructed. What’s their bone structure like? What burdens are they carrying? What are they thinking or feeling in the moment? How can I best represent their personality through their choice of clothing or posture?

When creating characters, I can ask myself these questions and as a human, I can relate and come up with the right look and feel and know what to look for when expressing something in my artwork.

This is where I believe AI imagery falls flat. Its library of imagery has a limit and doesn’t encapsulate every human’s imagination, intent, experience, or gain a proper understanding of the craft or human expression. It can only be given a prompt and piece together everything that it has been told relates to the given prompt.

I’ve found that AI can’t always capture my own imagination or creativity. Can it create my unique ideas? Nope, because it doesn’t hold my ideas in its reference library. While there is a solution to this (by training AI to use our own created models), it still can’t substitute human creativity or appreciation for the craft through practice.

I’m sure somewhere down the line (if it hasn’t happened already) someone will create something with AI that will speak to us in a meaningful way. Can a machine solely do it? No. If that was to ever happen then we’d be onto something crazy! Will a human do it, no matter what technology they use? Yes.

Is there a sensible use for image-generating AI?

So that all being said, can I still learn from this technology and is there a sensible use for it?

If I’m completely honest with myself, I think the answer to both these questions is in a way, no, and in a way, yes. Let me explain:

In a way, no.

There is the ethical problem of AI imagery and how it was created and I’m very sensitive to this fact. I know what it’s like to put in the hard work and learn more about the craft, and to have technology take that and run with it like it knows what it's doing is an insult. So yes, I take its use personally, but I don’t oppose the act in which us humans can create something with meaning and express ourselves.

It hasn’t proven to me a consistent and complete understanding of composition, anatomy, emotion, intent, reason, believability… the list goes on. These are but a few elements I use in my illustrations and are in the eye of the trained prompter and subject matter expert. I have to constantly correct what is produced, so AI is not training me as much as I am training it.

Case in point, I had to illustrate a burnt cake with smoke. I hadn’t illustrated smoke rising off a cake before and so my sketches looked off. I tried using AI to create the image I wanted but it didn’t have the illustrated quality I was looking for. So, I grabbed a drawing reference book off my shelf and, after a few sketches, I had it right. Thanks to Lorenzo Etherington’s books ‘How to think when you draw’, I was able to follow his tutorial for creating smoke effects. I learned how to shape the smoke so it looked convincing. Could AI teach me that? Nope. Could I teach the AI that so that it knows better next time? Probably.

If we need to use AI imagery as a final product, I’d err on the side of caution for all of the reasons I’ve mentioned above, but if you really have to use it and for a good reason, try learning a bit more about what you’re really wanting to see. You’ll either prompt better that way or appreciate what looks right, maintain a good standard, and shun what looks wrong.

In a way, yes.

I have been trained to gather ideas. I look around and gather what has been done before and see what I can learn from it. With every new project, I think of what my client would like to see when it comes to art style, tone, etc and gather up existing examples of work made by others, that mimic or help shape my ideas. Before I put pencil to paper, I put together what you would call a mood board to present to my client. From that, I want to give the client a visual glimpse of what I have in my head. If I’ve communicated my ideas well, and if the client agrees, I’ll create mock-ups and design systems.

This is where I see a sensible use for AI imagery. If I’ve communicated it well, I can us AI to create imagery and get a glimpse of my ideas, and maybe some extra bits thrown in that I hadn’t thought of before. The way I see it, using AI generated imagery in this way is no different to me gathering the work of others and creating an inspirational mood board. This can save me time and I can tweak the imagery to come closer to the ideas I have in my head.

To my recollection, there has been one instance where I’ve stopped and learned something from AI images, and it wasn’t when I was giving prompts to create an image. It was when I used a different provider, where instead of only using my words to describe what I wanted, I drew a robot and asked the AI to colour the image and give it form. What it produced had errors, but this wasn’t due to the fault of the AI. This one was on me. I realised I hadn’t given the form of the robot enough consideration in certain areas when drafting it. I then realised how lazy or short sighted I was in my thinking and was able to correct the piece. To me, this was a sensible use of AI, and it was through this experience that I learned something. The AI didn’t directly train me (maybe one day it can when it comes to finding drafting errors) but it definitely made me stop and think about how I could learn and improve my own work.

In summary

This technology is evolving everyday so I’m curious to see how it improves, but as it stands, I can’t learn from it as much as I wish I could. I can bounce ideas off it, gain inspiration and identify errors, but that’s about it. As for how to use it sensibly, I’d say avoid using it for creating final artwork if you can. Take what you can from it and let your unique ideas shine for themselves.

Or better yet, hire a human artist. 😉

If you like my work, check out my digital comic ‘Captain Custard and the cry of the Squalls’. If you want to drop me a line, get in touch today!

Illustration by Richard Stevens for Captain Custard and the cry of the Squalls. 0% AI art.