AI recommendations fail fans who like hard rock and hip hop, say scientists
- Reference: 1617188648
- News link: https://www.theregister.co.uk/2021/03/31/ai_music_recommendations/
- Source link:
The problem appears to be this, we're told:
[1]
People who prefer so-called hard music, which covers a wide set of genres from hard rock and punk to hip hop, aren't much interested in music outside of their niche.
People who prefer softer music, which is a limited set of genres, are happy to listen to artists outside of their niche.
Algorithms are better at recommending tracks for easygoing ambient fans than for choosy hard music fans, or in other words, worse at picking tracks for hard music fans than for ambient lovers.
In short, hard music fans are picky, and algorithms can't cope with these difficult-to-please folks. That might seem a trivially obvious conclusion but that's science, we guess: supporting a conclusion, even if obvious, and then finding an engineer to fix it.
[2]
“Some subgroups, like ambient listeners, seem to be more open to listening to music from other subgroups; plus, they are more similar to each other – all of this is great for recommendation algorithms and such users more likely accept recommendations from different groups,” Elizabeth Lex – co-author of this research, [3]published in EPJ Data Science this month – told The Register .
“In contrast, hard rock and hip-hop low-mainstream listeners are, in our data, the least open to music of other subgroups, and within themselves, much more diverse, and thus, harder to satisfy with recommendations.”
The researchers, led by boffins at the Graz University of Technology (TUG) in Austria, analysed the performance of multiple recommendation engines on the music listening histories of about 4,000 people scraped from Last.fm users; the code involved is on GitHub, [4]here .
Sure, check through my background records… but why are you looking at my record collection? [5]READ MORE
Given some of the tracks and the type of music each person listened to, could the models correctly predict what other tunes they would enjoy? A recommendation model's accuracy was measured by seeing if the software's suggestions overlapped with tracks a particular user had actually listened to and liked. When the team took an average of how well the recommendation systems tested, they found that their predictions were most accurate for ambient listeners, and least accurate for hard music fans.
Not only are fans of hard music less likely to listen to other genres, the songs they listen to within each genre are more likely to sound distinct from one another. Tracks that top mainstream charts do seem to sound the same. This makes it all the more difficult for recommender systems to find relevant music for hard music fans.
While the recommendation software deployed by the likes of Spotify and Pandora are secret, the algorithms will likely be based on collaborative filtering mechanism algorithms, just like the ones used in this experiment, we're told.
“Three of these algorithms are collaborative filtering-based, that means that similar users are detected and songs of these similar users are recommended. This is the most used type of recommender systems to date and thus, also reflects the primary ones implemented in Spotify or Pandora,” said Dominik Kowald, first author of the paper and a research area manager of the Social Computing team at the Know-Center, Austria.
These real-world systems will be more complex than the ones tested for this study, however, which is something to bear in mind.
“Our algorithms are more simplistic than what the streaming platforms use, particularly in terms of data they can exploit as naturally platform providers have complete access to users' data,” Lex added.
The team hopes that their study will improve music streaming services for people who prefer their particular musical niche.
“If we think about the problem from the perspective of artists, who produce low-mainstream music, if their work is recommended more often, they get more exposure and interactions – which is crucial in this business," Lex told us. "So, we hope that our research contributes to helping serve consumers better and to help low-mainstream artists get more exposure in music streaming platforms." ®
[6]
Editor's note: This article was updated after publication to include Dominik Kowald's quote, and to revise our framing of the paper.
Get our [7]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YGScnqOo1YVGDRU5ChipvAAAAEg&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YGScnqOo1YVGDRU5ChipvAAAAEg&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[3] https://epjdatascience.springeropen.com/articles/10.1140/epjds/s13688-021-00268-9#Sec11
[4] https://github.com/pmuellner/supporttheunderground
[5] https://www.theregister.com/2020/02/28/sure_check_my_background_records/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YGScnqOo1YVGDRU5ChipvAAAAEg&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://whitepapers.theregister.com/
Re: Why would it recommend?
As a fan of Hip Hop and Metal I disagree. In fact I completely disagree with this statement:
"People who prefer so-called hard music, which covers a wide set of genres from hard rock and punk to hip hop, aren't much interested in music outside of their niche."
Total bull. My favourite band is Motorhead but I regularly listen to everything from metal to house, classical, jazz, hip hop and folk music. I think I've seen Steve Earle perform as many times as I have Lemmy.
I'd suggest the problem AI has with recommending music is music taste is exceedingly personal and not explainable in machine terms. I couldn't explain why I like the bands I do but I definitely like them and dislike others.
Re: Why would it recommend?
I take more offence to the next paragraph, by calling that bland stuff "music". It is not, and they know it. It might be entertaining to some, I do not doubt that. There are all kinds of people in the world, and wouldn't it be boring if we were all the same etc.
Much like an operette not being great art, but entertaining for sure (a guilty pleasure, I guess).
