Hacker News Reader

Comment by Thegn | Hacker News Reader

One problem I've run into with the music services is with older styles. For example, I'm specifically into black musicians from the 1930s-1950s who played swing style music. I have yet to find a service that actually will play more than 2-3 songs in that style before deciding that what I really want to hear is rat pack or swing by white musicians. No matter how much I thumbs down it, I get Glenn Miller/Frank Sinatra/Benny Goodman instead of Count Basie/Duke Ellington/Slim Gaillard. The services have that music (I can find the songs and seed stations from it) but for whatever reason all of the ones I've tried (Spotify/Apple/Amazon/Pandora, and I have a feeling I've tried others and forgotten) just don't want me to listen to the style I'm looking for.

View on HN

The technical reason for this is just that the algorithm fails to distinguish between these two groups. Probably its just a hard problem to solve from a sample of what people listen to. I suspect a significant fraction of people that like black swing also like white swing and that results in the algorithm being unable to resolve that there are two features there and not one.

You really need a level of manual curation that a big data statistical model just can't provided at scale.

What I find curious about it though is that it's obviously recognizing the style - it plays 1-2 musicians from the "black" group, then circles into only playing Miller/Goodman/Rat pack and never comes back around to playing the music that I originally was trying to play. If it was behaving the way you are thinking, I'd expect it to mix the two styles.

I think the problem is that the algorithms are based on statistical probabilities from other users. I.e users who listen to X also like to listen to Y. So we’ll add Y to the queue. Then Y becomes the new reference point. I mean that is a gross simplification but essentially if your musical taste is outside of 2 standard deviations of the norm all the algorithms are gonna suck. For me they do.

So what you're saying is it's badly designed?

It's badly designed for your particular taste, but it probably works for most people which is why it's used.

I’ve definitely trained tidal that I prefer some pretty whacky sub genres (this Northern European country, but only metal with strong brass sections, or contemporary accordion, hurdy gurdy, and a dozen other clusters like that).

I’d guess if you created a profile and loaded it up with just black swing bands from the 30-50’s, it’d do OK.

If not, and I understand their algorithm correctly, it would not only be because no current listeners make that distinction (as discussed up thread).

It would also be because the metadata doesn’t give any signal for it. They seem to use information such as record labels, song writers, producers, guest musicians, etc.

If that metadata has no signal, then my guess is that you’re trying to get it to racially segregate music that was produced before the big interracial marriage scare.

People were worried that if their kids listened to the same musicians, then whites and blacks (or worse!) might marry, so they created white radio stations and black radio stations.

Before that, I imagine there was a lot more interracial collaboration, and the metadata wouldn’t find clean clusters along race boundaries.

It could also be that the old metadata was never digitized.

I wonder if this is the case, or if the model just has year as a heavily weighted factor, because bluntly, Ella and Louis were vastly superior musicians to Frank and Tony. I honestly can’t hear the similarity. It’s like thinking “Oh. You liked The Killers, here’s some One Direction”

The other replies are interpreting this as the algorithm failing, but I have interpreted these sorts of things as intentional design choices, wherein they want the recommender to keep trying to diversify your interests so it's harder for you to just quit the service and move to another one which might not have the same variety (or where you'd have to try to "teach" the recommender again). They've determined that the potential benefit is much better than somewhat annoying you.

This interpretation of their behavior is why I've stuck to buying my music (fortunately that's still common for the genres I'm into).

I’ve found that most algorithms tend to reinforce a taste by trying to provide more of the same. They rarely try to bring in something that diverges from the pattern. Of course, the libraries have limits and the algorithms will often match against characteristics that you do not consider relevant.

trying to diversify your interests

In this instance - and others in my experience listening to the recommendations on this and similar services - "diversify" is used when "dilute" would be more appropriate.

