Comment by smugglerFlynn
12 hours ago
> I make an effort to use Spotify to find and listen to albums, but it wasn’t built for this, and invariably find 90% of my listening happening on algo-generated playlists of songs that sound exactly like a song I like. I never learn the names of the songs or the names of the bands as the songs go by, and I fall in love with none of it… It just vaguely sounds like stuff I like. It sucks.
I don't think algorithms are to blame (bear with me).
It's that the word "discovery" internet platforms have started using for this kind of experience is very misleading.
What real discovery means:
- Spending more time and attention when selecting next artist
- Reflecting on what you like about the song/album, and why
- Taking time to curate your collection
- Exchanging thoughts with other people, and reflecting on their opinions
Platforms are selling you efficiency, in reality the've compressed above steps to minutes or even seconds.
This is not unique to music platforms by the way. Instagram spoon feeds you reels so you never actually reflect on anything - you don't have time for reflection, because content is coming. Instagram will say they've solved "content discovery" for you, which is good, right?
LLMs spoon feed you tons of data, leaving no room for reflection.
It is logical if you think about it: these platforms do solve accessibility, but they don't solve discovery, deep reflection or retrospection of the user. Why bother marketing things they _don't_ solve? So they oversell accessibility solution like they've solved everything else, while in reality their product teams spend literal zero time addressing the important things.
Unless you consciously prompt yourself to reflect and think (which takes x10 more time than just browsing content) you are missing out.
I've spent good 20 minutes reflecting while writing this comment. Could have been written by LLM based on a short prompt, right? But I write on HN not because I want for everyone to see my thoughts published out there - I write precisely because I want to _reflect on my own thoughts_.
I need help reflecting, not writing or discovering.
The idea of how algorithms and scrolling get in the way of contemplation is an intriguing idea.
I had some other similar ideas come up in that past that are similar to this concept.
https://aeon.co/essays/tech-vexed-how-digital-life-threatens...
I think I agree with this.
A possibly interesting quirk of it is that this is a fairly intellectualized description (specifically):
> - Spending more time and attention when selecting next artist
> - Reflecting on what you like about the song/album, and why
> - Taking time to curate your collection
> - Exchanging thoughts with other people, and reflecting on their opinions
Of a process that at the time could have been summed up as “chit-chatting with your friends and picking the next song.” I wonder what it costs us, that these sort of process have become something we have to actively reflect on and make an effort. In the past this didn’t feel at all effortful, it was just fun and the easiest way to get music.
This isn’t intended as a criticism of your line of thought, I think you’ve accurately described a good process. Just a thought about how the accurate current description somehow doesn’t quite match the feeling of the past.
Same thing happened with remote work. Things that were simple by-products of a face to face communication now need to be dissected and studied, and then carefully added back in, in order for remote environments not to suck. All the small talk, emotional check-ins, etc.
I agree, a focus on efficiency, immediacy, and quantity has lead us to a barren experience of discovery. Music streaming certainly has its virtues, it is a shame that they haven't made the discovery process better.
I wonder what it would look like to have a feature that elicited reflection, perhaps purely for its own sake but maybe also to help feed further discovery. You could have a player that didn't immediately start playing the next track but presented an interface where you could write notes or react to the song in a variety of ways. That reflection could deepen your appreciation for the song or help you put into words what you find missing. It would also be a much richer feedback for the system to understand what you are looking for and find the next song. We now have all these fancy tools and vector databases for a nuanced and meaningful search based on text content.
What I find most tiring about the status quo is that you have to skip through a bunch of tracks to find something that resonates. It seems mentally taxing and I can't help but think I may actually like a lot of these songs if I was in the right frame of mind to hear them.
At a verbal level LLMs are great: questions like "tell me about hip hop artists similar to MF doom" or "is there anything new like jefferson starship?" can be the start of great conversations. They will talk your ear off about what is going on with tracks like "Dangerous" off the Yes Union Album.
That's covered by What.cd's Gazelle "spider web of similar artists"
https://litter.catbox.moe/od8vcq.png
This is a great way to do it, especially when the LLM can actually get to know you. I've been working on a project that combines a persistent music expert LLM session with social listening, and gives the LLM access to YouTube so that it can find things and play them for you immediately. I've got it tuned pretty well now and I've made it available to the public at https://tunistry.com/
I haven't had much luck with LLMs. I can guess it works with famous artists but lots of other things work with famous artists. I asked it to find more tracks like "Hey Baby" (Deadmaus, Mellifresh) and they completely failed to even come close. I couldn't even get a similar vibe.
I asked Google's AI mode "a friend of mine likes "hey baby" by (deadmus/mellifresh) and wants to find similar tracks" and mainly suggested other deadmau5 and Melleefresh tracks (corrected our spelling) -- did recommended "Internet Friends" by Knife Part and "Exceeder" by Mason. I thought the first was a direct hit, the second is a little different but "sick" in a good way... It starts a little slow but the groove gets great once I get in.
2 replies →
[dead]