In the summer of 2025, Michael began a research project intended to investigate the potential of interacting with the Large Language Model (LLM), ChatGPT, toward a novel score synthesis methodology. He has been composing from this perspective since the year 2000 using an application called Cmask, which is a stochastic score generator for Csound scores, and sonifications. Cmask is based upon probabilistic weightings employing tendency masks. This generative compositional approach is considered a pre-cursor to the Artificial Intelligence (AI) systems currently in use that are rapidly increasing in capabilities. Providing such a functionality is easily achieved by the LLM. However, the powerful advantage of engaging with it as an expansion of these processes lies within their extensive learning capabilities.
The premise is that in training the LLM, as one might a very fast learning graduate student, a collaborative partner can be created. Further, as this partner learns and catalogs the results of the subsequent compositional interactions, it will become capable of not only synthesizing Csound scores that fall within the composer’s sensibilities but ideally surpass them into unprecedented musical explorations.
The links below are records of Michael’s interactions with ChatGPT and their results. They provide in-depth delineations of this superficial introduction to the project. The writing in them is also a collaborative effort between Michael and ChatGPT.
This project is an unexpected, and deeply interesting, interaction that reveals many of the unique characteristics of the LLM and ultimately of Michael’s compositional and programmatic expertise. It is a quintessential example of the results being greater than the sum of the inputs.
This project is currently a work in progress.