Everything in Music is Subjective, Except for This (Mostly)
Some of my music snob friends, whom I love dearly (sometimes), have mocked my musical taste over the years. While I could debate their reviews and credentials, I’ll admit that I do have very specific tastes at very specific times. Compounding the issue and the mockery, if something works, I’ll likely listen to it again and again...until it doesn’t.
I went through two years of writing new business pitches on behalf of a large ad agency that was pitching large accounts and I almost exclusively listened to Dave Mathews Band’s “Live at Red Rock” album while sitting in my office. Two years.
The team won a lot of new business during that timeframe and had fun mocking my musical choices. The amazing team and leadership deserve all the credit for these wins and by no means am I trying to suggest that my crappy music does anything other than spark innocuous conversation.
However, during that timeframe, I was remarkably focused and feeding off the team’s energy through insane workloads and ridiculous deadlines. That album for some reason always helped me through the chaos and write quickly and confidently.
Full disclosure, the album is on now, as I write this, and I am totally ok with you judging. I might judge me too.
This got me thinking…
I’ve always been curious about music and its impact on daily activities. I have even done research into beats per minute (BPM) and their effectiveness on mood and activity.
While it’s a fun hobby, I need someone to take the DJ helm at times. Enter YouTube Music (YTM).
Despite the diversity in my music taste, there is one streaming service I’ve stayed loyal to: YouTube Music.
To preface, I have a substantial library on Apple Music, with years of memories and “phases.” That has always limited me in expanding my tastes as well as situational wants and needs.
More to the point, they just haven’t kept up with how my tastes have evolved over the weeks, months, quarters, or even years. Manually feeding the system is a less-than-ideal situation.
I now have a library of songs, most of which I can not access, because I can not justify paying for multiple music services that don’t get me.
And to be completely honest, I can find my weird songs on YTM during those times that I really need to hear something like Saga or Donnie Iris.
But get this, while writing this I wanted to switch from DMB because it stopped working. I wanted to hear “On the Loose” by Saga (take a moment to judge, I’m doing it too) and YTM went ahead and followed it up with “Ah! Lea!” I did not tell it to play Donnie Iris. It knew and it’s freaky and awesome.
YouTube Music has an uncanny ability to read my musical mind, predicting not just what I want to hear, but when I want to hear it. This level of personalized music experience seems to be unparalleled.
It is and here’s why
Music and AI: A Personal Exploration
Music and AI had been on my brain quite a bit. I often wonder why certain music platforms that I listen to are so much better at learning about me than others. I honestly didn’t have any idea the providers are going as deep into the data as they are. The technology is there for many, but the levels of data are vastly different.
As a frame of reference, the services that I have devoted what I consider as significant time to learn my preferences and whims: Amazon Music, Apple Music, Pandora, YouTube Music, and Spotify.
To be clear my experience and anecdotes are not at all science. This is just me satiating my curiosity and trying to automate as much of my life as possible. A true test to verify actual data input and output is well beyond my modest capabilities and time.
I was on a quest to understand, in a layman’s capacity anyway, what service uses the best data and how close can it get to predicting my situational music needs — on demand.
It boiled down to that simple question that gave me some simple, yet compelling answers.
Advanced Personalization and Data Utilization
The effectiveness of YouTube Music’s personalization can be attributed to the vast and varied data Google collects. This made me curious about what data other major streaming services collect and how they use it to personalize their services.
The following is a comparison that I pulled verbatim from ChatGPT. Buyer beware, I have found that their specs are not always current. With limited resources, ChatGPT was the best solution for me with directional findings. I did clean up the formatting for Medium because I’m not a monster:
YouTube Music (Google): Data Sources: Search History: Google Search provides insights into user interests and queries. YouTube Activity: Viewing history, likes, comments, and subscriptions. Google Services: Data from Gmail, Google Calendar, Google Maps, and more. Device Usage: Data from Android devices, Google Home, and other smart devices. Ad Interactions: Engagement with Google Ads across various platforms.Actionability: Comprehensive user profiles allow for highly personalized recommendations across different contexts. Ability to predict user moods and preferences based on diverse data points.
Amazon Music: Data Sources: Purchase History: Data from Amazon’s vast e-commerce platform. Alexa Interactions: Voice commands and interactions with Alexa-enabled devices. Prime Video: Viewing habits from Prime Video. Amazon Services: Engagement with Kindle, Audible, and other Amazon services. Actionability:Utilizes purchase history to recommend related music and content. Tailors music recommendations based on user behavior with Alexa and other Amazon services.
Apple Music: Data Sources: Apple Ecosystem: Data from iPhone, iPad, Apple Watch, and other Apple devices.iTunes/Apple Music: Listening history, library, and playlist data.App Store: Download and usage data from apps.Siri Interactions: Voice commands and queries. Actionability:Integrates seamlessly with the Apple ecosystem for personalized recommendations. Uses Siri and device data to predict and suggest relevant music.
Spotify: Data Sources: Listening History: Songs, albums, playlists, and artists you listen to. User-Generated Content: Playlists created, followed, and shared. Social Data: Connections and interactions with other users. Third-Party Integrations: Data from connected apps and services. Actionability: sophisticated algorithms for music recommendations like Discover Weekly and Daily Mixes.Uses collaborative filtering and natural language processing to personalize music discovery.
Pandora:Data Sources: Music Genome Project: Detailed analysis of songs based on various attributes. User Interactions: Thumbs up/down ratings and listening behavior. Third-Party Data: Limited integration with external data sources Actionability:Personalized radio stations based on song attributes and user feedback.Less comprehensive but effective for a radio-style listening experience.
WOW and back to anecdotal…
This took me down several rabbit holes. ethical questions and considerations mainly. I will post that summary and this article on peterdawson.us soon.
I did have one, “what’s – maybe – in it for me” question for ChatCPT If Apple can use the troves of biometric data it has on me to predict what I want and need to listen to, it’s game over. You have won eternal loyalty. A cursory (again not scientific) glance points to this being possible with extreme opt-in/opt-out parameters to line up with privacy regulations.
For now, it’s YouTube Music, which somehow got moved from Saga to Brother’s Johnson as I’m wrapping this up. It gets me
Please check out my other articles and my website. If we can ever partner on something, I’d welcome the opportunity and will bring headphones, if needed.
