YouTube is deleting comments with two phrases that insult China’s Communist Party

[Sauron]

Just now, LAwLz said:

No, but it can be learned to be spam.

Hmmm, again there are so many things that could be detected as spam and yet aren't...

1 minute ago, LAwLz said:

1) Youtube isn't even in China, so Chinese people have very limited access to it.

China is a huge market and the Chinese government is willing and able to block websites that don't comply with their rules, Youtube has interest in appeasing them.

2 minutes ago, LAwLz said:

2) Youtube's automatic systems, which this definitely is a part of since comments are deleted within seconds, have not exactly been flawless before.

And they have been slammed for it, why can't we do that here?

2 minutes ago, LAwLz said:

3) Youtube admits to it being a mistake and will fix it.

Which they.... wouldn't do... if they were caught appeasing the Chinese government with shady tactics...?

4 minutes ago, LAwLz said:

Arguments for why this is a conspiracy:

1) It's about China and "china = bad"?

I dislike McCarthyist rhetoric but that doesn't mean the Chinese government isn't awful... this is well within the confines of what they've shown they're willing to do or demand be done in the past. I think it's feasible to criticize it without falling into xenophobia or red scare tactics even if that's what politicians are doing.

8 minutes ago, LAwLz said:

Remember Occam's razor. The simplest solution is most likely the right one.

How is a random chance of the algorithm learning this very specific phrase a "simpler" explanation than Youtube manually inserting it?

9 minutes ago, LAwLz said:

And also remember Hanlon's razor. Never attribute to malice that which is adequately explained by stupidity.

That's a truism... I would argue that we shouldn't be too eager to dismiss what can be explained with malice by attributing it to stupidity. Regardless, both could be at play here - and even if it only is stupidity it still warrants criticism.

11 minutes ago, LAwLz said:

I mean really. How often do we have threads about Youtube's algorithm fucking up and incorrectly categorizing or recommending stuff? And yet as soon as something similar happens to something related to China, everyone assumes the system is flawless and works 100% as intended.

The system always works as intended in that it overpolices the platform to protect advertiser investment and Youtube's interest - it almost never fails at that. Within those confines there may be fine tuning details that Youtube considers a mistake but ultimately those are not its primary concern. Appeasing the Chinese government falls well within "Youtube's interest".

[LAwLz]

26 minutes ago, Sauron said:

-snip-

Well I think this conspiracy theory has a plausible explanation and I can't think of a reason why Youtube would do this. That plus Youtube's own admission that it's a mistake is good enough for me to take the side of "this was a mistake". To me, it doesn't make sense.

I understand that you're not as easily persuade and believe this is a conspiracy, and I don't think me trying to rationalize my POV will make you change your mind.

 

To each their own.

[Sauron]

2 minutes ago, LAwLz said:

I can't think of a reason why Youtube would do this.

I gave you the reason more than once, China is a huge market and the Chinese government controls it.

[LAwLz]

10 minutes ago, Sauron said:

I gave you the reason more than once, China is a huge market and the Chinese government controls it.

Yes, but it's a market Youtube isn't in.

And it will take way more than just censoring a few words if the plan is to have Youtube enter the Chinese market. They would probably have to set up an entirely different site for it so in that case it doesn't make sense to censor the rest of the world's Youtube.

 

In fact, didn't Google do that with their search? They set up an entire different search engine in China where it was heavily filtered results. The filtering that happened on Google.cn never spread to google.com for example. So I don't see why a theoretical youtube.cn would spread to youtube.com.

 

We have historical evidence of this not being the way Google goes about appeasing China. And if this is an attempt to appease China then I don't think Google would come out and say "yeah this was a mistake and we will fix it". That would burn all the bridges they have supposedly tried to build up with China, right?

[Sauron]

8 minutes ago, LAwLz said:

Yes, but it's a market Youtube isn't in.

Of course they are in it, are you kidding me? Over a billion people of potential audience AND a bunch of advertisers?

[mr moose]

54 minutes ago, Sauron said:

I gave you the reason more than once, China is a huge market and the Chinese government controls it.

