Will I be replaced with a design robot?
Robot images courtesy of Andrew Szurley
When I first saw the illustrations in my LinkedIn feed, posted by my friend Andrew, I was immediately transported back to the 1980s when I spent hours pouring over fantasy illustration books featuring work from Peter Elson, Chris Moore, John Harris and others. But the colors, texture, and other-worldly details of Andrew’s fantasy illustrations are even more impressive because they were created entirely by artificial intelligence (AI).
AI platforms like MidJourney and DALL-E have been gaining more and more attention recently as folks have been playing around with their text prompt-based interfaces to create some truly impressive (and weird) images.
One of my first thoughts when I saw the AI-generated images was, “Could I use something like MidJourney to create custom illustrations for client work?” I admit this thought was partially motivated by my own tendency towards laziness but mostly I just wanted to see what the tools could do. So I signed up to be a MidJourney beta-tester.
My “MidJourney” into AI
In order to access the MidJourney app I had to sign up for a Discord account, which took about 15 minutes. My introductory MidJourney account would allow me to create up to 25 images before I had to pay a membership fee.
So, what would I “draw” first? After a minute of thought I settled on a pirate (big surprise). I noticed several other Discord users had created illustrations in the style of famous artists so I typed in “pirate portrait in Van Gogh style” and waited for the AI to produce the first images. The MidJourney app creates four variations of your prompt, of which you can select any one of them and ask it to produce four more variations. Unfortunately, the only control you have over the image outcomes are the text prompts you enter into the AI.
“Van Gogh” pirates
Photo-realistic pirates (according to MidJourney)
My Van Gogh pirates were interesting (not sure why a pirate would need two eye patches) but I wanted to see what the AI would do with the prompt “photorealistic pirate portrait.” The results didn’t exactly look photorealistic but they were definitely more “real” than my Van Gogh pirate. I selected one of the results and told the app to create variations of that one. Then, I selected one of those results and asked the app to render it in high detail.
The mixed Results
I’m impressed that the AI could come up with something recognizable based on a few simple reference words but I think it’s going to be a while before it can produce something that is indistinguishable from a photograph (or even a well executed illustration).
The algorithm seems a bit confused about how faces work. The “tattoo” the AI applied to my pirate’s face peels up in places and he also has a couple of small, auxiliary eyes — one of which is nestled in the corner of his right eye (not creepy at all). The upper lip seems to be made of non-lip like material and the visible ear has definitely gone a few rounds with Mike Tyson.
Some of the details are great, though, like the mustache hairs and the fabric folds in the hat. I’m also impressed with the lighting and how it defines the three-dimensional space the face occupies.
will Ai replace creatives?
AI technology has come a long way since the days of computer chess and the fact that you can sit down at your computer and within 90 minutes create a small portfolio of custom illustrations is impressive. Image generation is not the only area of creativity where AI is making inroads, though — there is intelligent software that will help you write articles or generate content for your social media channels.
Will AI tools replace creatives? In some respects, maybe. When I first started working as a designer, stock photography was in its infancy and very expensive. Now, inexpensive and free stock photography has become the norm for many marketing applications and hiring a photographer to create custom photography is the exception. The AI tools that are now appearing will likely capture a slice of the creative market but I don’t think I’m in danger of being replaced by a robot anytime soon. I will, however, continue to play with the image-generation AI and see if it can eventually render a pirate that I’d be willing to hang out with.