Viva La Radio

During the golden age of radio, Jocks ruled the world, airtime was King, and heavy rotation was the coveted prize of any artist.  The more exposure you got the more records you sold and the more money you made.  Then television came along and by most accounts the advent of MTV foretold the downfall of radio and in many ways those that made this claim were right.  Radio declined but even now decades later it still survives, although it is not what it was and for the most part it is now driven almost entirely by personalities, there is in some way a resurgence not in the conventional sense but in terms of digital services that attempt to mimic the behaviour of traditional radio.

Spotify, Apple Music, and Pandora among others attempt to provide their listeners with services akin to what Radio once was, but instead of playing music that is curated they play music they think their listeners will like.  Spotify and Apple use algorithms to achieve this, whilst Pandora does use curators, something which Spotify has attempted to mimic with curated playlists, yet in both cases there is still the absence of a Jock, there's no personality just music, which for most people is a good thing because they want to hear music they don't want to hear someone talking in between tracks especially when they have a personality they find personally grating.

Nevertheless this differentiation between traditional radio and these internet radio services is marked, the former had a degree of control and influence that was explicit whereas the latter has no explicit control at all, leaving the user to decide what they want to listen to, this makes marketing new music to them very difficult.  Indeed Spotify provides users with insights into which songs are played the most and which are played the least in each album they view, there's even a service called Forgotify [Not a sponsor] which randomly plays music from the millions of tracks on Spotify that no-one listens to.  This idea of forgotten content fascinates me, it's like forgotten books and movies too, there is now wealth of content that in practical terms will never be consumed simply because the rate of content production today is so fast that we cannot keep up with it.  There was a time when you could realistically set out to watch every video on YouTube, but at this point even if you spent your entire life from birth to death every waking minute watching YouTube you wouldn't even scratch the surface.

It is for this reason that I think curators are important and why I feel quite sad that Radio is not what it was.  There are still internet radio stations that specialise in genres of music that I still listen to but they aren't services that I would listen to as a matter of routine.  As for algorithms and their comparison in terms of effectiveness, the main problem with algorithms is that they all inevitably produce results that are the same or too similar, there is no diversity because they can't comprehend that two genres even though distinct and wildly different might be appreciated by the listener.  Algorithms inevitably send me down rabbit holes of content that all centre around the same ideas - this isn't exclusive to music streaming services either, it is also the case with YouTube and with sites like Amazon - the latter is the worst, it fails to comprehend that buying a set of earphones does not mean I would be interested in 50 more earphones, that need has been met, move on to something else.

This problem is caused by recursion, and by the fact that most of these algorithms rely on reinforcement learning, that is to say they take your feedback and use it to reshape their policy based on that feedback.  They don't understand the type of music you are listening to, they only understand how many connections to it have been established and how much weight they should place on those connections.  The end result is that the more you hear EDM tracks and enjoy them the more EDM tracks it feeds you - it might actually be possible to achieve something akin to curation if rather than pairing you with a single algorithm you were paired with about 20 that each started with different genres and excluded the genres the other algorithms started with that way each algorithm will in theory draw different conclusions.

It is also worth mentioning here that the concept of a Music podcast is something that is appealing to me, these however are rare as copyright basically makes them almost impossible for people to create without getting sued.