Spotify has a very cool feature called Community [2] where users can talk about what they want Spotify to do. The most upvoted ideas are things like "change my username" or "filter explicit songs." My personal wishlist is a "don't play this song again" button [3] and native Homepod support [4].
The point is that users aren't asking for Spotify to interpret their emotional state, they just want comparatively basic features.
Spotify would do better to get back to basics on song suggestion instead of trying to guess emotional state - Pandora still finds similar songs by musical elements better than Spotify. A good example to try is to select a song that is out of genre for an artist - an acoustic version of a song that usually doesn't do acoustic for example. The suggestions for similar songs from Spotify all have to do with the artist genre not the actual song.
I've given up even wanting extra features, I just want them to fix the glaring bugs they've had for years.
My current favourites:
- if you sort a playlist you're playing, then click the album art to find the currently playing song, it will skip to the position of the track in the _unsorted_ list
- playlists over a certain length just won't play via Spotify Connect. There's a couple of playlists I just can't play on my Sonos speakers.
- playing a daily mix hides the shuffle button, and yet it still uses the shuffle state you had selected previously. If you want to turn shuffle off you have to play something else, toggle shuffle, then play the daily mix again
I've tried reporting these. They made me jump through a thousand standard troubleshooting hoops despite them all happening on every device I own. Then a few days later they came back and said it was "working as intended". Mind blowing.
Agreed. I dont need Spotify to try to understand my mood, I actively dont want them to.
I DO need them to understand (or give me a manual way to tell them) which playist is for my 4yo and stop trying to play me their songs when I’ve selected a playlist for me. Drives me nuts.
How about this very reasonable request to "Remove current Song from Playlist" [1]? Filed in 2012, status shows as "Not Right Now", has collected 1343 votes and counting... They're even worse than Atlassian.
I really hate Spotify for everything except their pretty complete library and reasonable pricing. Discovery, searching, UI, everything just plain sucks, I miss 8tracks.
Yeah, I gave up on my subscription after I couldn't figure out how to list all the songs by a given musician. It's a simple feature that's part of all the music software that I've used before: Winamp, iTunes, Amarok, Gnome Music. ¯\_(ツ)_/¯
Sounds like they've got plenty of people to work on the shiny new things but nobody to maintain the boring stuff. It's a shame because there are loads of engineers out there who enjoy maintaining stuff over long periods of time and generally doing the boring stuff that makes customers happy. These engineers are really undervalued in companies like Spotify. I think the success of companies like Microsoft is due in a great deal to valuing engineers who do the boring stuff.
As a user, I agree that features which give users control and agency are preferable to increasingly opaque ML. But as business decisions go, this argument is terrible. The subset of users who bother to post new ideas to the community forum won't be large enough to affect Spotify's bottom line, and these users are almost certainly not representative of the average user (who probably doesn't realize the community feature exists).
Spotify isn't optimizing for placating the vocal minority of power users, they're trying to increase their paid subscription count. It would be great if we lived in a world where these were the same thing, but they clearly aren't.
Maybe because people who want this aren't their customers.
I've argued that recommendation system for music are near useless for years because they don't group music together by what I actually want to hear together, and that it needs to be more careful about interpreting my actions because what goes together varies by mood and purpose.
E.g. you can't just take genre,or artist, because the music I want before bed need to be slow, soothing and familiar. Transitions between tracks also need to be much less abrupt than other times of day.
Other times I want something new and surprising.
Or loud and upbeat.
Until they convince me they can do that they're not getting my money.
This trend towards Emotional and Sentiment based Analytics - the acquisition and monetizing every snapshot of our personality. This reduction of our humanity into a cluster of data points to be used against us by marketing and political campaigns - this is just gasoline on the flames of Silicon Valley techno backlash.
People like Affevtiva running facial micro expression algorithms etc. What happens when all this data is complied by your friendly neighborhood Data Broker in the name of a "Better Customer Experience" with realtime FitBit heartrates/sleep quality from Google with Amazon applying tone of voice stress analytics with Apple Pay's recent purchase history on top of .........
I am writing this inside a Facebook VR headset with eyetracking hardware and can't be sure my 'gaze' is not something I have agreed to have monitored in some EULA.
