alt Hacker NewsYouTube caught making AI-edits to videos and adding misleading AI summaries
Comments
The AI filter applied server-side to YouTube Shorts (and only shorts, not regular videos) is horrible, and it feels like it must be a case of deliberate boiling the frog. If everyone gets used to overly smooth skin, weirdly pronounced wrinkles, waxy hair, and strange ringing around moving objects, then AI-generated content will stand out less when they start injecting it into the feed. At first I thought this must be some client-side upscaling filter, but tragically it is not. There's no data savings at all, and there's no way for uploaders or viewers to turn it off. I guess I wasn't cynical enough.
An amateur tip that I sometimes use after I reencode something to check what i lost:
ffmpeg -i source.mkv -i suspect.mkv -filter_complex "blend=all_mode=difference" diff_output.mkv
I saw these claims before but still have not found someone to show a diff or post the source for comparison. It would be interesting.
I don't understand why Youtube would do this. Both applying these kinds of "enhancements" to video and to do so without consent or even informing people. How is this a smart move?
We need more people experimenting with creating a better platform for content creators. Not least so people like Beato, but not as well known, don't constantly get harassed by fraudulent and incorrect copyright infringement claims.
I learned to ignore the AI summaries after the first time I saw one that described the exact OPPOSITE conclusion/stance of the video it purported to summarize.
There are entire fake persona videos these days. Leading scientists, economists, politicians, tech guys, are being impersonated wholesale on youtube.
Are these AI filters, or just applying high compression/recompressing with new algorithms (which look like smoothing out details)?
edit: here's the effect I'm talking about with lossy compression and adaptive quantization: https://cloudinary.com/blog/what_to_focus_on_in_image_compre...
The result is smoothing of skin, and applied heavily on video (as Youtube does, just look for any old video that was HD years ago) would look this way
Youtube is an AI training data set.
There is no way Google thinks it's in their interest to serve up clean data to anyone but themselves.
Talking about AI, Google, and shady tactics, I wouldn't be surprised if soon we discover they are purposefully adding video glitches (deformed characters and so on) in the first handful of iterations when using Veo video generation just so people gets used to trying 3 or 4 times before they receive a good one.
I'm seeing Youtube summary pictures which seem to be AI-generated. I was looking at [1], which is someone in China rebuilding old machines, and some of the newer summary pictures are not frames from the video. They show machines which are the sort of thing you might get by asking a Stable Diffusion type generator to generate a picture from the description.
YouTube should keep their grubby hands off. And give that capability to us instead. I want the power to do personal AI edits built in. Give me a prompt line under each video. Like "replace English with Gaelic", "replace dad jokes with lorem ipsum", "make the narrator's face 25% more symmetrical", "replace the puppy with a xenomorph", "change the setting to Monument Valley", etc.
The citation chain for these mastodon reposts resolves to the Gamers Nexus piece on youtube https://www.youtube.com/watch?v=MrwJgDHJJoE
I really hate all the AI filters in videos. It makes everyone look like fake humans. I find it hard to believe that anyone would actually prefer this.
This story is several months old
FYI, I used to pay for YouTube Premium and have since stopped doing that. Deleting the app and letting ad blockers filter out this nonsense is a superior experience.
Strongly recommend. We’ll get local AIs that can skip the cruft soon enough anyway.
Every YT short looks AI-ified and creepy now
> "Machine learning is a subset of artificial intelligence,” he said.
No, gen AI is a subset of machine learning.
They're heating the garbage slightly before serving it? Oh no.
What’s the point of doing this?
I don't understand the justification for the expense or complexity.
This is wild.
I wonder if it will end up being treated as part of a codec instead of edits to the base film, and can then be re-run to undo the video's?
It feels like there needs to be a way to verify that what you uploaded is what's on the site.
> Samuel Woolley, a disinformation expert at the University of Pittsburgh, said the company’s wording was misleading. “Machine learning is a subset of artificial intelligence,” he said. “This is AI.”
It's the other way around isn't it? "AI" is a subset of ML.
What PM thought this was a good idea? This has to be the result of some braindead we need more AI in the product mandate
I’ve also noticed YouTube has unbanned many channels that were previously banned for overt supremacist and racist content. They get amplified a lot more now. Between that and AI slop, I feel like Google is speed running the changes X made over the last few years.
Someone mentioned Insta is doing this too in this comment section
I just completely despair. What the fuck happened to the internet? Absolutely none of these CEOs give a shit. People need to face real punishments
Being driven mad by conspiracy paranoia about 'face filters' (possible compression artifacts) is a great example of being AI-pilled.
And then the discourse is so riddled with misnomers and baited outrage that it goes nowhere.
The other example in submitted post isn't 'edits to videos' but rather the text descriptions of automated captions. The Gemini/AI engine not being very good at summarizing is a different issue.
This link is to a Mastodon thread which links to another blog post which links to an actual source on ynetnews.com which quotes another article that has a quote from a YouTube rep. Save yourself the trouble and go straight to that article (although it's not great either): https://www.ynetnews.com/tech-and-digital/article/bj1qbwcklg
The key section:
> Rene Ritchie, YouTube’s creator liaison, acknowledged in a post on X that the company was running “a small experiment on select Shorts, using traditional machine learning to clarify, reduce noise and improve overall video clarity—similar to what modern smartphones do when shooting video.”
So the "AI edits" are just a compression algorithm that is not that great.
[dead]
[dead]
It's fine, I can't watch youtube anyway with ublock origin on any more :)
"Making AI edits to videos" strikes me as as bit of an exaggeration; it might lead you to think they're actually editing videos rather than simply... post-processing them[1].
That being said, I don't believe they should be doing anything like this without the creator's explicit consent. I do personally think there's probably a good use case for machine learning / neural network tech applied to the clean up of low-quality sources (for better transcoding that doesn't accumulate errors & therefore wastes bitrate), in the same way that RTX Video Super Resolution can do some impressive deblocking & upscaling magic[2] on Windows. But clearly they are completely missing the mark with whatever experiment they were running there.
[1] https://www.ynetnews.com/tech-and-digital/article/bj1qbwcklg
[2] compare https://i.imgur.com/U6vzssS.png & https://i.imgur.com/x63o8WQ.jpeg (upscaled 360p)
A makeup influencer I follow noticed youtube and instagram are automatically adding filters to his face without permission to his videos. If his content was about lip makeup they make his lips enormous and if it was about eye makeup the filters make his eyes gigantic. They're having AI detecting the type of content and automatically applying filters.
https://www.instagram.com/reel/DO9MwTHCoR_/?igsh=MTZybml2NDB...
The screenshots/videos of them doing it are pretty wild, and insane they are editing creators' uploads without consent!