November 9, 2022
How an experience designer unlocked years of creative frustration with AI art
by Cory Cruser
I grew up with a cowboy father and knew how to work 200 head of cattle in a day before I knew my multiplication tables.
So, externally speaking, it’s somewhat odd that I loved tap dance, musical theatre, Mozart, Matisse and the Die Brüke art movement.
When I finally received my DALL-E invitation, I knew one of the first things I had to create was an homage to such a childhood: a cat cowboy herding cattle on a ranch in Montana, in the style of Ernst Ludwig Kirchner:
I have to say, I’m not mad about it. Sure, there’s so much wrong with the output, from the color palette to the brush strokes to the perspective. But there’s also so much right: it shows the foundations of what would become expressionism, which Die Brüke is credited with; it has a primitivist sense of place and a hyper unrealistic use of color; it’s literally a cat cowboy herding cats. And for the thirty seconds it took to type the prompt, it’s a quite remarkable output.
There will be enough thought pieces on the early ventures in AI Art centered on cats and astronauts, just as there will be on how everyone missed the potential of NFTs in the early days when they were centered on gorillas and 8-bit cryptopunks. I’m not going to contribute to that literary pantheon.
Instead, this article is about the freedom this experience designer has felt in the form of an AI art prompt field.
A life of artistic frustration
My hands were not meant for pen-to-paper. My handwriting looks as if two demons are fighting for control. In fact, famed facilitation illustrator, Dan Roam, who believes everyone can draw at some level, once looked at my version of a cat and said, “You had better stick to words.”
For most, not being able to draw isn’t that big a deal. For me, it is an almost lifelong struggle—especially because I’ve built a 20+ year career on being an ideas guy. Ideas are only successful if they can be easily communicated and understood. As the saying goes, a picture says a thousand words. In my professional life, as this article will attest, I’ve had to use thousands of words to compensate for my artistic shortcoming. Translation: I’ve had a very frustrating career.
Yet DALL-E and the arsenal of AI art tools like Midjourney, Stability, Neural Blender and DreamStudio are arming me with a new communication capability. If I can describe it, I can now create it visually. If I can create it visually, what was once a plead to read thousands of words just became moot… at least partially. Here’s how.
Ideas are only successful if they can be easily communicated and understood.
Experience at the heart of every brand
So much of branding can feel like theory. We stress over words and conceptual metaphors to manifest a promise that a company puts out to the world. But that promise is only as good as the experiences a customer has with the brand—in other words, how well the company delivers the promise. A brand that just creates a purpose or a new visual identity but doesn’t invest in the experience is comparable to fixing a house with a cracked foundation by putting up new drywall and paint.
That’s why it’s critically important to develop a view of the future experiences a business should deliver so a customer can “feel” the promise of a new brand positioning. It has also been notoriously expensive for brands to run an experience design stream in parallel to brand positioning and identity design.
Consider: much of brand strategy can be developed through primary research and distillation of business strategy. Similarly, much of brand identity—from visual to voice to name—is often best served through focused work between business leaders and brand experts.
But experience, especially for existing businesses, is very long tail with a great degree of ambiguity. From current journey mapping, concept design and refinement to building the business case for investment, training delivery of the experience and beyond, it requires a wide skill set requirement of designers, architects, digital, technology, service design, human resources, media buyers, data analysts, product owners, delivery experts, and more.
This is where AI comes in.
By visualizing the early concepts of an experience, there’s greater than 50 percent time savings on a typical project. Because we can see something as a collective, the team can agree on whether we are heading in the right direction or not. We can also place more fully fleshed-out concepts in front of customers early and often, getting better and more accurate feedback that drives our focus and helps weed out or refine poor concepts before major investment is made. This means not only cost savings, but quicker time to market.
To demonstrate the profound impact AI is already having in the field of experience design, let me show you a couple of use cases that make the theoretical real.
I had an idea.
In designing a proposal for a never-been-done hotel, I had the idea to build a hotel that floats on the sea. The details don’t matter, but what I wanted to display was the idea of the optical illusion—creating a glass hotel room pod that looked as though it were floating above the sea.
I started with DALL-E and typed in the following prompt: 8 floating glass pod luxury hotel rooms suspended over the ocean, golden hour sunset, architectural illustrations. What I got back was impressive, but missing the mark a bit.
So, I flipped over to Midjourney and tried a similar prompt: glass hotel room pods that are floating over the red sea, architectural illustration.
Immediately, I saw there was a more viable option: the lower left-hand side. Midjourney offers the ability to upscale an image and/or create derivatives of that image, so I began playing with that option. I still hadn’t landed on what I was looking for. Back to the prompt.
Next variation: Eight separate glass hotel room pods that float 5 meters over the sea, at sunset, architectural illustration. Clearly the AI was picking up on different parts of the prompt to show different variations. I upscaled a few of those but the closeness of the rooms still felt wrong. Back to the prompt.
This time I took the same prompt but added on one more piece of criteria: looks like an optical illusion. Boom! Now I was getting closer to what I had in mind. A little more tweaking and making it seem a bit more realistic and I had in 20 minutes created something that would have taken an illustrator and me 2-3 days and multiple iterations to get right, when all I needed was something sophisticated enough to stand up to a narrative I was building and elegant enough to bring a concept to life. Mission accomplished!
