YouTube is now messing with the Super Chat feature of live streams

[Master Disaster]

Video Source this one...

 

In an effort to cut down on harassing super chats YouTube is apparently changing the way it's super chat system works and I can see it pissing off a lot of live streamers.

 

Basically YouTube are using AI and algorithms (because that always works out well) to try and detect inappropriate or harassing super chats and if they detect one then they're going to take the money and give it to charity. Details are thin but it seems as though the streamer will never know the message has been sent and the sender won't be told what happened to their money.

 

Here's a quote from an article in the guardian where this story originated from.

Quote

The company has tightened the rules for which channels have access to monetisation features and deployed machine learning technology to identify hate speech in comment features, the spokeswoman added. If a user is caught sending abusive Super Chats, the revenue received will be donated to charity instead of being split between YouTube and the channel’s creator.

Now I should point out that the system has been created to apparently target alt right groups using YouTube to spread racism and hate speech (the guardian article I'll link at the end goes into much more details) however the worry is this will quickly spread and target all channels. Rich from Review Tech USA points out how his viewers like to troll him by sending words like Nipples & Cucumber, something thats 100% a joke but could easily be identified as hate speech by a bot and suddenly he's losing money without even knowing it.

 

Source (goes into alt right and right wing politics) - https://www.theguardian.com/media/2018/sep/18/report-youtubes-alternative-influence-network-breeds-rightwing-radicalisation

Video Source

 

Listen guys, please keep your political views to yourself and keep this about the super chat changes.

 

I can see where YouTube is coming from, I genuinely can but this is a dangerous road. Taking other peoples money away from its intended destination is never a good road for any corporation.

 

If YouTube is worried about making money out of hate speech how about they work of stopping the hate speech instead?

[jiyeon]

YouTube once again tries to fix something that isn't broken, and with AI once again.

 

7 minutes ago, Master Disaster said:

Taking other peoples money away from its intended destination is never a good road

You took the words right out of my mouth.

[porina]

Maybe a semi-manual way would be better? I would be concerned about false positives, in a way that friendly people will poke each other but they understand the context and are ok with it. To an outsider without context, it might not appear appropriate.

 

For example, "safe" comments could be processed as normal, but questionable ones might go into a hidden from public view for manual moderation by the streamer, who can then ok it as a one off, or even whitelist.

[D13H4RD]

Here's the problem. 

 

Does the AI know what is a joke and what is serious? 

[Master Disaster]

1 minute ago, porina said:

Maybe a semi-manual way would be better? I would be concerned about false positives, in a way that friendly people will poke each other but they understand the context and are ok with it. To an outsider without context, it might not appear appropriate.

 

For example, "safe" comments could be processed as normal, but questionable ones might go into a hidden from public view for manual moderation by the streamer, who can then ok it as a one off, or even whitelist.

The problem stems from the fact YouTube take a 30% cut of Super Chats, they're not happy making money off of hate speech (and I understand that) so by allowing manual review it doesn't fix the problem for them. The alt right racists will allow the messages through and YouTube still make money from them.

[Castdeath97]

3 hours ago, porina said:

Maybe a semi-manual way would be better? I would be concerned about false positives, in a way that friendly people will poke each other but they understand the context and are ok with it. To an outsider without context, it might not appear appropriate.

 

For example, "safe" comments could be processed as normal, but questionable ones might go into a hidden from public view for manual moderation by the streamer, who can then ok it as a one off, or even whitelist.

Manual work is a forbidden realm when it comes to Google. Won't be too long before they try putting algorithms everywhere.

 

3 hours ago, D13H4RD2L1V3 said:

Here's the problem. 

 

Does the AI know what is a joke and what is serious? 

Machine learning can struggle with relative things and context in general

[D13H4RD]

Just now, Castdeath97 said:

Machine learning can struggle with relative things and context in general

That's the issue. 

 

I've known the behavior of internet fans in general and there's always some sort of friendly banter going on. 

 

If the AI detects them as hate speech, that's going to be an issue. 

[Master Disaster]

1 minute ago, D13H4RD2L1V3 said:

That's the issue. 

 

I've known the behavior of internet fans in general and there's always some sort of friendly banter going on. 

 

If the AI detects them as hate speech, that's going to be an issue. 

Rich mentions this in his video, his regulars often send him "inappropriate" messages but because he knows the context and usually the sender he understands it's a joke. If this is actually true (and so far no one knows if it's happening or not) then he and other streamers are going to have money taken from them for no good reason.

[D13H4RD]

Just now, Master Disaster said:

If this is actually true (and so far no one knows if it's happening or not) then he and other streamers are going to have money taken from them for no good reason.

And herein lies the issue. 

 

I think this is a solution on the hunt for problems. 

 

And another question. Is this targeted towards hate speech in general regardless of political affiliation or just those in the right-wing? 

 

If it's the latter, then I have a big problem with that. 

[Master Disaster]

6 minutes ago, D13H4RD2L1V3 said:

And herein lies the issue. 

 

I think this is a solution on the hunt for problems. 

 

And another question. Is this targeted towards hate speech in general regardless of political affiliation or just those in the right-wing? 

 

If it's the latter, then I have a big problem with that. 

