YouTube admits 'wrong call' over deletion of Syrian war crime videos

http://www.middleeasteye.net/news/youtube-admits-wrong-call-over-deletion-syrian-war-crime-videos-1140126078

by jacobr • 9 years ago

241 135 9 years ago

itaris• 9 years ago

I'm much a proponent of automation as anyone else. But I think right now Google is trying to do something way too hard. By looking for "extremist" material, they are basically trying to determine the intention of a video. How can you expect an AI to do that?

sbov• 9 years ago

The AI doesn't seem to remove them though, it just flags them for human review. In theory, these humans should be the ones determining intent.

duskwuff• 9 years ago

And in this case, there should probably be a separately trained group of reviewers to carefully examine these videos. Not the same group that's quickly checking over videos to see if they're pornographic, for instance.

Raphmedia• 9 years ago

But is Youtube really the right platform for such videos? It's a platform made to host videos in order to put advertiser's ads on them. When I think "raw war videos", I think of Liveleak.

devrandomguy2• 9 years ago

For many users, Youtube is the only video publishing site, and many people get their news through YT. Very few people are aware of the existence of LiveLeak; it is an ineffective platform for spreading awareness.

pryelluw2• 9 years ago

YouTube is now what TV used to be. So yes, you should be able to show this content. However, the platform needs to provide betyer tools to users and producers to aide in categorizing content.

jmcdiesel3• 9 years ago

I dont think the problem is automation...

It's people's expectation for it to be perfect, and the egoic drive to blame someone when something goes wrong. There was no reason for the hype around this story... an AI determinator had a false positive. Thats not google attacking the videos, thats a technical issue and it needs to have zero feelings involved because the entire process happened in a damned computer incapable of feelings...

But everyone needs to feed their outrage porn addiction...

colordrops3• 9 years ago

It's not a technical issue. Software is not yet capable of accurate content detection, and even if it were, it's not clear whether this sort of thing should be automated. It's not like google can just change a few lines of code and the problem is gone.

megous2• 9 years ago

Read the article. It says that humans made the final decisions.

PhasmaFelis2• 9 years ago

Your whole premise is wrong, because the final decisions were made by humans. But even if they weren't, you're still mistaken. If you write a program to do an important task, it is your responsibility to see that it's both tested and supervised to make sure it does it properly. Google wasn't malicious here, but it was dangerously irresponsible.

firmgently• 9 years ago

From the article:

"Previously we used to rely on humans to flag content, now we're using machine learning to flag content which goes through to a team of trained policy specialists all around the world which will then make decisions," a spokeswoman said..."So it’s not that machines are striking videos, it’s that we are using machine learning to flag the content which then goes through to humans."

"MEE lodged an appeal with YouTube and received this response: 'After further review of the content, we've determined that your video does violate our Community Guidelines and have upheld our original decision. We appreciate your understanding.'

Humans at YouTube made the decisions about removing videos. Then, on appeal they had a chance to change their minds but instead confirmed those decisions. Then, because of public outcry, YouTube decided it had been a mistake. 'The entire process happened in a damned computer incapable of feelings' is inaccurate.

white-flame• 9 years ago

[Disclaimer: I'm not a youtuber, so my knowledge is only 2nd hand]

Aside from but related to this story, many people are making a living off of YouTube ad revenues, and the AI is unpredictable in how it will respond in terms of promoting your video content on the front page, as links from other popular videos, and so forth. I think it's also unknown how the AI categorizes the content appropriateness of videos to advertisers, which if categorized the wrong way leaves your stuff unmonetizable.

Basically, people are throwing video content up, but have no way to properly have a feedback loop to gauge whether or not they violate the "proper" protocols that the AI rewards. This really is a problem of automation using (presumably) trained statistical rules where nobody really knows what specifically influences the decisions about their videos.

nthcolumn• 9 years ago

It is the People's expectation that it be perfect. Once they have determined that there is something badly wrong going on in the video destroying it is a violation of U.S. Code § 1519 (destroying evidence with intent). They had better have backups.

