Support Overthinking It by becoming a member for $5/month!
Peter Fenzel, Mark Lee, and Matthew Wrather take another look at music generated by AI diffusion models, which have come a long way. They discuss the possible significance of this new technology and how it fits into our conception of creativity, and they play each other some tunes they’ve prompted.
Subscribe: iTunes Other Apps
Further Reading
- Dirt Road Brew
- Puppy Patrol
- Temple of Taste
- Podcast Theme
- Arcadia by Tom Stoppard (Wikiquote)
And here’s the last time we tackled AI music:
I really liked the section of the podcast that starts around 20 minutes in, when the Overthinkers consider the possibility that the vague dissatisfaction we feel regarding AI-generated music is a half-conscious sense that musical sounds ought to be produced by physical objects operated by human hands and/or mouths.
There’s a memorable passage somewhere in a classic sf book set in a distant future when all music is electronic (not necessarily composed by computers, but at any rate generated by telling computers what to do rather than by whacking bits of wood and leather), and some of the characters are amazed at the emotional depth of music made by on “primitive” instruments when they hear it for the first time. (Annoyingly, the scene was memorable but apparently not what book it was from. If anyone knows, please leave a comment because this is going to bug me. I thought it might have been from Ringworld but that doesn’t seem to be right.)
Anyway, so Fenzel is right that it’s been possible to make music without learning an instrument for a long time already now. Take K-pop, where stage performance is a huge part of the genre but (in most cases) you don’t see any instruments or musicians (other than vocalists) on the stage. Maybe there really has been a shift in what the popular culture wants when it wants what it calls “music,” in a way, a change in what music “is.”