Abstract
This talk explores the growing role of machine learning in contemporary musical creation, at the crossroads of artistic intuition and mathematical formalization. Drawing on my work on computational creativity, assisted orchestration and human-machine co-composition, I will examine how algorithms can go beyond imitation (mimesis) to become the vectors of a sensitive and cathartic transformation (katharsis).
The aim is to show how mathematical models - far from being limited to an optimization function - can structure aesthetic processes, reveal emerging forms and open up new spaces for listening and invention. In particular, I'll be looking at systems where neural architectures dialogue with deep musical logics, where symbolic and perceptual representation are articulated in the same creative gesture.
This epistemological shift, which delegates certain dimensions of judgment or style to computational agents, questions our usual frameworks of author, work and listening. Between science and artistic practice, I will defend the idea of a co-evolution between humans and machines, based on shared structures, languages in the making and the unfinished power of musical creation.