Utilizing OpenAI’s GPT-2 Text Generator to Foster the Creation of Fluxus-Inspired Event Scores

Cross-Disciplinary Strategies
Kean Ehsasi (Bachelor Thesis)
Betreuer*in: Adnan Balcinovic, Martin Gasser & Christine Böhler

Novel Fluxus-Inspired event scores are created with the help of OpenAI’s Generative Pre-trained Transformer 2 (GPT-2) language model. After a brief review of the recent developments in language modeling and a discussion of Fluxus, pre-existing Fluxus event scores are inputted into the language model in order to generate disparate output. The curated output is printed onto cards, resulting in new Fluxus-inspired event scores, which are intended to stimulate creative impulses, as well as induce new awareness of a performer’s setting. The collection of cards entitled Mud Bath 44 is created by a human and Artificial Intelligence hybrid writing team and offers scripts performable across disciplines, by anyone anywhere. By utilizing GPT-2, the process of creating Fluxus-inspired event scores becomes accessible and easy, providing plentiful resources for stimuli.