If they get these neuro brain machine interfaces to the point they work - I can only imagine the outcome without any established boundaries.
And just how can this industry establish boundaries ? Who decides and who decides who decides ?
The 'redeeming' aspect of this is that it is doomed to fail.
The machine learning systems here are just 'automation by averages' -- they are incapable of measuring any variables which could actually estimate your emotion.
Emotions are 'conceptualized bodily states' that the brain adopts in order to atune our motivation/etc. to our environment.
Indeed it isn't clear there are any devices we can attach to computer to even take measurements. At best, something like a continuous webcam feed would improve accuracy a little from "almost coinflip", and only for quite extreme emotions.
The problem with these systems is precisely that we are fed 'averages'. NOT that they are 'capturing sentiment' (ad-tech propaganda to sell their apps and ads).
My concern is that this propaganda is leading to people changing their behaviour: unaware they are being fed 'generic content' for some accidental, meaningless 'singanal' they sent, their habits/emotions/self-perception is being distorted to fit the stupidity of the average.
Stuff like this will just have me on bandcamp exclusively. Spotify has been a terrible experience, but for the sake of my teenage son, we decided to get a family account. Spotify is too obsessed with telling me how to consume music, I don’t need an album shuffled by default, the artist picked the track listing for a reason. The way Spotify cues music is bizarre. Their user interfaces are confusing and clunky. For a platform trying to go all in on podcast it’s near impossible to know when new episodes are released (unless I’m missing something). I don’t need something to spy on my mood and suggest music, I’ve figured that out - Dub techno while working Ambient while reading HipHop when hanging with the homies and Slayer when I’m working in the garage!
> I don’t need an album shuffled by default, the artist picked the track listing for a reason
I don't get you. Isn't the shuffle button on or off based on the last time you changed it? If the shuffle is on, and you play an album, then you'll play it shuffled. I'd much rather that than Spotify trying to predict whether I want to play something shuffled or not, and auto-changing the button status.
>> I don’t need something to spy on my mood and suggest music, I’ve figured that out - Dub techno while working Ambient while reading HipHop when hanging with the homies and Slayer when I’m working in the garage!
Out of topic a bit but this reminds me of the famous dictum, attributed to Charles V (H.R.E.): I speak in Latin to God, Italian to Women, French to my Men, and German to my Horse.
Honestly, I don't think this is a good idea. It opens one up for manipulation. You feel this way and Spotify will suggest you something that makes you feel rather that way.
It doesn't matter if it's one or more songs meant to support these feelings or counter them, I don't think this is a good for people in the long run. That's not even considering the fact that they want to basically spy on people.
My first thought was what accountability or required liability will there be? Laws are nowhere near caught up to be relevant or significant for the times, and so the answer is there will be no accountability, liability, understanding or responsibility for the impact of their algorithmic decisions; we know the experiment Facebook did with showing more negative or depressive posts had on people, yet nothing happened to them for this.
Unrelated to Spotify discussion, I remember quite a few years ago that there was research on algorithmically detected fatigue in human voice, the scope of the research was to implement a means of fatigue warning for airline pilots and ATC staff.
My point there is that there is absolutely intrinsic value in researching in this field, but it just seems to be rather misguided whenever companies suffer from a high VC valuation.
Beepity-boop beep! Emotional state detected: Depressed!
Suggesting playlist:
1. Nine Inch Nails – Hurt
2. Michael Andrews – Mad World
3. Klaus Nomi – The Cold Song
4. Johnny Mandel – Suicide is Painless
Did You Know: This playlist has been brought to you by analyzing frillions of Spotify users with advanced AI and selecting the optimal songs matching any emotional state.
When in reality it'd be arguably best to direct them to MDMA-assisted therapy in therapeutic setting and/or invite them to a medically supervised recreational dance party with MDMA and listing a selection of songs that may be played X/Y/Z DJs; not sure any nation's health system is anywhere nearly understanding or accommodating enough yet to offer our facilitate this, especially due to indoctrination and regulatory capture, for-profit interests, etc.
Urgh all music recommendation engines are broken as soon as you go out of like the top 10 genres and big labels. Listening to folk music is basically “7 degrees of Beyonce”. Start with any English folk song, start the radio feature on Spotify, and it won’t be long until it swaps to Taylor Swift, and then the next step is Beyonce. It’s like they forget about where you started and only base each suggestion off the last track.
