Loading…
This event has ended. Visit the official site or create your own event on Sched.
avatar for Glenn McDonald

Glenn McDonald

Glenn McDonald believes that music is the thing that humans do best, and lives in fear that the music he hasn’t heard yet is even more amazing than the music he has. At Spotify he channels this into genre taxonomizing, music discovery, personalization, gothic symphonic metal and fraud detection. He spoke about Every Noise at Once at EMP Pop Conference 2014, and about Satanic metal and the worst songs in the world in 2015.

“Can an Algorithm Have an Authorial Voice? Recommendation Systems and Applied Moral Philosophy in Playlist Generation”
Humans sometimes imagine that there is an ongoing contest, at once empirical and philosophical, between humans and machines to see who can do a better job of music recommendation. But the machines know that there is actually nothing on the “machines” side of this hypothetical opposition. There are only humans stuck behind varyingly thick layers of abstraction, trying to cut through them with varyingly complicated tools.

At Spotify I spend much of my time writing computer programs that use a lot of data and a lot of math to make a lot of playlists of many different sorts. All of them always end up involving non-trivial moral dilemmas and non-obvious elements of human art. I will talk about some of these, including:

  • When you ask what music is popular in a place, what do you mean by “in”? Or by “popular”?
  • How do you tell what’s actually popular when millions of people listen to your playlist of songs you say are popular?
  • How do you quantify the difference between the way you listen to CHVRCHES on a train ride home and the way you play Katy Perry songs to dance with your daughter and the way you decide which continent's rain noises to put on to go to sleep?
  • Does it actually matter what order the songs are in?
  • What power structures are you nominally questioning or implicitly reinforcing when you play a pop-punk cover of a Justin Bieber song?
  • If playlists are the new albums, how does individual artistic identity compete with playlist (or streaming-service) identity?
  • Where are the lines between description, extrapolation, and advocacy?
  • When should you play somebody their favorite songs again, and when can you try to convince them that their next favorite song could be something they don’t even know exists?

Twitter Feed