The Streaming Algorithm Paradigm
How I recently started using machines to grow musically.
Lately I've been working on multiple projects. In doing so, I needed to take a break and turned on an "80's Summer" playlist from my music streaming service. Some songs were familiar. Others were completely new to me. Thankfully, my wife was listening as well and pointed out that the percussion section of "Genius of Love" by Tom Tom Club is the same beat/groove as "Fantasy" by Mariah Carey.
Now keep that anecdote in mind because, prior to pulling up this 80's playlist, I was trying out different genres just to hear something new.
The next day I put on the same playlist hoping to hear more new songs from a familiar time period. That's when it hit me: the algorithmic format of streaming music is a library of blended topics. If you listen to a lot of jazz, rock, or pop music then you will likely be shown playlists that are similar. The same can be said of time periods like the 70's, 80's, and 90's. Yes, there are many artists and genres in a decade, but a time period can have many overlapping musical ideas. These overlaps help the machines to learn what you like based on the connections that we make in music, like how I was able to connect similar "Genius of Love" and "Fantasy" by the beat.
I used to look at recommendations from an algorithm in the same respect as a bad car salesman. I didn't even want to acknowledge any suggestions. I knew what I was looking for, so just give me that. Stop trying to sell me something I didn't ask for.
Now I'm starting to see recommendations as branches to what I already listen to. I could choose to try something similar and see where that takes me. If I do, and quickly change my mind, then the algorithm could view that as a dead end and avoid recommending other relevant playlists. This could be good or bad because it might help me go down the path I want without realizing that there were other positive options. Taking note of what an algorithm recommends could be a good way to keep new ideas available for another time, so use your device's screenshots to your advantage.
Another thing I considered is the idea of finding a song or artists I like and playing that a lot just to see what the algorithm recommends based on my current activity. That way I could get my copy of the Real Book out and listen to versions of songs by Chet Baker exclusively with the hopes that the algorithm will present new material that is similar to Baker's style. In this way the algorithms become tools for us to use. We can tell these machines what we want through our behavior and in turn allow the machines to behave for us.
A final thought on this topics is in regards to the act of conversation. When my wife pointed out that two songs have a similar "beat", the act of listening became part of the conversation. To fully use the give-and-take of interacting with an algorithm, I suggest bringing what you are interested in to other people. Try making your own suggestions and even inquiries like, "Hey, what do you think about this?" Allowing yourself to interpret music through others, and now machines, might be the best way to find new music because it connects us, others, and libraries of music as one system.
To find new music outside of the control of algorithms, keep an eye out for songs with only a few dozen likes or less. These songs might be new and have not had time to accumulate a vast quantity likes. You might even want to try looking for new music via forums, podcasts, and other periodicals. Unfortunately, I was unable to find a way to search for music based on like count, so if you have a way to find those great up-and-coming hits then please let everyone know in the comments section. When you do find something new, be sure to like it at the very least. If you want the algorithms to help a song out, then also share it and leave a comment. Any comment will do. For starters, check out this song that my Dad wrote. He’s got quite a few out there and I hear that some guys at one of the local music shops are into this one. Enjoy.
