Click-through rates are indeed very important, but that's not all they optimize for. They are also looking at watch time, what you do after watching the video (do you watch more videos from the same creator, or on the same subject, something totally different, or do you leave the site), whether you interact with the video by liking or commenting, channels you have subscribed to, things you have searched for, etc.
And I think that when you spend a significant amount of time watching videos on a certain subject or from a certain channel - or when you repeatedly decline to watch videos of a certain type when they are suggested - you are signaling a very clear preference.
Are the videos the algorithm serves up something that people will "derive lasting benefit, memory or happiness from"? Probably not - but I also don't think that's what most people are looking for from YouTube. Sometimes, maybe, but more often they are just looking for mindless entertainment. Engaging with media on a deeper level requires effort. YouTube is where people go when even finding something to watch on Netflix is too much effort, let alone doing something healthy.
To keep this all in context - the parent comment was complaining that the algorithm doesn't promote videos of a guy building bike jumps in his back yard enough. I like Berm Peak - but is that something that most people would "derive lasting benefit, memory or happiness from"? No, it's not.
Anyone who hasn't seen those videos hasn't lost out on very much. And for anyone who has spent any amount of time watching videos about bicycles on YouTube - they probably have been recommended Berm Peak videos on numerous occasions. I know I have. The guy has 2.77 million subscribers and 787 million total views on his videos. Whether or not people actually watched the videos when suggested is more likely a matter of personal choice than the algorithm doing him dirty.