Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

twitlonger.com

Post reply on HN
Print view View original

181–190 of 256 posts

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#181

Post by fortythirteen »

randyrand wrote: post #88

"Any language intended to spread hate." Could that be any more vague? Does encouraging people to hate DRM or hate cancer count under that? Clearly not. So what does 'hate' even have to do with it? Really its because "Hate speech" in and of itself is a dumb phrase. AFAIK It only refers to an incredibly small subset of hate. It does not refer to hate. And often refers to mean-speech instead of referring to the emotion…

"Hate speech" isn't a dumb phrase, it's a newspeak phrase perfectly designed for its purpose - to strike fear in the heart of anyone accused of it, and for it to be spoken with an undeserved sense of self-evidence.

You'll not find anyone outside of the "anyone who disagrees with my political opinions is inherently evil" group using it.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#182

Post by cnewey »

notyourday wrote: post #172
cnewey wrote: post #165

Earlier quoted context omitted.

You're implicitly making the assumption that an algorithm is unbiased and less likely to be fallible - and unlikely to be susceptible to bias or manipulation (at least, compared to a human). There is a considerable amount of academic and popular literature showing that these assumptions are false, at least at a technical level. The problem of vetting YouTube videos isn't necessarily something that can be solved by an…

> You're implicitly making the assumption that an algorithm is unbiased and less likely to be fallible - and unlikely to be susceptible to bias or manipulation (at least, compared to a human). There is a considerable amount of academic and popular literature showing that these assumptions are false, at least at a technical level. No, I'm arguing that algorithms are transparent, which means that they can be adjusted.…

Your argument that algorithms are transparent is equally flawed. There is an entire field of research dedicated to introspection of layers in convolutional neural networks! A system that is complex and nuanced enough to be able to deal with a problem such as auto-moderating YouTube videos will be phenomenally difficult to inspect (and nearly impossible to "adjust" in the way that you seem to be suggesting).

To clarify; I am not saying that it will never be possible to solve this sort of problem with an algorithm - just that doing so would require solving the small inconvenience of general artificial intelligence first... :-).

It certainly isn't beyond the realm of possibility in the (very distant) future, but to suggest that it's possible right now (and that it would be better than a human!) is to over-egg the pudding somewhat.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#183

Post by soundwave106 »

bufferoverflow wrote: post #38
amarkov wrote: post #34

Earlier quoted context omitted.

Strict rules for tagging would just run into the same problems as strict rules for demonetization. LGBT groups won't be any happier to see their content publicly branded "sexually explicit" than just demonetized.

By strict rules I also mean objective and unbiased. Penis/vagina -> nudity (even if it's medical or educational). "Fuck" -> cursing (even if it's about the etymology of the word). Advocating murder of group X -> ban (even if you're quoting a religious text). If these rules are strict and clear, we have a chance of creating a legal-like system. Otherwise it's subjective chaos with everyone pissed off.

Unfortunately, it is impossible to do this -- even strict objective and unbiased rules will piss people off.

The classic example I can think of, that has already generated headlines, concerns nudity policies in social media (Facebook in particular) that have flagged breastfeeding mothers for "nudity" (and thus ticked people off). A "by the letter" policy risks much more of this sort of thing... flagging something as "bad" that much of the social norm says isn't really a big deal.

One other thing I would worry about with "objective" rules is what culture determines the "watch list". The cultural norms for what is taboo actually does vary to some degree from culture to culture. Youtube is used heavily in so many countries, with diverse cultures. How can a single policy encompass the social norms of countries as diverse as, say, heavy Youtube using countries such as India, Japan, the UK, Germany, and the US? (And this isn't even accounting for the diversity of culture within these countries.)

As an example, the swastika, which is extremely taboo in Germany and triggers "hate speech" flags in much of the rest of the West, is (usually styled differently but still) a religious symbol in India. You could create a generic flag category for "swastika", but chances are this will eventually flag some, say, Hindu content from India and tick those people off.

I suppose you could structure the rules to account for as many cultural norms as possible, but that can get very complicated fast.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#185

Post by _dps »

majani wrote: post #128
jbardnz wrote: post #107

Maybe this is real, but it doesn't seem to line up with what I have heard from other Youtubers. Firstly it is almost certain that a video is initially demonetized by an algorithm. Only during a manual review is it looked at by a human. The main complaint from Youtubers is that the algorithm is way too strict and inconsistent. But the manual review process is pretty fair and the vast majority of videos get remonetized…

I wonder why you're the only skeptic in a forum that's usually skeptical about every post. Yes, you're right, this is not Google's MO. Google tend to look for software based solutions first and foremost before resorting to humans to solve problems. Recently Google have really been working on their voice transcription software and I imagine they could easily design a bot that scans subtitles and video frames for offen…

In the past I would have agreed with your skepticism, but they just announced a huge human-powered review team:

http://money.cnn.com/2017/12/05/technology/google-youtube-hi...

