This time it's another video, and another claimant, against a "song" that was released after the supposedly infringing video. It seems Content ID has a keyboard problem.
I'm pretty sure this instance isn't a lazy regex problem (though that is also a thing), it's straight-up fraud on the part of the party making these copyright claims. They're finding existing videos, uploading their own (infringing!) copies, then making copyright-infringement accusations against the originals.
Google should be requiring people making copyright-infringement claims to provide payment information or a bond, and investigate enough of the disputes to make this scam unprofitable. Allowing this sort of thing to happen unpoliced makes YouTube videos de facto ephemeral; one of the important use cases is that people upload-and-forget things like lectures, and those have a tendency to disappear when no one's keeping an eye on them.
It is not ripped from my video; it's just someone else typing on a keyboard. Content ID just seems prone to throwing false positives on non-musical noises, since it doesn't understand that the frequency spectrum over time of noise isn't copyrightable like the frequency spectrum over time of music is.
That said, yes, the system is broken in more weird ways. I have a video where I played back a B&W music video through a laser projector, and YouTube keeps offering me to make copyright claims against other videos of the same song/MV run through edge-detection algorithms and filters that look similar. The funny thing is that video is (correctly) Content ID copyright claimed/monetized by the song's owners, but I guess I have the right to place a sub-claim on copies of the video part? (even though none of the hits are actual rips of my video, just people doing similar stuff to the same source material, so they're all false positives anyway). These algorithms just don't understand derivative works; when two people use the same source material, it just thinks one ripped off the other.
I did get one actual wholesale match of a rip-and-reupload against an original video of mine though, claimed it, and the author did take it down, so it worked as intended in that one case. But the false positive rate is atrocious.
Edit: upon further inspection, the website of the distributor for the "song" in question says they specialize in Content ID and it's suspiciously generic: https://www.umag.io/. That makes me wonder if they're just a front for Content ID abuse. The "song" isn't stolen, but it might've been distributed as part of a collection of fluff to see if they get Content ID hits on noise. There are no hits for this "artist" other than the music stores themselves.
Yes... content providers like the MPAA, RIAA, and other corporate organizations. Essentially, the Youtube system was permanently wired to favor the same set of interests that have corrupted the US copyright system.
By automating their claim system and related customer service, Youtube achieves a high success rate for handling claims "properly" from the top few creators as organized by revenue. They probably give satisfactory service to the people who generate 90% of the money for the platform using their automated claim system, which is cheap to operate.
In other words, their system handles the people who make them money and screws over anyone who's not big enough or corporate connected. People who don't make them enough money don't matter.
I can hardly remember a time when Google's motto was "Don't be evil."
They can’t charge anything, they can’t audit anything, they must basically treat every DMCA claim as being legitimate or they open themselves to liability if they’re wrong.
This is not Google’s fault. It is the fault of the DMCA as it has essentially no penalty for incorrect claims. Even a deliberately fraudulent claim has no direct penalty - as I understand google would have to bring perjury charges (costing money) and likely not receive any compensation.
The DMCA is trivially a useable, and it isn’t possible for a company to defend themselves from being used to enable such abuse, without opening themselves up to legal liability.
These are not DMCA claims. These are automated Content ID claims, made possible by the automated system Google put in place for this purpose. They have nothing to do with any law. It is an extralegal mechanism. The DMCA only comes into place once disputes get rejected by the copyright holder. Google can and absolutely should police Content ID more, such as by penalizing groups that register non-music into the system.
I guess they simply don't have an incentive to do so - media companies immediately attack anything that "weakens" their tools, and their senators will act accordingly :-/
I think google could possibly watermark the original video upload somehow.
It could remove frames or other information from the uploaded video, or it could add some sort of invisible watermark.
Then it's possible that folks who download the video from google and try to submit it would have a less detailed video and could be easily seen as a copy.
> This is why I have a `sudo` password entry wrapper that mutes the mic (yes, really).
That’s a nifty concept for hardening security a bit. I imagine you might want to do the same for the camera (though that’s hopefully an even smaller risk than audio, unless you have a camera pointed at your hands or keyboard), blanking it briefly; I’m not intimately familiar with v4l2 or pipewire or whatever, but I don’t think video stuff has quite the same “mute” ability?
