Char Stiles is a regular at Algorave events across the East Coast of the United States. Based in Brooklyn, NY, the programmer and artist keeps busy with livecoding events, teaching workshops, and collaborating with Snapchat and artists of all kinds.
Interested in learning the math behind the code as well as exploring the philosophy of machine learning, Stiles is making their own path. I reached out to the creator after Lens Fest 2020 to learn more.
Q: How long have you been mixing traditional art and digital art?
A: I started mixing art and code in 2014. The first piece of code art I made is still up and running, astoundingly. It was before I learned about loops. I started actually using code in my fine art the same year.
Q: How do you describe yourself to people who don’t know the first thing about coding and XR?
A: I make art with the computer! No, not a graphics designer, no, not a 3D animator, but that’s closer! I work specifically with real-time graphics, so computer graphics that respond to your mouse input or any other external device that we can plug into the computer, like a webcam! Yes, like video games!
Q: What is your favorite medium to work in?
A: I like working with GLSL shaders. I want to make my own computer programming language for visuals — that is a project for far-in-the-future me.
Q: What inspired you to start the livecoding music events and how long have you been doing them?
A: I went to school for fine art and computer science, where I deemed computer graphics to be the intersection of the two. A piece that cemented this path for me was making a music video for Pittsburgh-based drag queen Moon Baby, “Everything is Cute to Me.” Here I worked on the volumetric data rendering and motion capture rigging.
I loved working with musicians and people outside of my domain. I was looking to do more work in this area, but I wanted to spend less time breaking my back in front of a computer screen. Then, I happened upon some Algorave in searching online. I thought it was cool so I decided to call myself a livecoder in 2017. I performed one basement show (drenched in sweat, praying my program wasn’t going to crash even though only six people showed up), then the local livecode scene picked me up and I performed in Pittsburgh’s first Algorave!
Q: In what ways does your art currently use machine learning?
To be honest, I am not sure what my relationship is with ML at the moment. I have been thinking about the power disparity that ML brings into the world, between the people training the algorithm and the people as the training data, or the people who are categorized and analyzed by machine learning. I have been wondering what it would be like to slow down the process of ML to the point where a human can fathom grasping what is happening.
I have been working on this sorta tongue-in-cheek piece called “GPU? No. Turtle.” where it’s this turtle that carries out the operations needed to create a neural network. It uses a visual representation of the inner state of the computer through cellular automata. While computing with a cellular automaton, it slows down processes immensely, so something like deep learning stretches out across years when you compute, say, 10 operations per second. This is a little bit of a response to the move to make machine learning faster and more efficient, “GPU? No. Turtle.” is just an absurd way to use machine learning! It’s still in the beginning stages, all it does right now is Wolfram CA.
Q: Do you build ML models yourself?
A: I enjoy learning about the math behind ML and I think that’s the best part of it all. I use easier to grasp concepts like Hebbian learning and use PCA to create small projects to just learn about the mechanisms. This helps me navigate using larger networks. The piece that I made for Snapchat, I used a famous model by Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A. Efros called CycleGAN. I trained a network using their model and added on my own end nodes to make it fit with Snapchat’s Lens Studio.
I would love to live a split life where I am better at math and muse all day on creative ways to use gradient descent to make my models more creative. We are standing on the shoulders of giants in machine learning. I think that’s why I am making “GPU? No.Turtle.” because I just want to understand the math more.
Q: Where do you see your career going — how important is it for you to be a digital creator?
A: I have been enjoying teaching creative code lately. I applied to grad school in hopes to be able to teach a class at an institution instead of being hired to teach one-off workshops. Wherever I end up, I will always be creating and iterating, that’s how you get all the bad ideas out of your system!
Q: If budget, time, and technology were all in place, what would be your dream project?
A: I think I would like to build a collective to share this perfect alignment with and then figure it out from there. Maybe a collective that is vaguely centered around expression of thought and meditation through digital art.
Learn more about Stiles at their website.