This is sad, I actually greatly miss the old homepage. Before personalized recommendations, when you were signed in, it used to show you what was up with channe

This is sad, I actually greatly miss the old homepage. Before personalized recommendations, when you were signed in, it used to show you what was up with channels you subscribed to; NOT just new videos they posted, but videos they commented on, liked, etc. It was sort of a social graph feed, but it worked very well for sort-of random-walking the site. The percentage of stuff you would see that is just random weird stuff was way higher and it made the experience a shit load more entertaining, even if the overall quality of the content was lower. It's pretty obvious of course, but even in the best case, the recommendation algorithm mostly just intensifies what's already popular anyways. That's all it really can do, because "personalized" recommendations are not magic, they're barely even clever. Everyone knows you watch one cat video and then suddenly you are bombarded with cat videos. Watch one SNL clip and it will throw more at you. It will do this for anything. The word "personalized" makes it sound super sophisticated, but no matter how sophisticated it is, the end result is the same: all it really does is show you popular content that is very closely adjacent to stuff you've already watched. It doesn't know why you've watched things, so if you don't carefully trim your YouTube History, the recommendations can sometimes just slowly clog with things you are absolutely disinterested in. That's really the best case scenario for the personalized recommendation algorithm though. The worst case is the other reality we face, which is the behaviors that it winds up rewarding. I mean let's face it, a major component to any recommendation algorithm is going to be your likelihood to click into the video. So is it any surprise that even some of the better channels on YouTube have fallen to horrid clickbait titles and thumbnails? "Top 5 worst didgeridoos (#3 shocked me!)" Terrible. You don't need a recommendation algorithm for clickbait to succeed, but recommendation algorithms amplify clickbait to the point where if you're not doing it to some degree you're literally getting penalized. No surprise that e.g. Linus Media Group has basically decided to constantly use terrible titles and thumbnails, it's literally what YouTube pays you to do. It's also clearly going to reward drama and other negative shit even more, as if people's natural tendency to gossip wasn't bad enough. In my opinion this is just another casualty of the way software developers have come to think about users and user interactions, not about quality but about quantity. Developers talk about increasing metrics, but they only have a slight connection to what this actually means. There's very little concern about whether these algorithms are actually improving the experience or not, only that it is increasing the numbers, so surely the experience is better. The only hint that a user even exists is the idea in here that "we're missing the opportunity to show users what they want." But did they even attempt to answer whether that was actually what they did? Now YouTube at least asks users about the recommendations, but I'm not very convinced they actually do anything interesting with the resulting data. I understand to some degree, though. I mean, YouTube in 2009 was almost certainly not on the path to solvency, and I am less sure every year if the web experiences I actually enjoyed could ever be sustainable. It's clear that websites which optimize for engagement and screen time are the ones that get all of the ad revenue, and the more shameless they are about it, the better it seems to work. I kind of hope some day, we can distill down what is wrong with these incentives so that we can penalize or even regulate them, because right now it just feels like websites are very manipulative to their users in ways that probably make their lives worse.