It is an amazing corporate touch to have celebratory swag, for having individually contributed to the company’s biggest yearly event, arrive AFTER you’ve been laid off from it. Wrapped, indeed.
The world listens to music, and thus I feel like it should be obvious that the collective insights from this listening also belong to the world. Every Noise at Once was a way of helping this knowledge self-organize, and giving it back.
Oh, I wrote a book! It's called "You Have Not Yet Heard Your Favorite Song: How Streaming Changes Music". It doesn't come out until June 2024, but it was announced to the UK trade press today, so it's finally no longer secret. There's more info here:
Honestly, at this point it would be a little reassuring to get a note from some decision-maker at Spotify telling me how much they always resented and distrusted my work, and how joyful and strategic it was to get rid of me. Give it some intent.
Getting laid off sucks, there's no avoiding that, and it doesn't really help to get angry about it yourself. But it's obliquely encouraging to have *other* people get angry about it on your behalf.
It's sad, but not too surprising, that a corporation might not see the value of this. But it has been reassuring to me to be reminded, here and elsewhere, that some people did.
A few more reminders, this morning, here:
It used to be my job to try to turn music-listening data into cultural knowledge. I think there are at least four fundamental forms of this: popularity, currency, similarity and community.
If you ended your year with
#SpotifyWrapped
, you more than likely discovered what your listening personality is. But how did your listening personality actually come to be?
@glenn_mcdonald
breaks it down:
I use Spotify data and math to see what songs spike up in popularity after notable world events. Here's a playlist of what did this weekend.
After the Counting
I don't have any news about New Releases by Genre, but if you're so interested in that topic that you would even read news about the lack of news about New Releases by Genre, here's some:
This is a brutal headline that should make everybody working on algorithms in music stop and think about what they're doing, and what powers it serves.
Decided to take a break from the string of music books and read one from my science-fiction stack instead. It opens with a guy contemplating his record collection. There is no escape.
“Mechanical” music royalties are a great example of obsolete legal formulation. They attach songwriter payment to the “reproduction” of a recording of a song, because in the physical era that’s all we could track. 1/9
I tried to get this out of my system with a short tweet-thread, but it didn't work. So here, if you're interested, is a much longer detailed analysis of the proposed Living Wage for Musicians Act.
@tedolsen
@EveryNoise
I keep opening the front door as if I'm going to check the mail or something, ready to pretend to be surprised at the cameras. Nothing yet.
Spending the day pleasantly reading angry tweets about all the genres we supposedly made up that we totally didn't make up. Plus Escape Room, of course.
I complain about AI because it ought to be good for a lot more than plausible-sounding chatbots you can't trust. What, exactly? Ask me again when I've been working at
@imbue_ai
for longer than 1 day.
If you're curious about how heavily the comparison of effective *average* per-stream rates between music services depends on plans and activity rates, you can do some illuminating math with Spotify public data.
New Particles is either paused or finished, I haven't decided which yet, but if you're one of the few who anticipates the Xmas edition instead of skipping it, here it is.
Math fun: public numbers suggest that for Spotify's global average per-stream rate across all tiers to reach $.01, the prices of premium plans would have to go up by a factor of ~3.5, and that's assuming no current subscribers would switch to free or a competitor as a result.
This is a very good article about the factors involved in streaming-music exposure and algorithmic recommendations in general. (In which I am quoted, but it's well informed from many directions.)
Today in music novelties, I set up an automatic playlist of the most popular songs with each numeric title from 1-10000 (no other characters allowed, and only 1/artist). So far only 4471 of 10000 slots are claimed, with 9 of the first 1000 still open.
If you care about the recorded-music business, you will definitely want to read the newest IFPI Global Music Report, published here:
and discussed here:
The good news is that I made a daily 5-hour query take 1 minute, so now the idiot who wrote that query has to do my job for the rest of the month. But you can probably guess the bad news.
Yes, please, governments, pass some laws requiring algorithmic transparency. I would love to tell you exactly how my algorithms work. You will have to pass even more laws to get me to *stop* telling you.
The structural problem with Twitter, reproduced exactly in all its clones, is that there is no shared experience. It’s only a side-effect of this that rebuilding ANY individual experience on a new clone is laborious.
@tedgioia
Seems like you're overlooking the biggest relevant change, which is that in streaming "old" listening is now tracked. Buy a new album on CD and listen to it for 20 years: nobody knew. Stream a 20yo song: we know.
