Hate Speech Laws and Social Media. The beginning of the end?

[Max_S]

@BigT

Video filters aren’t the only filters that can be used. Text filters can also help, assuming the video is described by either the person making it (which I believe it was in this case) or by people commenting. And, once they identify the video and block it, they can do better than hash algorithms by analyzing the actual frames as images.

There’s also something mentioned in the articles you linked. The ability to report is hidden behind a not always visible 3-dots menu. They could put it front and center. It’s rather strange that the police were called and Facebook had time to block the video before anyone reported it. Someone had to have seen it and called the police, but didn’t figure out how to report it.

I find it unlikely that Facebook would stop offering livestreams in Australia. Most likely, they will make changes like I described, and show they are doing what they can to make the process as expeditious as possible. The law doesn’t set some limit, so even if the result isn’t much better, it still shows they are following the law.

Text filters are a good idea but a bad actor could circumvent that by foregoing the text description and either recording a printed manifesto or reciting one. It would certainly help flag broadcasts by analyzing their comments, though written speech is difficult to analyze especially when dealing with in-jokes or abstract ideas.

Remember that there were bad actors aside from the perpetrator. Facebook says “we saw a core community of bad actors working together to continually re-upload edited versions of this video in ways designed to defeat our detection.”

And I don’t know if technology is fast enough to analyze live videos frame-by-frame. Facebook CTO Mike Schroepfer says they can distinguish two images in hundredths of a millisecond, or billions of times per day. If Facebook had 12.8 million Live broadcasts each lasting 10 minutes, that would be 7.68 billion seconds of video to process in real time. Let’s say the videos have 24 frames per second. That’s 184.32 billion frames to process per day. By comparison there are 8.64 billion “hundredths of a millisecond” in a day. If Facebook’s best technology needs a hundredth of a millisecond to distinguish two images, that is about 5% the speed required. This is assuming you wanted to take down videos by checking for one specific frame. The time required would grow exponentially if you needed to check for more than one specific frame.

I agree with you wholeheartedly about the user interface, though I think Facebook will streamline their “reporting experience” without regard to the Australian law. And on second thought, you are probably right about Facebook’s response to the Australian law. But then all we have is a law that doesn’t address the problem it was explicitly designed to address.

~Max

[Max_S]

@Sitnam

If YouTube can find a way to make sure their automated commercials, for say Coca Cola or Nike, aren’t promoted in the middle of a racist screed from some Oklahoma basement troll I think we can find a way to solve this.

There just needs to be a financial incentive for the business who runs the platform, carrot…stick…both…either.

My impression from this Washington Post article is that Youtube/Google, with all of their technology, could not handle the crisis.

The article explains that Youtube primarily blocked videos with content hashes; they reduced exposure by recommending videos from established news sources during the crisis; their artificial intelligence was confused because the shooter’s name was not publicized; and in the end it wasn’t enough.

@Elizabeth Dwoskin and Craig Timberg

As its efforts faltered, the team finally took unprecedented steps — including temporarily disabling several search functions and cutting off human review features to speed the removal of videos flagged by automated systems. Many of the new clips were altered in ways that outsmarted the company’s detection systems.

~Max

[Chronos]

Wait a minute. I’m sure I could go on YouTube right now and find a whole bunch of videos containing footage of the airplanes flying into the World Trade Center. Under this law, would YouTube be required to take down all of those videos?

[D_Anconia]

@Sitnam

If YouTube can find a way to make sure their automated commercials, for say Coca Cola or Nike, aren’t promoted in the middle of a racist screed from some Oklahoma basement troll I think we can find a way to solve this.

Why Oklahoma? There are basement trolls in NY and CA, and every other state.

[VinylTurnip]

Good to know that the Oklahoma basement troll can seek out-of-state help creating his racist screed, if needed.

[begbert2]

@Chronos

Wait a minute. I’m sure I could go on YouTube right now and find a whole bunch of videos containing footage of the airplanes flying into the World Trade Center. Under this law, would YouTube be required to take down all of those videos?

Only if it was recorded by the perpetrators or (more likely) one of their accomplices.

So, unless you’re a particularly arcane conspiracy theorist, those videos should be okay.