It’s not really though - if it was known in advance that you did not like a track then there would be no reason to recommend it. It’s the classic precision vs recall trade off: I can create a recommendation algorithm that only recommends your favourite song, forever, and that will have perfect precision but miserable recall. To increase recall we have to accept a drop in precision.

It's entirely possible to retain soul and widen reference.

Offering up bland imitations of authentic artists is hardly simply a drop in precision.

I would love diversifying, but it does not do that. It does "revert to mainstream" trying to push you toward the most generic thing accestible by association from what you like.

I encounter similar behaviour as you in an entirely different genre - I've long since suspected that Spotify keeps redirecting me back to songs that are either less royalties for them to play, or located closer to me on the CDN to save serving costs.

There are services like this that explain the behavior:

https://artistpush.me/collections/spotify-promotion

No idea if they work though.

Artists can also pay Spotify to promote their music:

https://support.spotify.com/us/artists/article/promoting-mus...

I imagine that boosts “organic” engagement, leading to the same symptoms.

You're probably wishing for a community playlist with people pitching in songs as they discover them ?

Would be great to have options to add stuff but keep it private while keeping in sync etc. Could be done with a meta layer on top of the Spotify player for instance ?

Reminds me of the old webring concept from long ago.

Or of the old mailing list concept from even longer ago.

I’d guess it’s just because the algorithm is not smart enough and is just looking at the category as a whole and then playing the most played. So Glenn Miller and Count Basie are in the same category, but more Spotify people who listen to that category listen to Glenn Miller.

Maybe one day, they’ll get smaller clusters and lump you in with other listeners who favor black musicians within that category.

This is my problem with these services in that they are very generic and smooth out the outliers. So it’s good for pleasing the 80%, but people with specific tastes are out of luck. Big time regression to the mean.

Which swing tracks do you recommend to a noob who just got introduced to the swing music and dance?

I'm personally a fan of the old stuff (Count Basie, Duke Ellington, Cab Calloway, Chick Webb, Jimmie Lunceford, Slim Gaillard, Ella Fitzgerald to name a few) but the best thing you can do is go to dances and lessons, and when you hear a song you like, go up to the DJ and ask them what it was. The best music is the music you like and that makes you want to dance, regardless of who made it.

On pandora I find every channel I make eventually slides into the nearest "standard" repetive theme (often abandoning the seed content entirely). I always assumed it was nudging me towards the content with the lowest licensing costs.

I'm a little puzzled as you're describing a music stylist on skin colour, black vs white. I presume you don't really mean that, so what distinguishes the two musical styles, black swing vs white swing? Serious question.

Not OP but I would assume that because they where still somewhat isolated groups in terms of directly lived culture, this would have influenced their works differently.

Like how the Blues didn't come out of a comfortable lifestyle.

Also isolated by force in many cases: even in states which didn’t have official segregation laws, things like redlining and police enforcement meant you had very distinct communities. This especially went for anything where alcohol is consumed (being drunk leads to deadly mistakes and could lead to crimes being ignored or minimized) or, especially, sexual contexts - if you’re a young black man, you’re probably not going to find it relaxing to be at a club where various white guys are stopping by to mention what’ll happen if you look at a white woman.

Looking of that period is a very sobering reminder of a very dark stain in our national history - and I’ve read too many stories about even well known performers being told they can’t play at certain venues or have to leave immediately afterwards to think everyone wasn’t aware of the stakes:

https://en.wikipedia.org/wiki/Mass_racial_violence_in_the_Un...

As an example of how widespread this was, it took Marilyn Monroe at the height of her fame intervening for _Ella Fitzgerald_ to be able to play at a club in Los Angeles! Not the Deep South, not 1917, but very modern California.

> In October 1957 Monroe made a call to the Mocambo nightclub in Los Angeles, on behalf of Fitzgerald. Monroe used her social status and popularity to make a deal with them. If they allowed Fitzgerald to perform, Monroe promised that she would take a front-row seat every night

https://americanhistory.si.edu/explore/stories/ella-fitzgera...

