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

YouTube now requires to label their realistic-looking videos made using AI

blog.google

Post reply on HN
Print view View original

481–490 of 517 posts

Re: YouTube now requires to label their realistic-looking videos made using AI

#481

Post by gretch »

alphazard wrote: post #188
the_duke wrote: post #31

They don't bother to mention it, but this is actually to comply with the the new EU AI act. > Providers will also have to ensure that AI-generated content is identifiable. Besides, AI-generated text published with the purpose to inform the public on matters of public interest must be labelled as artificially generated. This also applies to audio and video content constituting deep fakes https://digital-strategy.ec.eu…

Is anyone else worried about how naive this policy is? The solution here is for important institutions to get onboard with the public key infrastructure, and start signing anything they want to certify as authentic. The culture needs to shift from assuming video and pictures are real, to assuming they are made the easiest way possible. A signature means the signer wants you to know the content is theirs, nothing else…

Example: a video shows a group of police beating up a man for a minor crime (say littering). The video is signed by Michael Smith (the random passerby who filmed it on his phone). The video is published to Instagram and shared widely.

How do you expect people to take the authenticity of this video?

Re: YouTube now requires to label their realistic-looking videos made using AI

#482

Post by lucianbr »

AnthonyMouse wrote: post #204
lucianbr wrote: post #200

Earlier quoted context omitted.

> they just ban you if they don't like you, or at random for no apparent reason, and you have no recourse. The ban recourse problem is the opposite. This is the "keep recourse": "this video is obviously bad, but google doesn't feel like taking it down, and there is nothing I can do about it". Now there is, and it can actually go to a court with a judge in the end, if Google is obstinate. You didn't have a right to be…

> The thing is, they can't host you as they like, if you're breaking this rule. Except that the rule can be satisfied just by labeling it, and if there are penalties for not labeling but no penalties for labeling then the obvious incentive is to stick the label on everything just in case, causing it to become meaningless. To prevent that would require prohibiting the label from being applied to things that aren't AI-…

You could be right. But I wonder if truly nobody will want to claim their video is not AI-generated. Seems like some people will, and they would get an advantage out of it. Yes, Fox News claims they're entertainment and nobody reasonable would believe them. But not all news channels do this.

Did the California proposition 65 really result in cancer labels on everything? Or is it just hyperbole? I suppose having a lot of labels is still bad, even if they're not technically on everything.

Re: YouTube now requires to label their realistic-looking videos made using AI

#483

Post by BlueTemplar »

pojzon wrote: post #385
Gigachad wrote: post #376

Earlier quoted context omitted.

It just goes back to trusting the source. If 5 media orgs post different recordings of the same political speech, you can be reasonably sure it actually happened, or at least several orders of magnitude more sure than if it's one blurry video from a no name account.

And then you learn all of those media orgs are owned by the same billionare. There will be no way to say something is true beside seeing it with own eyes.

Then that's a single media org.

Re: YouTube now requires to label their realistic-looking videos made using AI

#484

Post by RandallBrown »

pixl97 wrote: post #463
RandallBrown wrote: post #312

Earlier quoted context omitted.

I imagine it would be something handled pretty automatically for everyone. If Apple wanted to sign every photo and document the iPhone they could probably make the whole user experience simple enough for most grandmas. Some people will certainly give away their keys, just like bank accounts and social security numbers today, but those people probably aren't terribly concerned with proving the ownership of their onlin…

>I imagine it would be something handled pretty automatically for everyone. Then your imagination fails you. If it is automatic/easy, then you have the 'easy key' problem, such as the key is easy to steal or copy. For example is it based on your apple account? Then what occurs with an account is stolen? Is it based on a device, what happens when the device is stolen? Who's doing the PKI? Is it going to be like https,…

When your account is stolen someone can create "official" documents in your name and impersonate you. There could be a system for invalidating your key after a certain date to help out with those situations.

For prominent people who actually have to worry about being impersonated they could provide their own keys.

The infrastructure could be managed by multiple groups or a singular one like the government. The point isn't to be a perfect system, it's to generate enough trust that what you're looking at is genuine and not a total fraud.

