New methods for music composition with AI: music synthesis and analysis with Periodic Spacetime Sequences

The research project investigates the role of Artificial Intelligence in the artistic practice of a human composer in which AI is used to compose new music. The outcomes of this research project are not only new music compositions, but also a set of new technologically-supported methods for music composition. For the project, a new software is developed: AI music composer’s assistant supporting human composer with the power of AI.

The research is especially focused on creating a general-purpose musical Environment which was initially inspired by notions coming from:

• Assemblage Theory and its notion of topological diagram (Gilles Deleuze, Felix Guattari, Manuel Delanda, and Paulo de Assis),
• the periodicity of the Shepard tone illusion,
• the concept of a strange loop by cognitive scientist Douglas Hofstadter.

The project proposes a post-human definition of a musical work as resulting from the feedback loop created by Observers (humans or AI agents) interacting with Environments. Environment is an n-dimensional space with a set of rules for changing it, for instance: a music score with rules for modifying pitch, duration and dynamics, or a virtual 3D space with physical rules for interacting with objects in virtual realities, such as the metaverse, and computer games). In my definition, inspired by the Reinforcement Learning branch of machine learning, interactions of Observers with Environments are named Performative Transactions. These interactions are recorded in blocks, and form blockchains (that become “compositions”). Blockchains are shareable between Observers enabling exploration, collaboration, and modification of the created sequences of events. During the course of the project different Environments are built. Observers (human researcher and developed AI agents) interact with them to create new music compositions.