In the audio play ‘The Life & Times of Aphrodite, Queen of Beauty’ FYTA employed the text-generating capabilities of contemporary AI to reflect on questions of beauty, art and materiality through a series of mythological genealogies, (anti-)humanist reflections and archetypal storytelling. The play follows goddess Aphrodite through the millennia and all the way to contemporary Athens, providing a philosophical account of the history of aesthetics. In this particular story, Aphrodite, the goddess of love and beauty, experiences multiple existential crises as historical definitions of love and beauty shift and collapse over time. In her attempt to stay relevant, Aphrodite has to re-configure her subjectivity and belief systems, a process of evolutionary adaptation that mirrors the function of AI’s generative networks.
The algorithm used to produce the text relies on a deep-learning model called a ‘recurrent neural network’ (RNN). This model is trained on a large corpus of text and can generate new text that is similar to the training data. In this work, we used Generative Pre-trained Transformer 3 aka GTP-3, which is trained on a vast number of online conversations and millions of digitised books. The model can learn the long-term dependencies in the text and can generate coherent and realistic outputs. As a starting point, it requires user-inputted prompts, in this case 10-15 word sentences. The prompts set the overall intention of the script, and the algorithm generated the different chapters based on them. For the audio composition AI generated noises were combined with stock compositions and folly FX, field recordings, 8-bit music, happy hardcore, a song by Dana International, and YouTube extracts.