Pilot: Assisted Musical Composition through Functional Scaffolding

The University Of Central Florida Board Of Trustees, Orlando FL

Investigators

Abstract

As a ubiquitous creative endeavor across all human cultures, musical composition is an effective microcosm for the study of creativity in general. This project will impact computer science by showing through assisted musical composition how computers can genuinely improve upon the creative capabilities of humans alone. By introducing an interactive framework that enables even inexperienced users to realize their creative vision, this project also helps pave the way for such systems to amplify our creative potential in other areas in the future, such as in engineering and design. The technology that will be developed, called Functional Scaffolding for Musical Composition (FSMC), takes the unique approach of computing accompaniment for existing musical tracks (called the ?scaffold?) by generating special functions that take the scaffold as input and output accompanying tracks. In this way, generated tracks are in effect transformations of the scaffold, allowing them to inherit the global structure and implicit nuance of the preexisting music. The implication for computational creativity in general is thus to harness the richness of preexisting human-generated content as a seed for further elaboration. Furthermore, the user will be provided an interactive evolutionary interface that makes it possible to search the space of such transforming functions, in effect allowing the user to continually breed and elaborate new concepts that build upon preexisting incomplete works. The primary target audience for FSMC as a practical technology will be musicians who lack the resources, collaborators, or expertise to produce complete musical compositions. For example, while a hobbyist with a keyboard might be able to compose a compelling melody, lack of expertise in other instruments may prohibit adding accompanying guitar or base. In addition, even more experienced musicians may benefit from the capability to quickly propose accompaniment as a new means of concept generation. In fact, existing computer programs that aid in musical composition often register millions of downloads online, demonstrating broad public interest in applications that enhance musical creativity. In addition to dissemination through scientific conferences focusing on computational creativity, the results of this research will be released in a form compatible with such programs, thereby directly impacting the public with a practical utility and consequently raising awareness of the potential for artificial intelligence and machine learning to enhance creativity in general.