One thing I think would be cool would be some sort of audio-generating device/software/thing that allows arbitrary levels of specificity. So, on one extreme, you could completely specify a fully deterministic stream of sound, and, on the other extreme, you could specify nothing and just say “make some sound”. Or you could go somewhere in between and specify something along the lines of “play music for X minutes, in a manner evoking emotion Y, using melody Z as the main theme of the piece”.
Now that you mention this, I do remember reading some years ago about a machine-learning composition project that had the algorithm generate random streams and learn what music people liked by crowd-sourcing feedback.
I think what you’ve described is a great idea, and I would pay for it.
Ideally, it would let me have different-styled streams dependent on what I want to do with the music / what activity I’m doing while listening. Triple bonus points if it can consume an existing piece of music to learn more about some particular style of stream that I want.
Now that you mention this, I do remember reading some years ago about a machine-learning composition project that had the algorithm generate random streams and learn what music people liked by crowd-sourcing feedback.
There have been a lot o’ such projects. I like some of the tracks produced by DarwinTunes.
One thing I think would be cool would be some sort of audio-generating device/software/thing that allows arbitrary levels of specificity. So, on one extreme, you could completely specify a fully deterministic stream of sound, and, on the other extreme, you could specify nothing and just say “make some sound”. Or you could go somewhere in between and specify something along the lines of “play music for X minutes, in a manner evoking emotion Y, using melody Z as the main theme of the piece”.
Now that you mention this, I do remember reading some years ago about a machine-learning composition project that had the algorithm generate random streams and learn what music people liked by crowd-sourcing feedback.
I think what you’ve described is a great idea, and I would pay for it.
Ideally, it would let me have different-styled streams dependent on what I want to do with the music / what activity I’m doing while listening. Triple bonus points if it can consume an existing piece of music to learn more about some particular style of stream that I want.
There have been a lot o’ such projects. I like some of the tracks produced by DarwinTunes.