Anyone remember Songza or Grooveshark? That was so beautiful. Makes Spotify and Pandora look a fool. What Spotify or Google Music Apple Music should be doing is what Songza was doing. Time of day, weather, genre, what do you want to do? -> Go
Companies file patents all the time; they aren't typically indicative of strategic direction. Having a strong patent portfolio is important for defense (it's like carrying a huge knife down a dark alley) so companies horde them. Getting support for filing patents is also a side perk for employee retention since most people don't have the resources or desire to file patents on their own dime but they look good on your LinkedIn page.
Well, the company who owns the rights to that song you skipped certainly has a financial motive to figuring out if there's somehow they can get you listening to it again so they can collect more royalties from it - so they may be willing to pay X for that chance, times X by Y for the number of profit-driven rights owners and that likely adds up to a decent, new revenue stream; industrial complexes are dead, long live industrial complexes!
This is insanity. I'm glad I gave up on this streaming nonsense a while ago when I realized it made me listen to music the platform wants me to listen, instead of music I want to listen.
Thank you for sharing the article. This sounds like an incredible disregard for privacy. Heavens, what happened to curating your own music playlists?
Most companies aren’t able to even keep my email safe.
This is not the kind of information I would entrust to anyone, let alone a music playing company.
Reminder to check your email periodically on https://haveibeenpwned.com/
Scary? Hardly. Most of the time Spotify's recommended music is just other songs by my favorite artists that I've already listened to 100x, on their platform. If they can't predict what I will like outside of the artists/genres I already listen to, I doubt they can make any predictions based on infinitely more noisy data points.
They can't even differentiate artists with the same fucking name.
The link returns a 500 Internal Server Error - maybe it was the HN hug of death, but this post doesn’t seem big enough yet for that? Either way, I hope to read this when its back online. Seems interesting, and also scary.
I read this on Pitchfork [1] couple of days ago, their source was Music Business Worldwide [2] and with a provided US patent: Identification of taste attributes from an audio signal [3]
Yet another automation that is not needed by the user. As a designer I get it - it is growth hacking. For some reason I think this will backfire, music is personal experience without clear emotional boundaries. Example: Current emotional state: Happy. Reason: Good memories with my girlfriend. Song: Hurt by NIN. If a song has emotional signature you cannot predict with certainty that it will be in unison with emotional state of the listener.
Link is dead for me. Does anyone know how this works? Is it: Spotify asks how I'm feeling, I click the frowny face, and then it recommends me some sad songs?
> Streaming music platform Spotify has won a patent enabling it to snoop on users’ speech and even background noise in order to gauge emotional state and location type to serve up the appropriate soundtrack. Not creepy at all!
> Spotify has received a patent that will allow it to use speech recognition and sound analysis to assess a user’s demographic attributes, determine their emotional state, and even glean insight into their location. The information will be used – hypothetically, at least – to pick the perfect song to play without requiring any conscious data input from the listener.
Straight out spying, basically. It's what people were warning about with Siri, Cortana or whatever it was called. And it was dismissed with "I don't care, I'm not doing anything illegal and I have nothing to hide".
This is one of those features that is good marketing because it makes headlines about their product.
It doesn't have to actually work particularly well (beyond what you can get with a pretty obvious guess), and probably won't. But it captures the media's attention and the public's imagination, just to claim that you're doing this.
[+] [-] millstone|5 years ago|reply
Spotify has a very cool feature called Community [2] where users can talk about what they want Spotify to do. The most upvoted ideas are things like "change my username" or "filter explicit songs." My personal wishlist is a "don't play this song again" button [3] and native Homepod support [4].
The point is that users aren't asking for Spotify to interpret their emotional state, they just want comparatively basic features.
1: https://www.researchgate.net/publication/342854806_Just_the_...
2: https://community.spotify.com/t5/Ideas/ct-p/newideas
3: https://tinyurl.com/y52qtavt
4: https://tinyurl.com/y3hlw6ky
[+] [-] dv_dt|5 years ago|reply
[+] [-] IneffablePigeon|5 years ago|reply
My current favourites:
- if you sort a playlist you're playing, then click the album art to find the currently playing song, it will skip to the position of the track in the _unsorted_ list
- playlists over a certain length just won't play via Spotify Connect. There's a couple of playlists I just can't play on my Sonos speakers.
