3 ms·
1) I think that YT could make a legitimate case that its recommendation algorithm is neutral (assuming that it's not actually biased by backroom payments etc.).
by PaulDavisThe1st 5y ago
1) I think that YT could make a legitimate case that its recommendation algorithm is neutral (assuming that it's not actually biased by backroom payments etc.). That is to say: it doesn't represent any point of view about anything, and simply tries to show a user "more videos related to the one you're watching". The devil, of course, is in the details of what "related" means.
2) I don't agree that not-on-YT means no wide audience. If I have 2M twitter followers, and I post a link to a video on (say) Vimeo with a sufficiently click-baity description, that video is going to get "a wide audience" (or at least, a large one). There are ways of alerting people to a video's existence beyond YT's own algorithm. The original (hah!) meaning of "went viral" didn't mean "YT recommended it to lots of people and they all clicked". It meant "link got shared by lots of people in an ever-expanding tree of contacts". That can still happen.
3) What's hard are the videos that fall in between the cracks. Not "my kid's 3rd birthday party singalong from last week", and not "the latest video from whomever the current k-pop phenom is". Videos that get, say, 200,000 - 2M views.
4) It's unclear how much view counts on YT are impacted by channel subscriptions vs the algorithm right now. Supposing for a moment that they are heavily correlated with subscriptions, YT could drop the algorithm and not see gigantic shifts in view counts. But I have no idea whether that's true, and I can think of lots of reasons why it may not be. A service that only shows/points you at videos from channels you've subscribed too is radically different from the YT of today.
- loup-vaillant 5y ago> I think that YT could make a legitimate case that its recommendation algorithm is neutral They’d have to remove shadow banning at the very least. I’m also virtually certain that their algorithms are machine-learning based, and as such very difficult to inspect, debug, and other ML based algorithms have proven to be biased in other contexts (like facial recognition). > I don't agree that not-on-YT means no wide audience. I… stand corrected, I guess. Also, I’ve just discussed it with my partner, and she pointed out a marginal exodus away from YouTube, which may amplify and chip away at their dominance. > A service that only shows/points you at videos from channels you've subscribed too is radically different from the YT of today. It would be indeed. YouTube is significant in the way it merges 3 tools together: hosting, search, and recommendations. They could be separate. (By the way, Google itself tends to merge search and recommendations, the famous "filter bubble".)