Salvaging Birds (2021-ongoing)
Watch:
About:
Over the last 50 years, at least 3 million birds have been lost in the U.S. and Canada alone, likely the result of human-driven causes. Increasingly, bird conservation efforts have turned to data and technology to better understand how such unprecedented losses could be mitigated. salvaging birds, a three movement, fifteen minute work for alto flute and AI-generated electronics, is about the logics of environmental data, and about how this data mirrors many of the same problems that we identify in human data, such as bias, and the illusions of control over systemic problems.
Specifically, this piece looks at birdsong recordings as a dataset. Research has demonstrated that birdsong datasets, like many other datasets, have a male bias, and they contain very little female birdsong. When I learned this, I began to wonder about all of the canonical classical music that was inspired by birds, and whether it also was based primarily on male birdsong and in my research I’ve found that it seems to be. For example, many of these pieces reference birds like the cuckoo and the nightingale, who are known for their male song.
For salvaging birds, we collected birdsong recording datasets from Cornell’s Macaulay Library and train AI-sound generation models (using the SampleRNN algorithm) on the birds who are less represented in bird data, as well as much of the classical music that references birdsong. Although they have been edited and rearranged, all of the electronics are generated by the AI, with no additional effects or processing.
Concept and direction by Maya Livio
Music by JP Merz
Machine learning programming with Raymond Finzel
Visuals by Cassie McQuater
Flute by Alex Sopp and Yoshi Weinberg
Commissioned with support from New Music USA Creator's Development Fund
Over the last 50 years, at least 3 million birds have been lost in the U.S. and Canada alone, likely the result of human-driven causes. Increasingly, bird conservation efforts have turned to data and technology to better understand how such unprecedented losses could be mitigated. salvaging birds, a three movement, fifteen minute work for alto flute and AI-generated electronics, is about the logics of environmental data, and about how this data mirrors many of the same problems that we identify in human data, such as bias, and the illusions of control over systemic problems.
Specifically, this piece looks at birdsong recordings as a dataset. Research has demonstrated that birdsong datasets, like many other datasets, have a male bias, and they contain very little female birdsong. When I learned this, I began to wonder about all of the canonical classical music that was inspired by birds, and whether it also was based primarily on male birdsong and in my research I’ve found that it seems to be. For example, many of these pieces reference birds like the cuckoo and the nightingale, who are known for their male song.
For salvaging birds, we collected birdsong recording datasets from Cornell’s Macaulay Library and train AI-sound generation models (using the SampleRNN algorithm) on the birds who are less represented in bird data, as well as much of the classical music that references birdsong. Although they have been edited and rearranged, all of the electronics are generated by the AI, with no additional effects or processing.
Concept and direction by Maya Livio
Music by JP Merz
Machine learning programming with Raymond Finzel
Visuals by Cassie McQuater
Flute by Alex Sopp and Yoshi Weinberg
Commissioned with support from New Music USA Creator's Development Fund