• YouTube is making it easier than ever to spot AI-generated videos.
• The company is rolling out new labels for both long-form videos and Shorts that make it clear which videos were produced using AI.
• YouTube is also rolling out a new automatic detection system to spot AI videos even if a creator doesn’t make a disclosure.
I quit watching YouTube, problem solved.
I just avoid videos with contrived scary headlines or the word “BANNED”.
The Youtube that secretly used AI to modify user’s videos without permission or notification ? That Youtube?
I still want to be able to block channels. A setting to filter our channels that is AI or just filter videos they use AI would be great but even then why can’t I block channels?
Click the 3-dots on a thumbnail and click “Don’t recommend channel.”
Or use PeerTube.
how about FILTERING if it has AI content in it? Like, I don’t want porn/risque content on my search, means they have the mean to tag videos and filter it.
How about filtering slop too?
I report those. I hope the automated systems just takes those channels down.
I hate AI videos as much as the next person, but why would YouTube take those down. As far as I know they aren’t against any guidelines.
There is a rule against low-quality or low-effort content which I believe is machine-generated content falls under.
Why? Because YT’s automation process is so regularly crap I assume it might work.
I can’t speak to guidelines, as I don’t regularly read those. But the “crying girl” AI that’s drop shipping things, or the AI girl talking about sex (in a non educational space), (and there’s more) and they all have links to external pages… feels like at best right now it’s exploiting a gray area. Some are just AI where they create situations for views, which, meh. But they’re not clearly labeled as “this is AI, this is fake, this is for entertainment only”. Not to use a slipper slope argument, but once it gets to the point it’s hard for people to quickly see it’s AI, this could easily (and pretty sure it’s already happened) be used to fake situations and get people riled up about literally nothing. Like we could start seeing videos of fake police violence. Which does nobody any good.
Also they need to ban Ai videos from YT Kids
Or just ban them all together
Or at least make the fucking “thumbs down” mean something.
Same goes for “do not recommend channel” and “not interested”
Absolutely!
will someone please think of the retention numbers
a measure that will actually help protecting kids? Not on their watch
I think a more aggressive approach is needed. Instead of simply tagging AI-generated videos that did not disclose the use of AI, those videos should be automatically unlisted. False positives may happen and they can be appealed. I’m not normally in favour of “guilty until proven innocent” policies, but in the case of deceptive AI videos (especially political ones), it’s what we need.
I actually support this as long as the false positives are not too bad
Good luck with that. YouTube content moderation is notoriously automated.
I don’t think Youtube’s history on false positives for things is all that stellar. Maybe if they get AI to help it will… oh, no, it will make it far worse than even Youtube can do alone.
The problem is most humans are horrible at it. youll see 10 year old videos reposted and people scream AI!!!
btw: my new mission is to not refer to anything as ai. its a bad blanket statement. so video generators or llms.
Will I be able to see that before clicking on the video? I just got an update and the video information is now squished to a single line. So if the channel name is long, I can’t see the age of the video. If they add it to the end, everyone will just have long names to hide it.
Let us filter out the AI videos cowards
This. For the love of God, this!!
I’d be all for YouTubeAI.com with only AI content there, then I can stop blocking that 1 asian dude with 500000 channels
If they’re adding a label you can just use that label to block them with ublock.
It kind of does it by itself if you use the algorithm correctly, I was seeing some AI slop in my shorts like a year ago but after having disliked it/reported it, it seems to have gone away. I haven’t seen a single AI video in many months
There’s plenty of long form AI content lately too. Wikipedia articles make for great YouTube videos, and there’s no shortage of them. I’ve run into several channels lately where it took me about 30 seconds to realize something was off, and checking the channel’s history you can usually tell by finding older videos which actually have a person in them who is clearly very different than the narrator in the newer videos, or by just not having any older videos at all.
I’ve ran into those too 🥲
What kills me is all the comments completely oblivious to the situation
That’s a trial and error approach for something that could easily be a simple setting.
It also doesn’t work if you use an alternative front end or aren’t signed in.
Additionally it also depends entirely on the content you view and your viewing enough of it to train the algorithm, which means new accounts or little used accounts have no recourse.
Every single AI short video that shows up I thumbs down and tell me never to show me that channel again. Doesn’t stop like 85% of my feed from being essentially the same AI short videos by users with names like agagin9927385, liafu63836662, or crazyvideos445442…I spend more times thumbs downing and saying never to show this again than I do watching the videos.
Same. I never see AI slop. Or politics. My algorithm is dialed in and my recommended is all bangers.
You mean the YouTube that dumps out AI summaries of the videos you’re on their platform specifically to watch directly? That YouTube?
Surely they aren’t a company composed of two-faced slimeballs like their parent. That would be unheard of.
Obligatory “fuck AI,” however I’ve reluctantly found that the AI summaries can be helpful on occasion, such as when deciding if a video with a clickbaity title is actually worth watching, or when I’m on YouTube looking for a solution to a problem (in which case the summary can sometimes get me the content way faster than watching the video). So of all the dumb shit YouTube has done to their platform in recent years, I’d argue the AI summaries fall on the “actually not that bad” end of the spectrum.
That’s very fair, it’s definitely in the vein of what LLMs are (theoretically) meant for, i.e. summarization and language parsing/analysis. I find that LLMs work pretty well for translation too. It’s just a shame that executives have run it to the ground like they usually do. And the people poisoning us are the same ones treating it in this case.
Sounds good now. Sounds like an excuse to ‘accidentally’ flag videos they don’t want getting clicks in the future.
That’s the Google YouTube way!
My bet is it will only check for SynthID
The problem isn’t spotting them. They’re fucking obvious slop. The problem is that they have inundated every fucking crevice of that website so you can’t watch a goddamn video without tripping over one.
Hello. Thank you for clicking on [this video]. Today. We will be looking at the topic of. Detecting eye videos on You Tube.
