Music - organised, comprehensible sound that can be captured, recorded, measured and transmitted - is data. Most forms of AI use structured manipulation of data to imitate human activities. This project turns to jazz in order to illuminate some problems in AI and data science. Jazz is made on the basis of "operations" (interactive improvisation) performed on "data" (song structures, rhythmic conventions etc.) We argue that jazz as a global practice is a "social machine" (e.g. Wikipedia). It aggregates the "energy" of networked participants - their - and converts this via interaction into meaning.
This project asks: How is it possible to discuss, let alone quantitatively measure, the 'energy' in jazz's interactions? Would AI-created jazz suffer from problems of 'explainability' (inability to account for the 'chains' that validate it) as other autonomous agents do? How can a jazz algorithm account for the mixed media (standards, lead sheets, performances, recordings) that make up its data set? How can programmed sequences be refined to account for 'the contributions of a crowd'? It is these contributions, sequences and the cocommitent sense of autonomy that account for jazz's political identity. By exploring them we can understand jazz as a social machine.
Principal Investigator: Dr Thomas Irvine
Co-Investigator: Dr Valentina Cardo