The following prompt (in bold) was input into OpenAI’s GPT-3 Beta Playground using the DaVinci model. The text in italics was the output for the prompt completed by GPT-3.
Instructions for a noise music performance for one person: (1) Invite a clown to perform with you. (2) Put on headphones. (3) Turn on loud noise. (4) Perform. (5) When the clown is finished, put on a record. (6) Turn off the noise
Production Date: January 23, 2021. Performed by: Emmett Palaima
When this text appeared out of the generation process, I was struck by the idea to perform it alongside an inflatable clown almost immediately, as though this idea were a part of the generated text itself. This method of interpretation seemed relevant to many of the themes and questions raised by the AI performance project, as well as the wider realities of digital performance in the year 2020. I thought a blow-up bozo doll was in many ways the perfect representation of digital elements replacing human interaction: a plasticky simulacrum, by equal measures humorous and unnerving, somewhat pathetic, yet also strangely indomitable by means of its absorptive passivity. I was furthermore attracted to the narrative structure contained in the instructions’ clear movements: the act of invitation, the performance of noise, the end of the clown’s action, the denouement of a record playing. The instructions were non-specific enough to allow for personal interpretation, but also provided a clear arc, a combination which I consider to be the hallmark of a successful alternative notation score.Given this form, the process of interpretation consisted of answering several questions: How is the clown invited? What is the noise? What, when, and how does the clown finish? What record is to be played, and in what way does it contribute to the overall performance? The clown would be made the offer of a synthesizer (Hammerhead Audio – Nuisance Engine NE-1) that it had no means to accept. In my frustration at this inevitability, I would initiate the slow deflation of my unwilling collaborator, which would in turn provide the central timing element of the piece. I would perform noise using a variety of modular synthesis equipment, including the refused synthesizer, until the clown was finished, at which point I would celebrate its demise to the tune of Antonio Carlos Jobim’s ‘So Danço Samba’ (recording by Stan Getz and João Gilberto), dancing (only samba!) atop its crumpled remains. This piece was recorded live, using the first complete take as the finished product. -Emmett Palaima