The closest I can come to a silver lining for this is that it allowed more artists to find a niche where they weren’t competing with the major national artists but that’s nowhere close to compensation for so much tragedy.

I admit I'm uncomfortable explaining it by skin color, but I've never found anybody who has been able to explain to me what the difference is, and I will admit lacking the musical knowledge to explain it. I fell into that style from dancing Lindy Hop; I was loving the music I was hearing when I was out dancing, so I went and bought an album from the only name I knew at the time - Glenn Miller. It was some of the most boring and trite music I'd ever heard in my life, and did not inspire me to dance.

For me (and I don't judge people for thinking differently) there's a certain joie de vivre in the music that is just lacking from what white musicians released commercially. I know they were capable of it (I once found a recording of Glenn Miller swinging it just as hard as anything Basie put out) but they were playing to their audience at the time. As I've learned more about the history of Swing and Lindy Hop, this was a specific choice made to "civilize" (as the white people of the time would have said) Jazz's savage rhythms. There's actually posters from the Arthur Murray school in the 1940s saying this exact thing.

(Aside, I had a friend who played a radio show in the 2000s playing old Jazz music. She told me once that if she put anything from a black musician on the show, she'd get hate mail from the listeners. Go figure.)

The simple explanation is that largely only white artists played on radio stations. Popular songs would be sanitized for white radio by recording what we would now consider as covers. However, it was also rather commonplace for a whole slew of artists to record a popular song at nearly the same time. The proliferation of covers wasn't so overtly motivated by bigotry since an original recording wasn't regarded with the same esteem as today.

>I'm a little puzzled as you're describing a music stylist on skin colour, black vs white

You may not realize this, but in the 1930-1950 era being described in America, there was something called "Segregation" where black people were considered legally inferior to white people. As such, there was a very hard line between "black" and "white", a line that was aggressively enforced by every level of society from lawmaking, policing and justice, to radio and TV access, to education, to neighborhoods, and frankly everything else.

With that context, I think it's very easy to see how there can be "black swing" and "white swing" -- it was in a society that forcibly separated everything into "black X" and "white X".

You should be able to describe the sound differences if the music is that distinct.

Go listen to Straight Otta Compton and then Vanilla Ice and you'll get it. Or for Swing Count Basie vs. the slicker more commercial Glenn Miller but it's subtle.

And then Coltrane and Orman and Davis come and change the whole jazz world.

I didn't ask for examples, I'm familiar with what race these artists are.

My point is if the music actually is so different, it should be noticable and describable without knowing the race of a particular artist. So how would you describe the black music that describes only black musicians but not white musicians from this period, and vice versa.

Sure. 99% of the white musicians at that time were total sellouts who played extremely straight, boring, conservative and no-frills music without any embellishment or soul.

The most obvious difference is the energy level and "rawness" of songs - those white bands had really carefully choreographed performances with minimal deviation, even solos were often written out in similar big bands.

Black musicians often shouted, yelled or mugged during performances - all of these are completely absent in white performances at that time.

Ya you're not being honest in the slightest. The OP likes the music made by the group he mentions. You instead say he's being racist for liking music of that group. I'm having a hard time discerning if you're trolling or serious.

So describe the difference between NWA and Vanilla ice to yourself while everyone else moves on.

"Should be describable" is a false metric. We can hear differences between categories and not be able to verbalize what we're hearing. It's the same for anything -- you could walk through a museum showing abstract expressionism and action painting, and feel that one of the styles speaks to you, and yet not be able to put into words how the two styles are different.

The brain can categorize much more easily than it can create a concrete definition for those categories.

> We can hear differences between categories and not be able to verbalize what we're hearing. It's the same for anything

For some reason, the view is widely held that internal thoughts are expressed in words. This would mean that anything you can think can easily be verbalized.

The fact that this view is quite obviously false seems to bother very few people.

