"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.
Please allow me "post-process" your comment a bit. Let me know if I'm doing this right.
> "Making AI edits to videos" strikes me as something particularly egregious; it leads a viewer to see a reality that never existed, and that the creator never intended.
Sure, but that's not YouTube. That's Instagram. He says so at 1:30.
YouTube is not applying any "face filters" or anything of the sort. They did however experiment with AI upscaling the entire image which is giving the classic "bad upscale" smeary look.
Like I said, I think that's still bad and they should have never done it without the clear explicit consent of the creator. But that is, IMO, very different and considerably less bad than changing someone's face specifically.
His followers also added screenshots of youtube shorts doing it. He says he reached out to both platforms and says he will be reporting back with an update from their customer service and is doing some compare an contrast testing for his audience.
> Here's some other creators also talking about it happening in youtube shorts (...)
If you open the context of the comment, they are specifically talking about the bad, entire-image upscaling that gives the entire picture the oily smeary look. NOT face filters.
EDIT : same thing with the two other links you edited into your comment while I was typing my reply.
Again, I'm not defending YouTube for this. But I also don't think they should be accused of doing something they're not doing. Face filters without consent are a far, far worse offense than bad upscaling.
I would like to urge you to be more cautious, and to actually read what you brandish as proof.
AV1 is the king of ultra-low bitrates, but as you go higher — and not even that much higher — HEVC becomes just as good, if not more. Publicly-available AV1 encoders (still) have a tendency to over-flatten anything that is low-contrast enough, while x265 is much better at preserving visual energy.
This problem is only just now starting to get solved in SVT-AV1 with the addition of community-created psychovisual optimizations... features that x264 had over 15 years ago!
I'd love to watch Netflix AV1 streams but they just straight up don't serve it to my smart TV or my Windows computers despite hardware acceleration support.
The only way I can get them to serve me an AV1 stream is if I block "protected content IDs" through browser site settings. Otherwise they're giving me an H.264 stream... It's really silly, to say the least
I can't say I understand why HEVC support being disabled would "prevent background blurring", especially because 1) the blur has nothing to do with HW decode (not even in weird unknown parts of the MPEG-4 specs like video object planes in part 2, or better yet: part 6 and part 16) — and 2), AVC HW encode is still there and is a completely acceptable fallback, so...?
It doesn’t. Disabling hardware acceleration does which they needed to do in order to play content.
“ needed to either have the HEVC codec from the Microsoft Store removed entirely from [Microsoft Media Foundation] or have hardware acceleration disabled in their web browser/web app, which causes a number of other problems / feature [degradations]. For example, no background blurring in conference programs”
The blur happens on the GPU. HEVC encode also happens on the GPU (or at least a GPU-adjacent device; it's rarely a full-shader affair). If you were to use HEVC software encode with GPU blur, you'd need to send the camera data to the GPU, pull it back to the CPU, and then software encode. Performant GPU readback is often cumbersome enough that developers won't bother.
But it is still more performant to do so in general. There are more image corrections of great quality happening than just background removal nowadays, like lighting improvements or sometimes upscaling, and you wouldn't want to do all that on the CPU.
But also, HW encoding of some codecs is not always of great quality and doesn't support the advanced features required for RTC, so the CPU encoding code-path is sometimes even forced! While it doesn't necessarily apply to HEVC as you'd need a license for it (and almost all apps rely on the system having one), it's happening for VP9 or AV1 occasionally more frequently.
I'm not surprised. I'm French and one thing I've consistently seen with Gemini is that it loves to use Title Case (Everything is Capitalized Except the Prepositions) even in French or other languages where there is no such thing. A 100% american thing getting applied to other languages by the sheer power of statistical correlation (and probably being overtrained on USA-centric data). At the very least it makes it easy to tell when someone is just copypasting LLM output into some other website.
If they realize the value of "sandboxing" something so insecure they should also be making it really easy for you to do the same with any app, or set of apps...
Technically not lying if this is NSA (non-standalone) 5G. The 5G band just comes on as an additional aggregated band. The icon just shows up because the tower is capable of supplying the band.
Really the bigger problem is that there's not enough distinction between SA and NSA
-140 dBm is far beyond no coverage, yeah. -120 dBm is pretty much when LTE stops working (sometimes it can painfully stretch to -123 to -125 but usually not because of noise etc)
Yes, even thermal background noise (the noise level that exists even in complete absence of RF) would be expected to be above -140dBm. It scales with channel size and temperature.
As near as I can tell, the smallest subcarrier 5G can use is 15kHz, the thermal noise floor for a 15kHz channel at room temp (300K) would be -132 dBm.
My guess is whatever chip doing the measurement simply couldn't measure that low accurately, or it reports "nothing detected" as -140 dBm.
GPS receivers can go down to around -160dBm, but that's a very low bitrate signal. Nonetheless it is possible to receive such signals via CDMA techniques.
It should be noted that most of EDF's massive losses are due to the ARENH.
The European Union insists that EDF must sell energy at very discounted prices, so that third-party "providers" can make an entry on the energy market. The idea was that they would eventually sell their own energy supply, but most just pocketed the difference between the dirt-cheap energy & what they charged customers, then ran away the moment there was any hint of change on the horizon.
Or, to put it in simpler, blunter terms: in the name of "competition", EDF was forced to heavily subsidize companies that turned out to be nothing more than rent-seekers that only sought to, effectively, grab free subsidy money.
So that's an European thing? huh. We have this in Romania - a couple years back when the war in Ukraine started just as the green deal took effect, the gov started spending like crazy on subsidizing energy. But they did it in a convoluted way with a layer of intermediaries that basically were allowed to invoice the state for price differences from arbitrary price levels. Almost "I'd like to sell at twice the price but you're not letting me, so gimme the difference" - if not exactly that.
I'm not sure if I'm feeling better or worse that it's a EU invention. Either way, it's hellof a corrupt practice.
I read your link and I don't see why you say originated. This is a French law. My understanding is that EDF wanted to take a stake in a German energy producer and to approve such a takeover the EU as the market authority required some type of market liberalization of the French energy market.
France chose to use the mechanism of ARENH. This isn't an EU thing.
I'm sure the court could have gotten him on other charges, but they went with the absolutely 100% safe one rather than the other 99% safe ones.
Sarkozy and all of his billionaire media allies are already trying their hardest to undermine the credibility of the justice system at every turn with extremely dangerous rhetoric; I dread to imagine what this would have been like had they gone with ever-so-slightly-less-safe charges
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)