- playing a daily mix hides the shuffle button, and yet it still uses the shuffle state you had selected previously. If you want to turn shuffle off you have to play something else, toggle shuffle, then play the daily mix again
I've tried reporting these. They made me jump through a thousand standard troubleshooting hoops despite them all happening on every device I own. Then a few days later they came back and said it was "working as intended". Mind blowing.
[+] [-] ssss11|5 years ago|reply
I DO need them to understand (or give me a manual way to tell them) which playist is for my 4yo and stop trying to play me their songs when I’ve selected a playlist for me. Drives me nuts.
[+] [-] airstrike|5 years ago|reply
Obligatory “If I’d asked customers what they want, they would have told me a faster horse.”
People may not realize the value of this new feature because they have not even imagined it.
[+] [-] tda|5 years ago|reply
I really hate Spotify for everything except their pretty complete library and reasonable pricing. Discovery, searching, UI, everything just plain sucks, I miss 8tracks.
1: https://community.spotify.com/t5/Live-Ideas/Mobile-Remove-cu...
Edit: I've been hoping for years 8tracks would be back, and low and behold, they have relaunched: https://blog.8tracks.com/2020/04/19/welcome-back-8tracks/
Edit2: not really, still not available outside of USA/Canada
[+] [-] highwind|5 years ago|reply
[+] [-] globular-toast|5 years ago|reply
[+] [-] dlkf|5 years ago|reply
Spotify isn't optimizing for placating the vocal minority of power users, they're trying to increase their paid subscription count. It would be great if we lived in a world where these were the same thing, but they clearly aren't.
[+] [-] vidarh|5 years ago|reply
I've argued that recommendation system for music are near useless for years because they don't group music together by what I actually want to hear together, and that it needs to be more careful about interpreting my actions because what goes together varies by mood and purpose.
E.g. you can't just take genre,or artist, because the music I want before bed need to be slow, soothing and familiar. Transitions between tracks also need to be much less abrupt than other times of day.
Other times I want something new and surprising.
Or loud and upbeat.
Until they convince me they can do that they're not getting my money.
[+] [-] rapnie|5 years ago|reply
[+] [-] milofeynman|5 years ago|reply
[+] [-] ridaj|5 years ago|reply
[+] [-] jimmySixDOF|5 years ago|reply
People like Affevtiva running facial micro expression algorithms etc. What happens when all this data is complied by your friendly neighborhood Data Broker in the name of a "Better Customer Experience" with realtime FitBit heartrates/sleep quality from Google with Amazon applying tone of voice stress analytics with Apple Pay's recent purchase history on top of .........
I am writing this inside a Facebook VR headset with eyetracking hardware and can't be sure my 'gaze' is not something I have agreed to have monitored in some EULA.
If they get these neuro brain machine interfaces to the point they work - I can only imagine the outcome without any established boundaries.
And just how can this industry establish boundaries ? Who decides and who decides who decides ?
[+] [-] mjburgess|5 years ago|reply
The machine learning systems here are just 'automation by averages' -- they are incapable of measuring any variables which could actually estimate your emotion.
Emotions are 'conceptualized bodily states' that the brain adopts in order to atune our motivation/etc. to our environment.
Indeed it isn't clear there are any devices we can attach to computer to even take measurements. At best, something like a continuous webcam feed would improve accuracy a little from "almost coinflip", and only for quite extreme emotions.
The problem with these systems is precisely that we are fed 'averages'. NOT that they are 'capturing sentiment' (ad-tech propaganda to sell their apps and ads).
My concern is that this propaganda is leading to people changing their behaviour: unaware they are being fed 'generic content' for some accidental, meaningless 'singanal' they sent, their habits/emotions/self-perception is being distorted to fit the stupidity of the average.
[+] [-] 12bits|5 years ago|reply
[+] [-] ttt0|5 years ago|reply
[+] [-] swilliamsio|5 years ago|reply
I don't get you. Isn't the shuffle button on or off based on the last time you changed it? If the shuffle is on, and you play an album, then you'll play it shuffled. I'd much rather that than Spotify trying to predict whether I want to play something shuffled or not, and auto-changing the button status.
