To make people more accustomed to the AI generated look so that when they release their next Veo integration to YouTube content creator tools, these videos will stand out less as unnatural.
Maybe it's to make it more difficult to train AI video models from YouTube.
Think about it, they have the raw footage so could use it if they want, but competitors using scrapers will have slightly distorted video sources.
I can’t think of a more dislike-able company than YouTube. I used to love youtube and watch it everyday and it would make me a happier, smarter person. Now youtube’s impact on their users is entirely negative and really the company needs to be destroyed. But they won’t be because they are now evil, and evil is profitable.
> "You know, YouTube is constantly working on new tools and experimenting with stuff," Beato says. "They're a best-in-class company, I've got nothing but good things to say. YouTube changed my life."
My despondent brain auto-translated that to: "My livelihood depends on Youtube"
And the other day he posted about the abusive copyright claims he has to deal with that cost him a lot of money and could maybe have his channels closed.
I push back on the idea there is anything despondent there. If YouTube was enabling my lifestyle I'd be pretty happy about the situation and certainly not about to start piling public pressure on them. These companies get enough hate from roving bands of angry internet denizens.
Touching up videos is bad but it is hardly material to break out the pitchforks compared to some of the political manoeuvres YouTube has been involved in.
As a consumer they are the most hostile platform to consume a video the way I want. Not the way they want me to. I am also required to use an adblocker to disable all shorts.
Beato is a musician and a producer. He just finds making YouTube videos an easier way to earn a living. He's said many times how frustrating it is as a producer to work with musicians.
I've noticed this for a while, when I accidentally click on YouTube Shorts. (I want to avoid it, because it's brain rot, but YouTube keeps enabling it and pushes it hard in notifications).
It's most glaringly obvious in TV shows. Scenes from The Big Bang Theory look like someone clumsily tries to paint over the scenes with oil paint. It's as if the actors are wearing an inch thick layer of poorly applied makeup.
It's far less glaring in Rick Beato's videos, but it's there if you pay attention. Jill Bearup wanted to see how bad it could get and reuploaded the "enhanced" videos a hundred times over until it became a horrifying mess of artifacts.
The question remains why YouTube would do this, and the only answers I can come up with are "because they can" and "they want to brainwash us into accepting uncanny valley AI slop as real".
Youtube says this was done for select Youtube Shorts as a denoising process. However most popular channels on Youtube, which seem to be the pool selected for this experiment, typically already have well lit and graded videos shouldn't benefit much from extra denoising from a visual point of view.
It's true though that aggressive denoising gives things an artificially generated look since both processes use denoising heavily.
Perhaps this was done to optimize video encoding, since the less noise/surface detail there is the easier it is to compress.
I've seem some game of thrones clips recently in youtube shorts which looked like they'd been generated by ai. I couldn't understand why anyone would have done that to the original good looking material. The only thing I could think was that it was some kind of copyright evasion.
I going to say something controversial, bu why is this even surprising?
Google and YouTube have been framing themself as the kind company that will appropriate your work and make make money out of you. "You are the product" is a repeated endlessy even on social media, and this it thier private paltform after all.
At this point getting involved with youtube is just the usual naive behaviour that somehow you are the exception and bad things won't happen to you.
Isn't this similar to what e.g. Instagram and co have done for ages? Even smartphones do it automatically for you, digital post-processing to compensate for the limitations of the cameras.
This is getting into conspiracy territory but my personal assumption that they're trying to gaslight people into thinking that these weird AI artifacts are just how videos work, so that it's harder to distinguish between real videos and AI generated ones.
How about they turn off their recent asinine title translation feature? Now every creator has to opt out of it manually - and the users have no recourse short of browser extensions.
The recent sinking in quality of youtube as a platform has been awful to watch.
Just a couple days ago I got an ad with a Ned Flanders singing about the causes of erectyle dysfunction (!), a huge cocktail of copyright infringement, dangerous medical advice and AI generated slop. Youtube answered the report telling me they've reviewed and found nothing wrong.
The constant low quality, extremely intertwined ads start to remind me of those of shady forums and porn pages of the nineties. I'm expecting them to start advertising heroine now they've decided short term profits trump everything else.
what it boils down to is that "enhancement" is one more pilon to swerve around in the ongoing persuit of reality.
given the overwhelming volume of media it's easy enough to modify ones asthetic monitoring and scan to simply ignore anything suspicious, or if it gets too bad, back right off
Unfortunately the article doesn't have an example, or a comparison image. Other reports are similarly useless as well. The most that seemed to happen is that the wrinkles in someone's ear changed. In case anyone else wants to see it in action:
I skimmed the videos as well, and there is much more talk about this thing, and barely any examples of it. As this is an experiment, I guess that all this noise serves as a feedback to YouTube.
A chill ran down my spine as I imagined this being applied to the written word online: my articles being automatically "corrected" or "improved" the moment I hit publish, any book manuscripts being sent to editors being similarly "polished" to a point that we humans start to lose our unique tone and everything we read falls into that strange uncanny valley where everything reads ok, you can't quite put your finger on it, but it feels like something is wearing the skin of what you wrote as a face.
This is why shadow banning rubbed people so wrong. I can't prove it, but i gave up on online dating a long time ago because i found a couple of automated systems would just not send messages and not tell you (in a middle of an already active conversation)
Reminder, there is no cloud, there is just computers of other people. And I for one support those other people's right to do on their computers what they want.
If AI is as wonderful and world-changing as people claim, it's odd that it's being inserted into products exactly like every other solution in search of a problem.
