YouTube made AI enhancements to videos without telling users or asking permission. As AI quietly mediates our world, what happens to our shared connections with real life?
It would not make any sense for them to be upscaled on the fly. It’s a computationally intensive operation, and storage space is cheap. Is there any evidence of it being done on the fly?
While it could theoretically be done on device, it would require the device to have dedicated hardware that is capable of doing the processing, so it would only work on a limited number of devices. It would be pretty easy to test this if a known modified video were available.
AI upscaling can be run on a ton of devices nowadays.
Also people are forgetting it’s not just storage, it’s bandwidth they save with this move. So even if they store both the low and high res copies they can save 4x the bandwidth (or more) serving to devices with upscaling capabilities.
it wouldn’t need dedicated hardware, it would just be slower on phones without that hardware. there’s nothing that AI does that can’t be done on any phone or PC.
same thing with ray tracing, it’s technically possible on cards that aren’t a part of the RTX line, they just can’t do it as fast as an RTX card (per NVIDIA).
That would depend entirely on WHAT its doing. I have not personally seen any of these videos yet, but based on what was described in the article, I would imagine that a typical CPU would not be able to handle it.
a typical CPU in a phone would do just fine. AI effects in photo and video started coming out in phones before new phones started having dedicated hardware to accelerate it. phones have been doing stuff as intensive as that for years. for example, iPhones have been able to make complex and precise full scale textured replicas of real world environments that you can then import into Blender using their lidar capabilities for years. that’s quite a bit more intensive of a process than using AI to edit a video.
and as for a PC, there isn’t anything you can do to edit a video using AI that a PC CPU would not be able to handle. if a 10 year old laptop can generate video out of thin air using genAI, then applying a sharpening effect would be a piece of cake. hell, I’ve done stable diffusion on a laptop with just 4GB of VRAM. it’s quite a bit slower than with a faster PC, but certainly doable.
It’s not that computationally intensive to upscale frames. TVs have been doing it algorithmically for ages and looking good doing it. Hell, nVidia graphics cards can do it for every single frame of high end games with DLSS. Calling it “AI” because the type of algorithm it’s using is just cashing in on the buzzword.
You are right that nvidia cards can do it for games using DLSS. Nvidia also has a version called RTX video that works for video. But are they could to be dedicating hardware for playback every single time a user requests to play a short? That is significantly different than just serving a file to the viewer. If they had all of these Nvidia cards laying around, they surely have better things that they could use them for. To be clear here, the ONLY thing I am taking issue with is a comment that it seems that youtube may be upscaling videos on the fly (as opposed to upscaling them once when they are uploaded, and then serving that file 1 million times). I’m simply saying that it makes a hell of a lot more sense any day of the week to upscale a file one time than to upscale it 1 million times.
They could do that without upscaling. Upscaling every video only fly would cost an absolute shit ton of money, probably more than they would be making from the ad. There is no scenario where they wouldn’t just upscale it one time and store it.
It would not make any sense for them to be upscaled on the fly. It’s a computationally intensive operation, and storage space is cheap. Is there any evidence of it being done on the fly?
It would if they can do it on your device.
While it could theoretically be done on device, it would require the device to have dedicated hardware that is capable of doing the processing, so it would only work on a limited number of devices. It would be pretty easy to test this if a known modified video were available.
AI upscaling can be run on a ton of devices nowadays.
Also people are forgetting it’s not just storage, it’s bandwidth they save with this move. So even if they store both the low and high res copies they can save 4x the bandwidth (or more) serving to devices with upscaling capabilities.
it wouldn’t need dedicated hardware, it would just be slower on phones without that hardware. there’s nothing that AI does that can’t be done on any phone or PC.
same thing with ray tracing, it’s technically possible on cards that aren’t a part of the RTX line, they just can’t do it as fast as an RTX card (per NVIDIA).
That would depend entirely on WHAT its doing. I have not personally seen any of these videos yet, but based on what was described in the article, I would imagine that a typical CPU would not be able to handle it.
a typical CPU in a phone would do just fine. AI effects in photo and video started coming out in phones before new phones started having dedicated hardware to accelerate it. phones have been doing stuff as intensive as that for years. for example, iPhones have been able to make complex and precise full scale textured replicas of real world environments that you can then import into Blender using their lidar capabilities for years. that’s quite a bit more intensive of a process than using AI to edit a video.
and as for a PC, there isn’t anything you can do to edit a video using AI that a PC CPU would not be able to handle. if a 10 year old laptop can generate video out of thin air using genAI, then applying a sharpening effect would be a piece of cake. hell, I’ve done stable diffusion on a laptop with just 4GB of VRAM. it’s quite a bit slower than with a faster PC, but certainly doable.
It’s not that computationally intensive to upscale frames. TVs have been doing it algorithmically for ages and looking good doing it. Hell, nVidia graphics cards can do it for every single frame of high end games with DLSS. Calling it “AI” because the type of algorithm it’s using is just cashing in on the buzzword.
(Unless I’m misunderstanding what’s going on.)
You are right that nvidia cards can do it for games using DLSS. Nvidia also has a version called RTX video that works for video. But are they could to be dedicating hardware for playback every single time a user requests to play a short? That is significantly different than just serving a file to the viewer. If they had all of these Nvidia cards laying around, they surely have better things that they could use them for. To be clear here, the ONLY thing I am taking issue with is a comment that it seems that youtube may be upscaling videos on the fly (as opposed to upscaling them once when they are uploaded, and then serving that file 1 million times). I’m simply saying that it makes a hell of a lot more sense any day of the week to upscale a file one time than to upscale it 1 million times.
My video card deffo heats up more when watching youtube over peertube. I’m pretty sure they’re using my graphics card for upscaling.
It would make sense if it’s a scheme to inject ads directly into the stream so adblockers wouldn’t work anymore.
They could do that without upscaling. Upscaling every video only fly would cost an absolute shit ton of money, probably more than they would be making from the ad. There is no scenario where they wouldn’t just upscale it one time and store it.