“Optimize This! Why Do We Care if an AI Can Write Songs?”

Andrew Goldman (Indiana University) Volume 12.2 (March 2026)

 

Link to bibliography

AI systems like Suno take in text prompts, and output audio of original songs that are compellingly human-sounding. Here, I describe three ways to compare human and AI generated music. To frame my discussion, I juxtapose my original song (“Optimize!”) with a song that Suno generated on the same topic. We can compare products, focusing on the musical work (as audio, or symbolic notation), and consider whether AI-generated music passes the Turing test, or whether there are features that sound artificial. I analyze my own song, and Suno’s. We can also compare processes. How did I produce a work of music compared to how Suno did it? Again, I compare my process with Suno’s. Both product- and process-based comparisons aim to explain features of musical scores and audio, but music is more than what is encoded in such representations. Thus, a third kind of comparison resists this work-based ontology of music: a comparison of musical practice. My song is more than the notes on the page; there was a social motivation to write about optimization. In contrast, AI music’s sociality is homuncular: it only has social purpose because the humans who use the technology do.

Keywords: Artificial intelligence; songwriting; music as practice; music cognition; creativity