> "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"
YouTube made AI enhancements to videos without warning or permission
41–50 of 292 posts
Re: YouTube made AI enhancements to videos without warning or permission
#42Isn'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.
The level of post-processing matters. There is a difference between color grading an image and removing wrinkles from a face.
The line is not cut clear but these companies are pushing the boundaries so we get used to fake imagery. That is not good.
Re: YouTube made AI enhancements to videos without warning or permission
#43> "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"
xkcd 2015 closer than you think due to the magical technology of money https://xkcd.com/2015/
Re: YouTube made AI enhancements to videos without warning or permission
#44Isn'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.
Maybe you’re thinking of TikTok and samsung facial smoothing filters? Those are a lot more subtle and can be turned off.
Re: YouTube made AI enhancements to videos without warning or permission
#45Earlier quoted context omitted.
Just don't smoke/eat junk/do drugs etc. They put addictive shit in your face and force you to use their bloated interface to access the service.
The worst thing for me is they don't show the channel names. So much of the channels pushing Star Wars shorts are quite obvious bot names, and it's hard to filter these from legitimate SW content creators who are, on top of that, all using the same damn AI voice.
The Venn diagram of AI voice users and good content creators is pretty close to two separate circles. I don't really care about the minority in the intersection.
Re: YouTube made AI enhancements to videos without warning or permission
#46Re: YouTube made AI enhancements to videos without warning or permission
#47Earlier quoted context omitted.
The "can not turn off" part is the most jarring. Seriously, did none of Californian PMs hear about the concept of "being multilingual" and not needing to translate non-English content? --- By the way, this reminds me also of another stupid Google thing related to languages: Say your Chrome is set to English. When encountering another language page, Chrome will (since a decade ago or so) helpfully ask you to auto-tran…
"Złomnik: Dodge Caravan Is a Large Wall Unit"
Re: YouTube made AI enhancements to videos without warning or permission
#48Re: YouTube made AI enhancements to videos without warning or permission
#49Youtube 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…
> Perhaps this was done to optimize video encoding, since the less noise/surface detail there is the easier it is to compress. If so it's really just another kind of lossy compression. No different in principle from encoding a video to AV-1 format.
Re: YouTube made AI enhancements to videos without warning or permission
#50So I feel like article doesn't address the "why" of it all. Why auto AI upscale?
1. 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.