Mycélium Garden is an ongoing research-creation project, led by Diane Schuh, where I act as machine learning designer.
The soil is an unthought entity: perceived as a surface, its inner life is ignored. Yet the mycelium, an underground vegetative organism, long thought to be inert, forms complex networks of inter-species interactions that are essential to life. Researchers have very recently discovered the complexity of the electrical signal transmission systems in these networks. The extreme diversity of these communications reveal the vital and dynamic entanglement of insects, plants and humans, sustained by the non-human, non-plant mycelium network.
We were interested in crafting a machine listening prototype to listen beyond sound to the mycelium’s electrical behaviour. We conducted a wavelet analysis, which is more sensitive to low-frequencies typical of mycelium’s electrical behaviour. We then used a shallow Gaussian Mixture Model to classify this frequency distribution in real-time. We decided to include machine learning uncertainty as a form of computational agency adding to other-than-human microperformativities of myceliums and their instrumental apparatus.
The prototype is implemented in Max/MSP; it is leveraged in the Mycélium Garden interspecific sound installation, which invites audiences to listen to the mycelium’s electrical behaviour.