Last week I went to buy a Philip K Dick eBook while on vacation. It was only $2 and my immediate thought was, “what are the odds this is some weird pirated version that’s full of errors? What if it’s some American version that’s been self-censored by Amazon to be approved by the government? What if it’s been AI enhanced in some way?”
Just the consideration of these possibilities was enough to shake the authenticity of my reality.
Even more unsettling is when I contemplate what could be done about data authenticity. There are some fairly useful practical answers such as an author sharing the official checksum for a book. But, ultimately, authenticity is a fleeting quality and I can’t stop time.
I suppose we should fire up those "AI" browsers and let them loose on YouTube in a while loop. They are just the right audience for "AI" enhanced content and YouTube's advertisers will be thrilled.
When you upload a video to YT, it's heavily compressed. Your pristine creation is converted 10 ways to Sunday, which then can be played back in a variety of formats, speeds, and platforms. Long before you even uploaded that video, for free, to give the world the chance to see your creative genius, you agreed to this process, by agreeing to Youtube's T&C's.
People may be upset, and I get that. But it's not like the videos were in their original format anyway. If you want to maintain perfect video fidelity, you wouldn't choose YouTube. You chose YouTube because it's the path of least resistance. You wanted massive reach and a dead simple monetization route.
AI fearmongering probably produces a lot of clicks if upscaling gets labeled as "might bend reality". YT shouldn't just be doing it without user's input, but pearl clutching is unproportiona.
Re: "without warning or permission". The YouTube Terms of Use require you to grant YouTube the (perpetual, worldwide, etc) right to prepare derivative works.
From the linked tweet from YouTube's head of editorial:
"No GenAI, no upscaling. We're running an experiment on select YouTube Shorts that uses traditional machine learning technology to unblur, denoise, and improve clarity in videos during processing (similar to what a modern smartphone does when you record a video)"
Considering how aggressive YouTube is with video compression anyways (which smooths your face and makes it blocky), this doesn't seem like a big deal. Maybe it overprocesses in some cases, but it's also an "experiment" they're testing on only a fraction of videos.
I watched the comparisons from the first video and the only difference I see is in resolution -- he compares the guitar video uploaded to YT vs IG, and the YT one is sharper. But for all we know the IG one is lower resolution, that's all it looks like to me.
45 comments
[ 2.7 ms ] story [ 51.2 ms ] thread1. See that AI upscaling works kinda well on certain illustrations.
2. Start a project to see if you can do the same with video.
3. Develop 15 different quality metrics, trying to capture what it means when "it looks a bit fake"
4. Project's results aren't very good, but it's embarrassing to admit failure.
5. Choose a metric which went up, declare victory, put it live in production.
My despondent brain auto-translated that to: "My livelihood depends on Youtube"
Touching up videos is bad but it is hardly material to break out the pitchforks compared to some of the political manoeuvres YouTube has been involved in.
It's most glaringly obvious in TV shows. Scenes from The Big Bang Theory look like someone clumsily tries to paint over the scenes with oil paint. It's as if the actors are wearing an inch thick layer of poorly applied makeup.
It's far less glaring in Rick Beato's videos, but it's there if you pay attention. Jill Bearup wanted to see how bad it could get and reuploaded the "enhanced" videos a hundred times over until it became a horrifying mess of artifacts.
The question remains why YouTube would do this, and the only answers I can come up with are "because they can" and "they want to brainwash us into accepting uncanny valley AI slop as real".
It's true though that aggressive denoising gives things an artificially generated look since both processes use denoising heavily.
Perhaps this was done to optimize video encoding, since the less noise/surface detail there is the easier it is to compress.
Says everything. Hey PM at YouTube: How about you think stuff through before even starting to waste time on stuff like this?
At this point getting involved with youtube is just the usual naive behaviour that somehow you are the exception and bad things won't happen to you.
Just a couple days ago I got an ad with a Ned Flanders singing about the causes of erectyle dysfunction (!), a huge cocktail of copyright infringement, dangerous medical advice and AI generated slop. Youtube answered the report telling me they've reviewed and found nothing wrong.
The constant low quality, extremely intertwined ads start to remind me of those of shady forums and porn pages of the nineties. I'm expecting them to start advertising heroine now they've decided short term profits trump everything else.
https://www.reddit.com/r/youtube/comments/1lllnse/youtube_sh...
I skimmed the videos as well, and there is much more talk about this thing, and barely any examples of it. As this is an experiment, I guess that all this noise serves as a feedback to YouTube.
This is a contradiction in terms.
Just the consideration of these possibilities was enough to shake the authenticity of my reality.
Even more unsettling is when I contemplate what could be done about data authenticity. There are some fairly useful practical answers such as an author sharing the official checksum for a book. But, ultimately, authenticity is a fleeting quality and I can’t stop time.
People may be upset, and I get that. But it's not like the videos were in their original format anyway. If you want to maintain perfect video fidelity, you wouldn't choose YouTube. You chose YouTube because it's the path of least resistance. You wanted massive reach and a dead simple monetization route.
"No GenAI, no upscaling. We're running an experiment on select YouTube Shorts that uses traditional machine learning technology to unblur, denoise, and improve clarity in videos during processing (similar to what a modern smartphone does when you record a video)"
https://x.com/youtubeinsider/status/1958199532363317467?s=46
Considering how aggressive YouTube is with video compression anyways (which smooths your face and makes it blocky), this doesn't seem like a big deal. Maybe it overprocesses in some cases, but it's also an "experiment" they're testing on only a fraction of videos.
I watched the comparisons from the first video and the only difference I see is in resolution -- he compares the guitar video uploaded to YT vs IG, and the YT one is sharper. But for all we know the IG one is lower resolution, that's all it looks like to me.