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

141–150 of 517 posts

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

#141

Post by idatum »

> Altering footage of real events or places: Such as making it appear as if a real building caught fire, or altering a real cityscape to make it appear different than in reality.

What about the picture you see before clicking on the actual video? This article of course is addressing the content of the videos, but I can't help but look at the comically cartoonish, overly dramatic -- clickbait -- picture preview of the video.

For example, there is a video about a tornado that passed close to a content author and the author posts video captured by their phone. In the preview image, you see the author "literally getting sucked into a tornado". Is that "altered and synthetic content"?

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

#142

Post by IshKebab »

lp0_on_fire wrote: post #121
bgirard wrote: post #97

I wouldn't be surprised if this ends up like prop 65 cancer warnings, or cookie banners. The intention might be to separate believable but low quality hallucinated AI content spam from high quality manual content. But it will backfire like prop 65. You'll see notices everywhere because increasingly AI will be used in all parts of the content creation pipeline. I see YouTube's own guidelines in the article and they se…

Am I the only one who is bothered by calling this phenomenon "hallucinating"? It's marketing-speak and corporate buzzwords to cover for the fact that their LLMs often produced wrong information because they aren't capable of understanding your request, nuance, or the training data it used is wrong, or the model just plain sucks. Would we tolerate such doublespeak it were anything else? "Well, you ordered a side of fr…

Nonsense. It isn't marketing speak to cover for anything. It's a pretty good description of what is happening.

The reason models hallucinate is because we train them to produce linguistically plausible output, which usually overlaps well with factually correct output (because it wouldn't be plausible to say e.g. "Barack Obama is white"). But when there isn't much data to show that something that is totally made up is implausible then there's no penalty to the model for it.

It's nothing to do with not being able to understand your request, and it's rarely because the training data is wrong.

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

#143

Post by jahewson »

mock-possum wrote: post #103
AnthonyMouse wrote: post #46

These rules have really nothing to do with AI. They're trying to impose a ban on deceit.

Requiring labeling / disclosure is not the same as banning.

In this case it is, because labelled deceit is not deceptive.

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

#144

Post by samatman »

arduanika wrote: post #49

Is there a word missing from the title here? Requires whom?

The title was editorialized, which people do far more often than they should. The original title, with the domain name next to it, would have been fine.

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

#145

Post by orbital-decay »

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…

Labeling AI-generated content (assuming it works) is beneficial for Google, as they can avoid some dataset contamination.

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

#146

Post by AnthonyMouse »

ajross wrote: post #122
AnthonyMouse wrote: post #94

Earlier quoted context omitted.

They don't have anything specifically to do with cars. They apply equally to motorcycles, trucks and anything else that could go that fast. Get pulled over in a tank or a hovercraft and try to tell the officer that you can't have been speeding because it isn't a car. Should we like deepfakes any better if they're created by a nation state using pre-AI Hollywood production technology, or "by hand" with Photoshop etc.?…

You missed the analogy, so I'll spell it out: before we had cars[1], we couldn't go fast on roads, and there were no speed limit signs. Before we had AI, we couldn't deceive people with easy fakes, and so there was no need to regulate it. Now we do, and there is, and YouTube did. Trying to characterize this as not related to AI just isn't adding to the discussion. Clearly it is a response to the emergence of AI fakes…

Trying to shovel in "and all of the other stuff" breaks the analogy though. Misinformation isn't new. Image gen is hardly the first time you could create a fictional depiction of something. It's not even the first time you could do it with commonly available tools. It's just the moral panic du jour.

YouTube did this because the EU passed a law about it. The EU passed a law about it because of the moral panic, not because the abstract concept of deception was only recently invented.

It's like having cars already, and speed limits, and then someone invents an electric car that can accelerate from 0 to 200 MPH in 5 seconds, so the government passes a new law with some arbitrary registration requirements to satisfy Something Must Be Done.

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

#147

Post by jprete »

Aardwolf wrote: post #63
Jensson wrote: post #56

Earlier quoted context omitted.

If you manually did enough work have the copyright it is fine. But since AI can't legally have copyright to their music Google probably wants to know for that reason.

I hope there is some kind of middle ground, legally, here? Like say you use a piano that uses AI to generate artificial piano sounds, but you create and play the melody yourself: can you get copyright or not?

IANAL. I think you'd get copyright on the melody and the recording, but not the sound font that the AI created.

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

#149

Post by samatman »

cush wrote: post #81
lampiaio wrote: post #62

AI that is indistinguishable from reality is a certainty for the not-so-distant future. That future will come, and it will come sooner than anyone's expecting. Yet all I see is society trying to prevent the inevitable from installing itself (because it's "scary", "dangerous", "undermines the very pillars of society" etc.), instead of preparing itself for when the inevitable occurs. People seem to have finally accepte…

There are some interesting hardware solutions from camera makers that provide provably authentic metadata and watermarks to videos and images - mostly useful for journalists, but soon consumers will expecting this to be exposed on social media platforms and those they follow on social media. There really are genuinely valuable things happening in this space.

This will always be spoofable by projecting the AI content onto the sensor and playing it to the microphone. Which will give the spurious content a veneer of authenticity, this is within reach of a talented malicious amateur, and would be trivial for nationa-state actors to do at scale.

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

#150

Post by speff »

AnthonyMouse wrote: post #132
speff wrote: post #120

Earlier quoted context omitted.

From the report button on a youtube video: Misinformation - Content that is misleading or deceptive with serious risk of egregious harm. That sounds like a quagmire of subjectivity to enforce. You can argue whether generated content was created to mislead or whether it would cause _egregious_ harm. Now there's no more arguing. Is it generated or is it real - and is it marked if generated? I still fail to see the down…

You're acting like this is a court with a judge. The company makes up subjective rules and then subjectively enforces them. There was never any arguing to begin with, they just ban you if they don't like you, or at random for no apparent reason, and you have no recourse. > Now there's no more arguing. Is it generated or is it real - and is it marked if generated? I still fail to see the downside here. How is this sup…

Youtube isn't some mom n' pop operation - they do have a review process and make an attempt at following the rules they set. They can ban you for any reason...but they generally don't unless you're clearly breaking a rule.

I'm not getting into the rabbit hole of finding out if it was actually generated - once again, that's a different problem. One that I already mentioned 2 comments back. My point was that there is less subjectivity with this rule. If the content is found to have been generated and isn't marked, then there are clear grounds to remove the video.

What youtube does when there is doubt is not known yet. I don't deal in "well this _could_ lead to this".

Post reply on HN