3 ms·

I find YT does a gradient ascent into more popular videos. I start out listening to a song with 1000 views, I get recommended 10k view songs, then 100k, and be

I find YT does a gradient ascent into more popular videos. I start out listening to a song with 1000 views, I get recommended 10k view songs, then 100k, and before you know it it's back in the mainstream. There's no way to surf a niche, to hunt in a pile of bin ends. I imagine it's similar for conspiracy stuff. It'll point you in the direction of the most convincing nonsense.

082349872349872 6y ago
If you search in non-latin scripts, maybe it still ascends into a mainstream, but it probably won't be your mainstream. (YT does much more cross-script indexing now than it did when I first started playing around with this, but it still tends to silo by writing system, and if it's relevant search works much better in the original script than relying upon auto-transliteration.)
read_if_gay_ 6y ago
If conspiracy stuff were the same way you describe music, you would eventually end up with the mainstream opinion, no?
- harry8 6y ago
  end up with the currently really popular conspiracy stuff. As opposed the nuanced, analysed, subtle thorough and less excitingly paced stuff. 1) CIA is interefering in US elections with evidence-free accuasation of Russia helping candidates who might reign them in to follow the law. Conspiracy? Or something to it? 2) Russia is paying bounties to kill American soldiers. Conspiracy? Or something to it? 3) Hunter Biden might be a bit corrupt. Conspiracy? or something to it? 4) The President has been compromised by Russia. Conspiracy? or something to it? There will be incendiary stuff on all of the above absolutely full of utter nonsense. There will be reasonable stuff also assessing evidence and may even not come up with a clear and definite conclusion on one or another. Which of these will YT reccomendation push you to? I don't know.
- fabatka 6y ago
  Perhaps in that case the algorithm finds the local extremum, not the global.
- pjc50 6y ago
  Reality is too boring to film.
tartoran 6y ago
I'm big into discovering obscure music and unfortunately it's become very hard on YT, you're right. If I don't constantly search for stuff I end up on the mainstream over listened playlists which I really don't want.
dijksterhuis 6y ago
This is my biggest pet hate of all of YouTube. I remember waaaaaay back in 2009-ish hunting for some new electronic music. \begin{old_man_better_in_my_day} First video I went on had about 1000 views. Every "related" video had around the same (sometimes less). I spent 4 hours finding a bunch of new and relatively unknown artists because, rather than dragging me back to the 100k aggregate, I was allowed to explore the graph of "other people went on to watch this". These days I'm lucky to see any sidebar suggestions that have fewer than 100k views. \end{old_man_better_in_my_day} I think the problem is the "one-size-fits-all" approach can't possibly capture how I would find it useful (music discovery via graph/tree traversal) vs. someone else (e.g. regular sky sports football highlights).
amadeuspagel 6y ago
> I find YT does a gradient ascent into more popular videos. Doesn't this totally contradict the claim that youtube pushes people to the extremes?
- barrkel 6y ago
  It's within a category. I mean I don't get recommended music videos when I'm watching an episodic video podcast, nor vice versa. But within those categories I get heavily pointed at the most popular for the category. I hypothesize it's similar for conspiracy content.

082349872349872 6y ago

If you search in non-latin scripts, maybe it still ascends into a mainstream, but it probably won't be your mainstream. (YT does much more cross-script indexing now than it did when I first started playing around with this, but it still tends to silo by writing system, and if it's relevant search works much better in the original script than relying upon auto-transliteration.)

read_if_gay_ 6y ago

If conspiracy stuff were the same way you describe music, you would eventually end up with the mainstream opinion, no?

harry8 6y ago
end up with the currently really popular conspiracy stuff. As opposed the nuanced, analysed, subtle thorough and less excitingly paced stuff. 1) CIA is interefering in US elections with evidence-free accuasation of Russia helping candidates who might reign them in to follow the law. Conspiracy? Or something to it? 2) Russia is paying bounties to kill American soldiers. Conspiracy? Or something to it? 3) Hunter Biden might be a bit corrupt. Conspiracy? or something to it? 4) The President has been compromised by Russia. Conspiracy? or something to it? There will be incendiary stuff on all of the above absolutely full of utter nonsense. There will be reasonable stuff also assessing evidence and may even not come up with a clear and definite conclusion on one or another. Which of these will YT reccomendation push you to? I don't know.
fabatka 6y ago
Perhaps in that case the algorithm finds the local extremum, not the global.
pjc50 6y ago
Reality is too boring to film.

tartoran 6y ago

I'm big into discovering obscure music and unfortunately it's become very hard on YT, you're right. If I don't constantly search for stuff I end up on the mainstream over listened playlists which I really don't want.

dijksterhuis 6y ago

This is my biggest pet hate of all of YouTube. I remember waaaaaay back in 2009-ish hunting for some new electronic music. \begin{old_man_better_in_my_day} First video I went on had about 1000 views. Every "related" video had around the same (sometimes less). I spent 4 hours finding a bunch of new and relatively unknown artists because, rather than dragging me back to the 100k aggregate, I was allowed to explore the graph of "other people went on to watch this". These days I'm lucky to see any sidebar suggestions that have fewer than 100k views. \end{old_man_better_in_my_day} I think the problem is the "one-size-fits-all" approach can't possibly capture how I would find it useful (music discovery via graph/tree traversal) vs. someone else (e.g. regular sky sports football highlights).

amadeuspagel 6y ago

> I find YT does a gradient ascent into more popular videos. Doesn't this totally contradict the claim that youtube pushes people to the extremes?

barrkel 6y ago
It's within a category. I mean I don't get recommended music videos when I'm watching an episodic video podcast, nor vice versa. But within those categories I get heavily pointed at the most popular for the category. I hypothesize it's similar for conspiracy content.