Right Now in the ATA Window Gallery, AP/SE 7 (Video Edition) concludes with Mirabelle Jones in What Happens Next Will Be Documented by Images and Texts:
What Happens Next Will Be Documented by Images and Text is a series of performances composed by an artificial intelligence and performed by a human. In this twelve hour durational performance, Jones will live stream a series of impromptu performances written by GPT-3: the largest language model in existence. The performances were generated in response to the prompt “An idea for a performance art piece:”. OpenAI’s GPT-3 uses machine learning to produce texts which are difficult to distinguish from those by humans. The language model was trained on the largest dataset of any to-date including articles, books, blog entries, and other texts sourced from the internet. As a result, however, GPT-3’s text sometimes reflects the statements and ideas of the internet and its human authors including biases, prejudices, and sometimes hateful or toxic output. This has resulted in some controversy around GPT-3 adding to existing concerns about artificial intelligence’s increasing presence in our lives. This series of performances touches upon these issues as well as the relationship between technology and the body, authorship and identity, and AI creativity vs. reproduction ultimately asking: what makes performance art performance art? What works of performance art might an artificial intelligence generate? And what can those performances tell us about ourselves and the words we immortalize online?
