YouTube is still restricting and demonetizing LGBT videos | Hacker News Reader

YouTube is still restricting and demonetizing LGBT videos

theverge.com

> We use machine learning to evaluate content against our advertiser guidelines. Sometimes our systems get it wrong, which is why we’ve encouraged creators to appeal. Successful appeals ensure that our systems get better and better

There's been bad press about this for months now. People with conservative viewpoints claimed to be facing similar issues

The cost of a false positive seems extremely high. Why would they depend on ML for high influencers?

Even the 'high influencer' subset is a massive quantity of videos.

As Facebook learned recently, people aren't willing to give much benefit of the doubt to solutions involving human review. Can you imagine a creator saying "I don’t think YouTube is doing this on purpose", as they did in this article, if some actual Youtube employee had clicked a button to demonetize?

That same quote jumped out at me too. In another sentence they say they don’t have a list of words that trigger demonetization.

Well, YT, you do.

It’s just in your ML model. You let a pattern recognition algorithm create the list for you.

(Yes it’s not quite just a list, but still.)

Ya, under 100k subs or some sort of viewer metric should be a completely different system. Perhaps a similar system to twitter, who uses hybrid components for different scale users.

> We use machine learning to evaluate content against our advertiser guidelines.

This means Google violates civil rights law by maintaining a list of phrenological behavior based on racial, sexual, and demographic data.

Advertisers choosing videos to show up in also means selecting which videos to filter out. This means Google has a list of acceptable behavior for blacks, gays, and women to perform such filtering.

>This means Google violates civil rights law

Correct me if I'm wrong, but I don't believe sexual orientation is a protected class. Are there other statuses which would make this behavior illegal? Obviously IANAL and don't really know what I'm talking about here.

I think OPs jump in logic means if you can filter ads on sexual orientation then the same tooling can be use to attack protected classes.

Perhaps, though they are lumping in "sexual [orientation]" with race and gender. I also unaware of any cases of potential violations against a protected class.

"Our machine learning algorithm got it wrong" is the corporate version of when a kid blames their own bad behavior on an imaginary friend.

Sorry but you're the one running the ML algo. And if it fucks up, you're responsible.

The ones that stay monetized are getting targeted by anti-lgbt group advertisements. Which of course spawns "let's react" videos, which are hilarious in a life gives you lemons kind of way.

Obviously there needs to be more evidence but it seems quite possible the word transgender triggers demonetization, at least in these cases.

I agree there's a good chance that's true, but probably not because somebody at YouTube flagged the word 'transgender'. But instead by some convoluted second-order effects of whatever algorithm they're using to automatically flag stuff.

IMHO, that doesn't matter much. Algorithms shouldn't be above the law. If a company creates an algorithm, it needs to make sure it follows the law. Companies use this excuse on ML we don't know why it is doing what it is doing is BS. Machine Learning, despite what the popular media might like to say, is not sentient AI. It does not have free will. It does what it is trained to do. It is predictable. It is testable. I understand that AI allows companies to chew through quantities of data that would cost thousands of human operators' salaries otherwise. But if the company gets to reap the profits of the AI, then it needs to take responsibility for it too.

The only reasonable thing would be to only restrict things that are actively breaking the law. Of course, that runs into breaking whose law, but I don't want to digress.

Maybe filter your advertising platform better, if your advertisors have oppinions. That's not maybe the easiest thing, but YouTube has hundreds of people to throw at forced requirements, and I'm a single person at a small company that gets these requests daily, and my give-a-damn is busted

To be honest it's a business move lots of the LGBT Pro or Contra videos have a lot of say the least "controversial" comments on youtube.

You can justify any shitty thing by saying "It's just business."

I think we all know it's a business move. Doesn't make it right. When Milton Friedman said "the business of business is business" or something to that effect, he was wrong.

You can. It is their business to satisfy their customers (advertisers) and the advertisers don't want to be seen in a political light or near hateful comments and trolls. It is totally understandable why they would do it.

Totally understandable. If you have no sense of social responsibility.

Edit: On second thought, you may have a point, and this is more nuanced than I thought. There are a lot of topics advertisers don't want to be associated with (violence or terrorism) and we expect youtube to filter these out. We don't expect the advertiser to filter these on their own.

If you ask me, transgenderism is not anything like violence or terrorism, in terms of its controversy. But I realize some people disagree.

I think it's impossible for youtube to accomplish this without having some kind of political stance. That would piss people off. But considering Google has already stated they are pro LGBT, you'd think they would avoid penalizing the transgender videos this way. Let advertisers filter out transgender videos on their own.

There is no substitute for judgement, however. Youtube can't be impartial.

Yeah, it's so sad that an important topic like that is overrun by trolls.

Reply on news.ycombinator.com