Confabulation
2024

4.0 ch Composition

12'00"

Dimension Varies

In psychology, confabulation typically refers to a phenomenon where a person involuntarily fabricates memories and experiences that never took place. In machine learning, confabulation occurs when a machine learning program, trained with a specific data set, generates outputs that are incorrect or nonsensical. The compositional system of this piece consists of a reciprocal process between machine learning programs and human performers. The programs generate non-existing words in different languages, which are later interpreted using the phonetics of a language other than the one the word was initially constructed in. The human performers bring these non-existing words and phonetics into existence through their vocal interpretations. The result of this process is played through loudspeakers in the exhibition space. In this piece, the gaps of interpretation are not only filled by the performers and the machine learning programs, but also by the visitors of the exhibition space, as they attempt to make sense of the words or sounds. Confabulation highlights the personal nature of the listening experience, that intertwines what we hear and what we remember, leading to unique, subjective interpretations within the listeners’ familiar languages.

Voices: Amelie Baier, Takuji Kawai, Quentin Tolimieri, Marion Wörle

Commissioned for the festival, Extended Spaces - Resonant Bodies: Alvin Lucier, by singuhr - projekte