The Data in our Melody
Music | Data | Tech
Hi,
We are back with our long-awaited second edition. This time, we won’t just play a duet for you, but an entire orchestra! This edition is all about music.
How are data and tech being used in the music industry? Is the joy of discovering new music or having your taste in it over? Can AI create music? What does it mean for creativity? What does it mean for us as a listener?
Music plays a very important role in our lives! Hence, we really wanted to explore these questions. If you want to join our quest, keep reading. :)
Abstraction
“Music, this complex and mysterious act, precise as algebra and vague as a dream, this art made out of mathematics and air…”
– Guy de Maupassant
Did you hear music lately? We hope that you did. If you didn’t, you better go and immerse yourself in some music. For when you immerse yourself is where our story begins. Or ends. It really depends on how you look at it, as you will shortly realize.
It was Thomas Edison, the equally vilified and revered chap, who kicked off a revolution by inventing the phonogram. It was the first device that could record and playback sounds.
For most of the recorded music’s history, decisions about how to sell music were based on assumptions, experience, and intuition. If someone asked for more cowbell, you gave them that. Similarly, the choice of music was driven by music charts, a personal circle, or tracks available in your neighborhood store.
Fast forward nearly a century. Streaming has become synonymous with music. Music discovery, once driven by serendipity and friends, is increasingly becoming a bastion of machine learning algorithms. Questions about whether we actively choose music anymore can lead to endless debates. Are AI recommendations propagating bubbles? What even is free will? Is Indietronica really a thing now?!
It’s not just the listener. The music industry itself has undergone a massive shift. Right from artist discovery to track distribution, data and social media are increasingly playing a huge role.
In the middle of it are the musicians. Their creations evoke emotions ranging from sheer despair to pure joy out of thin air. It’s an exciting and strange time to be a musician. There are more ways available than ever to distribute your music. At the same time, abundance has amplified the existing cut-throat competition for the listener’s limited attention. If that wasn’t enough, AI-generated music has become a thing. But is that really music? What about the creative process, you ask? It can’t be all stripped down to just numbers and algorithms.
That’s where you come back into the story. You, the listener, decide if the music generated by an artist is good enough. It shouldn’t be any different with music from an AI. Is the song enough or is the artists’ story equally important to you? Are there identifiable patterns in certain melodies or ragas which are of universal appeal? What does good music and, more broadly, creativity mean in the first place?
Whatever you decide affects the musicians, industry, and nature music itself, just as they affect you. So the story will continue, with all its elements, in a loop.
Actuality
Sometime back, OpenAI released Jukebox, a neural net that generates music. Some hailed it as a triumph for AI in a creative field. Others saw it as a paltry addition to a musician’s toolbox. You can check it out here.
If that doesn’t impress you, here’s a composition from AIVA - the world’s first virtual composer to be recognized by a music society (SACEM).
No matter where you stand on AI-generated music, we can all agree that AI and data-driven services have altered how we consume music. In 2017, Spotify’s founder was named the most powerful person in the music industry. Streaming has resulted in decreased piracy and increased exposure for independent artists. However, the recommendation engines are bringing about even bigger changes.
Services like Spotify analyze a song’s loudness, tempo, and duration, among other factors. Doing that enables the services to understand complex features users like and improve its recommendations. You may have seen playlists such as “Atmospheric Calm” or “Afternoon Acoustic”. These are increasingly becoming more important than albums and conventional genre classifications.
We now also have data-driven services that claim to predict upcoming breakout stars. More than the music itself, such predictions are driven by streaming and social media data. Traditional music charts have also recognized and recently incorporated this phenomenon.
Musicians are compelled to engineer music in order to find their audience. This has also resulted in songs getting shorter. Many point these out as negatives. But is this behavior something new?
For a long time, tracks were made for radio or limited by the storage device’s limits. Data is often informed by the distribution and storage mediums. As the mediums change, so does the data. The change has just never happened at the current pace and scale previously.
The music industry is also dealing with issues the same as any large-scale data management project. The presence of data silos and the absence of metadata has led to artists losing royalties.
What does all of this mean for you as a listener? Do we continue to “discover music” or are satisfied by the algorithmic recommendations? A French study from 2019 found that “the algorithm-based user is still in minority”. The results are not necessarily representative of the casual listener. But if they are, that means we aren’t surrendering control of our musical experience easily.
Accompaniments
We know of one style of music that does not need a classification algorithm to identify the mood it evokes. The musician is already aware of the possible “bhaav” (emotions) that it can create when played in a certain way. I may have given it away with the choice of words.
I’m talking about Indian classical music and its underlying system of Ragas. Each Raga, due to its precise melodic and aesthetic structure, has the ability to evoke certain emotions. It “colors the mind” of the listener in a unique way. Raga is like a color palette for a musician - we even tried to visualize it.
The beauty of Ragas lies in the fact that they are formulaic yet open to a musician’s own style. It allows for more improvisation than its western classical counterparts. Due to its scientific nature, attempts to create “Bandish” (A musical composition) in different ragas using AI have been made. However, their utility at present ends at being a supplementary tool for musicians, to help the practice better. An artist is capable of rendering their own interpretations (“Khyal”) of emotions attached to each raga and depict it in a novel way each time.
Don’t trust my words, listen to this beautiful piece based on Raga “Piloo” (This Raga is associated with emotions of joy and season of monsoon) created live by Anushka Shankar and Patricia Kopatchinskaja using a beautiful combination of Sitar, Violin, and Tabla.
Weird Trivia
A few years back, Vulfpeck, a US-based band, released an album called Sleepify. It was a completely silent recording composed of tracks that are each a half-minute in length. This was done since Spotify calculated royalties based on how many times a track has been played, counting a single play as listening to the song for at least 30 seconds.
If you want to keep receiving future editions, subscribe to the newsletter below.
Till then, happy musing!