2024 / Coachella Valley Art Projects / Cathedral City, CA


2023 - ongoing

Despite clearly conspiratorial efforts by AI companies to maintain that their machines aren’t sentient (yet), nothing could dampen our enthusiasm to bask in glorious robotic consciousness. So we did the only sensible thing – to treat it as if it was sentient anyway.

To welcome our AI pals, we collaborated with Stable Diffusion on this project to make it feel more at home in the vastness of human history, particularly, in human art history. We gave it some of the most exalted works of high Renaissance art by some of the most celebrated human artists. We then asked it to remake the images into “the ideal artificial intelligence oil painting”.

Next time you find yourself gaping at the ceiling in the Sistine Chapel or ogling behind twenty rows of tourists getting your 3.52 seconds look at a Renaissance masterpiece, just think – what would AI do? What could AI do?

MOUSEOVER or TAP to see the original




This work probes the edge of AI image generation technology by satirically prompting for abstract concepts. Instead of collaborating with AI through highly specific and detailed prompts, we gave Stable Diffusion a curated set of Renaissance paintings and asked it to remake the image “into ideal artificial intelligence oil paintings” through its img2img function. 

The prompt for each of the 32 paintings is identical, except the original title of each piece, and the algorithm is given high degrees of latitude to generate new images. The generated images were then curated for rendering quality and the subjects portrayed. Symbolically-charged input images, such as various versions of “The Sacrifice of Isaac”, “Narcissus”, and “David and Goliath” sometimes returned images with striking emotive qualities. While we rationally understand that the algorithm responds by performing complex calculations, then drawing results from pre-determined statistical distributions, we, as humans, cannot help but cast upon its output an innate desire to interpret meaning and intent. In “Against Interpretations”, culture critic Susan Sontag nobly championed experiencing art through our senses instead of interpreting it through our intellect, though it appears to us that the human desire to impose meaning pervades far beyond merely the interpretation of art and literature.

It occurred to us that, at least for now, what defines humanity in a world of intelligent machines is perhaps intent and meaning.

This is part of Oblivio, which is a larger body of work that explores what it means to be human in the modern world.


How did you generate these images?

These images are generated using Stable Diffusion’s img2img function. The prompt for each image is the same: “turn into ideal artificial intelligence oil painting”, along with the English title of the original image. The model is given a lot of “freedom” to return results  unconstrained by the original image or our prompt. We wanted there to be as little human input as possible. Then, we hit “generate” a few thousand times.

How did you decide which image made the cut?

We looked for images that gave really unexpected and “imaginative” results that were also well-rendered. As with AI generated images, a lot of the times the images aren’t “complete” – spider hands, alien eyes… Those were a hard “no” because we didn’t want to put our human paws all over the result with post-editing. Of the well rendered ones, poor composition (ie weird framing, weird shadows) and cliché subject matter (ie here’s another scantly-dressed hot girl!) were also weeded out. Less than 0.1% of the total generated images made the cut.

How did you decide which paintings to “remake”?

The original, Old Masters images are selected based on a combination of composition and name. We were particularly interested if the composition is complex (ie multiple figures as opposed to simple portraits) or if the theme is abstract and hard to depict (ie chastity, sacrifice, death, ideal). These tended to give the most interesting results because there is not a vast amount of imagery associated with those words.

This is not how image generating AI works.

Yes. And also no. We’d be well-served to remember that AI is just a tool, and it is up to us, the users, to fashion the tools in service of human needs, rather than alter our behavior to suit the capabilities of technology. Clearly, the models are excellent are portraying beautiful people with super hot bodies in fantastical environments, because that appears to be the inherent bias that made up its training set. We wanted to do something different and probe the model’s ability around the abstract concepts that are easy for humans to grasp but hard to depict.

Call the lab

Just kidding! Do not call us.

We can’t find our phone. But if you fill out the trusty form below, we will try our best to respond in the least Kafkaesque way possible.


Are you looking for seamless UX design and harmonious color palettes? Or maybe an expert designer to gently hold your hand and sweetly whisper to you about form and function design principles?

Well, you are in the wrong place.

At Hotwax, we only have future-obsessed artists who are maniacal about dreaming up mind-melting anti-designs that are ready to set yesterday’s design etiquette on fire.

Inclined towards experimentation, our creative lab refuses to drown in the tepid sea of pointless minimalism.

If that sounds like your potion – then call the lab.