For example, do you think someone like DJ shadow who has spent tens of thousands of hours digging in empty, old, dusty basements finding and curating music over a lifetime has the same insight as a dicovery algorithm?
It's not even close to a comparison, the algorithms are lame compared to actual experience and expertise.
I spent 10+ years in the music business as a "buyer" and was later part of a successful music start-up, I can tell you with a resounding "fuck that" to auto discovery unless you like water downed with a side of boring.
Radio DJs, on the other hand, have any number of forces affecting their song choices (to the degree the role exists at all anymore), and all in all they're just as much a black box as an algorithm. Sure, we all had radio shows we liked for some time, but those DJs changed and we found something else (notwithstanding people who have been listening to "Renee and Crazy Pete in the morning" for 20 years).
Even with all of this, I suspect there's a discrete economic or mathematical reason that profit-oriented discovery engines regress to a mean and will always be trying to recommend the latest AdeleBieberNational. They may not, right away, but a few clicks deep in the sidebar and you find LL Cool J or Foo Fighters popping up.
But if you want to find hits in a genre + time period that you don't know very well, recommendation engines are pretty good, and you don't need to have DJs for that.