“It is not acceptable for a few major players to meet in secret, agree a self-serving business model and then present it to the world as a fait accompli.”
Cogent that this is the first point. Resistance must confront the power-structures first.
Every once in a while, go through and listen to the whole top 50 as if it's your only advance knowledge of a planet you're about to land on in a ship almost out of fuel.
If you run a company that has something to do with music, you should probably order copies of this book for all your employees, and a couple extras for yourself. E.g.
@eldsjal
I don't believe that genres are dead, but I did just go from Sudanese-inspired violin-rap into a Romanian-named Swedish post-metal dirge without immediately realizing the song had changed.
1000 True Fans sounds easier until you realize that most people won’t have the capacity to be True to more than one artist, so you need 1000 people who like you more than absolutely anybody else, including all the artists who work harder than you unless you are Taylor.
Young putting his music back on Spotify now that Joe Rogan is available everywhere again is a pretty good snapshot of the state of the business. It’s hard for anybody embedded in these power-structures to do much better than anybody else, but it’s also hard to do worse for long.
4 moral imperatives for personalization systems:
Consent: personalization must be optional
Transparency: personalized results must be labeled and explained
Control: any implicit factor must be explicitly overridable
Accountability: the collective results must be accessible
The contention that Trump is better able to handle COVID policy than Biden because he HAS it is idiotic, but it also explains why Trump thinks he's better at handling racism.
My PopCon contribution this year is not a talk, it's a web explorer for the music that defines, unites and distinguishes kids around the world. It's called The Aqueduct of Youth. You can be young at any moment.
This is a really good piece about fear and nostalgia and human complexity and the difference between "algorithms" and control. I don't say this because my own work (sic) is mentioned, but it's obviously not a coincidence.
I affirm some and take issue with other parts of
@newyorker
writer Kyle Chayka's case on taste and recommendation algorithms, w/ particular nods to
@glenn_mcdonald
, in the jam-packed new issue of
@Bookforum
.
J’ai le plaisir de lancer ce 6ème épisode de
@DataBroadcast
, dans lequel j’interroge
@glenn_mcdonald
,
#Data
Scientist chez
@Spotify
et créateur du formidable site “Every Noise at Once”
🎵Et vous pouvez même jouer à des blind tests entre les séquences
2 days until my book’s official publication date. Delivery reports have already begun, but technically you could still say you pre-ordered it if you did so right now.
All I meant to do is sit down and listen to one Big Bopper song, because I wasn't sure I knew any. But apparently I can't even do a simple thing without it becoming an essay about algorithms and hubris.
In listening communities as in geographic ones, the ability to move between places and across borders doesn't cause places or borders to cease to exist...
Postscriptural notes: I work at Spotify, so I have both interests and data, but this theory is just mine. And I spent $1000s/year on CDs, so if I'm right, it's precisely people like me who no longer have disproportionate power.
Using AI to add the "missing" parts of a photograph is an amazing example of not understanding an art form at all. I assume generating clothes into a porn scene will be next.
Go home and stay home. Otherwise yesterday's pictures of no toilet paper in supermarkets are three-weeks-from-now's pictures of no ventilators in hospitals, and you not laughing because your lungs can't breathe on their own. The math is the same.
2 upcoming appearances in LA:
Talking about music data complications @ Music + Data Symposium, UCLA 3/8
Talking about streaming and memory @ the Pop Conference, USC 3/9 9:00am
Both are free but limited; register in advance!
Hint: if you see anybody cite the amount that a music-streaming service PAYS "per stream", without also citing the amount it EARNS "per stream", then you know they have no idea what that amount means.
But if your PDF reader is broken, I'll summarize it for you. Pages 1-5: Hi. Pages 6&7: Streaming continues to grow, and historical graphs always look better if we pretend inflation doesn't exist. Pages 8-17: IFPI represents labels. Pages 18-: Did we say labels? We did? OK.
Seeing
@ZOLAJESUS
retweet
@tagaq
, then listening to Tanya covering Iron Maiden and Zola covering Black Sabbath, then wanting to know what a collaboration would sound like...
When I take over completely I will enforce a policy of never listing songs without artists. "Wonderwall" here is Violet Orlandi's fantastic metal cover, not the excruciating original.