Vera Molnar, John Cage, Sol Lewitt, Casey Reas and Sasha Stiles have explored the frontiers of an art practice known as generative art. The artists prepare a set of formal instructions, which are interpreted to perform an artwork. The instructions can be written on paper, and then the performers read them to build a sculpture, to paint on a canvas or to perform a musical piece. The instructions can also be formalized as a computer program. The computer interprets the instructions to generate a digital image, synthesize sound live, read poetry or perform choreography fully autonomously. That’s the modern version of generative art.

Generative AI emerged in the past year, as a form of AI that generates images (MidJourney) or text (CHatGPT) in response to a set of instructions passed as a “prompt”. Artists have rapidly embraced the generative power of these new AIs to produce animations, comic books and video clips.

In this post, we discuss a collision that is happening between generative AI and generative art. We’re interviewing Perpetual Motion (@perpetualsthlm), digital artist based in Stockholm, about the recent piece, “Emergent Dichotomy” (April 2023). This piece is a unique mix of generative AI and generative art, off the beaten paths of both worlds. @perpetualsthlm has prompted a generative AI to generate a program that generates art, and minted as NFT.

‍

BB & MM: Hi Perpetual Motion! Good to see you today. You’ve recently published your first generative art NFT on fxhash, could you tell us more about the fxhash platform?

‍

Perpetual Motion: fxhash was launched in 2021, it’s a free platform dedicated to generative art, hosted on the Tezos blockchain. Most of my friends have pushed something there, including @almyre and Jaime Reyes (@_jreyes79). After a few months as colector. I was hit by FOMO and in the midst of the intense ChatGPT energy, the piece crystallized.

‍

BB & MM: Your piece mixes generative AI and generative art. What does this mean?

‍

Perpetual Motion: Generative Art is the idea of generating artworks with code, it usually does not involve AI. Generative AI, with world famous ChatGPT is about generating text from prompts. It is usually not used for art. In short, generative art and generative AI are two distinct worlds with no overlap. It has been a long time that my thoughts have been wandering at the frontiers of both worlds. In “Emerging Dichotomy”, I use generative AI to generate code that generates art.  So the idea is to write a prompt, to get source code for the prompt, to execute this code, and all of this eventually results in visual art. That fits in: Prompt -> Source Code -> Visual Art. Emergent Dichotomy, fxhash, 2023‍

That’s a bit meta, because the prompt does not directly generate the visual. So, it’s not using AI to generate an image, as Midjourney and sisters would do. It’s much more sophisticated. I call this generative-generative art.

‍

‍BB & MM: Interesting. Are you the only artist exploring this idea of generative AI and generative ART?

‍

Perpetual Motion: I’ve been following the generative AI scene for some time. Nobody does this. When I did my piece, I had never heard about this and was deep in my flow. Only at the time of minting on fxhash, I realized that a few others had explored similar concepts, including Chris Wallace in “The Geometry of Rain”. As pure coincidence, a few days after minting, kaloh, our loved critic, wrote a full piece about ChatGPT and generative art, see https://www.kaloh.xyz/p/using-chatgpt-to-create-generative 

‍

BB & MM: What’s the unique features of “Emergent Dichotomy”? What makes it notable in the field of generative-generative art?

‍

Perpetual Motion: A lot of things! First, the generative art code is 100% generated by ChatGPT. In many related pieces, only one single function is generated, or the generated code is heavily reworked by the human artist. The code of Emergent Dichotomy is totally generated by the generative AI and I did not alter a single character to make it work.

Second, generative AI has also been used to generate the title of the piece, “Emergent Dichotomy” and to generate the notice of the piece, 

> "Emergent Dichotomy" is a groundbreaking piece of generative art created through the collaboration between a human artist and artificial intelligence. It explores the dynamic interplay between the starkness of black and the passion of red, representing the dual nature of human experience and emotions.

‍

BB & MM: That’s super cool! Could you tell us more about the prompts you used for both?

‍

Perpetual Motion: No way, that’s my artist secret sauce :) Yeah, OK, let me give a few bits. For generating the code, you have to explicitly  tell the language and framework you’d like to use. You can also refer to specific API method names to steer the generative AI. For generating the notice, you have to put the AI in a specific mindset, with long prompts, you can get inspiration at https://github.com/f/awesome-chatgpt-prompts. I cannot tell more because a few other pieces are in the works now.

‍

‍BB & MM: Did you iterate a lot on the prompts before getting the results you wanted?

‍

Perpetual Motion: Not so much! When I prompted for the code, my #1 constraint was that I wanted the code to work out of the box, without changing a single character, that’s the core concept of the piece. To my surprise, this was the case quite fast, because ChatGPT is very good at generating fully functional code! For the notice, I contemplated the prompt for a long time, and the notice really looked how I wanted it to look in the first trials.

‍

BB & MM: What are your next steps? What will be your next piece? 

‍

Perpetual Motion: “Emergent Dichotomy” is pure Javascript, drawing on the canvas. Next, I want to use generative AI to generate code that uses a creative coding library, such as p5 or hydra. Also, I want to use generative AI to generate generative music code. Still very meta (laugh). This requires quite advanced prompts. Another topic close to my heart is that fxhash does not make the code very visible. Some artists even use minified code in their artworks. I strongly believe that reading, looking, and admiring the code of generative art is as important as looking at the resulting visuals. Especially if the code is a piece per se, which is the essence of the concept of Emergent Dichotomy. I’m thinking about how to make this happen, in fxhash or in an alternative platform.

‍

BB & MM: Thank you very much Perpetual Motion! How can people follow your work?

‍

Perpetual Motion: @perpetualsthlm on Twitter, and follow me on FxHash https://www.fxhash.xyz/u/perpetual-motion. Thanks you for reaching out to me, I love your work at re|thread by the way.

‍

‍