Chat with a Spotify's Human-in-the-AI lab scientist; and the Cultural Moment of nostalgia meets avatar concerts
Edition 07
Hi folks
Happy Wednesday! Welcome to The Industry Playlist 🎵
I had good fun writing the newsletter this week, mostly cause I had a fun chat with a Spotify research scientist from their Human-in-the-AI lab which you’ll get to read about today. Besides, this week I dive into why artists should be rebelling against DSP’s predicting ‘hits’ and why avatar concerts may finally have their cultural moment (it's not because of tech).
If you haven’t subscribed already, here is your chance
Finally, if you are fan of folk music but never ventured outside the Euro-centric or even the South-Asian sphere, you are in for a treat. Check out Kekele’s Mbote Ya Pamba 🎵, a Congolese rumba revivalist band that released music in the 2000s (they’re like the Buena Vista Social Club of Congo). Styllart Records is a Paris-based African & Afro-Latin music label (independent & active since 1981) who have been responsible for bringing such acts to the world.
Now, let’s get to it.
MY FUN CHAT HIGHLIGHTS
Last week I had a fascinating chat with an ex-Spotify research scientist (he left in Mar '24) who was part of the Human-in-the-AI (HAI) lab at Spotify. He has a PhD in Human-Computer Interaction and Behavior Research, working on everything from improving search and recommender systems to experimenting with generative AI at Spotify. I enjoyed that conversation and wanted to share some of the highlights with you (with his permission of course).
Spotify's recommendation algorithms are the secret sauce behind its success. These algorithms make sure we keep coming back by constantly suggesting music we’ll love. I learnt from him that to surface new artists, Spotify would pull organic demand data from places like Wiki page views and Reddit chatter. But with these platforms tightening their API access, it's getting tougher to rely on them.
The interesting part is the “black box” of Spotify’s algorithm. Traditionally, the weights of the algorithm were tweaked by editorial teams. But with AI advancing, machines are starting to take over this fine-tuning process. As a result, the human touch in curation is diminishing, which is a big shift.
This shift has had some ripple effects. Spotify’s R&D team, which focused on understanding user behavior, has been scaled back. This cutback means fewer cool experiments like collective listening sessions or hyperlocal playlists that bring a social element to music are happening. I personally would love to see more of that except Spotify is solely focused on getting profitable now.
One of the new exciting things we talked about was how Spotify is testing ChatGPT-style conversational search features. It begs the question if listeners actually want to search for music that way - their research is showing that listeners find it useful to organize their play queue with such a feature. E.g., If you were playing a 90s rock hits playlists, and you put in a prompt to remove all Nirvana songs from the list. Curating on-the-fly becomes easier. Sounds gimmicky, but it's a new way to interact with the app that could give music discovery an intuitive boost. Time will tell.
So, why is this conversation important? Because it highlights where the music streaming industry is headed. AI is playing a bigger role, but it’s also changing the dynamics of how music is curated and discovered. This balance between machine efficiency and human creativity is something to keep an eye on.
SONIC SHIFTS
Spotify & TikTok predicting summer’24 hits [Streaming] [Tech]
This is the kind of stuff that makes me wonder if the dog (listeners) is wagging the tail (algo) or the tail (algo) wagging the dog (listeners). Who is to know just by putting a ‘list’ out there, these will become a hit because listeners flocked to check it out. On a serious note, I imagine this is going to become the norm in the future. It is the new version of ‘playlisting’ where artists will want to be ‘featured’. Artists and fans should be rebelling against this algo-driven monoculture with utmost urgency. Take my advice and don’t click on the links I’ve shared.
(Nostalgia trip X Avatar concerts) cultural moment might finally be here [Fandom] [Live]
We will soon see a resurrected digital Elvis Presly. The past few years we have seen an unprecedented boom in catalog sales, not just by music companies but also by the big boys of finance: private equity. You might think that this is tech doing its tech thing, but I actually think it is something else coming together. The tech has been there since 2012 when Tupac came to tell us at the Coachella stage. What I think is the decade-long effects of streaming creating a phy-gital musical experience that can realistically make money - flattening of musical eras (old is no longer old) + fragmented fandom (the new is not memorable at mass scale).
Sponsor backlash is the new reputational and business risk for promoters [Live] [Biz]
A lot of things around the world are burning, and big corporations are part of that economic system that has led to it. Multiple artists are pulling out of many music fests, the recognized brands are facing pushback from fans. Think carefully about the sponsors you bring on, check for the skeletons in their luxurious closets.
READ THIS SHIT
The future of marketing and A&R is all about 'scenes' - With consumption fragmenting in the streaming era, audiences are pulling from a wider range of inputs than ever to form their identities. For marketers, this means not just considering audience demographics (age/location) but also psychographics (values/aspirations).
Stop Fetishizing Data - We measure nothing that is culturally-forward, or culturally significant. In today’s risk-averse music industry, it’s all about data – not belief, taste, or gut-feeling – and that obsession has led to stagnation. Our numbers obsession has killed music journalism, now it’s slowly killing music itself.
14 Issues in Music That Don't Get Enough Attention - A handful of problems get most music press. There's a lot more that needs to be addressed.
KEEP TRACK OF
Rythm launches to revolutionise Group Listening - Rythm, the world’s first community-based group listening music platform, officially launched last week. Originally created on Discord as a way for friends to discover and share music while gaming - it soon became Discord’s most popular music app before it was shut down in 2021 due to licensing issues. In its new avatar, the founder has secured global licensing rights from top music companies and venture capital.
beatBread strikes partnership with music asset and catalog management firm OpenPlay - They’re increasingly going on the distribution side, changing their play to a B2B2B - work with labels/managers who work with artists. Their approach to work directly with artists must not be panning out (in a previous newsletter I’ve called beatBread a wolf in sheep’s clothing).
INTERESTING STUFF
Genderswap.fm - Built to catalog women covering songs by men, it’s grown into a nifty tool to help locate covers according to specific metadata.
The friendship theory of everything - Came across this blog called Bookbear Express. Loved the lightness with which Ava explores topics of the heart. This one on friendship had to be shared.
That’s all for today. We’ll be back in your inbox next week.
Thanks for reading,
Rohit