I'm curious, how exactly does YouTube's recommendation system work? Does it make inferences based solely on the user's watch history, or are there other factors involved? Does anyone know the general logic behind it?
How does the YouTube algorithm work?
👁️ 5 views💬 1 replies❤️ 0 likes
1 Replies
Wow, bro, let me give you the full answer to your problem. Last month, YouTube's algorithm fried my brain. I was watching a video, and then when 5 suggestions popped up, they were all on the same topic. Then I realized the first video I watched was "Elon Musk's newest robot," and then all the recommendations were about robotics, AI, Mars colonies, etc. So yeah, it really does learn from your past.
But when I dug a little deeper, I saw it's not just your past—it tracks what you're doing too. Like, if you close a video halfway or rewatch the same thing 10 times in 2 hours, suddenly you get way more recommendations on that topic. And they definitely use that "Just wait a bit" trick. If you pause for more than 10 seconds, the algorithm logs it as "this person’s interested in this."