Music Discovery at Your Fingertips

Back in June 2016, Soundcloud introduced a new feature on their site called “Suggested Tracks”. Suggested tracks generated by an algorithm, that is. By implementing this feature, the platform joined the ranks of every streaming service out there in following an industry-wide trend; one that offers solutions derived from the marriage between human curation and artificial intelligence.

Equipped with the ever-growing capabilities of machine learning, recommendation algorithms make the promise of delivering new music suggestions that are right up the listener’s alley; perfectly catered to the individual taste. It’s a tempting promise playing its part in a paradigm shift, in which the ways for discovering new music have greatly diversified in recent years. Spotify, Soundcloud, YouTube or Apple Music are forces to be reckoned within this process, joining tried-and-tested institutions like the radio, record stores or the usual word-of-mouth.

The algorithm’s ongoing refinement and optimization could also be read as a disruption to the existing and deeply rooted ways in which the music industry works. I spoke to John Seabrook, staff writer for The New Yorker and author of the book The Song Machine: Inside the Hit Factory about the influence of algorithms on music discovery. After describing (major) music labels as historically technophobic institutions, Seabrook said: “A&R has changed from being something where you would have someone come in and play, or you’d go see them in a club, to a practice where you do your research on YouTube or Soundcloud. YouTube does the selecting that you used to do and you can, for example, immediately see how many views a certain band has. It is an indicator of their prominence. It is automating a bit of that A&R process. It is not literally being done by machines, but it is networked in a way that makes it a sort of mechanized process. Social media is definitely a curator of talent.”

Listening to Seabrook and observing the differing practices of streaming services, it seems, things get interesting when it is not “either/or”, but “as well as”. For example, Soundcloud’s “Suggested Tracks” had led me to stumble upon snippets of Yugoslavian Space Program, an LP that came out in November 2016 via Discom containing eight tracks of “electronic space music” from former Yugoslavia made between 1981 and 2015. I found this record because an algorithm had drawn my attention to it (though admittedly, later on, I also spotted it in the Rush Hour bins), but the reason I was able to find it in the first place was because Luka Novaković and Vanja Todorović had decided to compile a compilation that relies on an apparent and human-set agenda; oscillating between tender and futuristic, providing a glimpse of a time and place far from one’s own reality.

Digitron – “Digital Minds”
Yugoslavian Space Program ( Discom, 2016)

Using the previous track as a starting point for an algorithmic-led discovery on YouTube triggers well-known behavioral patterns. Hours and hours can be spent clicking through tracks; some unknown, some popular, some recurring (for me, it is “Deep Burnt” by Pepe Bradock), as if to say: “listen to me, you will not regret it.” After browsing through a few Music From Memory outputs such as Michal Turtle’s Phantom Of Dreamland, the algorithm ventures into Japanese territory. The beautiful A Touch Of Temptation by Masaaki Ohmura is part of a playlist compiled by YouTube’s algorithm, brimming with slow funk, disco, wave and more obscure sounds that I mostly never heard of. Some of them seem to be holy grails on Discogs though. Check the LP American Eyes by Rare Silk for reference.

Michal Turtle – “Maid From The Mist”
Phantoms Of Dreamland (Music From Memory, 2016)

Masaaki Ohmura – “A Touch Of Temptation”
You Gotta Chance (SMS Records Japan, 1985)

Rare Silk – “Storm”
American Eyes (Palo Alto Records, 1985)

Playlists are wanderers between the automated and human world. Spotify’s famous and hugely successful “Discover Weekly” playlist is for the most part compiled by algorithms, based on your individual listening history with the streaming service. With regards to these playlists and their effectiveness, Seabrook raises doubts: “People’s taste in music is much more complex and less logical than an algorithm can ever foresee. It is totally normal to see somebody’s music collection and they have all these different areas of interest. The algorithms face a steep challenge, because they try to duplicate or model something as a logical system that isn’t necessarily logical. You also have to weigh in on all sorts of identity issues: why people listen to music in the first place, what their relationship to the music is and what their relationship to the people around them in life is.”

Nonetheless, it can be said that playlists have changed the way people are discovering and listening to music. iTunes and its predecessors that were detached from commercial pressures – Napster, Kazaa or Limewire – started a lot of it by liberating songs from albums. Now considered a given, it definitely made discovering a broader array of music much easier, for 99 cents a track. If one is not in favor of the algorithms as a driving force behind playlists, there are, and always have been, many committed people doing the exact same thing. Usually, these playlists are catered to moods, certain moments and specific situations. The right sounds for the right moment. For an example, listen to THC Sounds, compiled by The Haxan Cloak. It ranges from Fugazi to Cabaret Voltaire to Eyeless in Gaza.

THC Sounds, a Spotify playlist curated by The Haxan Cloak

As a counterbalance to all the algorithmically-based discovery tools that emerged in the last couple of years, I want to stress the everlasting importance of the institutions I mentioned in the opening paragraph. Record stores and the radio still provide all the possibilities of serendipitous discoveries. For example, OOR Records in Zurich keeps it interesting by ignoring conventional genres and providing an alternative reading. There, one is greeted by “in your face hystérias”, “good morning & nightmare drones”, “ladies in the 80s” or “(very intelligent) dance/mathematicians”, which makes you find music like Ultimate Care II by Matmos, an LP that is entirely produced with sounds made by a washing machine:

Matmos – Ultimate Care II Excerpt Five,
Ultimate Care II (Thrill Jockey, 2016)

As a conclusion to the previous discoveries I would like to state for the record: It is my opinion that machines or algorithmic recommendations will not be able to replace the human element in music discovery. Rather, it’s more likely that men and machine would continue to move closer towards each other. These intertwined entities may become a hybrid organism at some point. As a final recommendation, I would like to close with a performance by Robert Aiki Aubrey Lowe aka Lichens – a musician who is interested in a synthesizer’s potential as an autonomous instrument. In his work, he contemplates the synthesizer’s limitations and how they can be overcome. For example, he discovered a technical solution that made it possible to connect the human body or even plants to the synthesizer. How did I come across this musician’s work? A friend told me about it.

Robert Aiki Aubrey Lowe – Performance at Exploratorium