Curated playlists have become the staple of everyone’s digital lives. Being a music editor for Spotify was almost as cool as being a cosmonaut back in the 60s. But as AI tunes up to take center stage, are we approaching the end of the human touch in the music curation?
The curated playlist, once a personal mixtape's digital descendant, has evolved into the soundtrack of our lives, echoing through headphones and speakers worldwide. Streaming giants, armed with thousands of human curators, have fine-tuned these collections to resonate with every conceivable mood and moment, turning them into a cultural phenomenon.
The wonder of curated playlist
The power of playlists extends beyond mere convenience. That’s why streaming platforms now employ diverse cohorts of playlist curators who possess an intricate understanding of music genres, cultural trends, and the emotional landscapes of their listeners. These curators select tracks for thematic playlists, such as workout anthems, party mixes, or genre-specific compilations, offering a tailored listening experience that traditional album formats can't match.
For artists, especially the up-and-coming, playlists offer a golden ticket to exposure. A spot on a prominent playlist like RapCaviar or Mint can catapult an unknown artist to stardom overnight. These placements are crucial, not just for the streams they bring but for the attention they harvest across the industry and among listeners.
Julia Trainor, a music supervisor and Head of Sync A&R for ALIBI Music, has first-hand experience with that: “Now, an artist can get exposure to a greater number of people by getting on a popular playlist, and audiences can discover a wider range of music they may not have discovered otherwise.”
Listeners, on the flip side, benefit from a curated experience that brings a sense of discovery and excitement. "Playlists are a very low friction way to share music, which is a fun activity. Back in the late 90s, we used to go to the trouble of going to the store to buy physical CDRs to burn mixes for friends, so playlisting is a natural extension of that, but much more convenient and free," reflects Emilio Guarino, music producer and engineer at Glitchmagic.
Julia Trainor, also believes that music playlists have exploded in popularity out of sheer necessity: “We’ve never had so much music available to us (a good thing), but it is overwhelming to navigate and impossible to know where to start (a bad thing). Having someone whose tastes we trust curate a mix of music provides the soundtrack to our day that we so desperately need (but have no time to figure out how to do ourselves).”
Playlists are everywhere, from the quiet corners of a cozy cafe to the pulsating heart of a nightclub. They've become the lifeblood of our digital music experience. But there's another side to it…
The dark side of playlists’ popularity
The dark side of the playlist phenomenon is a slightly shady business, where artistry and ethics sometimes falter at the altar of commerce. The allure of "large curated playlists" has led to practices that blur the lines between meritocracy and marketeering.
And now the placements on these coveted lists can, allegedly, be bought. Prices for these placements have soared as the market has become more competitive. According to a Daily Dot investigation, initial modest payments for playlist additions have escalated dramatically. What began as a $5 investment for a spot on a lesser-known list has ballooned to fees ranging from $100 to $200 for playlists with 90,000 followers and up to $2,000 for those with half a million followers.
Spotify has taken steps to combat these practices, updating its terms of service to prohibit the sale of playlist spots. Yet, the effectiveness of these measures remains a topic of debate.
This commodification of music curation begs the question of whether AI could offer a solution to these ethical dilemmas. By removing the human element susceptible to bribery and corruption, could AI curation restore the integrity of music discovery? And maybe that's one of the reasons why big companies are so excited about AI curation?
When AI steps in…
AI playlists are reshaping how we discover music. Examples include Spotify's Discovery Mode and AI DJ, along with YouTube's radio station feature and Amazon Music's partnership with Endel. All of these show a shift towards algorithm-driven curation.
“Generative text models have made it possible to build a new type of human-computer interface, almost like the introduction of the computer mouse or touch screens in the 1960s. It is now much easier for users to interface with products by expressing what they want in a natural way and having the machine do the complex work. You can simply say, “I want a playlist for a birthday party, with children's music,” and get back a playlist of family-friendly pop and birthday music. It is definitely changing how users and curators will be doing many tasks,” explains Omar Marzouk, VP of Music Experience at Soundtrack Your Brand.
And that’s not just for business benefits. Users seem to be enjoying the renaissance of AI in music curation, too: “The latest generation of music fans seems to consume music more based on the moods, activities, or genres playlists help target instead of seeking out a specific artist. This is where AI-generated playlists can help, as the algorithms can access user data to make specific recommendations,” according to music composer and producer Tero Potila.
Since Spotify is the leader of the music streaming market with more than 600 million users and 30.5% of streaming subscribers worldwide (as of the second quarter of 2022), let’s focus on them for a second. For a comprehensive understanding of how Spotify's recommendation works, including the use of collaborative filtering and content-based filtering, you can watch this video by The Wall Street Journal.
And as Spotify's curated playlists like RapCaviar see a decline in streams, while Discovery Mode potentially prioritizes algorithmic promotion over editorial insight, the role of human curators is increasingly under scrutiny. These, combined with three rounds of layoffs, prompt a critical inquiry: is everything that’s going on merely a reflection of Spotify's strategic realignment, or do they herald the end of an era where music curation is based on human creativity and understanding?
Layoffs and the algorithmic changes
In the echo chamber of Spotify's boardroom, the sound of layoffs reverberated three times in 2023. First, 600 employees were shown the door, then another 200, and finally, a staggering 1,500. Among them was Glenn McDonald, a data alchemist and the creator of Every Noise at Once, a music exploration site adored by anyone who loves discovering new music. Years ago, Glenn was also a part of the music intelligence start-up Echo Nest, which was eventually acquired by Spotify in 2014 to build a database of songs and artists. So basically, the whole Spotify discovery as we know it is based on McDonald’s work. The outcry was loud. Fans clamored for Every Noise at Once, McDonald's brainchild, to continue. Spotify marked it as a "Good Suggestion," yet nothing changed.
