Teach yourself to sing (2023) // for vocal ensemble and interactive music system
In Teach yourself to sing the computer music system uses machine learning and data extraction processes in order to simulate and "replace" six vocal performers. During the first half of the performance the computer collects sound material and data pertaining to how the performers navigate 4 types of vocal utterances. Later during the performance, these data are used to train a Recurrent Neural Network (RNN) that builds a model of the performers' trajectory through the vocal utterances. Based on predictions made by this RNN and using the recorded sound material the computer eventually takes over the performance, rendering the performers redundant. The piece is a comment on data extractivism, data ownership and the relationship between data and labor, inviting the listener to reflect on the labor implications of Generative AI and its definition of "data".