Live data from Hacker News

YouTube to automatically label AI-generated videos

blog.youtube

361–370 of 881 posts

Re: YouTube to automatically label AI-generated videos

#361
post #332

Earlier quoted context omitted.

Right now the problem is the flood of low-quality AI spam that might (or might not) be low hanging fruit. We can worry about high quality AI artifacts later if that becomes a problem. (and yes, there is no guarantee that YouTube won't fail due to these spammers)

But is an algorithmic AI detector really a thing? I get the idea: get 10k each samples of human data and AI data, train a simple classifier until it gets 99.9999% accuracy or Is such tool feasible at all with current state of AI technology, or is it just a reasonable take from the past that may not be so reasonable anymore?

We will find out shortly? YouTube is the one saying they are going to implement this:

"If a creator doesn’t specify whether or not they used AI, but our systems detect significant photorealistic AI use, we will now automatically apply a label."

Re: YouTube to automatically label AI-generated videos

#362
post #12

> “If a creator doesn’t specify whether or not they used AI, but our systems detect significant photorealistic AI use, we will now automatically apply a label,” YouTube said. > YouTube creators who believe their content was incorrectly flagged as AI-generated can modify the disclosure status using the YouTube Studio tool. What's the general overall state of AI-based AI flagging tools development? They seemed to have…

That's not that helpful. For example there's thousands of 'psychology' videos with descriptive bait titles like 'the psychology of people who don't click on video recommendations' which are AI-screipted, narrated, and animated. If you look at the channels the same creator will have hundreds of videos that are micro-targeted to absurdity, like 'the psychology of women who spend too much time on youtube' or 'the psychology of people who don't watch big sports games'. Most of these seem to have tens or hundreds of thousands of views. Maybe the views are botted as well, but clop producers have every incentive to spend a few hundred $ in tokens and generate 50 or 100 new videos to just blanket the recommender engine and shut out regular content that takes time and effort to produce.

I've drastically cut my use of YouTube (even though there are creators I like and wish to support) because I am so tired of wading through all the junk.

Re: YouTube to automatically label AI-generated videos

#363

I have a hard time believing that AI can be used to label AI-generated videos without there being a significant number of false positives/negatives. I think back to ZeroGPT and it labeling the Declaration of Independence as AI-generated.

videos and images have non-ai methods for detecting if its ai, its easier than text. There are some image artifacts in AI that can be found statistically.

Re: YouTube to automatically label AI-generated videos

#364
post #242

Earlier quoted context omitted.

There are millions of people making music in an ever-expanding set of genres. The idea that no one is making 80s or 90s style music is absurd. I guess she can listen to slop but maybe just look around a little instead? Edit: slop not slob

> The idea that no one is making 80s or 90s style music is absurd. The idea that only humans can make music is absurd. > I guess she can listen to slob but maybe just look around a little instead? The idea that AI generated = slop is absurd. Humans create just as much, if not more slop. Look at 99% of "professional" output in creative fields. It's awful. A human with taste steering AI tools can be better than a "clas…

Huh, I wonder what happens when people stop making real music for AIs to train on, then.

Re: YouTube to automatically label AI-generated videos

#365
post #190

Earlier quoted context omitted.

> People praising it in the comments seem none the wiser Or.. they simply like it? Regardless of what we think about it

I think the op mean people writing stuff like: "Amazing what a human soul can create", "This is such a beautiful song. I'm so happy it's not another AI slop" type of comments. I have a fairly popular youtube channel with AI generated music, I make it very obvious that it's AI, yet I still get hundreds of those comments a month.

I make it very obvious that it's AI, yet I still get hundreds of those comments a month.

That suggests you've done a good job of directing the AI to generate what people like.

Re: YouTube to automatically label AI-generated videos

#366

Curious to see if this will apply to music. YouTube seems to be filled with AI music these days - just do a search for "focus music" or the like, and you'll see creators pushing new 1-hr tracks every few days with no mention of where the music came from or the fact it is AI generated. People praising it in the comments seem none the wiser (or perhaps they're also bots).

That's actually how I fell down the rabbithole --- found these long mixes and started listening, enjoyed what I heard, then tried to find out more about the author and realised it was AI-generated. People praising it in the comments seem none the wiser (or perhaps they're also bots). I admit to leaving praise on some of them, because they do sound really good, much better than what I thought AI music could be. Someon…

Yeah, I had the same experience, and it makes sense for companies like Spotify. I do hope that this doesn’t hinder “innovation” in music (eg people being creative and introducing new types of music), because as usual with AI, it’s just really good at imitation, not necessarily creating new things.

To be honest, as long as the music is to my liking, I don’t really care all that much.

Re: YouTube to automatically label AI-generated videos

#369

Curious to see if this will apply to music. YouTube seems to be filled with AI music these days - just do a search for "focus music" or the like, and you'll see creators pushing new 1-hr tracks every few days with no mention of where the music came from or the fact it is AI generated. People praising it in the comments seem none the wiser (or perhaps they're also bots).

That's actually how I fell down the rabbithole --- found these long mixes and started listening, enjoyed what I heard, then tried to find out more about the author and realised it was AI-generated. People praising it in the comments seem none the wiser (or perhaps they're also bots). I admit to leaving praise on some of them, because they do sound really good, much better than what I thought AI music could be. Someon…

It also seems that there is a big percentage of people who are completely against AI music for a multitude of reasons. Even if they liked the sound they would still hate it if it was AI generated.

But to me this seems silly. Yes I want real artists to make music and be able to make a living not some faceless company spitting out endless music until something works. However at the end of the day if something sounds good then one should enjoy it not refuse to accept it simply because it is AI.

Because how far does their stance against AI go? They won't accept music. What about if AI created a cure that could save their child? Or what if AI could could sort through a massive backlog of evidence in unsolved murders and other violent crimes giving new leads previously missed? I am just curious if some people will simply be against it no matter what the use is. As for myself I think it has it's uses but also think it comes at a heavy price as in massive power and water consumption and other issues it comes with. Anyways

Re: YouTube to automatically label AI-generated videos

#370

Earlier quoted context omitted.

You still ultimately have the analogue hole here - pull the camera apart, splice your own hardware somewhere between the sensor and the thing that adds the signatures (or in front of the sensor).

Or just straight up point the camera at a computer monitor, without even trying to hide it. Most of the security camera footage online is already uploaded this way.

I intentionally didn't say that because I feel like people might dismiss that with "oh but you can tell the difference with sufficient analysis etc" whereas literally sending data directly through the same path as the real sensor would be potentially less detectable (or more, if the sensor itself has some kind of noticeable fingerprint)
Post reply on HN