The firing spree and decline in streams from human-curated playlists sparked debate. Are these all signs of AI's growing dominance in the music curation space, as some media suggest? Was this the end of human curators?
Not exactly. The simple truth? Management fumbled. The CEO's strategy, not human creativity, was flawed. Spotify had binged on cheap money, bloating its staff. Now, cutting costs meant cutting people indiscriminately.
A former playlist curator, once managing Spotify's largest instrumental playlist, shares, "I’m not convinced AI is the cause for layoffs. Tech companies got too fat. It's about cutting SG&A (Selling, General and Administrative) expenses. As an independent curator, I see AI making independent curation more crucial."
Spotify's layoffs aimed to slash operational losses. Yet, this was part of a broader, unpopular strategy. The platform stopped paying royalties to lesser-known artists and pushed fully automated playlists. These moves, far from saving pennies, alienated artists and fans alike. Automated playlists, criticized for their lack of depth, hurt artist discoverability. AI-driven features like AI DJ have been met with mixed reviews; some would even go so far to say it has no soul.
"Anyone else's Discover Weekly completely suck?" yet another user lamented on Reddit. Only to learn that the feature is solely usable to discover a lot of stuff you won’t like.
So, despite technological strides, Spotify's AI-generated features don’t seem to be loved as much as intended. The question now looms: Can AI ever match human curators in crafting playlists that truly resonate?
Can AI even find new music?
One of the best people to answer the question of how AI discovery on Spotify works and who’s better at making playlists is Glenn McDonald since he is the one who contributed a lot to make it happen in the first place.
“Asking whether AI or people are better at making playlists is, I fear, a little like asking whether servers or chefs are better at making a meal. AI doesn't "make" playlists in the sense that humans do, it just interpolates patterns from the human playlists that already exist. If AI _replaced_ people, it would have nothing new to learn from,” explains McDonald.
And many experts from the industry agree with him.
Answering the question of whether AI playlists give new artists a fair shot, Randall Foster, Chief Creative Officer at Symphonic Distribution, shares that it depends on who wrote the AI and the algorithm. “A simple search using PlaylistAI, for instance, for "Independent Nashville Singer-Songwriters," yielded results like Sam Hunt, Chris Stapleton, Miranda Lambert, and Zac Brown Band. In some cases, the artists represented were not really singer-songwriters at all, and in all cases, they were not independent (they included Niall Horan, for goodness sake). This is an isolated search obviously, but I feel that as we utilize AI if we don’t get the results we are seeking, we are going to be missing the human curators more than ever,” she says.
“Just like in music creation by AI - AI can only work based on existing human work. If music in a fresh, new genre shows up, I’m not sure AI would pick up on the importance of that style until humans have already made it popular. Human ears and an emotional connection are still needed to spot the ‘next big sound,’” emphasizes the point musician Tero Potila.
Emilio Guarino from Glitchmagic adds another nail to “AI-is-better-then-humans” coffin: “AI is very good at making similar recommendations sonically. If you like Bach's piano music, decent shot you like Beethoven or other classical music. AI can do this pretty well. What it can't do yet is make big intuitive leaps. For example, if you had a friend who enjoyed an alt-metal band like TOOL, they might be curious about a modern orchestra piece like The Rite of Spring. Both contain lots of driving, heavy odd meters that have a primal feel to them. But genre-wise, they couldn't be further apart. Currently, a human could make this connection easily, but AIs can't do it.”
And I say those smart people can’t be ALL wrong, right? So human curators are still needed. At least for now…
The future of curated playlists with or without AI
Now let’s think about the future. In the not-so-distant time, where the line between AI and human creativity blurs into a cosmic jam session, we're hurtling toward a musical renaissance that smells less like silicon and more like teen spirit.
Glenn McDonald, a music data analysis visionary, offers a hopeful perspective on the potential synergy between AI and human creativity. "For me, it's much more interesting to think about AI-powered tools, or algorithmic tools in general, that can a) help human curators by expanding their awareness, b) amplify human curators' efforts by allowing them to get started with criteria instead of just individual songs, or c) collate collective listening behavior to make playlists that are effectively crowdsourced charts," McDonald muses.
“One example of (c) is this set of best-of-2023 playlists I produced using a human-curated genre system to categorize artists, and then those categorized artists to find a genre's audience computationally, and then that audience's listening to algorithmically find the new songs from 2023 that each genre-audience disproportionately liked compared to the rest of the world: https://everynoise.com/2023_around_the_world.cgi. In this case, I produced the playlists with code directly. Still, you could also imagine treating these playlists as starting points that a human editor could either adjust (that would be a case (b) example) or pull from to supplement their own knowledge (that would be a case (a) example).”
Then there's Julia Trainor, a music supervisor from ALIBI Music, peering over the digital horizon with a mix of skepticism.
“Would we even know if AI replacement of human music curators already happening? That's the real question. How many “curated” playlists on Spotify are actually created by a bot pushing the “recommended” button 20 times? I mean, we can't tell what news articles are fake news, which images are deep fakes, so how do we really know who's behind a music playlist?” she wonders.
Now, let me lay down my own prediction. Picture this: a world where Spotify's automated playlists have finally hit a sour note with the masses. Fed up with algorithmic DJs that can't tell Billie Eilish from Billy Joel, the people cry out for a hero, the underdog—a scrappy new open-source streaming service powered by AI and humans, locking arms in harmony, not competing.
This brave new service is so in tune with your tastes it predicts your next music crush before you've even heard of it. This future streaming giant plays matchmaker, connecting songs with souls in a musical love story for the ages. I can’t wait! (It’s probably not going to happen, but can we just dream about it for a second?)
