Listening Through: Machine Learning for Sound-Makers – A workshop led by Luke Fischbeck
Published
This will be a short, intensive workshop exploring the current state of machine learning techniques for working creatively with audio. As a set of approaches for predictive pattern-matching—classifying, categorizing, translating, generating, etc—machine learning has been applied to everything from captions to gestures to maps to MIDI. Audio, with its intensely complex set of perceptual and contextual resonances, provides a tricky and illuminating challenge for machine learning.
Participating artists are Niklas Adam, Magnus Bugge, Toril Johannessen, Zoe Efstathiou, Mattias Arvastsson, Roel Heremans, Elise Macmillan and Øyvind Torvund. From BEK, Espen Sommer Eide and Sindre Sørensen will participate.
The focus of this workshop will be on identifying the poetic possibilities inherent to this set of approaches, with the goal of initiating our own creative projects. Additionally, attention will be paid to philosophical and political questions embedded in the idea of ‘creative machine intelligence.’ What does it mean to work collaboratively with computers? Moreover, considering the resources required to develop this technology, what is lost when a small set of corporate research groups becomes responsible for the data collected, the tools developed, and the techniques shared? What strategies of resistance might be found in exploiting these techniques, data, and concepts for artistic purposes? What alternative tools are at our disposal for working (playing!) creatively using machine learning?
We will be following a hands-on approach, using publicly accessible, state-of-the-art techniques to consider what new critical, conceptual, and aesthetic strategies are made possible through machine learning. The participants have prior experience working with digital audio processing, machine learning and/or experience writing code (python or javascript).
Luke Fischbeck
Luke Fischbeck is an artist, composer, and organizer based in Los Angeles; a PhD candidate in the Interdisciplinary Media Art and Practice program at the University of Southern California, researching music, affect, and machine intelligence; and a contributing member of the collaborative group ‘lucky dragons’.
Image: ‘Rainbowgrams’ of a series of notes reconstructed by the WaveNet autoencoder. “Neural Audio Synthesis of Musical Notes with WaveNet Autoencoders” Engel et al, 2017.
The workshop is produced in collaboration with Notam, as part of a workshop series that we jointly arrange for experienced artists. The workshops will be held in English, and will be arranged alternately in Oslo and Bergen between 2017 and 2020. There is a limited amount of places and highly qualified participants will therefore be prioritized. The series is supported by Arts Council Norway.