Genrematic

An experiment with generative AI and popular music

Genrematic is a hobby project I completed towards the end of 2023. It was an experiment with Chat GPT, and a means of investigating some of the possibilities, problems and questions related to artificial intelligence.

I’ve been meaning to write about this process for some time but have so far not had the time. This post is a quick introduction to what Genrematic is, and why I undertook it. I will follow it up with a more detailed series over the next month or two.

Like many people, I was fascinated by the emergence of Chat GPT in November 2022. As many others did, I played around with AI tools and explored some of the possibilities using the various free web interfaces provided by vendors. This got me thinking about doing something a little more systematic.

Combining my interests in popular music, programming in R, broader questions around the nature of knowledge creation, and an ongoing interest in the social and cultural impact of emerging technologies, I set about creating something using Chat GPT, and Genrematic was what emerged.

It is an end-to-end workflow created in R that uses the Chat GPT API to produce ~3,000 word essays about musical sub-genres. From Acid House to Zydeco, via Dub, Outlaw Country, Memphis Soul, Emo, and more.

The workflow produced the following things:

  • Essays for 534 musical sub-genres, providing a short history, some cultural context, an exploration of major figures, and a list of records.
  • A bookdown output that is now for sale on Amazon as both an ebook and print-on-demand paperback
  • A blogdown website that presented a random sample of the ~500 essays.
  • A randomly generated image for each of the blog posts on the website.

My motivation for completing this project was first of all to see if I could produce an end-to-end process that produced the above. After a few weeks of experimenting, I was able to do that. Interestingly, the process was considerably sped up by the use of Chat GPT as a coding helper.

Secondly, I wanted to explore some of the possibilities and problems associated with using generative AI. Popular music is something I am highly knowledgeable about, so it was the ideal subject. I knew that I would be able to ‘fact-check’ the results.

Thirdly, I wanted to think about some of the broader implications of using generative AI in this way. In running this project, I essentially developed a replicable workflow that is able to produce a commercial product. It could easily be adapted to produce other works - a book of Italian food recipes, or a book of places to visit in South America, and so on. There are moral and ethical questions around doing this that are at the heart of many concerns around the implications of AI.

My aim with this coming series of posts is to discuss the process in more detail, and to reflect on the motivations above. Ultimately I want to think through what I have produced, and discuss the broader implications of ‘creating content’ in this way.

More to follow in due course. In the meantime, you can explore the various outputs of the project through the following links:

Dr Craig Hamilton
Dr Craig Hamilton

My research interests include popular music, digital humanities and online cultures.