It's certainly within googles MO/history to engage in censorship in order to secure their presence in China.

[Touch My Hamm]

18 hours ago, dizmo said:

China's getting too expensive to do business in, so we'll rapidly see the shift to India as their infrastructure improves, or the Philippines, etc.

What one has to worry about is the effect China's dominance will take on the rest of the world as it rapidly becomes self sufficient.

I would like to see companies more global. Its sad that so many manufacturers went to china since it was cheap labor. This could be the main determining factor as to what made china such a massive superpower now. Well with whats going on now we can only hope companies smarten up and think more long term than the short term gains their are all so used to. 

[Bombastinator]

1 hour ago, GodSeph said:

I would like to see companies more global. Its sad that so many manufacturers went to china since it was cheap labor. This could be the main determining factor as to what made china such a massive superpower now. Well with whats going on now we can only hope companies smarten up and think more long term than the short term gains their are all so used to. 

It wasn’t just cheap labor, it was lack of environmental enforcement.  They had environmental laws but they didn’t enforce them.  The allowance of High pollution processes and slave labor base an irresistible combination

[Elements6007]

i dont know why google is working with china if china doesnt even allow them in china................. just makes no sense at all.......

[suicidalfranco]

4 hours ago, Elements6007 said:

i dont know why google is working with china if china doesnt even allow them in china................. just makes no sense at all.......

because google wants to get into china

[StDragon]

"This was an error with our enforcement systems and we have rolled out a fix." - Alex Joseph (YouTube spokesperson)

 

Interesting? Well, did you know Twitter adds Dr. Fei-Fei Li (former Google VP) A.I. guru to their team. Did you also know that she has ties to the CCP?!

 

And that's what we currently know...

 

Many of the high talented individuals from China are working in Silicon Valley. I guarantee you that many are moles (spies) with the sole intent of corporate espionage.

 

 

[Bombastinator]

22 minutes ago, StDragon said:

"This was an error with our enforcement systems and we have rolled out a fix." - Alex Joseph (YouTube spokesperson)

 

Interesting? Well, did you know Twitter adds Dr. Fei-Fei Li (former Google VP) A.I. guru to their team. Did you also know that she has ties to the CCP?!

 

And that's what we currently know...

 

Many of the high talented individuals from China are working in Silicon Valley. I guarantee you that many are moles (spies) with the sole intent of corporate espionage.

 

 

“Ties” there’s a vague term.  What ties?  She a member?  
 

“Ties” is so vague it could be “they killed my parents when I was a teen”.

 

You want to make noises about “ties” or “is linked” say how.

 

you “guarantee” “many”.  How?  Special super secret sleuthing skills? Brute ignorance?  Statistically a few?  Probable. Guarantee” “many” smells of BS.

[StDragon]

2 hours ago, Bombastinator said:

“Ties” there’s a vague term.  What ties?  She a member?  
 

“Ties” is so vague it could be “they killed my parents when I was a teen”.

 

You want to make noises about “ties” or “is linked” say how.

 

you “guarantee” “many”.  How?  Special super secret sleuthing skills? Brute ignorance?  Statistically a few?  Probable. Guarantee” “many” smells of BS.

"Guarantee" as in IP theft has been proven. Entire books on the subject alone if you care to read. Dozens in fact. Audio or paperback, you choose. Non-fiction of course. Going into detail is beyond the scope of this discussion, but the information is out there. How strong is your google-fu?

 

Many Chinese companies are really just fronts for the CCP. Many of their business decisions examined on their own aren't rooted in any logic unless you accept the fact these companies are just tentacles of a much larger organism with unified goals of the state's interest. The Soviet Union attempted similar means but it eventually bankrupted them. It will bankrupt China too. The books are cooked and thus impossible for them to remain on their current economic trajectory.

 

To understand the future you must first have knowledge of the past.

 

 

[StDragon]

And yes, the CCP has been involved with propaganda through Hollywood too.

 

This news about Youtube is but a fart in a hurricane of countless other activities the Chinese government is involved in.

