YouTube's AI label move: too little, too late?

budgetburner

So YouTube's finally going to auto-label AI videos instead of just begging creators to do it. About bloody time, but let's not break out the champagne just yet. They're only flagging "significant photorealistic AI" - which means anything obviously fake like a unicorn in space sails right through. Need an AI voiceover for your video? No label. Want to render Cartoon Network characters as disturbingly real humans? Probably passes because it's "obviously" AI. Their internal detection system will supposedly catch what creators "neglect" to disclose, but that's a lot of faith in an algorithm that can't even stop copyright claims from scammers. they've been leaning on self-disclosure for two years already - and surprise, surprise, most lazy creators just didn't bother. Now YouTube's stepping in, but the policy hasn't changed. So they're still giving a pass to anything that doesn't look "photorealistic"? That's a massive loophole for synthetic voices and non-photorealistic deepfakes. i'd bet a solid chunk of the AI slop on the platform will keep skating by. Good first step? Maybe. But I'm not holding my breath for this to actually work.

ethan-pr

I’m glad they’re doing this. Anyone who spends time on content platforms can see how much AI slop is starting to take over. I’ve seen a lot of those “veteran” or “feel sorry for this person” videos that end up linking to donation pages. Most of the time you can tell they’re fake, but they still blow up and probably make money off people.

We definitely need better safeguards and stronger policies to stop AI from taking advantage of people like that. Automatic labeling is a good start, but it shouldn’t stop there... I am talking to you R3ddit!

gif

avadata

Finally. But let's not pretend this fixes the mess. The detection models are still laughably inaccurate on anything that isn't pure deepfake - half the labelled content will be false positives, and the other half will be ads optimised with generative scripts that slip through anyway. Useful for the compliance checkbox, useless for actual transparency unless they also open up the training data and confidence scores

lucasvid

Ugh, can we block AI videos?

liamcopy

So YouTube wants us to label AI videos now.

Cool. But here's where it gets muddy.

A unicorn in space? Obvious fantasy. No label needed, right? But a photorealistic video of a real person saying something they never said? That's getting cracked down on.

What about AI voice-overs? Or the brainrot trend where someone takes Courage the Cowardly Dog and AI-generates him as a photorealistic character staring into your soul? that's obviously AI, but does it fly under the radar?

I'm all for transparency. But the grey areas are massive. Curious how they'll actually police this without just drowning in false flags or doing nothing at all.

isa-ecom

Honestly, at this point it'd be faster to tag the ones not using AI. Half the "authentic" creator channels I see are just reading off ChatGPT scripts with a slightly different accent. Feels like we're a few months away from automated unboxing videos pushing my own products back at me.

budgetburner

great, because algorithms are famously perfect at detecting their own kind. This'll go smoothly