This is more about getting feedback/criticism for the idea that fedizens should run programs that monitor what they engage with similar to how Facebook,YouTube, or any corporate online service would simply so you can use it to run algorithms for find content you might like.
The key thing is that you only share the your data consentual at this point the data will stay locally stored.
I try to get a pulse anti algorithms crowd and it’s hard to figure out if they’re against algorithms generically or just specific platforms. People leave twitch because that also pushes people to be more commercial or capitalist.
The idea came up when I built peertube recommendation algorithm I got the part built for tracking watch time on peertube videos and I thought “track YouTube watch time and engagement to make peertube suggestions”.
I like the idea. I’ve even experimented with some on device algorithms by basically weaving together feeds, even something simple like makes a noticeable improvement
I prefer actual recommendations from real people. That’s how I’ve found most of my favorite books and videogames.
Music is a little different. I do occasionally hop on Pandora to let their algo recommend new things. I also listen to indie radio and pull from there. But I still rely on stealing songs I hear friends play, if not specific recommendations.
Its true. But there was this brief, beautiful moment, during the early internet, where you could get good recommendations from YouTube…
Why is it when I miss a day of Daily Show it’s eviscerated from my feed entirely, but no matter how many Jordan Peterson thumbnails I hastily pass over, they still appear week in week out?
To date I’ve watched one JP video: Matt Dillahunty verbally tearing him a new asshole.
The right wing rabbit hole is real and, even if you’re explicitly against it, YouTube is like “if I can get him to watch just one …”, because they know they’ll have you watching videos like the guy from Clockwork Orange in no time. It’s disgusting.
Pro tip: never press “do not show me this” on right wing videos, just skip over them. That engagement causes the algorithm to mark you as emotionally triggered by politics and they’ll just shovel more of the same in your face. For them it’s just a roll of the dice, and the value of them flipping you is too high to avoid.
A yo dawg algorithm?
Hey I’ve seen you watched 1 reaction video, I’ve made your entire feed reaction videos!
I don’t like recommendation algorithms. They’re like processed food for the mind.
The idea that you could do this isn’t a bad one.
It could be interesting, to build a generic recommendation algorithm. Would it be like a browser add-on?
Something like a news feed, but you have all the data under your control, so you could see why something was recommended.
I tried the Reuters app a while back, I couldn’t deselect sports as a category of interest… Why the fuck not? Why do they care of I like sports or not… Uninstalled pretty quick.
It’s a browser extension for chromium based browsers like brave.
https://github.com/solidheron/peertube_recomendation_algorythm
I tried to make the alfo as simple as possible. It matches the words of the videos descriptions title and tags to other other videos words and ranks them via cosine similarity (de-ranks then via if the video has been seen before)
I should add a blacklist for keywords because I’m kinda sick of all the Linux content and there’s a lot of Linux content on peertube
I think that’s called “having a personality.”
I’m lost
