Proving the Human Innovation of Art

by Douglas Clarke

As the world enters a new age of generative AI we are faced with technical innovation’s role in the arts crossing a significant boundary. Technology has offered advancements for the artist’s toolset. With each advancement comes a leap in efficiency of creation. What hadn’t changed until very recently is the act of artistic creation itself; a human process with its own innovations. Innovations in art are generally driven and limited by an individual’s imagination, effectively making innovation in art as boundless and infinite as human consciousness itself. 

When technical innovation in the field of machine learning yielded Generative Artificial Intelligence, the boundary between technical innovation and artistic creation was breached. Now an AI can create art. However, I propose that artistic innovation shall remain a human process. Technological innovation has given artists yet another tool, another efficiency; and that is creation itself. 

I am both a technical innovator and an artistic innovator. These two sides to my life and work are complementary, but I have had to reconcile what sets apart artistic innovation from artistic creation; especially when that creation is a product of technology. The difference is the human imagination, powered by human emotion, and its application in the form of human intention. 

In order to prove this reconciliation, I decided to take a temporary leave of my 26-year career as a technical innovator to conduct an experiment being a full-time artistic innovator. I only had a small budget for this experiment. It was August 23rd, 2024 and I decided that I would write and direct an original theater play. I would additionally create the play’s entire soundtrack. I promptly booked a two weekend run with a local theater group, Zombie Joe’s Underground in North Hollywood, CA, who would co-produce the play. The play would open January 10th, 2025. I had exactly 20 weeks to do something not only artistically creative, but artistically innovative; something I had never before attempted. 

I set some ground rules for myself. I would allow myself to use generative AI, but only as a utility. Artistic innovation had to come first, and it had to come directly from my heart. In the same way the human intelligence helps to articulate the heart’s expression, so may, in this case, an artificial intelligence. In this way, creation can be a product of innovation expressed through intention. 

The creation of the play commenced. The play would be called Soulmate, an immediate recognition of what characteristics we define as most human; the soul and love. To sell tickets online, the theater needed a play poster. 

I had a very specific key image in my mind’s eye, but with very limited drawing skills, no way to create it. By using an AI image generator (DALL-E 2), I could begin to have my thoughts materialize in an image creation. 

The primary frustration that I have had with any generative AI utility, is that if you are imagining something very specific, it is nearly impossible to produce a satisfactory result with a one-shot prompt. While more advanced pixel-level ai tools are beginning to emerge which will let the user fine-tine an image, all I had available at this time was the general context window, and my prompts. While the generated images vaguely resembled the image in my mind, several aspects would be very wrong. To improve the result, I used several prompt engineering techniques that I had learned in the many ML and AI courses I had taken. Results were slightly better, but the result was still way off. I started realizing that using an image generator alone would take too much work and too much time. I needed a solution. 

I thought back to paintings I had made years before. Having no drawing or painting skill, I would photograph people, places, and objects that resembled elements that I imagined. I then brought those elements into an image editor (Adobe Photoshop in this case) and would create a crude collage that placed each visual element where I would want it on the canvas. Then I would project the image onto the canvas. That way I could just trace the lines, fill in the colors, and complete my painting. 

So this is what I did with these AI outputs. I put them into Photoshop, clipped the elements I desired; resized, positioned, and combined them into a crude collage, and then projected the result on a canvas, painting over the lines and filling color in between with acrylic paints. I then painted my own free-form background to fill the space around the key image. I photographed the canvas, and that photograph became the basis for my play poster. 

 There is only one more time that I used generative AI. I realized that I needed a much larger digital canvas with an extension of the painted background, so that I could fill in all of the lettering and information, so I used Adobe Firefly to expand the background. I then completed the poster in Photoshop. I had budgeted 3 days for this process but it took me an entire week. The image generation was slightly helpful as an initial step, but created several difficult additional steps for me to mitigate its own shortcomings. 

With the poster complete, tickets on sale, and 19 weeks to go it was time to write my play, write the music (playing and recording each instrument in real-time), cast the play, create promos, and do everything else I needed to do to open on time. I had AI tools lined up for writing and for generating music, but I didn’t touch them. I simply didn’t have the time to. I realized that I had such a strong vision, stemming from my heart and interpreted and visualized by my brain, that all of the artistic innovation I needed was already there. The generative AI technology may have helped, but would only ultimately slow me down. 

With the Los Angeles fires raging the week the play was to open, we pushed the opening performance one night to January 11th. The play went on for 5 performances total over 2 weekends. Each house was full and the last two performances sold out. So we extended the play for 6 more performances over 2 more weekends. The extension was so successful that I am now taking the play off-broadway in NYC, opening in late May 2025.

Connect with Douglas

 www.linkedin.com/in/douglasaclarke