# Music is a search problem now Created: [[2026_04_26]] 12:22 Tags: [[Blog]] [[Technology]] We think of music and art as the artistic expression of the artist. But, what is strange is that because we can encode that into a long string of numbers (essentially what an MP3 is); it can also be thought of in terms of other computer science problems. Which is what I'm trying to explore here: For example, if you think of a music file as just being a very long set of numbers. You can conceptually think that a random number generator could produce any work of music (that is encodable into an MP3 for example). So, another way of looking at music is the process of _searching for that number_, when you tune a vocal, or add a chord, you are just manipulating that number. Of course, music is _more than just a number_, but because it can be encoded as a number, it is reduced to just that number. Putting other factors aside like being a fan of the personality behind the music, and just thinking of the sound itself, theoretically, you can have any song you want if you generate enough numbers. I believe that we will get to a point where you can have your music tuned to your preferences in real-time. It will get to the point that music which is made by a person will be seen as a curiosity, even a bit strange: > Why try to make the music, when you can just generate it? Wait, she actually knows how to play a guitar? Who wastes their time with that, she isn't even that good. Now, this is rather strange of course, because we think that there is so much more behind music, how it resonates with us, how it relates to the time period, the style that it is, etc. But, if you think about it, those are just constraints right? Sort of like filters on what you expect the top result to be. I want a song that is punk, 1980s, by David Guetta; and you could see potential results of that - probably isn't good though. Which is why I framed this as a _search problem_. In a [[Sea of Information]], you can theoretically have any combination you want, but now it is generating the "right" subset that is most likely good for that combination. I see there being a sort of continuous Spotify, which is always taking feedback from you on whether you enjoyed A or B, and from that can build a dataset of parameters of peoples preferences for certain kinds of music. The value of any one song drops towards zero, whereas theoretically you'd enjoy them more than ever. It is similar to what happened to the clothes industry, where the loom dropped the value of clothes so much that it became quaint to ever make your own. ## References - [[Sea of Information]]