“I trust my preferences more than AI” – Spotify’s new playlist creation tool is fun, but it can’t beat human interaction
About a year ago, Spotify introduced AI playlist editing to its Premium subscribers, using the technology to create any kind of playlist that will match your mood, the weather, any activity, your current problem… you name it.
Enter your information, ChatGPT style, and voila – a playlist of songs matching your parameters will be created for you. My partner Harry McKerrell had a great time testing it, with his encouragement to get more advanced and crazy to push the limits of what this technology could do, he ended up with playlists titled “Startled Chick’s Dubstep Journey”, “Abyss Stares Back” (there’s that problem) and just a song about cheese.
As the sun finally broke through the dreary, rainy weather we’ve had for months in the UK, I started with the command Spotify showed in the demo: “Uplifting songs for now when it stops raining in London.”
It takes a few seconds (sometimes a minute or more for more complex commands) to start generating a playlist, and I get a 12-strong list of songs like Here comes the Sun, Walking in the Sunlight, Levitating, Uptown Funkand so on. Great playlist, all in context. As I was enjoying the early March sunshine with the playlist running, it certainly put a bit of luck in my step.
If you like your resulting playlist, you can set it to update daily or weekly on specific days, like your chosen Discover Weekly or Daily Mix. You can also add songs, edit the list, change the name of the playlist and organize the information itself as you wish.
I’m willing to ask it for a dynamic, dark educational-inspired playlist to help me write (something I often look for in existing playlists), and it comes with tunes from Lana Del Rey, Noah Kahan, Of Mice And Men – they’re very intelligent, gentle and indie compared to the deep tones I’ve been after, so I’m including a little music, and the music played “vocals”.
This changed the entire playlist to include Arvo Part, Max Richter, Ludovico Einaudi, Nils Frahm, Clint Mansell – which was very close to what I had in mind. However, the tone of the songs was still on the light side, and I realized that what I really wanted was a Ramid Djawadi-produced one. Game of Thrones song, full of amazing cellos and powerful disturbances, and I found myself changing the playlist to songs that I knew would hit the spot better.
I’ve done the hard work for you next: “10 songs that sound like an Elliott Smith song. But don’t add Elliott Smith songs.” This turned out to be very complex and showed the flaws in how the Al algorithm “thinks” compared to how a person coordinates communication. Suggested Playlists for “Elliott Smith Soundalikes” include Leonard Cohen, Yo La Tengo, Jeff Mangum, and Spotify even tried to cheat by putting Smith on guitars, but not singing. I get it; it’s hard to emulate the brilliance of his technical guitar playing and softly sung confessional lyrics.
But the selection didn’t seem close to what I consider Smith’s sound. Personally, I would have expected to see a few tracks from, say, Death Cab For Cutie or early Jeffrey Lewis, while my husband (a big fan of Elliott Smith) pointed out that tracks from Fionn Regan’s. The End of History would have been a better shout than most of what Spotify’s AI has revealed here.
Where Ordered Playlists come into play, however, is when you’re working with more detailed information. I’m asking for the first 10 songs I ever played on Spotify, then the top tracks from every year I started using Spotify (since 2011), in ascending track length. These parameters have proven to be very successful, bringing back musical memories in the same way that you get snapshots of a photo memory from 10 years ago on your social media channels.
There are a lot of songs that I completely forget that I played so much – especially songs that I used over and over again as test tracks when I joined. What is Hi-Fi? in 2012. I have absolutely no idea why I’m streaming Van Morrison Sweet Thing 36 times that year, while I had forgotten I was obsessed with pianist James Rhodes’ albums – they were great in the background when I wrote my original review.
Everything that comes in comes with little snippets, which I love. From statistics on how many plays, track length and year, to even a little blurb of what the song contains and how it relates to your content.
Getting more specific with realistic parameters rather than trying to mimic my mood continued to be more successful: I asked it to compile a playlist of songs used in TV shows. Detectorists again Minor Prophets – and this is the kind of data-driven intelligence work that AI is best used for, saving me the time to go through each song carefully and manually add it to a playlist (which I’ve done before with other TV shows).
While that time-saving nature is certainly appealing, the process of creating a playlist when it needs to fit a specific situation or purpose is a human, collaborative effort. Every song you choose, the connection between them, and the way you place them, tells a story.
Sometimes the choices and connections you make are confusing, chaotic and make no sense to anyone but you – and it’s these connections that not only make playlist editing rewarding, but also completely different from an AI algorithm.
I spent a week creating a Motivational Playlist when the situation came up, and I found the results to be immediate. You can be as precise as you like with your instructions, but it’s not as tangible or permanent as choosing the music yourself with intention and the feeling behind it. After a week, it was really over Minor Detectives/Prophets a playlist I kept coming back to.
One of my colleagues, Robyn Quick, made the point that Spotify’s recommended playlists already do a good job of providing playlists you’ll like, but she notes that “I trust my preferences more than AI”, and I can’t help but agree.
Spotify’s Promoted Playlists feature is undoubtedly clever and the scope of possibilities is huge. While it can’t replace the emotional connection that a human-curated playlist can provide, if you’re an avid playlist builder, especially for a specific activity or flight status, it’s worth a spin.
MORE:
Top music tracks to test your hi-fi system
Qobuz releases a tool to protect us all from AI-generated music
I went from a major streaming service to this niche platform for audiophiles: that’s why I’m glad I made the switch.
