Atmospheres of Inference

There is a persisting affection for analog audio equipment developed over the second half of the 20th century. Unlike instruments based on digital technologies, early analog instruments exhibit unpredictable behaviours contingent on their environmental conditions and the components which furnish their circuit boards. These components and their acute material properties, drastically affect an instruments' performance and functionality. In a world calibrated to the binary logic of digital technologies (discrete processes of quantification, capture, validation, and storage), the analog (continuous, unstable, and entropic) demarcates an exceptional space laden with contingency. This project instrumentalizes recursive computing strategies, specifically deep learning (DL), as a means to suggest that the phenomenological distinctions between the analog and the digital are becoming increasingly blurred. Through a musical rhetoric, Atmospheres of Inference attempts to dramatise these two technical modalities by performing a 1976 Serge Modular alongside an inferred abstraction of the same instrument: a variational autoencoder trained on a corpus of Serge recordings.