Dirlewanger2• 9 years ago

Doesn't matter, it's already out in the wild. This year so far, tons of channels whose videos have had ads for years are being instantly demonetized without explanation. If even one word from a video's title, one tag is on their "controversial" shitlist, it's SOL. Knowing YouTube's track record with this stuff, they will continue to be silent and not give a shit.

Content creators that don't produce content for 5 year-olds need to start looking somewhere else than YouTube.

Fej2• 9 years ago

YouTube is the only option. Smart businesspeople (i.e. most channels over 500k or even 200k subs) have alternate sources of revenue, mainly sponsorships but also Patreon.

notananthem• 9 years ago

I mean its also youtube so .. who cares

johansch• 9 years ago

Case in point:

https://twitter.com/lazygamereviews/status/89825992922771046...

CydeWeys• 9 years ago

Demonetization as a feature comes from the advertisers' desire, not from YouTube itself. Advertisers are very sensitive about what they will allow their brands to be associated with, and they demand the tools to configure their ad campaigns to not display on broad swaths of potentially offensive content. YouTube has had some huge issues with advertiser complaints of ads on offensive content in the past. If anything, YouTube should be praised for continuing to host demonetized content at a loss, since it's not making any advertising revenue for them.

People who make non-advertiser-friendly videos need to figure out some other kind of way to make their money. Patreon is excellent for that.

anotherbrownguy5• 9 years ago

The demonetization effort is too targeted and obvious to be explained by "AI did it". If AI did it, they could undo it after appeal which they don't.

nsxwolf• 9 years ago

Basically every YouTuber I follow has complained about having videos demonetized this week. Subjects ranging from video game reviews to body dysmorphic disorder.

It really seems they've bitten off more than their machine learning algorithms can chew here.

DINKDINK2• 9 years ago

How can you expect an AI to do that?

Machine Learning is just a way to launder bias. What will be defined as extremist by US companies will favor the West but discounting Western extremism and overplay non-Western extremism.

norea-armozel2• 9 years ago

I think YouTube really needs to hire more humans to review flagging of videos rather than leave it to a loose set of algorithms and swarming behavior of viewers. They assume wrongly that anyone who flags a video is honest. They should always assume the opposite and err on the side of caution. And this should also apply to any Content ID flagging. It should be the obligation of accusers to present evidence before taking content down.

schoen• 9 years ago

They assume wrongly that anyone who flags a video is honest.

I don't think they assume that at all. If they did, you'd see at least an order of magnitude more videos removed.

I agree with the sentiment of your criticism, but I think we could phrase it more in terms of prior probabilities or something about the false positive and false negative rate in their review process. Flagging of videos is extremely common and even a small amount of unreliability in the review process translates into a huge number of mistakes.

Also, users of the site don't actually agree with each other much at all about which removals were in error; we could say that there's absolutely abysmal inter-rater reliability if the end-users of the site are the "raters" of the quality of content removal decisions.

Also, most people who flag things don't necessarily know much at all about YouTube's terms of service or how YouTube has interpreted or applied them in the past, so it's hard to be clear on what it means for flaggers to be honest or dishonest. Probably the most common meaning of flagging is "ugh, I'm upset that this video is up on YouTube".

norea-armozel2• 9 years ago

The biggest problem, imo, isn't the random flagger but rather the concerted actions of groups to flag videos. This is obvious in terms of reddit or 4chan users swarming a channel they don't like. This kind of behavior needs to be mitigated in some way. I think a quick solution would be to force a cool down timer on flagging of 24-48 hours for all users to ensure they're not abusing the system. That should include random users who file DCMA takedowns that aren't partnered with Youtube in some way.

7622362• 9 years ago

Automation is the only real solution. These types of conversations seem to always overlook how normal people don't want to watch such videos. Do you want to spend your day watching this stuff to grade them?

EpicEng2• 9 years ago

Yet youtube is admitting that the videos should not have been pulled, and there's no AI in the world that could have made the right call here. So... what sort of automation are you suggesting? It seems as though the real solution is the exact opposite of what you're proposing; human review by better trained personnel with clearly defined criteria.

7622362• 9 years ago

This is irresponsible to the trained personnel. We have job safety requirements for people working on assembly lines. We should also have psychological safety for people, and there are lots of stories of employees that are paid to view the toxic videos suffering (even developing PTSD).

qbrass• 9 years ago

The real solution is to continue to blame the machines whenever an angry mob decides to show up at their doorstep instead of admitting it was their clearly defined criteria that was at fault.

em3rgent0rdr• 9 years ago

"Do you want to spend your day watching this stuff to grade them?"

What about providing more tools for community to categorize disturbing videos other than simply "flag".

ben_w3• 9 years ago

I will literally never choose to watch a disturbing but real event. War crimes, either in the form of documenting them or promoting them, will be flagged as "people like Ben never click on previews of this video, don't waste time suggesting it to them in future". I won't even get as far as the page the flag button is on, never mind any other options.

paganel• 9 years ago

These types of conversations seem to always overlook how normal people don't want to watch such videos

That makes me not normal, I guess. Plus, the NSFL videos were almost instantly taken off YT immediately after having been uploaded, what had remained was really interesting stuff documenting the war in Syria (or at least interesting for people such as myself, a guy interested in wars and conflicts in general).

vacri4• 9 years ago

People also don't want to spend their day scrubbing toilets, but there are plenty of janitors out there.

hnaccy• 9 years ago

So hire abnormal people?

I'm pretty sure there's a healthy chunk of population who would be unfazed and would love a cushy job judging videos.

ezoe• 9 years ago

What I don't like about those web giant services is, to get a human support, it requires to start social pressure like this.

If they fucked up something by automation, contacting to human support is hopeless unless you have very influential SNS status or something.

balozi• 9 years ago

Well, the AI did such a bang–up job sorting out the mess in comment section that it got promoted to sorting out the videos themselves.

charlesism3• 9 years ago

5 minutes reading YouTube comments is enough to make me ill. I don't know what a couple hours a day after school would do to a person after ten years. We'll all find out, I guess, once this generation of kids reaches adulthood.

dilemma• 9 years ago

Probably has the same effect as computer games did on us - little.

goatsi• 9 years ago

The AI isn't the only problem, Syrian government supporters are actively reporting evidence of war crimes in an attempt to get it removed.

https://twitter.com/EliotHiggins/status/898496719125700608

RandVal30142• 9 years ago

Something people need to keep in mind when parsing this story is that many of the effected channels were not about militancy, they were local media outlets. Local outlets that only gained historical note due to what they documented as it was unfolding.

In Syria outlets like Sham News Network have posted thousands upon thousands of clips. Everything from stories on civilian infrastructure under war, spots on mental health, live broadcasts of demonstrations.

Everything.

Including documenting attacks as they happen and after they have happened. Some of the effected accounts were ones that documented the regime's early chemical weapons attacks. These videos are literally cited in investigations.

All that is needed to get thousands upon thousands of hours of documentation going back half a decade deleted is three strikes.

Liveleak is not a good host for such outlets because it is not what these media outlets are about. Liveleak themselves delete content as well so even if the outlets fit the community it would not be a 'fix.'

osteele• 9 years ago

HN discussion of deletion event: https://news.ycombinator.com/item?id=14998429

alexandercrohde2• 9 years ago

I think youtube needs to consider backing off regulating political content.

