Hacker News new | past | comments | ask | show | jobs | submit login

YouTube recommendations suck so hard.

I'm subscribed to all this stuff but it never seems to give me similar content, it recommends all the videos I've already watched.

If I happen to watch something new, like a People's Court segment, then all I see is People's Court stuff.

There never seems to be balance.

Then there's the problem where a new video (like Gourmet Makes from BA) gets posted on a channel I like. I'll watch it with my partner on his machine, but then YouTube just chokes up, wondering why I never click on it, and then shoves it in my face ad nauseum.

Like, their search engine doesn't even seem to scratch the surface of content they're holding. It seems to be really stupid machine learning or AI.

Why can't it show me balanced recommendations from everything I watch or subscribe to, toss in some new things which are similar/popular, and maybe differentiate between content likely to be consumed multiple times (music videos) vs stuff people don't often watch a rerun of?

They seem to have so much data they could work with. They have amazing engineers and Google's expertise in algorithms.

Facebook even had problems with news feed being overrun with low quality content early on but seems to have figured it our fairly well- at least in my experience on the site (I know there are big echo chamber problems over there as well, don't get me wrong).

Even a lot of my non-techie friends seem to complain often about how terrible YouTube's recommendations are.




I feels like YouTube's recommendations were better around 2010 to 2015. I think back then they recommended videos based on the tags a video had, seemed to work much better


I'm assuming there were less people adding fake tags to earn more views then as well.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: