India
13h
645
Why Worshipping Lord Ram Important in Hinduism?
Tech Industry
7h
1004
Racism towards Indians
Tech Industry
Yesterday
522
Do we really need Product Managers? 🧐
Tech Industry
Yesterday
1428
Switch to Roblox or stay at Google
Tech Industry
4h
641
Am I racist if I don’t want to marry a Southeast Asian? (Korean here)
It just gets worse and worse. I heard some significant changes were made. Can anyone share details?
I have no idea, but It has been recommending me to watch this guy in england who resoles dress shoes 👞. I have been barefoot at home for a year and obviously not interested in dress shoes.
Ah I know right. Recently he done something on Gucci shoes and people in comments were crazy like they can just listen him and it's a dream come true for them
The ads are starting to feel like the same weird ass sky mall products catalog that Instagram turned into. Like I’m looking at a Russian/Chinese product catalog of crap.
I am getting recommendations of life is a Russian village. Time to leave city?
You should check out Life of Boris instead if you haven't already. 😂 it's hilarious. Going by your username you might be familiar https://youtu.be/9X8VJvoo6y0
It’s optimized for clicks and not interesting content. So they just keep recommending the same garbage they think you will click on. Tragedy, it could be so great to surface novel content
Mine just keeps recommending same channels again and again. No diversity. Search is useless.
It’s a misinformation and radicalization engine on top of the other crap.
I feel their algorithms are optimized for improving centrality/locality of the video content . Make people watch videos which are already to the nearest edge server rather than pull interesting stuff from long tail distribution of the videos which might suit you but would be relatively expensive to google in terms of bandwidth. Most of the video platforms make highest profit on viral videos rather than one off videos customized for targeted audience.
It’s messed up n so out of way for me.. I am on the thinner side n all ads n recommendations r for obese n fat people n how to reduce weight ..
I think they still use co-visitation and not even like collaborative filtering. This is why I don't open random YouTube links people send me. You watch one, and your page is filled with the same garbage and you have to religiously click "don't recommend channel" or "not interested". Of all the things out there YouTube has the worst recommendation system and Tiktok has the best one from my experience Right now to compete with tiktok and reels they are pushing short vertical videos or older videos that are like 10-20 seconds long.
It’s a lot easier to train a recommendation model on 10 second TikToks with music genres than 30 minute complex YouTube videos
Not really. Live streams or smaller videos have way different content each time. So it's harder. That's why twitch uses Hypar as a combination of collaborative and content based recsys. They use preference ratios over weekly and monthly cycles. YouTube is the easier one to train on. Plus you have rating systems and viewer engagement to avoid cold start. On small ones everyday is a cold start. Channels pretty much have same user base , very less discovery, high loyalty and same established content. Longer duration of videos and huge history makes it even easier to specify which parts of the video are to be highlighted for ads. They just don't care. Google publishes interesting cutting edge stuff but they don't put any of that into production. Best example is Google translate. They still use 10 year old stuff instead of more recent DL approaches, which might be for scaling constraints but still, production level is ancient. Theoretical research is novel