The fact is politics and morality are inherently intermingled. One can use words like extremist, but sometimes the extremists are the "correct" ones (like our founding fathers who orchestrated a revolution). How could any system consistently categorize "appropriate" videos without making moral judgements?

secfirstmd• 9 years ago

Agree but it's far easier for politicians (May, Rudd, Trump, Turnball, Putin, Netanyahu, Xi Jinping, Mugabe, Zuma, Chan-ocha, el-Sisi, Erdoğan, Khamenei, Maduro - the list is endless) to blame videos on Youtube for radicalising people then it is to tackle the long running political, historical and socio-economic grievances that fuel the fire.

jjoonathan• 9 years ago

Not only politicians, unfortunately.

anotherbrownguy5• 9 years ago

Do you think any rich established British company supported the American revolution? Why do you expect Google to help anything new or different regardless of how "correct" it is? Big established players can't afford revolutions, too risky for them.

syshum5• 9 years ago

Rich Established French Companies did, as did Nations and people outside the British Empire.

Your comments would have to mean YouTube/Goolge is a Syrian Company not a American Company when it pertains to the Topic at hand.

Your Analogy is simply defunct on many levels, further your switch from British Companies supporting the revolution, to "Google Supporting anything new or different" is nonsensical. Google supports all kinds of "New and Different" things that will likely change the face of society in profound ways, things like AI, Self Driving Cars and other Moonshot Programs.

No I really do not understand what point your driving at here

sqeaky2• 9 years ago

A decision not to decide is still a choice.

A decision to attempt to remain neutral should be carefully considered.

Spivak4• 9 years ago

Sure, but whether the choice has any political content depends on whether YT considers themselves publishers or infrastructure. If they just want to be the world's video hosting site then they're not taking any political stance by staying out of it and just hosting what people upload.

kenjackson• 9 years ago

Aren't you always making moral judgments on non-political videos too? Which videos are "adult", which videos are scams, which videos are inappropriate, etc...

alexandercrohde2• 9 years ago

So I think the distinction here is that "what is porn?" (again, a deep question that the supreme court has addressed, considering how much nudity there is in art) does have a non-moral answer, even if it's arbitrary (such as whatever arouses 10% of people).

The other distinction is that with political speech is that if you screw up that line you hamper criticism of authoritarian regimes (for example). If you set the bar too tight on your porn filter, it's much less likely you stifling political progress.

thaumasiotes3• 9 years ago

The fact is politics and morality are inherently intermingled.
How could any system consistently categorize "appropriate" videos without making moral judgements?

Politics and morality are inherently intermingled, I agree.

Appropriateness and morality are the same concept.

Spivak4• 9 years ago

I disagree with your last statement. A video featuring gratuitous violence is not appropriate for YouTube Kids but there's no moral statement there.

clarkmoody• 9 years ago

Interesting that we have a whole vocabulary of words reserved for those on the "wrong side": insurrection, sedition, traitor, deserter, criminal, smuggler, terrorist, extremist, rebel, revolutionary, etc.

But from the "right side" of an ideology, those are the freedom fighters, visionaries, defenders, Founding Fathers, Underground Railroad, idealists, etc.

Controlling vocabulary is a very powerful tactic in politics, as illustrated in 1984. People respond in very predictable ways to certain words, hence their power.

3131s• 9 years ago

Why doesn't Youtube remove this "extremist" content I wonder...

https://www.youtube.com/watch?v=BWsnYxJ1ghA

baybal2• 9 years ago

As I understand, their "machine learning" grep "stuff in arabic"| send to people in places like Tunis, Albania and Morocco (true, both Google and Facebook run their "AI flagging" shops there, blogs giving account of that are well known)

tenpies• 9 years ago

I think youtube needs to consider backing off regulating political content.

They can't. We've entered an era where a disagreeable tweet is - in the mind of some people - a good reason to dox you and destroy your life. This extends to companies. Your ad shows up next to an opinion piece that they disagree with? Time to start a #activist campaign to boycott your company and mention how you would probably sell your goods to Hitler and your board is full of white patriarchs who are probably racist, anti-Muslim, anti-immigrants, anti-women, and closet Neo Nazis.

