The Singularity Symphony

About

Hello, my name is Kayla Gray! I am a senior double-majoring in Computer Science B.S. and Mathematics B.A. at SUNY Oswego. I am graduating in May 2023. This website displays supporting documents and audio files chronicaling my work on my honors thesis, "The Singularity Symphony."

Inspiration

Music has been an important factor in maintaining my sanity over the years. As a professional overthinker, I oftentimes find myself oscillating wildly between past and present - music has a special way of clearing my mind and tethering myself back to the present. Around 7 years ago, I decided that I wanted to learn how to play my favorite song at the time, "Outer Space/Carry On" by 5 Seconds of Summer. I started off by teaching myself how to play bass using tabs online. From there, I moved to learning acoustic guitar using the help of YouTube. In 2017, my parents gave me an electric guitar for Christmas. I fell in love with that instrument, playing it nearly everyday and using it has an outlet for channeling anxiety. Since then, I've learned to play drums (I'd say I'm subpar, at best) and have been teaching myself the basics of music production by creating covers of my favorite songs (here are some if you're interested). When I was deciding on my honors thesis topic, I had a hard time thinking of ideas that would be worthy of such a long and dedicated project. I eventually stumbled upon OpenAI Jukebox and was intrigued by the musical calamity it had produced. Given my musical background and my interest in computer science, it felt right to pursue this curiousity about algorithmic composition to create the Singularity Symphony.

Abstract

This project is an exploratory study on algorithmic composition in the context of the ongoing debate regarding artificial intelligence and the singularity. Music is a juxtaposition of creativity and emotion, two considerably human qualities; if a machine can simulate musical pieces featuring these facets of humanity, do humans have a place in the future of music? This project provides insight and use case analyses on several algorithmic composition techniques: rule-based systems, constraint-based systems, Markov chains, genetic algorithms, grammars, neural networks, and generative adversarial networks. Upon building foundational knowledge of algorithmic composition, I present my own creations in algorithmic composition under the umbrella band name of “Dystopian Tuesday.” These creations feature the use of constraint-based systems and genetic algorithms to create music ranging from simple melodies to music samples featuring multiple instruments. This project concludes with a discussion of the future of this technology. Ultimately, I predict that artificial intelligence will revolutionize the music industry by amplifying the stories of experienced and inexperienced musicians alike via musical collaboration afforded by algorithmic composition.

Dystopian Tuesday

In an attempt to poke fun at the singularity debate, I decided to house all of my algorithmic composition projects under the fictional band name "Dystopian Tuesday." Band members compose of the artificial intelligence I created as well as myself. I relish the opportunity to declare that I was apart of a band consisting of AI. Dystopian Tuesday went through three iterations: a melody-making genetic algorithm, a constraint-based system that creates music samples featuring multiple instrumental parts, and a hybridized genetic algorithm and constraint system that uses user input to create music samples featuring two melodies. These are denoted by their version numbers in chronological order.

Dystopian Tuesday v1

Dystopian Tuesday v1 is a simple melody-maker written in CLisp that uses a genetic algorithm to create melodies of different styles. It was created for CSC 416 - Artificial Intelligence I as the "wild card" genetic algorithm assignment.

Dystopian Tuesday v2

Dystopian Tuesday v2 is the second iteration of my algorithmic composition exploratory programs. It uses a constraint-based system written in Java to produce music samples featuring 0-4 instrument parts. I wanted this one to model its songs after how a traditional band - rhythm guitar, lead guitar, bass, and drums - sounds. I tried to add structure by following an intro-verse-chorus-verse-chorus-bridge-chorus structure. In each part, instruments have a chance of not playing to add variation. The first four samples do not have the aforementioned song structure, as they were created while as the program was being developed. The last two samples are songs created using the final result of Dystopian Tuesday v2. The inclusion of all of the songs below is meant to highlight the progression of development. All names are randomly generated adjectives and nouns chosen by myself.

Dystopian Tuesday v3

This iteration of Dystopian Tuesday seeks to increase user collaboration with algorithmic composition programs. Dystopian Tuesday v3 uses a genetic algorithm and a constraint-based system to generate music samples with two melodies. It uses user rankings of music samples as the fitness metric to increase user participation in the creation of the music samples. This project was created for CSC 466 - Artificial Intelligence II.

Interactive Examples of Algorithmic Composition