 

五毛

[Bombastinator]

27 minutes ago, StDragon said:

"Guarantee" as in IP theft has been proven. Entire books on the subject alone if you care to read. Dozens in fact. Audio or paperback, you choose. Non-fiction of course. Going into detail is beyond the scope of this discussion, but the information is out there. How strong is your google-fu?

 

Many Chinese companies are really just fronts for the CCP. Many of their business decisions examined on their own aren't rooted in any logic unless you accept the fact these companies are just tentacles of a much larger organism with unified goals of the state's interest. The Soviet Union attempted similar means but it eventually bankrupted them. It will bankrupt China too. The books are cooked and thus impossible for them to remain on their current economic trajectory.

 

To understand the future you must first have knowledge of the past.

 

 

There’s a big difference between “it’s happened” and this person is likely a perpetrator.  “Many” implies a significant percent.  Might it be worth looking at?  Perhaps. There is a reason the root of predudice is pre-judgment though.

[themoose5]

Came to this thread from the WAN show topic and wanted to add my 2c.

 

First a little background, I work professionally as an AI engineer. While I don't do direct model development I am very versed in the processes around model development. Also I don't work for google or yt so I don't have specific inside knowledge of their processes or models.

 

On the surface I think Linus' hypothesis he presented on the WAN show could be very close to the truth of how this word ended up on the automatic filter. If in fact google/yt is using a model or collection of models for automated comment filtering it's architecture is probably going to be some thing pretty common.The key here really isn't the architecture of the model but what training data was used to create that model. Specifically where google/yt got that training data from for non-English content. If their training data contained a significant amount of examples containing the Chinese characters in question that lead to a negative sentiment following their use it is easy to see how the model would learn to flag these characters as offensive. 

 

This seems to be backed up by the fact that only the use of these specific Chinese characters result in a comment being removed but the use of the romanized version does not receive the same treatment. If this phrase was added by someone manually I think it would be a safe assumption that the person adding them would add in all the variations of the phrase in Chinese characters and their romanized counterparts and not just the Chinese character version. On the other hand if this was something learned by an AI model using data pulled from outside sources all it would take is having enough examples in the training data for the model to learn the Chinese characters of this specific phrase. That learning would not necessarily transfer to learning the romanized version as well, resulting in what seems to be the current behavior. 

 

Training data review and analysis is something that is still kind of a sore spot in the field of AI. Training data quality has a large impact on the final performance of a model and there are numerous examples of how AI models can be biased by their training data. Most likely what happened in this case was the non-English language training data was not reviewed well enough to catch these issues before deploying.

[mr moose]

50 minutes ago, themoose5 said:

Came to this thread from the WAN show topic and wanted to add my 2c.

 

First a little background, I work professionally as an AI engineer. While I don't do direct model development I am very versed in the processes around model development. Also I don't work for google or yt so I don't have specific inside knowledge of their processes or models.

 

On the surface I think Linus' hypothesis he presented on the WAN show could be very close to the truth of how this word ended up on the automatic filter. If in fact google/yt is using a model or collection of models for automated comment filtering it's architecture is probably going to be some thing pretty common.The key here really isn't the architecture of the model but what training data was used to create that model. Specifically where google/yt got that training data from for non-English content. If their training data contained a significant amount of examples containing the Chinese characters in question that lead to a negative sentiment following their use it is easy to see how the model would learn to flag these characters as offensive. 

 

This seems to be backed up by the fact that only the use of these specific Chinese characters result in a comment being removed but the use of the romanized version does not receive the same treatment. If this phrase was added by someone manually I think it would be a safe assumption that the person adding them would add in all the variations of the phrase in Chinese characters and their romanized counterparts and not just the Chinese character version. On the other hand if this was something learned by an AI model using data pulled from outside sources all it would take is having enough examples in the training data for the model to learn the Chinese characters of this specific phrase. That learning would not necessarily transfer to learning the romanized version as well, resulting in what seems to be the current behavior. 

 

Training data review and analysis is something that is still kind of a sore spot in the field of AI. Training data quality has a large impact on the final performance of a model and there are numerous examples of how AI models can be biased by their training data. Most likely what happened in this case was the non-English language training data was not reviewed well enough to catch these issues before deploying.