It’s not me, it’s the AI.
While there’s a lot of fun to be had in this digital AI art world, there is a learning curve to get good at it. It’s like HTML: a lot of people can learn it, but that doesn’t mean they can build an app.
The heart of AI art is the prompt which falls into three parts:
- The main idea: what you are trying to communicate narratively
- The setting: the surrounding world building you describe; the clearer a description, the more resonant and accurate the output becomes
- The treatment: the artistic medium and style you seek to produce; here knowing art history, art production, photography terminology, etc. is valuable if not essential
Most people are fine with the main idea. Few really go on to refine the setting and fewer still go on to build out the treatment beyond “photo real,” “by Picasso,” or some similarly simple terms. The setting and treatment, from my observations and experience, are what separate AI play from true AI expression. Those mastering the craft of defining and refining the setting and treatment are also those that are taking AI Art from play into the real world, for better or worse.
In my explorations, I found that the times I was getting the poorest results were not because of an AI failure, but because I was either being overly functional in my descriptions or on the other end, being too florid with my language. Sometimes, however, the results of an abstract prompt were devastatingly beautiful and surprising.
What was most profound and unexpected, however, is that a text-based prompt entry was forcing me to think through my ideas and concepts in ways I hadn’t before. Ultimately doing so meant my personal understanding, ownership and ability to convey and more accurately improved.
As an ideas gent…
…I can’t begin to count the number of game-changing ideas I have left on the cutting-room floor largely because, in the moment, I couldn’t convey them with enough clarity. So you can imagine how satisfying it felt when late last week I was asked if I could create a quick storyboard for an airline of the future to help in a proposal process and I was able to say, without reservation, “Yes.” In an hour’s work I was able to draft a narrative story. In the next hour, I was able to bring that story to life in a fully visual way.
And as an experience designer…
I’m just at the beginning of employing AI art in my work. I’m arming my teams with this toolbox and encouraging them to do the same, living in a place of experimentation with this new technology.
The road forward is slowly coming into focus but there are still myriad use cases that have yet to be discovered. No doubt, just as there were no end to horrid websites at the dawn of the .com age, so too will there be no end to bad AI art over the next several years. But it won’t be long before those of us in the ideas business have perfected how to bring this budding capability into our work, making us better custodians of immersive experience building, better partners to design crafts people, and able to get more stuff from brains to paper to the real world – the real goal of all of this.
At the end of the day, craft will always win.
With every new artistically inclined technology, there is immediate uproar from dyed-in-the-wool practitioners. With AI Art the debate is really two-fold as I see it:
- If AI art is so quick and so good, what is the value of the artist?
- Since AI is trained on publicly available art, isn’t that really pirating other artists work?
Starting with the second point, as someone who studied art in many forms during university, I would argue that all art is derivative. I watched masters like Bob Fosse and Jerome Robbins to learn choreography and staging. I copied lighting from David Nancarrow. I acquired acting skills from John Malkovich. Vivian Westwood inspired costume designs.
In short, all artists are a product of their influences. One of the first questions in most interviews with celebrities is, “Who are your influences (do you listen to, etc.)?” It’s expected that we learn from what came before.
Without Mae West there wouldn’t have been a Marilyn Monroe; without Marilyn there wouldn’t be a Madonna; without Madonna there wouldn’t be a Britney Spears; without Britney there wouldn’t be a Lady Gaga, and so on.
Just like Google has organized the world’s information, it feels like AI is organizing the world’s art and empowering the artist in us all to create like those who so inspire us, even if that creation is simply derivatives of cats and astronauts in the style of Caravaggio.
To the first point, history always shuns new artistic-enabling tech. Painters hated photography, stage actors hated film, authors hated word processors, photographers hated photo software, etc. Yet history has shown that no matter what new technology emerges, it only seems to increase the value of the artistic foundation that came before it. Because streaming has made film so ubiquitous, going to the theatre feels special and actors on stage seem elite. Because photography is so pervasive, painted mediums are treated with greater prestige and professional photographers more easily stand out from the crowd. In essence, because every grocery has shelf stable bread, it makes a fresh-based sourdough loaf a thing to relish.
The same will hold true for AI art. Those who are true craftspeople will always rise to the top. Will it make the field more competitive, weeding out lessor, mediocre talent? Most likely. Will it shrink the size of the pool for those who can call themselves artists (meaning make a living doing it)?
Again, probably. Will it create a whole market of AI art by glorified amateurs that ends up filling online stores only to be placed above toilets in households around the world? Undoubtedly. But will it replace the true artisan craftwork? Definitely not.
I never understood the value of true design craftsmanship until I worked with the extraordinary talent at Lippincott. They design logos and visual identities with such a degree of strategic craft that even if a machine could do it with exactly the same precision and insight, I expect most people would still opt for the dynamics and rigor of humans bring to the process. Why? Because creating something that emotionally connects to humans isn’t just about precision; it’s as much about sharing the story of how you get to a place, building a share ethos around the work, and creating consensus as to the promise we want to put out into the world. All of these things are innately human and nuanced such that machines may try to mimic it, but will be hard-pressed to ever fully react that level of sentience.
Let’s see if the last 200 years proves me right 200 years from now.