Google: YouTube is so overloaded staff cannot filter content | Hacker News Reader

Google: YouTube is so overloaded staff cannot filter content

hosted.ap.org

35 pointstshtf47 comments

"Despite the serious threat posed by extreme and violent videos"

Wait... what exactly is the THREAT of violent videos? The videos themselves aren't dangerous, and removing them won't stop the violence... what is portrayed in the video is the danger, not the fact that it was videoed.

I suppose they would say that the threat is that the videos serve as propaganda pieces for these organizations, including use as recruitment tools.

The threat may be to the standards and community that have been part of YouTube's success. Sites like Dailymotion and LiveLeaks are a totally different world for a reason.

>The videos themselves aren't dangerous, and removing them won't stop the violence

I actually take some issue with that. I think publicizing violence can be a cause of violence (the same critique could be applied to 24/7 media coverage of a shooting tragedy, which is possibly just what the shooter wanted). That's really not a point worth having an argument about, I just thought I'd throw it out.

I think what that sentence actually means is "it's a threat to the edifice of safety we've created and potentially to the usability of our service", i.e. "People want to pretend this kind of stuff doesn't exist and certainly don't want to see it when they are watching cat videos." Honestly, I somewhat agree. If I thought I would stumble across a beheading video when watching YouTube, I'd probably skip it entirely.

Yeah, I might have been a little bit flippant in my dismissal any danger at all, I just found it a bit melodramatic to act like the video is some huge danger.

I agree, I don't want to stumble on a violent video while browsing youtube. I just don't think it is actually a threat to our safety.

If you've ever submitted an iOS app and had a youtube frame/etc that allowed unfiltered searches, the app submission team will nearly always look up "Violent," and reject the app based on the ability to look up Violent videos.

Violent videos just seems to be a trigger point for a lot of politics/companies/etc.

I remember the first time I got that, and went, "well, yeah, you can look up puppies too." But, hey, maybe someone is downloading our app just to get through some youtube filters their parents set up.

I guess Liveleak is the most dangerous place on the Internet then.

The real threat is people finding out how barbaric our "allies" are, then realising we are the barbarians by proxy.

Seeing things going on in Syria from both factions - which ones would you support round the table? I watched a civilian get slowly bayoneted to death and I'm probably funding that.

Watching some Mexican cartel beheading & torture videos had quite an impact on my views on recreational drugs.

Frankly, I think the world would be quite a different place if Saudi state beheadings were on the BBC evening news.

A lot of ISIS recruitment is through propaganda videos, and a video on Youtube has a larger reach then one hosted on a more obscure site. It is considered a major issue here in the UK, although it is often overblown like such things always are.

Was curious so worked it out.

Assuming filter people work 40 hour week 50 weeks a year with near perfect efficiency.

(300(60 24 * 365))/2000 = 78840 people.

Of course it's nowhere near that intractable as a simple reputation/time based filter (is this account more than a month old, skip, does this user have a good reputation, skip) would bring that down hugely (and this assumes they watch 100% of each video).

In any case you are still at a good number of people.

So why don't Google just hire 100 000 people to do it? Assuming a generous 30k/yr salary, that is only 3mln/yr.

Because they're better at math than that.

Because that would be 3 billion dollars plus office space, computers, power and HR for 100,000 people.

Call it 4 billion with overhead.

Not a winning idea when they only have 55,000 now.

$3b != $3m

When you're systemically banning speech because it's a "threat"... isn't that literally, exactly, the same argument that every censorship regime in history has used in order to manipulate the public to support things it wouldn't otherwise?

Isn't China, Russia, Singpore, and everyone else pushing that same line? Are we going to tolerate it here?

It's illegal to incite people to violence and make death threats in the USA, even on YouTube.

Hitler's speeches did an amazing job of inciting people to violence, guess we should ban any videos of that, you know just in case.

Google is a private company not a state.

Google is a public company.

Disingenuous use of a different use of the words public/private.

I meant that they are not necessarily expected to treat Youtube as we would expect, say, the US Government to treat a piece of public land like a park (ignoring of course the recent phenomenon of "free speech zones" and such.) You are referring to the fact that fractional ownership of the company is available to nearly any member of the public, as opposed to a company held by a small set of owners.

When does a video do harm? How can you be harmed by watching a video? If you don't like it, close your browser. If you don't want your kids seeing it, using parental controls or watch them when they use the computer.

This just reeks of governments trying to impose censorship by using recent terror events as leverage.

I saw an animal torture video once. It went very fast and because I was in both a state of shock/fear/sadness/whatever, I couldn't even get myself to stop, pause or close it in time. The torture just happened too fast.

I own and care a great deal for my own pets. So what happened to the same animals in the video.. it haunted me for weeks and was a really depressing experience.

Sure, I'm more careful now.. heck, I'm even outright censoring a lot of stuff from myself, because of this experience.

But it sounds like you are saying "videos can never do harm" and I just really disagree with this, because of how I was affected by this particular one.

I'd rather people were able to post animal cruelty videos. That shit needs to be stopped and people need to be aware that this stuff happens.

