Chicago Creative Machines

Molly Jones, School of Music, Theatre & Dance

Collaborators: Julie Zhu, Presidential Postdoctoral Fellow, Assistant Professor, Performing Arts Technology, SMTD

Almost every interaction we have with a screen in 2024 is mediated by machine learning algorithms. Machine learning influences what soup we make for dinner, our purchase habits, and our children’s social media presences. We spend hours each day in the company of various AIs, and many of us don’t know how they work.

The popularity of proprietary, closed-source deep learning models like DALL-E and GPT has provoked critical public dialogues around originality, intellectual property, language, and a dozen other topics, but fewer voices are clearly explaining machine learning fundamentals and promoting engagement with ML as a set of accessible tools that can be used ethically.

Artists, who develop virtuosic mastery of advanced techniques in the service of creative work, are uniquely positioned to develop novel, personal, and radically new ways to use machine learning, but first they need to know that it’s possible and how to approach it. Artists should actively guide the direction of generative AI development, not be steamrolled by it.

Chicago Creative Machines is an effort toward artist education and empowerment around machine learning. I found my own way into machine learning during the early months of the pandemic, when I began a data science bootcamp in the hope of pivoting from experimental-music-plus-dog-walking to a more lucrative job. I didn’t expect to uncover a world of fascinating, non-deterministic techniques for making audio, visual, and conceptual art because I hadn’t heard of most of these techniques nor of artists using them.

After finishing the bootcamp, I spent two weeks as an artist in residence at Ragdale in Illinois building and training adversarial audio-generating networks with Tensorflow. The outcomes weren’t particularly interesting, mostly artificial voices saying words like “yes” and “dog,” but the confidence of knowing that I had built and trained my own generative models convinced me that these techniques could become a robust part of a compositional practice. Artists dedicate years of our lives to honing specific creative skills; why not machine learning?

Two years later, as a Composer in Residence at the University of Toronto, I was premiering an electroacoustic composition for two scratch-trained neural networks and the performers on whose sounds they had been trained, and OpenAI was releasing GPT-4. My PhD research in Performing Arts Technology now centers on ethical, original approaches to building and using machine learning models for music and sound art.

Many incredible artists already code and train their own machine learning models, but they work across a spectrum of disciplines and rarely have opportunities to come together to share their knowledge. Musicians and composers (Hunter Brown, Sam Pluta, Ted Moore, Ethan Manilow, Hugo Flores García), video artists (X. Alice Li), visual artists (Mansi Shah), and text artists (Kate Compton, Julie Zhu) in the Midwest and beyond have much to impart from their personal journeys of engineering, learning how to integrate machine learning into their work, and creating compelling art with the results. They have insight as artists, interdisciplinary workers, and digital citizens.

Chicago Creative Machines harnesses the energy and innovation of four of these artists for a free, public lecture/performance series at Experimental Sound Studio, an internationally renowned art space in Chicago. From late February through March of 2024, each artist will present a technical lecture — accessible for a general audience — followed by a performance of work they have created using their own machine learning tools.

The lectures and performances will be video and audio documented by Experimental Sound Studio, archived, and disseminated free online. The four presenting artists include X. Alice Li, a video artist and data scientist at Github Copilot; Hunter Brown, a laptop improviser teaching at the University of Chicago; Hugo Flores García, a sound artist and computer science researcher at Northwestern University; and Julie Zhu, a transdisciplinary artist working with text and sound and completing a postdoctoral fellowship at the University of Michigan.

After the completion of the lecture/performance series, during Michigan’s summer break, I will partner with Cory Sibu Tripathy, a machine learning educator and composer teaching at Michigan State’s Apple Developer Academy, to create educational screencasts and collect resources related to machine learning education and host them on the Chicago Creative Machines website.

Providing ongoing, free access to these resources will require web hosting and some web design. Three Jetson Nano Developer kits produced by NVIDIA will be acquired, and these small GPU-loaded computers will be provided on a project-by-project basis for students at the School of Music, Theatre, and Dance desiring to begin a machine learning practice. This will offer music students at U-M an opportunity to experiment with machine learning without requiring the financial resources required for a Colab Pro subscription (Google’s cloud-hosted in-browser machine learning platform) or any hardware of their own.

Machine learning, data science, and adjacent tools contain awesome potential for beauty and affirming connection. I want artists to gain an understanding of these tools, build practices with them, and contribute to dialogues around the most constructive uses of AI. As a PhD student in Performing Arts Technology, my own research revolves around building models ethically, and in my first semester, I developed a syllabus for a Machine Learning for Artists course.

I see machine learning education and creativity as the core of my doctoral work, and Chicago Creative Machines is an outgrowth of that. As the PAT department completes a faculty search for a faculty member focused on AI, Chicago Creative Machines can become a connection between the PAT department’s growing work in this area and artists elsewhere.