In my opinion, this is the best 30 seconds of music ever composed,

Or maybe this,

Seriously, what’s the difference? :)

A take: In Music AI, regardless of autonomy/co-creation, local/non-local models, real-time/offline, single-push generation/text-to-music, or controllable generation, the main issue is aesthetics, and the cost of being unopinionated is too much, and the current music AI landscape is not doing well at all along these lines. I’ve come across this view a lot in musical communities; people who have soloed on stage at paid gigs in front of a couple of hundred people mostly agree, based on my conversations. Sorry not sorry, I have to be strongly opinionated here, for the greater good :)

Regardless of where the system sits in terms of these axes, it should have serious autonomy to be respected and at the same time that autonomy should be well-utilised. The system should make decisions and they’d better be good ones, this is an AI-based setting anyway. The 808-effect analogy mostly doesn’t hold here; machines should be much more involved now, doing much more heavy lifting in terms of creativity.