My experience has generally been that any song recommendations by the algorithm are completely wrong. Even when I build the playlist with a certain vibe and then request suggestions from the algorithm, it will go into left field and pick the most awful choices to fill out the list.
You would think that, with all this data they are keen to collect, they would have some level of understanding of vibes. Not direct understanding, obviously, but I would have thought them capable of cross-referencing various listening profiles to suggest better music.
It’s legit one of the things I could see AI actually being good at and used ethically for. I guess the fact that it’s not used for better taste profiling is a small mercy, considering the technology would be used for more nefarious stuff elsewhere if it was any good.
A prime example: the song “The Hoodin’ of Miss Fannie Deberry” by Kenny Rogers is decidedly different from the rest of his catalog. No matter how hard I try to get Spotify to find songs with similar vibes, I can’t escape Kenny Rogers and general country tunes. It’s maddening.
Any given song radio will trap you in a decade of music/genre, and pay no attention to the vibe of a song.
It has gotten slightly better recently with making custom playlists for me, but once my hyper-focus changes, I’m sure it will be out of sync with me again. Frontier Psychiatrist Radio slaps, but I’m still trapped in the '90s-'00s.
This has also been my experience (hence the meme). I’ve taken to finding music via human recommendations, as well as finding music that is related to another artist (for example, songs featuring another artist, or recommendations from artist, like their inspirations and stuff).
Another thing I’ve taken to is listening to artists’ entire discographies, which has been really great for me so far.
Oh, and internet radios are cool. Especially really niche and indie ones.
That was largely my experience as well. Things like last.FM or listen brains would tend to give more relevant results at first. It certainly didn’t happen overnight or even soon. But the algorithm through YouTube music has gotten really good for me. I have a wide listing range but all of it still pretty niche and eclectic.
But I’m not going to lie that took it a few years to be able to do. I don’t know if they just improved their algorithm in that time. Or if my listening habits have helped it. But I’m pretty happy with the recommendations and things it brings up. The thing I try not to do as a rule of thumb. Is to thumbs down or down vote something. I have found that on tracks I don’t like if I skip them early and often. It picks up on the fact.
Again this may have changed I remember Pandora back in the early 2010s. Up voting the stuff I liked, downvoting the stuff I didn’t. It definitely got to a very narrow feedback loop with extremely poor discovery.
My experience has generally been that any song recommendations by the algorithm are completely wrong. Even when I build the playlist with a certain vibe and then request suggestions from the algorithm, it will go into left field and pick the most awful choices to fill out the list.
You would think that, with all this data they are keen to collect, they would have some level of understanding of vibes. Not direct understanding, obviously, but I would have thought them capable of cross-referencing various listening profiles to suggest better music.
It’s legit one of the things I could see AI actually being good at and used ethically for. I guess the fact that it’s not used for better taste profiling is a small mercy, considering the technology would be used for more nefarious stuff elsewhere if it was any good.
A prime example: the song “The Hoodin’ of Miss Fannie Deberry” by Kenny Rogers is decidedly different from the rest of his catalog. No matter how hard I try to get Spotify to find songs with similar vibes, I can’t escape Kenny Rogers and general country tunes. It’s maddening.
Any given song radio will trap you in a decade of music/genre, and pay no attention to the vibe of a song.
It has gotten slightly better recently with making custom playlists for me, but once my hyper-focus changes, I’m sure it will be out of sync with me again. Frontier Psychiatrist Radio slaps, but I’m still trapped in the '90s-'00s.
This has also been my experience (hence the meme). I’ve taken to finding music via human recommendations, as well as finding music that is related to another artist (for example, songs featuring another artist, or recommendations from artist, like their inspirations and stuff).
Another thing I’ve taken to is listening to artists’ entire discographies, which has been really great for me so far.
Oh, and internet radios are cool. Especially really niche and indie ones.
That was largely my experience as well. Things like last.FM or listen brains would tend to give more relevant results at first. It certainly didn’t happen overnight or even soon. But the algorithm through YouTube music has gotten really good for me. I have a wide listing range but all of it still pretty niche and eclectic.
But I’m not going to lie that took it a few years to be able to do. I don’t know if they just improved their algorithm in that time. Or if my listening habits have helped it. But I’m pretty happy with the recommendations and things it brings up. The thing I try not to do as a rule of thumb. Is to thumbs down or down vote something. I have found that on tracks I don’t like if I skip them early and often. It picks up on the fact.
Again this may have changed I remember Pandora back in the early 2010s. Up voting the stuff I liked, downvoting the stuff I didn’t. It definitely got to a very narrow feedback loop with extremely poor discovery.