How are you meant to expose animal cruelty if youtube censors it?

I doubt what you were watching was an exposure video but the content you want filtered is the same.

Why are you trying to "expose" animal cruelty? That's a crime. Give your video to the police. Uploading it to YouTube so the general public can see it not helpful.

And if the police decide to do nothing?

And what of states where creating that video in the first place is itself a crime?

What if the abuse is not criminal, but you wish to convince the public that it should be?

> And if the police decide to do nothing?

What is uploading it to YouTube going to do? The only motive I can think of there is to try and shame someone for something, but that's rarely applicable. Besides, nobody is saying "the police did nothing, therefore I uploaded it to YouTube".

The best reason I can think of for sharing an animal cruelty video with the public (or at least doing so before exhausting any effective strategy) is when you need help identifying the culprit, so you can properly report it. But YouTube isn't the right community for that sort of thing.

> And what of states where creating that video in the first place is itself a crime?

How does that excuse uploading the video to YouTube? If making the video itself is a crime, then sharing it with the public seems like a great way to get yourself in trouble.

If your intention was to try and distribute the video anonymously, surely there's ways to anonymously submit complaints + evidence to the police.

> What if the abuse is not criminal, but you wish to convince the public that it should be?

Good question, perhaps that situation does warrant uploading to YouTube and spreading publicly. Although you'd need to be careful of exposing yourself to a defamation lawsuit there depending on whether the perpetrator in the video is identifiable. Although even in this case I'm skeptical as to the efficacy of this action.

Related to this, I'd say that the best justification for uploading such a video would be if the video is considered to be in the public interest. This would cover situations such as a government-funded institution engaging in unethical/criminal behavior (I'm thinking here of things like Abu Ghraib). But in such a situation I still think uploading to YouTube is probably the wrong approach; I would expect you could find journalists that would be interested in a story like that and would be much better-positioned to bring attention to the unethical/criminal behavior.

Propaganda, especially recruiting. http://www.nbcnews.com/storyline/isis-terror/isis-releases-h...

However, if you were to post a clip from a copyright infringing song -- it gets removed immediately.

It's much easier for a computer to recognize audio than video content. I'm sure in a couple of years, YouTube will automatically flag violent terrorist propaganda.

https://xkcd.com/1425/

Not just any audio but known audio. Finding known video is something they can do already too. That's why some people upload clips from their favorite shows and flip them horizontally to avoid automatic detection. Think of it like a hash for the audio/video portion. Pretty easy to check if it's in the database.

I'm surprised whatever feature detection they use isn't robust against trivial changes like that. Rotation, flipping, and scaling/cropping should all result in the same hash.

We only brushed over how hashes work in college, so I may be totally ignorant, but-- wouldn't rearranging the order of the data (rotate, flip, etc) necessarily change the resulting hash?

They don't use those hashes, they use content fingerprinting: http://en.m.wikipedia.org/wiki/Digital_video_fingerprinting

And why would they? Is this because they are (legally) obliged to filter content? If this is an actual problem then I suppose one way to deal with it is to either hire more staff or restrain the number of videos users can upload -- except the first one doesn't scale and the second one will not be popular with YouTube users. And given that Google is all about doing everything they do at a massive scale, neither of the two suggestions fit their weltanschauung.

> But when individuals flag up problems only a third of it is taken down.

Because they don't give enough importance to these requests, or because many of the requests are invalid? If it's the former, that's something that content hosts should be held accountable for. If people flag stuff for you for free, you better use that information...

Flagging as a means of shutting down opinions people disagree with or to target groups they don't like is unfortunately very common. Sometimes even in organized fashion, where a group will urge its members to flag content from someone they oppose. There have also been instances where a group tried to invoke copyright infringement because someone said their name in a video.

It's assumed this is just an issue of manpower, but it's a much thornier problem. There's a point where a video isn't obviously in violation of community standards and at that point, you're asking one of 100,000 different people to make a judgement call. You're bound to get some wrong and the someone will create a whole other post on HN about it.

How can the host be accountable for not having enough manpower to deal with abusive videos people upload to the service?

"Can't", I think they meant "won't". Google could hire a few more people easily.

According to calculations, they need approximately 70k people to screen every minute of video uploaded onto Youtube.

Even if ML could cut this down by 85%, that is still 10k people. Not a brilliant idea when Google only employs 50k people worldwide.

This is an effort to fight ideas by suppression. My preference would be to fight ideas using other ideas.

Would s/filter/pre-filter/ make for a more accurate headline?

No, they wait for viewers to flag content before they review it. But even reviewing only flagged content is still overwhelming, because the flag function gets abused so much. The new idea is to have a separate channel for known-good agencies to have a priority-flag function that gets reviewed first. Since that feature will hopefully be abused less, it will have less traffic.

I'm not sure the article is saying that; see https://news.ycombinator.com/item?id=8961944

Google is not pre-filtering videos. But it is already "tough to catch all terror related content".

lol they're "overloaded" yet they still find time to screw over musicians? stop being so evil google.

Wow, censorship is hard and has many related risks and costs to society.

I wonder if it's worth it.

Reply on news.ycombinator.com