4 ms·
When the Pandora music service launched, I remember reading about how it worked (predicting what you might like next) and it explained how they don't need to kn
by hunter2_ 2mo ago
When the Pandora music service launched, I remember reading about how it worked (predicting what you might like next) and it explained how they don't need to know anything about the songs at all (such as genre, tempo, instrumentation, place of origin, year of release, language, etc.) to predict what else you'll like. They simply identify other users with a similar history to you, and deliver to you some of the stuff that those other users had listened to that you hadn't listened to yet. If you abandon a song (suggesting that you don't like it much) then you become a bit less associated with the users who play that song a lot.
- dgellow 2mo agoI don’t think that works well for movies and tv shows because there is way less datapoints than music (they are way longer, meaning we don’t consume that many films/episodes). A song is also generally more focused, you generally have one tempo, level of energy, and limited instruments. A move on the other hand is a mix of so many art forms. If you only rely on what your cohort voted for you’re missing so much information to make an actual good recommendation. Netflix system seems to work for their goal, which is to increase revenue, but it’s not a great system for the user satisfaction