Investigating Style Forming and Twisting by Adapting AntsOMG for Composing Second Species Counterpoint

Human creators are often nourished from explicit knowledge while developing their own styles with creation, and such knowledge often comprises a multi-layered or certain much more complex structure, in which learners are guided to establish their creative vocabulary. Taking music creation as an example, countless composers have built their own composition skills by learning Species Counterpoint, while it is often ignored in current machine-learning-based creative AI works, resulting in some music generators that may be fraught with problems in their outputs when further inspected. In this study, we adapt an existing creative framework, AntsOMG, for composing Second Species Counterpoint in order to conduct an investigation on music style forming and twisting, thereby gaining insights into the crucial mechanisms and essential components of a creation process that leverages the white-box property, potentially enabling machines to incrementally learn complex, multilayered structures from a computational perspective. Step by step, we demonstrate how the target task is accomplished by expanding the base framework and exploiting its white-box property. Our implementation is capable of producing the outcomes that exhibit the desired effect, indicating a promising line of research on investigating into the use of white-box AI methodologies in the field of computational creativity, along with the forming of machines' own styles, perhaps even aesthetics in the future, based on autonomous development of complex knowledge.