Echo Chamber

Monica Lim, Jarrod Knibbe, Bingqing Chen, Ying Sima, and Melanie Huang

Proceedings of the International Conference on New Interfaces for Musical Expression

Abstract

Echo Chamber is a participatory sound installation using MusicGen’s audio generative AI. Participants play a short melody on a piano, which is then used as a reference for MusicGen to generate multiple versions of itself. These evolving echoes of the original are played back through a multi-channel speaker installation. The work alludes to concerns about generative AI creating a monoculture through a feedback cycle of data scraping, copying and regenerating. By deliberately engaging with (and simultaneously serving as a critique of) MusicGen’s innate tendency to stay within predictable melodic and harmonic structures, we can layer and loop multiple versions of the AI-generated audio while staying musically coherent. The layering of multiple samples also allows us to produce a real-time participatory work despite the time required for the AI generations (which currently take longer to generate than the length of the audio samples generated). A piano was chosen as the interface through which an audio reference is created by participants, as it plays on the familiarity and ubiquity of the instrument, which has become both a symbol and a tool of colonisation. Not only has the keyboard become a dominant interface for music-making, the equal-tempered scale has become pervasive across the globe in most popular music cultures, relegating other modes and scales to the fringe – a pattern repeated by generative AI amplifying what is already amplified.

Citation

Monica Lim, Jarrod Knibbe, Bingqing Chen, Ying Sima, and Melanie Huang. 2024. Echo Chamber. Proceedings of the International Conference on New Interfaces for Musical Expression. DOI: 10.5281/zenodo.15027150

BibTeX Entry

@article{nime2024_installations_1,
 abstract = {Echo Chamber is a participatory sound installation using MusicGen’s audio generative AI. Participants play a short melody on a piano, which is then used as a reference for MusicGen to generate multiple versions of itself. These evolving echoes of the original are played back through a multi-channel speaker installation. The work alludes to concerns about generative AI creating a monoculture through a feedback cycle of data scraping, copying and regenerating. By deliberately engaging with (and simultaneously serving as a critique of) MusicGen’s innate tendency to stay within predictable melodic and harmonic structures, we can layer and loop multiple versions of the AI-generated audio while staying musically coherent. The layering of multiple samples also allows us to produce a real-time participatory work despite the time required for the AI generations (which currently take longer to generate than the length of the audio samples generated). A piano was chosen as the interface through which an audio reference is created by participants, as it plays on the familiarity and ubiquity of the instrument, which has become both a symbol and a tool of colonisation. Not only has the keyboard become a dominant interface for music-making, the equal-tempered scale has become pervasive across the globe in most popular music cultures, relegating other modes and scales to the fringe – a pattern repeated by generative AI amplifying what is already amplified.},
 address = {Utrecht, Netherlands},
 articleno = {1},
 author = {Monica Lim and Jarrod Knibbe and Bingqing Chen and Ying Sima and Melanie Huang},
 booktitle = {Proceedings of the International Conference on New Interfaces for Musical Expression},
 doi = {10.5281/zenodo.15027150},
 editor = {Laurel Smith Pardue and Palle Dahlstedt},
 issn = {2220-4806},
 month = {September},
 numpages = {3},
 pages = {1--3},
 presentation-video = {https://vimeo.com/938949011},
 title = {Echo Chamber},
 track = {Installations},
 url = {http://nime.org/proceedings/2024/nime2024_installations_1.pdf},
 year = {2024}
}