The problem is absolutely the YT algorithm. You can start them on decent stuff but if they follow the gradient they will end up in a gutter of absolute trash.

The problem is absolutely the YT algorithm. You can start them on decent stuff but if they follow the gradient they will end up in a gutter of absolute trash. I feel like this experiment in rapid feedback optimization for engagement has yielded a truth about the human condition that ought to become one of those "laws": trash maximizes engagement. It sounds pessimistic and awful but I think it's one of those things that makes sense in retrospect. Consider the following: Let's say two people walk past you on the street. One of them says 'hi' and you have a brief light chat with them. The other smears themselves with peanut butter and starts clucking like a chicken and saluting Hitler. Which maximizes engagement? This implies that any system that optimized for maximum engagement will converge on some kind of trash. This is interesting: https://www.reddit.com/r/science/comments/1bn8gfq/researchers_uncover_pornification_trend_among/ https://www.reddit.com/r/science/comments/1bn8gfq/researcher...