How Algorithms Are Homogenizing Music (and the World)
Вставка
- Опубліковано 28 бер 2021
- ►Support us on Patreon: / thegoodstuff
►Subscribe: / thegoodstuff
►Follow us on Twitter: / goodstuffshow
►Follow Matt on Twitter: / mjosefweber
►Follow us on instagram: goodstuffshow
►Like us on facebook: / thegoodstuffshow
►Sign up for our mailing list: eepurl.com/bnSOcH
This is the story of two songs. They are by different bands that don't exist anymore. Both of them are over 20 years old. Neither song was particularly popular in their time, but in the last few years, they have become these bands' most streamed songs by far.
What's going on?
A seemingly innocuous change on Spotify might be the culprit, and it belies a larger problem with the mysterious algorithms that choose which songs we listen to. And it isn't limited to music.
Enjoy!
On the other hand, everyone's Spotify algorithm is uniquely tailored to their previous listening habits. I've found that my own Spotify is becoming more niche, without much overlap with other people's music. Possibly because I use the Discover playlist a lot. My music is all "samey", but only in relation to my own music. In my opinion algorithms are disconnecting us even as they simplify our tastes.
This works on the assumption that you'll like what it gives you. Common denominator's taste likes common denominator music, it's just made more immediately accessible.
The algorithm trains people to match with the algorithm, scary!!
Pick your own music. Form you own opinions. You can do it, kids. 👍
Good point, although algorithms and recommendation systems can be powerful tools to do this. They are like the guy who sorts the shelves in the store, creating a preselection, so you don't have to grab thousands of random discs until you find something. The problem is that users are rarely allowed to control the decisions and emphases the system makes for them.
I have a few distinct playlists like obscure 80s songs. The playlist is actually labelled obscure and when it runs out, I don't think I can recall it ever recommending popular 80s songs instead. It manages to understand what I want from it.
But if I listen to my modern indie playlist, radio then it does start sounding the same. But maybe that's a fault of that specific genre instead?
I probably have been doing this subconsciously all my life. Since I'm a senior citizen, I don't have a spotify or pandora or something like that. I play music on CDs (wish I still had all my old albums!) . I can say that songs with a fast tempo, specifically in the key of D Major, can do no wrong. If I ever wrote out a "top 100 favorite songs" I bet at least half would be faster tempo, key of D Major.
There's a jump in logic here, it assumes that those songs get shown equally to all users, but I doubt that they'd show up after someone listening to classical music or heavy metal. It might encourage homogeneity within genre, but not of music overall. (also doesn't some degree of homogeneity define what genre is)
I definitely saw this short film
I don't think its keeping things the same, it's just making it harder to make big changes. The algorithm had decided everything has to happen gradually.
This is why a good DJ is still important and invaluable. A DJ will make things more organic and less samey!
DJs still exist? I thought radio stations just let a list of 20 songs shuffle play for a year at a time.
Until the make an algorithm to do that...
@@breadman32398 twitch, internet radio and after the pandemic physical spaces still have lots of DJs
I'm sure Spotify also tries to figure out the quality of songs, as measured by a low skip-rate, or number of "like"s.
Back then being unique was rewarding. People wanted new fresh and good material to consume. Now thanks to the algorithm, everything original is buried. That includes many things really and not just music. Horrible...
First: good job guys making interesting videos without the 10 min mark to achieve, that feels pre 2015 hahaa but its good, and second: thats the reason i loved pandora, they work in a different way, but its not available in my country meh
"similar, in style, to the music _-as-_ you had been listening to"
I'm not sure that I can agree that it's inherently true that old bands having obscure songs becoming moderately more popular means that the algorithm is inherently homogenizing. Given that it likely has a large number of variables that it uses to recommend something you'll like based upon what you're already listening to, which include information about how similar users truly liked a song such as number of listens or skips, it is often going to play you a song which meshes with what you're currently listening to and that may include songs by relatively unknown artists or back catalog music which you are better able to appreciate because it's of a similar style or genre to what you were listening to before.