Again, as is typical with YouTube, we simply don't know. They love to keep things enigmatic.

 

7 minutes ago, VegetableStu said:

staying about the idea of moderating Superchats for a bit without going to Google's AI intervention:

 

say the channel is extremely uncomfortable or (people) universally disagrees with a superchat. should it be removed? should the money be rejected? if it remains, does the channel have to comment on it with a message of disapproval? should the channel be able to block senders? should Superchats be non-refundable from the poster's side?

 

(asking the questions for a baseline here o_o)

1) I'd assume the people watching the stream would allign with the streamers views.

2) If the streamer takes issue with a chat then I think yes, he should be able to return the money

3) Streamers don't have to acknowledge super chats at all

4) yes

5) I don't understand the question

[Master Disaster]

Just now, VegetableStu said:

like if someone sends a superchat can that someone retract (and should they get a refund on the way)

Not as far as I know, there's no way of retracting. Such a system would be open to massive abuse.

[Bouzoo]

24 minutes ago, D13H4RD2L1V3 said:

Here's the problem. 

 

Does the AI know what is a joke and what is serious? 

Yes / No/s

Yes/s / No

 

You try and guess what the author wants to say, let alone an AI. 

[D13H4RD]

1 minute ago, Bouzoo said:

Yes / No/s

Yes/s / No

 

You try and guess what the author wants to say, let alone an AI. 

I don't think an AI knows what is sarcasm, let alone the /s tag

[Master Disaster]

Just now, D13H4RD2L1V3 said:

I don't think an AI knows what is sarcasm, let alone the /s tag

Sarcasm is impossible to convey using text, even for humans who understand what it is totally.

[Bouzoo]

2 minutes ago, D13H4RD2L1V3 said:

I don't think an AI knows what is sarcasm, let alone the /s tag

 

1 minute ago, Master Disaster said:

Sarcasm is impossible to convey using text, even for humans who understand what it is totally.

Pretty much what I meant. Just a simple demonstration of things that can "mean" the same thing that confuse the shit out of us. 

[hestolemydog]

Can you just imagine how much money would have been donated to charity if this feature would have ran on Twitch. I'm pretty sure cancer would have been cured by now.

[Master Disaster]

1 minute ago, Bouzoo said:

 

Pretty much what I meant. Just a simple demonstration of things that can "mean" the same thing that confuse the shit out of us. 

It's been proven fairly consistently on YouTube that AIs suck at understanding human concepts and constructs yet here we are again.

[Misanthrope]

Youtube will continue to apply tech and fail miserably since there are no tech solution for a site that has grown that much that expects to be an advertising platform while still relying almost exclusively on user generated content.

 

Until then we will keep hearing about the false positives instead of the (imho justified and rightful) removal of actually shit content like Nazi propaganda. So you know who's going to be happy that the conversation about youtube revolves around technical issues and false positives? The guys organizing armed protests to support white nationalism and killing people with their cars.

[Spotty]

If Youtube want to do this to take some sort of moral high ground, then it should only be the cut that Youtube takes (~30%??) from the super chat that goes to charity. The Youtuber should still get their cut.

[Mihle]

33 minutes ago, Spotty said:

If Youtube want to do this to take some sort of moral high ground, then it should only be the cut that Youtube takes (~30%??) from the super chat that goes to charity. The Youtuber should still get their cut.

And the youtuber can decide himself if his parts go to charity or not.

[Jito463]

2 hours ago, Master Disaster said:

Now I should point out that the system has been created to apparently target alt right groups using YouTube to spread racism and hate speech

2 hours ago, Master Disaster said:

Listen guys, please keep your political views to yourself

1 hour ago, Master Disaster said:

The alt right racists

Oh, the irony.  For the record, the term "alt-right" is a fictitious one (just like "assault weapon"), made up by the ctrl-left to smear anyone they disagree with.  I'm not defending racism, just disputing the use of that made up word.  I won't make it any more political than that, but given your repeating of this mantra - after telling people to leave politics out - I feel justified in posting this correction.

 

With that out of the way, this is just another example of Youtube trying to play mommy/daddy for people using their platform.  If there weren't a good number of conservative voices still uploading to YT, I wouldn't even use it.  I already use private mode for certain videos, just to mess with their algorithms.

[Spotty]

29 minutes ago, Mihle said:

And the youtuber can decide himself if his parts go to charity or not.

Could be an option where the individual youtuber can toggle this automoderation charity system on or off. Such as if they disable automoderation and flag their channel/stream as possibly containing mature content, they keep their cut. If they have automoderation enabled and aren't flagged as mature content, then any inappropriate donations go to charity.

(This doesn't mean that I agree with it moderating donations and donating the money to charity)

[Doobeedoo]

Are they still going to be surprised why their livestream will never kick off huh. 

[asus killer]

is it just me that think this is fine? if people are using YT to be assholes then YT gives the money to charity. How can this be a bad thing?

 

 

[Jito463]

3 minutes ago, asus killer said:

is it just me that think this is fine? if people are using YT to be ******** then YT gives the money to charity. How can this be a bad thing?

You don't see an issue with this?

2 hours ago, Master Disaster said:

Details are thin but it seems as though the streamer will never know the message has been sent and the sender won't be told what happened to their money.