In a world where AI bots are generating fake information about everyone in the world, that kind of system could certainly be built and be useful.

Re: YouTube now requires to label their realistic-looking videos made using AI

#485

Post by dataflow »

michaelt wrote: post #465
dataflow wrote: post #358

Earlier quoted context omitted.

That wouldn't look nearly realistic. And it would be significantly harder to achieve for most people anyway.

It's true that 1990s pirated videos where someone snuck a handheld camera into the cinema were often very low quality. But did you know large portions of The Mandalorian were produced with the actors acting in front of an enormous, high-resolution LED screen [1] instead of building a set, or using greenscreen? It turns out pointing a camera at a screen can actually be pretty realistic, if you know what you're doing.…

That's a freaking massive LED wall... with professional cinematography on top. If you believed my comment was intended to imply that I believed that's somehow impossible, well... you and I have a very different understanding of what it means to "just take a picture of a high-resolution display"...

Re: YouTube now requires to label their realistic-looking videos made using AI

#486

Post by pctf »

gretch wrote: post #481
alphazard wrote: post #188

Earlier quoted context omitted.

Is anyone else worried about how naive this policy is? The solution here is for important institutions to get onboard with the public key infrastructure, and start signing anything they want to certify as authentic. The culture needs to shift from assuming video and pictures are real, to assuming they are made the easiest way possible. A signature means the signer wants you to know the content is theirs, nothing else…

Example: a video shows a group of police beating up a man for a minor crime (say littering). The video is signed by Michael Smith (the random passerby who filmed it on his phone). The video is published to Instagram and shared widely. How do you expect people to take the authenticity of this video?

The signing happens using hardware inside the phone or camera. It implies that it came from a real device and thus was not AI created or modified. Check out https://c2pa.org/specifications/specifications/1.4/attestati... for more details

Re: YouTube now requires to label their realistic-looking videos made using AI

#488

Post by dev1ycan »

barfbagginus wrote: post #322
tkiolp4 wrote: post #284

Earlier quoted context omitted.

Basically, Google decides what’s real and what’s not. Cool.

Don't worry Google has become incompetent - it is in the "Fading AOL" portion of its life cycle. Do you remember how incontinent AOL became in the final minutes before it became completely irrelevant? Sure it's taking longer with Google. But it's not a process that they can reverse. That means the system will be really really awful. So challengers can arise - maybe a challenger that YOU build, and open source!

Eh, I think their control of Google, Youtube and Android is enough to keep them afloat while being as you said, incompetent.

Re: YouTube now requires to label their realistic-looking videos made using AI

#489

Post by dev1ycan »

sigmoid10 wrote: post #3

>Some examples of content that require disclosure include: [...] Generating realistic scenes: Showing a realistic depiction of fictional major events, like a tornado moving toward a real town. This sounds like every thumbnail on youtube these days. It's good that this is not limited to AI, but it also means this will be a nightmare to police.

You see thumbnails? I haven't seen a thumbnail in years, I use both DeArrow and Sponsorblock, youtube is very watchable.

Re: YouTube now requires to label their realistic-looking videos made using AI

#490

Post by hobo_in_library »

HlessClaudesman wrote: post #414
jjcm wrote: post #37

I think it's smart to start trying things here. This has infinite flaws with it, but from a business and learnings standpoint it's a step toward the right direction. Over time we're going to both learn and decide what is and isn't important to designate as "AI" - Google's approach here at least breaks this into rules of what "AI" things are important to label: • Makes a real person appear to say or do something they…

Porn classification / regulation boils down to: "I'll know it when I see it." Implying the existence of some hyper vigilant seer who can heroically determine what we should keep behind the video storre curtain of dencey, as if no grey areas exist. This also has the problem of requiring actual unbiased humans to view and accurately assess everything, which of course does not scale. Perhaps AI classification is the mir…

> We could conceivably use good AI to spot the bad AI

I suspect this is Google's actual goal with the tagging system. It's not so much about helping users, rather it's a way to collect labeled data which they can later use to train their own "AI detection" algorithms

Post reply on HN