|
|
Post by Baron von Lotsov on May 5, 2024 9:23:41 GMT
I'm starting to see a trend on Youtube where often if you leave a comment which is several sentences long it will often get deleted on a contentious thread because the AI is statistically more likely to spot a word it sees as a non-Newspeak term. It's just another way that today there is a trend towards short comments. They are also liked a great deal more. A comment consisting of 3-4 lines would be lucky to see one or two other people like it, suggesting longer comments are rarely read. Since on Youtube you get channels which are mainly orientated towards certain nationalities you can see which nationalities do the short ones and which do the longer comments. I can't say for all nationalities, but in my experience I listen to a lot of stuff that comes from the UK, the US and China. Out of those three, the UK is shortest, and China is the longest. Indeed the Chinese prefer longer comments where they can get their teeth into them. Short throw-away slogans leave me unfulfilled. It feels like if you went to the bar to buy a pint and the barmaid poured you an inch in the bottom. One gulp and it is gone, leaving you thinking where is the beer.
|
|
|
|
Post by Orac on May 5, 2024 10:12:24 GMT
You used to easily be able to have a discussion with another user on Youtube.
However, from the google takeover onwards it became more and more difficult to have a simple back and forth
It was quite deliberate - you don't break something that works fine - to that degree - by accident.
The war was lost when people acquiesced and tried to fit their communication around these prize anuses. We now have a new generation who think opinions are things limited to 280 chars
|
|
|
|
Post by Baron von Lotsov on May 5, 2024 13:20:44 GMT
You used to easily be able to have a discussion with another user on Youtube. However, from the google takeover onwards it became more and more difficult to have a simple back and forth It was quite deliberate - you don't break something that works fine - to that degree - by accident. The war was lost when people acquiesced and tried to fit their communication around these prize anuses. We now have a new generation who think opinions are things limited to 280 chars It really is pissing me off that I am automatically prevented by the AI from making any intelligent point to the ones the media denounce as "Far Right".
I'm trying to reverse engineer the censor system, and have found a way to track it thanks to a bug in their system. It's a kind of Bayesian prediction system and it uses multiple proxies. I don't even blame YouTube for it because I know they have to follow British legislation, which is around non-moderated comments. It's legal on the part of Youtube to operate in Britain as long as they take "all reasonable steps to prevent" someone being unlawfully insulted, e.g. by using an abbreviation for Pakistani and that kind of thing. Then they are legally able to keep that comment published if it goes undetected until it is brought to their attention via the report button, whereupon if it is an illegal comment then it must be removed in a reasonable time after reporting it.
Now if you were Mr Youtube you would know pre-moderated comments would never work. If they did it that way the subscription costs would be so high the platform would be like an elite forum, like the WEF types. So in practice the only way they can operate is by AI, and the way you do that is you have to get the AI to chase a goal. You are free to set the goal or even multiple goals and how much each goal should be influencing the decisions, but one goal you can not skimp on is the law. It's a matter of probability. If you set the AI only just strict enough normally it will leak some illegal ones through and then Youtube would be up on the firing line. They could end up being banned from trading in the UK. For this reason they have to set the AI extra strict , as per look at it like a Gaussian. If you are say 3-4 standard deviations then you are OK, but one is no good.
So how would you do this. You would need a proxy for illegal comments. The AI can learn the report button is a good proxy. The more the posts are reported the more the content of the reported ones is marked as potentially illegal. It will cross-reference the words in one reported post to another and see which words cause the best predictions of illegal, i.e. how good each word is as a proxy. Likewise it can also use the channel as a proxy to illegal content.
Now suppose you have a group like Stonewall who go around reporting anything negative said about LGBT, such as mother writes, I feel uncomfortable my child is taught LGBT at school. This is a perfectly legal comment, but Stonewall won't like it so they report it. The 50c/hour mod from India hasn't a clue about British law so lets that get deleted. Now the AI is learning to delete political content.
|
|
|
|
Post by Bentley on May 5, 2024 16:36:17 GMT
YouTube doesn’t exist to serve you chaps . You are the commodity.
|
|
|
|
Post by Baron von Lotsov on May 5, 2024 17:45:13 GMT
YouTube doesn’t exist to serve you chaps . You are the commodity. They tweak the AI to throw out or demote negative comments.
It's that American commercial thing. All punters must be blissfully ignorant, much like it is at Mc D's.
|
|
|
|
Post by Bentley on May 5, 2024 17:48:47 GMT
YouTube doesn’t exist to serve you chaps . You are the commodity. They tweak the AI to throw out or demote negative comments.
It's that American commercial thing. All punters must be blissfully ignorant, much like it is at Mc D's.
They do whatever they want to do , to suit them . YouTube is not their for your benefit. |
|
|
|
Post by Baron von Lotsov on May 5, 2024 18:09:51 GMT
They tweak the AI to throw out or demote negative comments.
It's that American commercial thing. All punters must be blissfully ignorant, much like it is at Mc D's.
They do whatever they want to do , to suit them . YouTube is not their for your benefit. If they had it their way it would not be half as bad as it is. This shit has come about in the changes in the law which have quietly been put in place. Youtube's motive is to generate mass appeal to their site, to keep people on there as long as possible and for them to watch as many adverts as possible. They have to censor videos which put advertisers off, known as demonetization. This cuts the advert feed to keep advertisers as customers but then it cuts the earnings as well from the video. I think they have done a reasonable job given the constrains they have to work in. Woke enforcement comes from what is called ESGs. I can explain them if you are unfamiliar with how it works. |
|
|
|
Post by Bentley on May 5, 2024 18:20:35 GMT
They do whatever they want to do , to suit them . YouTube is not their for your benefit. If they had it their way it would not be half as bad as it is. This shit has come about in the changes in the law which have quietly been put in place. Youtube's motive is to generate mass appeal to their site, to keep people on there as long as possible and for them to watch as many adverts as possible. They have to censor videos which put advertisers off, known as demonetization. This cuts the advert feed to keep advertisers as customers but then it cuts the earnings as well from the video. I think they have done a reasonable job given the constrains they have to work in. Woke enforcement comes from what is called ESGs. I can explain them if you are unfamiliar with how it works. I’m familiar with the fact that you are a commodity to YouTube . |
|
|
|
Post by seniorcitizen007 on May 5, 2024 18:36:44 GMT
Youtube noted that I frequently watched videos about police interrogations of suspects (who had sometimes been accused of sexual crimes) ... so they started recommending videos about POLICE who had committed sexual crimes.
|
|
|
|
Post by Baron von Lotsov on May 5, 2024 18:52:42 GMT
Youtube noted that I frequently watched videos about police interrogations of suspects (who had sometimes been accused of sexual crimes) ... so they started recommending videos about POLICE who had committed sexual crimes. As I was suggesting above, it is a Bayesian system, so it thinks you are a member of the group who watch this video, so for all those who watched this one, which other one did they go to next. There are all these other weighting conditions as well, so someone's preference who watched two of the ones you watched is weighted higher. What you may find is there are few people who exhibit close behaviour to yourself, so it is a case of trying to find a trend on a graph of points where the points are sparse. |
|
