Art Through the Lens of AI: A Mid Journey Experiment at Art Omi
As an artist in residence at the Catwalk Artist Residency, I was granted the opportunity to visit Art Omi, a beautiful sculpture garden nestled in upstate New York. During my visit, I was particularly captivated by a sculpture that appeared as a tree composed of basketball hoops. This unique piece was crafted by Alexandre Arrechea and is known as ‘Orange Functional’¹.
Fascinated by this unusual and interactive sculpture, I set out to test the capabilities of a modern artificial intelligence platform, Mid Journey. This platform is equipped with a feature that allows users to upload an image and outputs four text-based descriptions of that image. These descriptions can then be fed back into Mid Journey to generate visuals based on those descriptions. My objective was to assess how closely the AI-generated visuals would align with the original photograph of the ‘Orange Functional’ sculpture.
Mid Journey’s /describe command is designed to analyze and potentially reverse-engineer images into text prompts. Given an image input, the tool generates four rough guesses of prompts that might produce a similar image. However, it's crucial to understand that these “guesses” are not necessarily aligned with what experts would consider a “good” or “well-formed” prompt².
Interestingly, the third description generated a community guidelines violation when passed back into Mid Journey. However, upon clicking the dispute button, I was allowed to generate the images.
Below, you’ll find the four descriptions generated by Mid Journey when I uploaded a photo of the ‘Orange Functional’ sculpture. Accompanying each description are four images, produced when I fed each description back into the platform. The outcomes were quite fascinating. For each descriptor, Mid Journey produced a set of visuals that mirrored elements of the original sculpture, yet also added fresh interpretations that strayed from the initial photograph.
“Basketball trees in a field with some basketball hoop’s in them, in the style of public art, light red, midwest gothic, made of plastic, metallic rotation, mid-century, teethcore — ar 3:4”
“Basketball tree with basketballs in it, in the style of large-scale public art, midwest gothic, mushroomcore, light red, frontal perspective, whiplash curves, pastoral — ar 3:4”
“Arnold Fuchs basketball tree 2015 Laurie Vrienden, in the style of Jeppe Hein, Soviet realism, George Catlin, ornamental structures, dark red and light indigo, grandma Moses, low-angle — ar 3:4”
“A sculpture sits in the middle of a field with basketballs thrown in, in the style of light red and dark bronze, George Ault, ethereal trees, Svetlin Velinov, extruded design, realistic blue skies, public art — ar 3:4”
This experiment demonstrates the creative potential of artificial intelligence in the realm of visual art. While AI tools like Mid Journey may not precisely replicate a given image, they offer artists new ways to engage with their work, inspire fresh ideas, and explore unique aesthetics.
I’m currently unable to find any specific information about why the third description may have led to a community guidelines violation. Further research would be required to understand the rules and nuances of Mid Journey’s community guidelines and how they apply in this case. Despite this minor hiccup, my experience with Mid Journey was a fascinating dive into the intersection of art and artificial intelligence. It underscored the fact that AI tools are not just about replication, but also about transformation and reinterpretation, offering a new lens through which we can engage with the world around us.
Stay tuned for more AI art explorations!