The companies are also to blame since they've been mostly spineless. Save a few companies, they fold and pull their advertising the minute these #activists put them in their sights.

To an advertising delivery company like Youtube, which is probably barely at breakeven, losing advertising dollars is crushing. Over the years they've been slowly reducing how much money they bleed, so a reversal of the trend is grounds for alarm.

So Youtube has to either ween itself off advertising (virtually impossible, since Google is an advertising company) or make sure the content is advertiser friendly.

bogomipz• 9 years ago

I think you bring up an interesting point. Big corporate advertisers are demanding of youtube the same predictability and sanity checks which they have with traditional media buys such as television, print etc. However part of the appeal of a medium like youtube is in much of the unscripted, random, spontaneous, esoteric content. And these two interests seem to be at odds with each other.

jimmy2020• 9 years ago

i really don't know how to describe my feeling as a syrian when i know the most important evidence that witnessed the regime crimes were deleted because of wrong call. And it's really confusing how artificial algorithm get confused between what is is obvious as isis propaganda and a family buried under the rubble and this statement makes things even worse. mistakenly? because there is so many videos? just imagine that may happen to any celebs channel. Will youtube issue the same statement? dont think so.

tdurden• 9 years ago

Google/YouTube needs to admit defeat in this area and stop trying to censor, they are doing more harm than good.

tzs• 9 years ago

Is there sufficient data available to actually draw that conclusion? All we generally hear about are videos that are taken down because they are false positives, and usually those are just down temporarily. We'd have to know how many are taken down that are not false positives, and how harmful those were, in order to begin to estimate net harm or good.

Buge• 9 years ago

Just admit defeat and let all the advertisers pull out?[1] Then go bankrupt and (like Soundcloud was on the verge of) let everything get deleted?

[1] https://www.theguardian.com/technology/2017/mar/25/google-yo...

CaptSpify4• 9 years ago

Maybe the advertisement-funded content needs to be pulled into question then

pgnas• 9 years ago

YouTube (google) has become the EXACT opposite of what they said they were not going to do.

They are evil.

zeep• 9 years ago

At least, someone at Google was good (honest) enough to drop that motto...

williamle8300• 9 years ago

Google (parent company of YouTube) already sees itself as the protector of the public's eyes and ears. They might be contrite now but they behave as a censorshipping organization.

DINKDINK2• 9 years ago

What about all the speech that's censored that doesn't have enough interest or political clout to make people aware of the injustice of its censoring.

molszanski2• 9 years ago

Let's look at the bigger picture. First, in March some newspapers find an extremist video. It has ~14 views and YT advertising all over it. They make a a big deal out of it. As a result YouTube looses ad clients and tons of money.

Then, as a response, they make an alg. They don't want people to call them a "terrorist platform" ever again. Hence they take down the videos.

Now, this algorithm is hurting the bystanders. IMO the real problem is a public and business reaction to the initial event.

And this peace of news is an inevitable consequence.

chii• 9 years ago

if you look at this sequence of events, the deeper trouble is really that of the newspapers (or any other media outlet). Their incentive is to find things people are going to get shocked by, and report it to get views/clicks (that fuel advertising). They aren't going to fully explore the pros/cons, like long form journalism, since that's expensive, and they get paid peanuts from advertisers.

The public's reaction is going to be the one of outrage, because the news article was _designed_ to evoke such feelings! If the public sentiment was more liberal, then the news article would've picked an even more extreme event, and evoke the same reaction.

Fix the problem at its root - that of advertising funded news and media.

molszanski2• 9 years ago

Couldn't agree more. In my opinion, advertising is not the _only_ problem. Remember, YT is a competitor to the "old media". And hurting them is in their best interest. Hope that initiatives like wikitribune will pave the way to a better future

miklax• 9 years ago

Bellingcat account should be removed, I agree on that with YT.

YouTube admits 'wrong call' over deletion of Syrian war crime videos

Related Stories