Re: Why would it recommend?
It's all right, apparently hip-hop is "music" too, so their definition of the word is obviously a lot broader than we're used to.
This reminds me of this investigation by Stereogum - why have some tracks that weren't even singles, B-sides in some cases, become the most listened to for some indie bands?
[1]https://www.stereogum.com/2105993/pavement-harness-your-hopes-spotify/columns/sounding-board/
Quick answer is.. "auto-play" picking these because they sound more like mainstream songs..so yeah it totally chimes with this article.
[1] https://www.stereogum.com/2105993/pavement-harness-your-hopes-spotify/columns/sounding-board/
Not necessarily bland, although ...
Is this strictly a "hard music" issue, or is that judgement merely based on the limited musical horizons of the researchers? Suppose I like 15th century motets of the Flemish but not the Italian schools, some but by far not all by Schubert and Joni Mitchell but only early Joni Mitchell (these not being being "bland" but certainly not "hard" either), would the AI be able to cope?
all of the recommendation algorithms are a failure
Amazon always thinks I need more of the same that I already bought. Netflix cannot understand that I want to watch different genre movies and TV series, not more of the same I watched latest. Spotify recommendations are also bollocks as my taste spans a whole lot of music styles.
Re: all of the recommendation algorithms are a failure
Agreed. Spotify things I like "Fallout! Boy" because I like "My Chemical Romance". They are sort of audibly similar, to a dumb computer, but not to squishy humans.
However, given enough time and enough data I suppose they might feasibly be able to find some great music for me. But not there yet.
TL:DR
People who like Lift Muzak will also listen to Hold Musak.
Not so much fans of Body Count (Ice T's Heavy Rock band for those who don't know).
Re: TL:DR
"Oh, we have both kinds of music, Country and Western..."
Re: TL:DR
"... Country music played on the radio so I turned it off... "
42 ?
No matter how many times you multiple the matrices together, the tautology applies : garbage in -> garbage out.
How many iterations does it take to converge upon "Celine Dione"?
False assumptions
I like Alt-J, Metallica, Tchaikovsky, Caro and hundreds (or thousands?) of other things.
Some of the things I like are "hard" Queens of the Stone Age (and I even like some Hip Hop from Public Enemy to NF).
Saying I don't like to listen outside my genre is utter rubbish.
Just recommend stuff that's "Good" because that's all I like to listen to.
But it has to be subjectively good (to me I mean) and not necessarily what the next person likes.
And some people (many people I think) don't like something because it doesn't fit into their self perception. I happen to also like some Britney Spears songs, in addition to the above... so good luck recommending things to me. But I know if I like it when I hear it.
Re: False assumptions
All of those composers / bands of Troubadours names contain the letter `a`. You are more predictable than you think. You probably really like AABA.
The study should rather say that people who listen to "hard music" don't like the yoghurt garbage that other people are fed with by mainstream media. These people are often much more open to different styles of music, they just don't like the commercial $hit produced by the so-called "music" business.
So it's racist
> AI recommendations fail fans who like hard rock and hip hop, say scientists
Leaving aside the hard rock, given the majority of hip hop listeners are non-white, I think it's clear this AI is just racist, like the people who programmed it.
Re: So it's racist
That might have been true when hip hop first emerged, but today I think the dividing line between those who like it and those who don't is probably one of age rather than race.
Re: So it's racist
As above ageist more like.
Napalm Death
Do you actually need anything else?
Re: Napalm Death - Do you actually need anything else?
Somewhere nice to go in the afterlife?
Spotify's spotty recommendations
I never bother with recommendations. Spotify (which I don't use any more) used to recommend either music that I had never heard of, or music that I did know but hated. Given the latter recommendations, I never dared listen to the former.
Well DUH!
So it's easier to recommend songs to people who like the dull manufactured tripe that passes for pop music? Well DUH! You just recommend another dull manufactured piece of tripe and there is plenty of that available. It's much easier to recommend stuff to people who like all their music to sound the same.
A decent play list for me would be Peter Gabriel, Bolt Thrower, Celine Dion (in French), Bach, Motörhead, Rush, Hans Zimmer, Hawkwind, The Prodigy, Yes, and Nile's excellent Rameses - Bringer of War to finish, but all that disparity is just going to give the AI a nervous breakdown.
The problem is lazily relying on AI playlists
If you're happy to be spoonfed "recommendations" generated by an algorithm for your listening choices, you'll have to be happy with whatever you're given.
If however you make a bit of an effort, you'll be able to find the good stuff, whatever your tastes.
Frankly the simple Youtube recommendations are better, because they're generated by seeing what other people who listen to a piece of music on there also choose and seek out, so the elements of personal taste and music knowledge get incorporated in the data.
Why would it recommend?
Hip hop to a metal fan?
Totally different genres.
The fan might like the odd tune, but then NWOBHM fans often are not keen on thrash.