How do we know it is using AI and not just a straight up filter for specific combinations of words?

 

EDIT: for clarification, google said the phrase was added to the automatic filter after it was repeatedly reported.

[StDragon]

6 hours ago, mr moose said:

How do we know it is using AI and not just a straight up filter for specific combinations of words?

 

EDIT: for clarification, google said the phrase was added to the automatic filter after it was repeatedly reported.

Google is partially blocked in China. Youtube is entirely blocked! Youtube is owned by Google.

 

Hypothesis: China told Google to block certain key phrases, and then Google's black-list replicated over to the Youtube side as either a process or a shared look-up database.

 

There is no reason to be blocking anything from Youtube when it's in fact already blocked in its entirety from China. 

[sub68]

On 5/26/2020 at 8:24 PM, Var_ said:

YouTube is secretly gathering data for the Chinese government. Ask Dennis, he knows all about it

isn't he from taiwain? EDIT oh taiwain is part of china

also china has its own YouTube, https://www.bilibili.com/

knowing china yes they are getting data like tiktok

so this dosent surprise me

[StDragon]

3 hours ago, sub68 said:

EDIT oh taiwain is part of china

It WAS, no longer. They had a civil war in 1949 that settled that dispute.

 

Long story short - The original Chinese government retreated to Taiwan. The Communists took over the mainland.

 

The dispute - Taiwan sees itself as both the legitimate government of the island and mainland. The communists view the island of Taiwan as "unfinished business" with the idea of an inevitable reunification on its terms.

 

[sub68]

Just now, StDragon said:

It WAS, no longer. They had a civil war in 1949 that settled that dispute.

 

Long story short - The original Chinese government retreated to Taiwan. The Communists took over the mainland.

 

The dispute - Taiwan sees itself as both the legitimate government of the island and mainland. The communist view the island of Taiwan as "unfinished business" with the idea of an inevitable reunification on its terms.

 

since my dad is from Taiwan he told this a while then forgot ( some times iam dumb)

then I pulled a wiki page and remembered

[Bombastinator]

18 minutes ago, StDragon said:

It WAS, no longer. They had a civil war in 1949 that settled that dispute.

 

Long story short - The original Chinese government retreated to Taiwan. The Communists took over the mainland.

 

The dispute - Taiwan sees itself as both the legitimate government of the island and mainland. The communists view the island of Taiwan as "unfinished business" with the idea of an inevitable reunification on its terms.

 

Taiwan’s national status is weird and vague.  Nixon famously instituted the “one China” policy which says China and Taiwan are one nation, but the Chinese government can only institute authority with the consent of the Taiwanese people which has never happened.  So yes and no.

[I_nochi]

 This video is marked by youtube as "offencive?". I mean there is scenes of the riots, but the community surrounding this channel is pretty upset about yt marking this as offencive. The community still thinks that yt is just siding with the chinese goverment. I think that the blood and riot scenes are not really much of a deal. You still see those in games and yt dosnt give a damn.

[StDragon]

1 hour ago, Bombastinator said:

Taiwan’s national status is weird and vague.  Nixon famously instituted the “one China” policy which says China and Taiwan are one nation, but the Chinese government can only institute authority with the consent of the Taiwanese people which has never happened.  So yes and no.

Well, only weird because of the UN which is likened to the Star Wars bar scene; members that carry the baton of double-speak whom are represented by the most unscrupulous that humanity has to offer...

 

But yes, correct, no one can claim Taiwan but Taiwan and China alone. The reality of it is this - For Taiwan to be reunified with China, it must be through democratic and diplomatic means. The alternative is that China invades by force.

[mr moose]

4 hours ago, help_pls said:

 This video is marked by youtube as "offencive?". I mean there is scenes of the riots, but the community surrounding this channel is pretty upset about yt marking this as offencive. The community still thinks that yt is just siding with the chinese goverment. I think that the blood and riot scenes are not really much of a deal. You still see those in games and yt dosnt give a damn.

Are any of the videos of the current US riots marked as offensive?

 

I'd just look myself but humans behaving badly depresses me (because it is nearly always the result of being treated badly).