It'd be awesome if we could create a station based on a description of the type of music we want it to include (examples: "80's and 90's rock and roll"; "hard-driving Viking or similar chant with heavy percussion"; "easy listening jazz with no vocals", etc.), rather than knowing a few examples already, creating a station from those, and then spending time curating it with thumbs to try to teach the system what the station is supposed to sound like.
Obviously, some descriptions might be too specific (or general), but modern LLMs should be able to handle most requests, I'd imagine.
