¶ Unfancy Machines · 10 June 2015 listen/tech
I spend a lot of time fiddling obsessively with the knobs on complicated music-discovery machines. I wouldn't do it if I didn't think it helped, and I spend a fair amount of that time pondering whether I'm confident that it's helping.
Your streaming music world is also about to be extra-complicated by more machinery for personalization, and I work on some of this, too. One clattering armada of machinery for trying to find music worth being discovered by somebody, and another one for trying to figure out which of that music is the music for which you are that somebody. It's kind of a wonder we can still hear over all that grinding.
Although, actually, it's less of a rhetorical wonder than an actual question. Can we hear over it? It's not inherently bad if your complicated thing is complicated, but it's bad if it's complicated and not better than something simpler. So in this spirit, I've added one more playlist to the end of The Needle, my array of emerging-song supposedly-discovering lists.
It is called The Straw. It is assembled by programmatically taking one track each from a random sample of 1000 or so of the both obscure and uncategorized releases from the current week's Spotify New-Release Sorting Hat.
There is no complicated scoring or optimizing or clustering, there is no personalization, there is no iteration or feedback or machine learning. Some of it probably isn't even new, or shouldn't have been uncategorized, or maybe shouldn't have been made. If my fancy machines are working, and if the underlying collectivist moral premise they implicitly represent -- that the world's listening patterns distinguish memorable music from forgettable -- has significant validity, then listening to this playlist should be a poor use of your time.
But those are non-trivial "if"s. If anybody claims that they do something mysterious and yet valuable, based on appealing but abstract assumptions, you ought to be able to demand that they, at least, have actually contemplated how the world would be if they didn't.
And this "they" is me, so here is my professional memento mori, an unfancy discovery engine that believes wholeheartedly, even when I have the courage to think I'm only pretending to, that to find amazing music you've never heard, all you need is a way to find music you've never heard.
Your streaming music world is also about to be extra-complicated by more machinery for personalization, and I work on some of this, too. One clattering armada of machinery for trying to find music worth being discovered by somebody, and another one for trying to figure out which of that music is the music for which you are that somebody. It's kind of a wonder we can still hear over all that grinding.
Although, actually, it's less of a rhetorical wonder than an actual question. Can we hear over it? It's not inherently bad if your complicated thing is complicated, but it's bad if it's complicated and not better than something simpler. So in this spirit, I've added one more playlist to the end of The Needle, my array of emerging-song supposedly-discovering lists.
It is called The Straw. It is assembled by programmatically taking one track each from a random sample of 1000 or so of the both obscure and uncategorized releases from the current week's Spotify New-Release Sorting Hat.
There is no complicated scoring or optimizing or clustering, there is no personalization, there is no iteration or feedback or machine learning. Some of it probably isn't even new, or shouldn't have been uncategorized, or maybe shouldn't have been made. If my fancy machines are working, and if the underlying collectivist moral premise they implicitly represent -- that the world's listening patterns distinguish memorable music from forgettable -- has significant validity, then listening to this playlist should be a poor use of your time.
But those are non-trivial "if"s. If anybody claims that they do something mysterious and yet valuable, based on appealing but abstract assumptions, you ought to be able to demand that they, at least, have actually contemplated how the world would be if they didn't.
And this "they" is me, so here is my professional memento mori, an unfancy discovery engine that believes wholeheartedly, even when I have the courage to think I'm only pretending to, that to find amazing music you've never heard, all you need is a way to find music you've never heard.