These existential themes comprised the AI-scripted and human-performed musical production Channelers, an interdisciplinary art project funded by the Dietrich Inchworm Grant and headed by Assistant Professor of Digital Arts Anna Huff.
The project began years ago when Huff’s friend Claire Evans, a Grammy-nominated musician and tech writer, visited her class to explain the process of creating an AI musical album. Evans mentioned the charming blurbs her AI would spit out, reminiscent of a 3-year-old before rules of language took form.
Huff’s mind caught on this piece of information, giving rise to questions about how AI and children learn to process language similarly, how we decide what language counts, and how we use it to tell stories. Using these questions as inspiration, Huff and former Assistant Professor of Instruction in Theatre Jeff Larson began writing down quotes from their young children. These quotes became the first pieces of information used to train Samantha, the child AI that created the child character’s script for the musical.
“The project has evolved into a conversation about the relationship with the AI almost as an oracle and how we are essentially replacing elders with algorithms to ask these important questions,” Huff said.
In addition to Samantha, other AIs were created, including a feminist AI trained on feminist theory and an oracle AI trained on the prophecies of famous oracles. These AIs can then produce dialogues, scripts, and poems based on the essence of their training input information. When put into plagiarism checks, the AI-produced works will pass by roughly 98%, if not more. The production team developed scripts using the works generated by the AIs, allowing the actors to perform a thematic production.
Kate Bondarenko ’24 figured out how to make the AI characters using a sophisticated network that involves the programming language Python and GPT-3 on Open AI. This summer, she and Maheen Masoud ’23 worked to fine-tune the AI characters, including Moody Anthony, an AI trained on the creative writing pieces of Anthony Christiana ’22. Christiana also created AI-generated music and sound design for the production. Using a custom AI created with the software program MAX MSP, he’s reprocessed original music written for the piece by Huff.
“The AI essentially rewrites the music I put in,” Christiana said. “I input some dancey disco songs and get a weird, mirror image of disco music in return.”
Digital 51ÁÔÆæ
Digital 51ÁÔÆæ encompasses learning opportunities using advanced technologies to explore and create research projects across the College’s disciplines.
The summer work culminated in an invited draft viewing of the work in late June. An audience saw the production, which included both student and professional actors, as it was at that stage of development. Audience members had an opportunity to ask the AIs questions through the actors. Ear pieces connected the actors with the AIs, which allowed the actors to hear the AIs spontaneous answers and relay them to the audience.
“I think this production is meant to be about our own kind of human relationship to these machines, and how we function in a world where so much of our relationship to knowledge, our sense of self, and sometimes even spirituality is generated through these AIs and algorithmic interfaces,” Huff said.
