Colloquium Archive

All songs can be distilled to a progression of chords and lyrics. Songwriters almost always start their creative process melding these two elements, harmonic changes and words. On top of this foundation, they layer melody, voice, and emotion.

In this presentation, I will detail a proposed project in which students will use machine learning and artificial intelligence to develop a web-facing platform that allows people to collaborate with the best musicians and lyricists of the past to generate compelling new creative works.  As a proof of concept, I will share a demo song that I created using a low-tech prototype of this software, built and programmed entirely using functions in Google Sheets.  I will show my process of how I created a small database (Bob Dylan, Kate Tempest, Jay-Z, and the Beatles), “farmed” results generated by machine learning, and crafted them into a coherent new song.

I hope that this prototype and example will inspire students to imagine the possibilities of how artificial intelligence might be used to enhance our most human of qualities, like creativity, and how working together to build a more sophisticated, easier to use platform with a larger database could meaningfully contribute to the expression and understanding of our collective experience.

Soundscapes to Landscapes (S2L) is a NASA-funded project led by Dr. Matthew Clark of the Center for Interdisciplinary Geospatial Analysis (CIGA), in Geography, Environment and Planning. The broad goal of the project is to advance animal diversity monitoring from the next generation of Earth-observing satellites. The project uses sounds recorded from low-cost recorders placed in the field (i.e., soundscapes) and bioacoustics analysis to identify bird species by their calls and measure overall avian diversity. Bird diversity data are used by the science team to explore the benefits and trade-offs in using new and existing sensors in space for mapping of bird diversity and conservation planning. One new sensor the team is using is NASA’s Global Ecosystem Dynamics Investigation (GEDI), docked on the International Space Station, which uses a laser to provide detailed measurements of vegetation canopy height and internal structure that can be related to bird habitat. A critical component of S2L is a partnership with “citizen scientists”, or volunteers in the community, that help with placement of recorders on public and private lands, assist with data management, and identify bird calls in sound recordings. The project facilitates citizen science work with a web-based bioacoustics analysis platform (Arbimon) that permits rapid bird-call validation. These validation data are then used to train and test convolutional neural network classifiers for automated bird-call detections in recordings.