You already have two key things ML requires: lots of data and the user feedback (skipping songs, liking songs, search, etc.. ) which allows ML to learn and create personalized playlists
I believe this is a no-brainer, outperforming human curation of playlists should already be…
Stijn Coolbrandt

Well, you do… but it’s actually not trivial to go from that data to recommendations. If you’ve listened to something one time, should the algorithm assume you like it more than what you’ve listened to zero times? Is number of plays a proxy for how much you like it? (If so, you may come to curse your party playlists.)

Also, you need to model people — the variation in people’s tastes — simultaneously with the variation in music that’s relevant to it. On top of that, you have situational context.

Machine learning can answer lots of questions with enough data, but in this case, it’s hard to figure out the right questions to ask.

Like what you read? Give Harald K. a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.