[+] [-] YeGoblynQueenne|5 years ago|reply
Out of topic a bit but this reminds me of the famous dictum, attributed to Charles V (H.R.E.): I speak in Latin to God, Italian to Women, French to my Men, and German to my Horse.
https://en.wikiquote.org/wiki/Charles_V,_Holy_Roman_Emperor
[+] [-] stjohnswarts|5 years ago|reply
[+] [-] MrYellowP|5 years ago|reply
Because many don't know the actual term for what they're analysing, I thought I'll post it here:
https://en.wikipedia.org/wiki/Emotional_prosody
Honestly, I don't think this is a good idea. It opens one up for manipulation. You feel this way and Spotify will suggest you something that makes you feel rather that way.
It doesn't matter if it's one or more songs meant to support these feelings or counter them, I don't think this is a good for people in the long run. That's not even considering the fact that they want to basically spy on people.
[+] [-] loceng|5 years ago|reply
[+] [-] DoingIsLearning|5 years ago|reply
My point there is that there is absolutely intrinsic value in researching in this field, but it just seems to be rather misguided whenever companies suffer from a high VC valuation.
[+] [-] mackrevinack|5 years ago|reply
[+] [-] unhammer|5 years ago|reply
Suggesting playlist:
Did You Know: This playlist has been brought to you by analyzing frillions of Spotify users with advanced AI and selecting the optimal songs matching any emotional state.Enjoy! :)
[+] [-] loceng|5 years ago|reply
[+] [-] robotmay|5 years ago|reply
[+] [-] blondie9x|5 years ago|reply
I miss this! http://appadvice.com/appnn/2012/06/songza-goes-universal-wit...
[+] [-] theklr|5 years ago|reply
[+] [-] approxim8ion|5 years ago|reply
[+] [-] hackitup7|5 years ago|reply
[+] [-] jerzyt|5 years ago|reply
[+] [-] loceng|5 years ago|reply
[+] [-] Kaze404|5 years ago|reply
[+] [-] dntrkv|5 years ago|reply
[+] [-] chordol|5 years ago|reply
[+] [-] designatedtruth|5 years ago|reply
[+] [-] dntrkv|5 years ago|reply
They can't even differentiate artists with the same fucking name.
[+] [-] spacephysics|5 years ago|reply
[+] [-] TomGullen|5 years ago|reply
[+] [-] spacefiish|5 years ago|reply
[+] [-] boraoztunc|5 years ago|reply
[1] https://pitchfork.com/news/new-spotify-patent-involves-monit...
[2] https://www.musicbusinessworldwide.com/spotifys-latest-inven...
[3] https://patents.justia.com/patent/10891948
[+] [-] stayux|5 years ago|reply
[+] [-] dorkwood|5 years ago|reply
[+] [-] ttt0|5 years ago|reply
> Streaming music platform Spotify has won a patent enabling it to snoop on users’ speech and even background noise in order to gauge emotional state and location type to serve up the appropriate soundtrack. Not creepy at all!
> Spotify has received a patent that will allow it to use speech recognition and sound analysis to assess a user’s demographic attributes, determine their emotional state, and even glean insight into their location. The information will be used – hypothetically, at least – to pick the perfect song to play without requiring any conscious data input from the listener.
Straight out spying, basically. It's what people were warning about with Siri, Cortana or whatever it was called. And it was dismissed with "I don't care, I'm not doing anything illegal and I have nothing to hide".
[+] [-] stjohnswarts|5 years ago|reply
[+] [-] nfoz|5 years ago|reply
It doesn't have to actually work particularly well (beyond what you can get with a pretty obvious guess), and probably won't. But it captures the media's attention and the public's imagination, just to claim that you're doing this.
[+] [-] anigbrowl|5 years ago|reply
[+] [-] bdibs|5 years ago|reply
https://pitchfork.com/news/new-spotify-patent-involves-monit...
[+] [-] pgt|5 years ago|reply
Customers don’t like having their speech recorded to be recommended “relevant content”.
What is Daniel Ek doing? Who files surveillance patents in the midst of the biggest privacy-oriented comms migration in history?