There’s an assumption to your argument that I don’t believe holds true, that there is such a significant difference that it should be easily describable.

Talking about the arts is difficult. In everyday conversation, well known phrases for describing the arts include “I know it when I see it” and “if you have to ask, then you’ll never know”. I’m a fan of “writing about music is like dancing about architecture”. It just isn’t easy to describing differences in performance and interpretation.

Since it’s widely recognised that describing music is difficult, and since you’re familiar with the artists in question, perhaps you could accept the point in good faith or put into words why you don’t think they are any different?

>You should be able to describe the sound differences if the music is that distinct.

They are and I can, however the existence of Black Swing is in no way predicated on a difference in sound only.

Consider this: white culture in America continually stole from a legally repressed black culture, including white swing which stole the black art and commercialized it. Even if a 1950 white swing song sounds similar to a 1940 black swing song, there is still a "black swing" and a "white swing".

Frankly, I think trying to reduce history of music down to "the sounds themselves" is a way to whitewash the history and destroy the true knowledge of what happened and why. The context is very important.

People not deep into music have hard time verbalize difference between metal and rock or hiphop and rap. Despite differences being super obvious when listening.

That's extremely patronising, I know well about segregation of that era, and that segregation functionally continued far longer than the 1950s. Blatantly there's a difference, that was what I was asking about the effect of, which you ignored. At least @ Pannoniae provided an answer.

Back in those period it's still very distinct / segregated black vs white culture.

Hey! I have a quite similar music taste. (my favs definitely include Slim Gaillard and Fats Waller, that kind of stuff)

If you have Discord, I've been curating a musicbot with a similar music rotation since I've also been completely fed up with streaming services pushing more Glenn Miller/similar straight bands, not the hip ones :) If you are interested, join my server, the bot is running 24/7: https://discord.gg/wjsC2TUZPK

Thanks, I'll check it out.

I recently discovered everynoise.com, it can make a playlist of a genre for you, it has a lot of black*

Edit: spelling

everynoise.com

Last FM and music neighbors could have solved this.

It is also my experience that Spotify reverts to what's popular and last.fm doesn't.

I haven't used last.fm for a long time and it seems like a shadow of its former self, pre acquisition. I've discovered so much music on there, and I'm getting really disappointed by spotifys repetitiveness. Is last.fm still good to discover new music or is it just harvesting scrobbles?

Last FM was so good for this.

I'm sorry, but wanting your music algorithm to key in on the composer's skin color is a ridiculous expectation. Listen to albums or make a playlist.

In this case it's not that ridiculous because "black" here isn't just a skin color, but primarily a subculture/subgenre with some distinct musical attributes.

Nobody finds separating French electronic music into its own subgenre ridiculous. Same with Italian Disco.

Such distinct movements are quite usual, so dissatisfaction about Black Swing on streaming services is understandable.

Also, the subculture exists in the US because of hundreds of years of intentional effort by the majority to destroy any preexisting cultures among black people and prevent any integration with the mainstream.

That's funny, because a lot of 'black' music in that period came from irish musical influences and totally did not come from what you are claiming.

I think many here are missing the point being made. Of course there are stylistic differences between some groups of artists. The thing is that they probably aren't coded by skin color let alone period location etc, so of course it will bleed. Playing some swing and expecting it to continue to stay within very blurry racial lines is unrealistic, silly, and maybe irresponsible for a recommendation algorithm.

As it's been said, there are better methods of discovery for this purpose. In your example, I'm sure there are Spotify playlists for Italian disco that have been curated.

But these were different music styles.

Spotify won't be able to keep those genres separate either.

No, he's saying that within the genre of mid-20th century Swing there are distinct musical traditions found in black vs white bands, which he wishes he could partition against. What's ridiculous about that?

Because they all borrowed from each other and the line is blurry if it even exists at all.

Reply on news.ycombinator.com