Embeddings II

Embeddings II  continues the exploration begun in Embeddings I, a work that examines sound through the lens of information theory and its capacity to construct perceived realities—and perhaps even material ones.

As vector embeddings encode contextual nuances in textual meaning, Embeddings II investigates how perceptual meaning can be modulated through sound and other induced informational elements.

The installation features intentionally ambiguous speech, whose interpretation shifts depending on accompanying “extrinsic” sonic cues. Through this dynamic interplay, Embeddings II draws parallels between machine learning representations, neural correlates, and the intrinsic informational content of sound—framed within a materialist perspective.

image generated by the artist

This talk is part of IRCAM Forum Workshops Hors-les-Murs 2025 Rīga-Liepāja (Latvia)