My research is about using the potential of free improvisation and neural networks in a musical and intermedia context.

Free improvisation is a form of expression in which participants attempt to leave behind the artistic traditions, assumptions, or preconceived biases they previously held, and to develop a language that is unique to the occasion through real-time artistic collaboration. The aim is not to find an intersection of the participants’ past and knowledge, but to communicate at a high level while searching for the path itself. Learning based on the continuous interaction of participants is therefore a key component of the process. The use of neural nets that mimic the workings of the human brain seems like a convenient way to simulate this process. In a field as shrouded in near esoteric concepts as free improvisation, inevitably a need for a more rational, scientific mathematical or algorithmic description arises. During my earlier studies, it was a revelation that even concepts taught as almost mystical - such as agogics, intonation, pulsation - can be accurately expressed in the language of mathematics.

The core product of my research is a software that is capable of generating a response datastream in real time to an incoming flow of parameters in an unpredictable way, based on the principles of free improvisation.

Through the development of the algorithm, which is the larger format process of machine learning, research itself becomes a protracted meta-learning, a type of meta free improvisation. This means that not only the end-product of the research (i.e. the ‘finished’ algorithm), but also the succession of intermediate states can be interpreted as a work of art.

A new possibility emerges for crossing between media by not adapting the source material to the material created (medial transcription), but adding new, complex layers in response to it (metamedia).

In the spectrum of human-to-machine interaction, the majority of research is development created for a specific, rational, and predefined purpose, but here the irrational character of the task means that the new questions that arise can be much greater than those that are answered.