3 ms·

I think the "oh no human moderation would cost us too much" defense is a distraction. It's a good one, because it is working, but it is still a distraction. L

I think the "oh no human moderation would cost us too much" defense is a distraction. It's a good one, because it is working, but it is still a distraction. Let us suppose we have a program which checks for orthodoxy according to the YouTube Terms of Service against each video submitted. Nothing but a lack of having done the work prevents YouTube from adding to its "rejected" flag the following: 1) A list of timestarts and durations of the portions of the video violating the ToS. 2) For each item in that list, make it a tuple and add a reason -- which clause was violated? Then outputting that to the owner of the rejected video. Nothing stops them from doing this. Some set of words or images happened somewhere in the timestream -- the orthodoxy program has that timestamp as it churns through the video. And a specific clause was violated -- a particular word in a wordlist as an example, because there's no simple, non-composite function that just says "good or not good." This is really not a particularly high bar as an ask.

flavius29663 6y ago
That will never work. If you tell the user exactly what they did wrong, they will keep trying by changing very little in the original video until it gets past bots, but still keeps the illegal or "forbidden" message. I think youtube should not be allowed to remove videos for their content unless for legal reasons. Meaning it's a violation of the law, in which case they have to send the information to the police as well. And that, of course, should require a human stamp of approval. Of course, Youtube should be allowed to remove videos if they are too long, wrong format etc., but NOT for their content.
- tumetab1 6y ago
  > If you tell the user exactly what they did wrong, they will keep trying by changing very little in the original video until it gets past bots, but still keeps the illegal or "forbidden" message. I think there's a possibility to implement that handles the issue you describe. It's trivial to detect an user trying to circumvent content controls if he's trying that from the same account. When that is detected the user account can be put into manual mode review or automatically block new uploads for a while. If circumvent is tried with different accounts the current system already doesn't prevent that so nothing is lost.
- Mehdi2277 6y ago
  My experience having worked at content platforms on ml (so fb, tiktok, snap) the largest motivation for content removal is content that general audience would be uncomfortable with. Pornographic content is the easiest example. Should a video platform be required to show porn? Some platforms allow it with tagging others avoid it entirely and often try to remove gray things like promiscuous videos. Another big area is political content and how much of it to allow. A lot of people complain on hyperpartisan news content and fake news. How to moderate it is very platform dependent with some wanting to restrict a lot of it and others leaning to allow almost anything. Also I’d be extremely surprised if youtube has no human moderation. My experience has always been a mix of human and ai moderation. Facebook has like tens of thousands of contractors paid for moderation. They still use a lot of ai on top of that.
  - flavius29663 6y ago
    I think Reddit handles porn well. If you don't want to see porn images, they are just a NSFW picture in the thumbnail, and nobody forces you to click on it. If you want stricter parent controls, that is ok to censor much more aggressively for porn. As to political speeches, I find it completely unacceptable that a private company gets to decide what stays on and what not. For example just the other day Youtube threatened to shut down the Senate channel in Romania because of the speeches of one of the senators. She is a wacko, true, but she is a Senator in the Romanian senate and didn't violate any law, it's unacceptable that private companies get to decide what is free speech nowadays. The Romanian MPs have special protections in the law, and the most important one is the free speech. They have even more free speech that any citizen of Romania. But hey, youtube is above the Romanian Constitution. Who is going to hold Youtube employees accountable? What democratic or judicial process is there to have checks and balances on that? Just saying: move to another platform is not good enough. You can't cut off the phones of your political adversaries and just say: buy a cell phone with another carrier. If the platform is becoming overwhelmed with extremist content, then it's on to the platform to find ways to better show the most relevant content to each user (e.g. burying it), but outright banning non-illegal content should be illegal.
    - Mehdi2277 6y ago
slt2021 6y ago
this will make it very easy to develop adversarial algorithm and avoid detection by youtube's system
- at_a_remove 6y ago
  That would literally beat the current situation by a country mile. The goal isn't to perfect the YouTube flagging algo, after all ... an adversarial algo is a small price to pay.
  - colinmhayes 6y ago
    >That would literally beat the current situation by a country mile. I don't think google agrees.
    - pixl97 6y ago
      Google likely only agrees with whatever choice makes them the most money. Whatever decision they make will always be jaded by how they think it will affect their stock price.

flavius29663 6y ago

That will never work. If you tell the user exactly what they did wrong, they will keep trying by changing very little in the original video until it gets past bots, but still keeps the illegal or "forbidden" message. I think youtube should not be allowed to remove videos for their content unless for legal reasons. Meaning it's a violation of the law, in which case they have to send the information to the police as well. And that, of course, should require a human stamp of approval. Of course, Youtube should be allowed to remove videos if they are too long, wrong format etc., but NOT for their content.

tumetab1 6y ago
> If you tell the user exactly what they did wrong, they will keep trying by changing very little in the original video until it gets past bots, but still keeps the illegal or "forbidden" message. I think there's a possibility to implement that handles the issue you describe. It's trivial to detect an user trying to circumvent content controls if he's trying that from the same account. When that is detected the user account can be put into manual mode review or automatically block new uploads for a while. If circumvent is tried with different accounts the current system already doesn't prevent that so nothing is lost.
Mehdi2277 6y ago
My experience having worked at content platforms on ml (so fb, tiktok, snap) the largest motivation for content removal is content that general audience would be uncomfortable with. Pornographic content is the easiest example. Should a video platform be required to show porn? Some platforms allow it with tagging others avoid it entirely and often try to remove gray things like promiscuous videos. Another big area is political content and how much of it to allow. A lot of people complain on hyperpartisan news content and fake news. How to moderate it is very platform dependent with some wanting to restrict a lot of it and others leaning to allow almost anything. Also I’d be extremely surprised if youtube has no human moderation. My experience has always been a mix of human and ai moderation. Facebook has like tens of thousands of contractors paid for moderation. They still use a lot of ai on top of that.
- flavius29663 6y ago
  I think Reddit handles porn well. If you don't want to see porn images, they are just a NSFW picture in the thumbnail, and nobody forces you to click on it. If you want stricter parent controls, that is ok to censor much more aggressively for porn. As to political speeches, I find it completely unacceptable that a private company gets to decide what stays on and what not. For example just the other day Youtube threatened to shut down the Senate channel in Romania because of the speeches of one of the senators. She is a wacko, true, but she is a Senator in the Romanian senate and didn't violate any law, it's unacceptable that private companies get to decide what is free speech nowadays. The Romanian MPs have special protections in the law, and the most important one is the free speech. They have even more free speech that any citizen of Romania. But hey, youtube is above the Romanian Constitution. Who is going to hold Youtube employees accountable? What democratic or judicial process is there to have checks and balances on that? Just saying: move to another platform is not good enough. You can't cut off the phones of your political adversaries and just say: buy a cell phone with another carrier. If the platform is becoming overwhelmed with extremist content, then it's on to the platform to find ways to better show the most relevant content to each user (e.g. burying it), but outright banning non-illegal content should be illegal.
  - Mehdi2277 6y ago

slt2021 6y ago

this will make it very easy to develop adversarial algorithm and avoid detection by youtube's system

at_a_remove 6y ago
That would literally beat the current situation by a country mile. The goal isn't to perfect the YouTube flagging algo, after all ... an adversarial algo is a small price to pay.
- colinmhayes 6y ago
  >That would literally beat the current situation by a country mile. I don't think google agrees.
  - pixl97 6y ago
    Google likely only agrees with whatever choice makes them the most money. Whatever decision they make will always be jaded by how they think it will affect their stock price.