I don't know based on that article if it's directly related, but they've obviously decided that they need serious human review teams.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#186

Post by dogma1138 »

NietTim wrote: post #59
PeterStuer wrote: post #23

Google is an American company. Their "rules" are extremely embedded in a US culture. Unfortunately YouTube is a product that is consumed worldwide. Many of the regions where it has a dominant market share for web VoD do have cultural palates that are very different from the US. I feel Google should embrace their international position and be less biased in it's reasoning.

Yeah having country specific moderation teams with people form those countries would be great. But I won't hold my breath..

It would be great until you reach Saudi Arabia or Iran, or heck don't need to go that far since Venezuela is just near by...

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#187

Post by blurbleblurble »

musage wrote: post #66
bufferoverflow wrote: post #25

Earlier quoted context omitted.

The solution is obvious - tag/categorize the videos, according to very strict rules, let the advertisers decide which tags to serve. Allow the publishers to dispute the tags. For instance, "nudity" can be defined as "visible genitals, female nipples" in the US. "Controversial" tag should not even exist, it makes no sense, any non-trivial subject is controversial. And honestly, YouTube can benefit from a more precise…

Where would this fit in? https://www.youtube.com/watch?v=9R5w-PIzlUk It's a clip-art style "coloring video" drawing a baby with a lot of syringes, which then each empty their colored liquid into the baby. Technically, perfectly innocent. In context of a positive flood of videos with an undertone ranging from brain damaging to abusive targeted at little kids it doesn't seem that innocent to me. The sum isn't larger th…

Okay so I've watched quite a few of those particular videos now to make sense of them... And I've come to the conclusion that the reaction to the syringes is cultural misunderstanding. My hunch is that the purpose of the syringes is to get children comfortable with the vaccination process. There are number of other videos where the syringes are being demonstrated on animals to make the animals feel better after getting sick. In this case, it seems like US Americans are projecting their particular sensibilities onto videos made by some people in an entirely different part of the world. Yes, they're weird to me too, and definitely not high quality, but I've actually started to appreciate some of those videos as being well intentioned propaganda, as opposed to some of the other videos that are literally showing children rape scenes and violence. It's important that we don't just lump everything that makes us uncomfortable into the same category.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#188

Post by notyourday »

cnewey wrote: post #182
notyourday wrote: post #172

Earlier quoted context omitted.

> You're implicitly making the assumption that an algorithm is unbiased and less likely to be fallible - and unlikely to be susceptible to bias or manipulation (at least, compared to a human). There is a considerable amount of academic and popular literature showing that these assumptions are false, at least at a technical level. No, I'm arguing that algorithms are transparent, which means that they can be adjusted.…

Your argument that algorithms are transparent is equally flawed. There is an entire field of research dedicated to introspection of layers in convolutional neural networks! A system that is complex and nuanced enough to be able to deal with a problem such as auto-moderating YouTube videos will be phenomenally difficult to inspect (and nearly impossible to "adjust" in the way that you seem to be suggesting). To clarif…

It is not flawed at all. If you cannot provide a validation for your algorithm decisions when challenged then you algorithm belongs in a research lab or on your laptop, not in production.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#189

Post by ChicagoDave »

As I just explained to my 16 year old daughter about YouTube policies. YouTube is a business. Businesses will do whatever they need to do to get customers and make money. If anyone was under some strange illusion that YouTube was a public and free platform to make money and/or create any content they wanted was naive at best.

If you want to create a channel for users to view your content, purchase a domain, buy a VM on some cloud service, and set up your own video streaming service for your own channel.

It would be interesting to see how much it cost one person versus the income generated. I'm pretty sure that anyone with millions of subscribers would be bankrupt instantly.

I think the one reasonable avenue would be to unionize all of the major content creators and invite any creator to be a part of the union. Then threaten a strike. But the content creators are probably like a herd of cats. It's unlikely any coalition is remotely possible.

Re: YouTube Demonetization Screenshot Leaks and Secret YouTube Meeting

#190

Post by dsfyu404ed »

rkapsoro wrote: post #173

My take on the demonetization controversy is that it is too coarse: the output is a boolean, is it advertiser friendly or not so, to say nothing of the preferences of individual advertisers. To me it seems that even assuming the current single set of criteria for advertiser friendliness (with all of the other problems discussed in the other comments regarding subjectivity) that different advertisers might or might no…

>My take on the demonetization controversy is that it is too coarse: the output is a boolean, is it advertiser friendly or not so, to say nothing of the preferences of individual advertisers.

Different metrics matter more/less to different advertisers/industries/audiences and the more niche it is the less it matches up with the global average that is a "general audience"

This. If there's a video of some guy laying in the mud welding some piece of heavy equipment back together in the middle of nowhere whether or not he's swearing like a sailor probably doesn't matter at all to Hobart or Lincoln or some other company that might be advertising their line of welding consumables with the video.

Likewise I'm sure some companies that sell products that make us less afraid of unlikely events (fire extinguishers, life insurance, etc) would have little problem with their ads appearing in conjunction with violent content.

Post reply on HN