That being said I think that music streaming platforms also clearly undercompensate artists and that's certainly reason enough to be wary of them
There are two sides to every story. You make it sound like we don't have a choice when spotify presents similar music, but that's pretty much the main reason I use it in the first place. Spotify just wouldn't be the same without the auto-generated radio, and I would almost never discover new bands. If the algorithm is homogenizing the music world, it's because we're telling it to, not because all the world's music is becoming the same.
... But maybe I'm just weird, because I like to listen to entire genres, rather than anything specific, most of the time.
Ye I listen music depending on the mood im in, and diffrent genres have variety of BPM and moods. I like spotify because its like genre detector. I don't want it to play drum and bass when im in relaxing music mood, when im gaming maybe I do. One thing bad about it would be harder to discover diffrent styles of music if you not looking for them. Algorithm is a tool, it's up to you how you use it I guess. This video lacks a bit, seems rushed.
But when I start radio in spotify based on a movie score song, star wars for example. It doesn't give me songs that sound like john Williams it just give me every movie score ever shuffled. That's even worse than what you presented I feel. The algo is just based on genre, not subgenres or sound like pandora seems to use
I miss the Rhapsody app.
I am extremely skeptical that this is actually how it works. In particular, your assumption that it recommends songs based on pure resemblance is dubious. It's more likely that it works along the same lines as most other recommendation systems that you'll find on youtube, amazon, etc. by recommending content that similar users have also liked in the past (similar users in this case being the ones that have a large overlap in the kinds of music they listen to). This kind of algorithm doesn't necessarily lead to homogenization.
As far as I know, Spotify and Pandora Radio for example are both using multi-criteria recommender systems, based on content, metadata, similarity of musical attributes as well as collaborative filtering like you explained.
I prefer to find music that is unlike anything I've heard before and hoo boy is that difficult on Spotify
This assumes that the algorithm only takes into account the similarity of music. I'm pretty sure Sportify's algorithm won't be that stupid. Most likely it's a combination of similarity and, more importantly, non-direct feedback from users. E.g. if a person changes away from a song, they probably didn't like it. If this happens a lot with different people who have similar tastes, you can pretty confidently say that people with this kind of taste don't like this song.
I don't use Spotify nor do I let any other source decided what I might like.
Algorithms, the religion of the online world, you must conform.
Kinda unsettling ngl
It play songs similar to the music you listen to, i for one don't really listen to only one type of music and have gotten several new bands recommended.
So if you don't want "only the same kind of music" don't only listen to the same kind of music?
Some might only like that kind of same music so i really don't see the point here, its not like they recommend and play me awful pop music that is the most popular but music that is similar to the current play list i had on....
Oh great, all songs sound the same so more of that is coming? O_O Yikes!
This reminds me of the story of the Twitter bot that quickly became a horrid racist. Algorithms have no morals.
Speaking of algorithms, it’s tragic how your videos have fallen behind on the UA-cam one.
if only the algorithms had taste right?
Hehee
Good stuff
And yet you keep producing videos in the same style and I very much doubt your using a computer algorithm for your style.
Depressive :(
ah deep learning.... no one knows why it does what it does, but it prints money, and that earns its proponents social credit, so they can finally stick it to all those stodgy AI people.
that is how every minecraft channel became full of yelling and emotional extrems: algorithms feeding kids!
But... music isn't homogenous? Like, that sounds like "back in my day" boomer crap but you aren't giving any evidence for music becoming more similar.
When you autoplay you're given music you don't know about that is similar to what you've been listening and that's... uuuh... normal? I dunno, if I'm listening to some metal it's safer for the algorithm to give me more metal, or at least rock or classical music than say, jazz or rap.
The fact that it gives financial incentives for artist to create homogenous music isn't an algorithm problem though. It's a capitalism problem. Just like the unammanned drone isn't a technology problem, it's a war problem, yes it's more dangerous than shooting people with cannons but the problem is that you are shooting people to begin with, the technology just expose the problem, but you don't solve the issue of war by saying drones are forbidden and you don't solve the issue of art being guided by profit by removing or modifying algorithm.