5 ms·
I just notice how messed up the search is very recently. One of the things that it does is that it uses search terms from previous searches in the results for a
by fimdomeio 6y ago
I just notice how messed up the search is very recently. One of the things that it does is that it uses search terms from previous searches in the results for a specific search, it's awful for people with diverse weird interests.
My example I was searching for "primitive garden irrigation systems" after searching for some "city building games". Half of my results were "water canals in city skylines". It's both useless for what I was searching before and what I'm searching now.
- dmos62 6y agoBut that's probably also how you get really interesting, yet somewhat-random-somewhat-related suggestions.
- Fiveplus 6y agoAgreed, I've discovered some really nice tangentially related videos/channels that way. That said, it should be a strict opt-in model presented on the settings dropdown not a Here-is-a-feature-we-think-you-will-like-so-you-have-to-use-it.
- doliveira 6y agoYouTube recommendations have gone to shit lately. I can't even find anything new, they always recommend me the same old channels over and over
- throwaway2048 6y agoIts even worse for me, they recommend me the same videos over and over again, even ones I've watched that very same day.
- retox 6y agoI don't have any proof but I believe it's because of all the stories in the press about scary "alt-right/conspiracy rabbit-holes" and how something _must_ be done. A YT rabbit-hole is exactly how the product should work; you watched these videos so I'll show you closely related videos because you're more likely to stay on the site and generate ad revenue.
- grawprog 6y agoThis comment prompted me to actually look because i've noticed the same kind of thing and it's frustrating as hell. It seems like a hassle, but I did manage to find the youtube watch and search history where you can remove individual videos and searches from their algorithms. It's not very user friendly and doesn't give a ton of control, but it looks like it could help at least keep those weird one time searches from cluttering things up.
- sshagent 6y agoThere does seem a lot of that kinda shit nowadays. Websites thinking they know what you mean. Just search what i typed pleased, maybe if its got zero results then sure...have a guess what you think i intended. So often you can't search entire phrases of things, it'll arbitarily pick one word from your search and show you results on that. So annoying
- itsoktocry 6y ago>There does seem a lot of that kinda shit nowadays. Websites thinking they know what you mean. And we pretend it's "AI". It's horrible.
- Jommi 6y agoWell I think you both need to take a step back and be more objective. How would treat such an occurrence of it was your product? Surely they would look at the data and supposedly, it could be that 90% times it actually is a great predictor of intentions, and these are just edge cases. Edge cases will always happen. ALWAYS.
- kohtatsu 6y agoYes, I imagine most people will become engaged in a lot of videos that are offered, even if those videos aren't what they intended to watch. The edge case is not being directed and consumed by an algorithm for profit.
- ornornor 6y agoI rarely search YouTube directly so I haven’t noticed that. Instead I search for « whatever I’m looking for YouTube » on ddg and I don’t get these irrelevant results because YT doesn’t have my search history to mess with.
- robotnikman 6y agoI usually always open youtube in a private window or container tab when I need to search for something specific because of this.
- X6S1x6Okd1st 6y agoIt's likely a success if you only measure click though rate, i.e. someone loses focus of what they were trying to do because they are presented something similar that is very salient and switches to that context. I've certainly noticed that more and more with features like "frequently opened sites" YouTube's recommendations, a highly targeted ad etc
- jeppesen-io 6y agoYouTube is the least likely to look at just click rates How long someone watches a video is deeply tracked and even impacts creators revenue
- X6S1x6Okd1st 6y agoThat's fair, I think the scenario I'm laying out still holds though, i.e. that it's a win in E(sum(watch_time)|impression)