5 ms·
That is not quite what I was thinking. More like the industry is selecting for the optimum combination of common factors. For an overly simple example: - perso
by Rury 3y ago
That is not quite what I was thinking. More like the industry is selecting for the optimum combination of common factors. For an overly simple example:
- person A likes musical features W,X,Y,Z
- person B likes musical features I,J,K,Z
- person C likes musical features A,B,K,Z
Z is a common feature they all like, so it's very popular. K is the next most common feature, but not quite as popular as Z. Making a song that only has features K+Z is bound to be more popular than one that is solely made of Q+D (which no one likes). Mind you, features could be anything from rhythm, time, timbre, melodies, as well as other emergent stuff.
But having song with features K+Z is not Person C's favorite music, and rather a song which has features A+B+K+Z would be Person C's preferred music, but the industry doesn't select for that since A+B aren't common features everyone likes. And a song consisting solely of K+Z is simpler than one which has A+B+K+Z.
However, making a song of solely of feature Z is too simple, and hence doesn't work. As if there's a minimum complexity requirement to meet in order to be a song. Hence K+Z works but a song solely of Z doesn't. And K+Z isn't necessarily anyone's ideal/preferred music, even though it's popular.
Also, making a song of A+Z would probably stand to be somewhat popular, but not as much as K+Z. And so, a lot of popular songs seem repetitive, because they all use common features, such as feature Z in this example.
- binary132 3y agoThen the cardinality of the marketable featureset is inversely proportional to the feature diversity of market preferences, yeah?