Long-form on youtube in general is buttblasted by the algorithm in general, especially if it's politics related. It's not just the length but also the frequency of uploads. Youtube really likes daily/weekly/consistent upload type thing.
Also, I'm pretty sure the fact that lefty stuff have more chances of saying touchy/bad words in an effort to show societal problem XYZ also has a negative effect.
Another thing is that, unlike the quartening and shit who have the typical BIG LETTERS + FACE + SUBJECT down to an art, a lot of lefties really aren't on top of their thumbnail game. Especially long-form dudes who generally do their content solo without much outside help.

Not all long-form gets buried though. Joseph Anderson's stuff gets on my feed when he release new stuff it appears to me as well, even when I'm not subscribed. His last two have all released on my feed when they went live and they're 4 hours and 5 hours long for Witcher 1 and 2 respectively. Although the combination of video game discourse + Witcher might just be too stronk, youtube has no choice but to stan or something.

Not a contradiction. Thor was a bit clumsy with his wording, what he meant is that modern AI and algorithms are based on complex machine learning procedures which do not produce any readily interpretable output anymore, as opposed to old-school models like, e.g., parametric regression or decision trees.

These days, some variation of Deep Neural Networks is typically used. In these algorithms, you have multiple hidden layers stacked on top of each other and it's really difficult to understand how exactly information is processed between the input layer (data google/youtube have on you) and the output layer (which search results or youtube videos are recommended to you). The inner workings of these algorithms are most commonly black boxes. However, this is a feature of the method and does not mean that the researcher is unable to understand the way his model is set up or which effect the choice of a particular structure or hyperparameter will have on the outcome produced by the algorithm.

So, for example, a google engineer could very well tell you "if I increase this hyperparameter, customers are more quickly classified and then more aggressively steered toward radical content within their category. And this other parameter right there controls whether popular or controversial content is preferred." But those things require an actual understanding of the underlying learning algorithms, you won't find a line in the source code which reads "toggle CensorConservatives TRUE".

wait
u think sandy hook was legit?
LOL?

i bet you think planes went into the towers, too
wow hahahahaha

My YouTube remains based and chalk full of furries.

Lucky

I just had to go through and find as many Dave Rubin videos on my feed as possible and click "do not recommend this channel"

I avoid all murdoch media like the plague and my youtube is never free of sky news clips, ever.

Yeah lol

I have to actively seek out leftist media and I cant avoid right wing media no matter what.

Clicked on my you tube, scrolled a bit.

3 sky news, one abc, i actually look at abc links now and then, i NEVER click murdoch media.

Ruschia... Tchina... Everybody is envious you know ...