My streams don't have camera feeds at all, so that's not a problem; I stream a full screen share only (except when I'm showing a camera pointed at a laptop screen in an ffplay window, but that's rare).
For v4l2, I imagine the simplest hacky way to do it would be to change the camera settings to result in a pure black or otherwise unusable image (e.g. zero exposure, zero brightness). That can be done with v4l2-ctl. But at that point you're probably using OBS for streaming anyway, so you might as well just use the OBS-websocket API to hide or censor the camera frame instead (and mute the audio that way while you're at it).
Sorry, for the slower of us, what does that mean? Muting a microphone with a script? Would the poster effectively keep their mic muted until they want to speak (and issue sudo commands two times every time they speak)?
I think they mean mute the mic, do the actual sudo then unmute. A proof of concept script that I have hacked together from some other things I have (untested because I'm on my mac not my linux laptop atm) would look like this for linux
#!/bin/bash
set -euo pipefail
#Get current mute status
SINK=$( pactl list sinks | sed 's/Sink #//;q' )
current_pa=$(pactl list sinks | awk '/Mute/{print $2}' )
current_cap=$(amixer cset iface=MIXER,name='Capture Switch' | awk -F= '/: values/{print $2}')
#mute it
pactl set-sink-mute "${SINK}" 1
amixer cset iface=MIXER,name='Capture Switch' off,off > /dev/null
/usr/bin/sudo "$@"
#put it back the way it was before
pactl set-sink-mute "${SINK}" $current_pa
amixer cset iface=MIXER,name='Capture Switch' "$current_cap" > /dev/null
I think you might only need the amixer part to mute the mic but I'm muting the speaker as well with pactl.
Close. That wouldn't work for long-running commands, since they'd keep the mic muted. Instead I use a sudo askpass wrapper to only mute during the password prompt bit.
Initially I thought they were muting the mic so that the key strokes would not be heard/recorded.
There has been talk in the past of inferring passwords based on the unique sounds of each key, time between strokes, etc... Moreover, the more frequently you do broadcast your password keystrokes, combined with all the other key stroking you do throughout the rest of your stream, the easier it becomes to come up with the password.
Muting the microphone while you type the password. That is, you run `sudo foo`, and it mutes the microphone before giving you the “[sudo] password for ‹username›: ” prompt, then unmutes it when you have entered the password and pressed Enter, and then runs `foo` as root.
Corporations like AdRev have as their whole business model trolling small content creators for copyright. Every channel on YouTube beyond the smallest seems to have the same story - corporations claiming that content created by the channel belongs to them.
Even the "legitimate" claims assume guilt on the part of the creator and disable monetization immediately until the claim is resolved. Then there are the false claimants, like third world channels copying popular videos from successful channels, then either appending a tiny "comment" section or flipping the video side to side and claiming it as their own. Heaven help the channels who then ask for credit or revenue from their own stolen content.. the thieves use the copyright claim system to abuse and torment the creators they stole from.
Until Youtube/Google are forced to disable their cheap AI based content claim system and implement something smarter that does the job right (and which must involve humans), small creators are going to suffer. Youtube holds a de facto monopoly on its niche, it needs to be forced to play fair, and either regulated or broken up.
Just to make it clear: are small creators going to pay Youtube enough to keep an army of human claims checkers?
Youtube, being an effective monopoly due to the network effect, has no incentive to lower its profits by paying the (large) amount of salary to the prospective claims checkers.
You don't have very good household finances, do you? Because that logic sounds conducive to only fiscal ruin. Can and should are very different notions. Just because you can remortgage your home to buy a luxury car doesn't mean you should.
The whole Karen notion that "Big Tech should hire armies for my whims" is like saying Amazon should be exact solving Traveling Salesman because it has so much processing power. It fundamentally willfully ignores the scale.
Who says they aren't already? If Youtube had no more small creators from on, I have my suspicions about what their revenue would look like 1, 5, 10.. years out.
The amount of money that Youtube spends on content and copyright management will probably have to increase, and to be fair the cost might well be shared across the board evenly among all creators as a percentage of income. YouTube will also have to adjust its spending, the creators can't/won't bear all the cost.
That may mean that the present business model of YouTube isn't viable. If not enough revenue comes in to pay for managing copyrights, providing hosting, corporate management, marketing, etc then a new business model may be needed.
Or whether YouTube exists or not as a business may become a philosophical question - is it better to have a YouTube that is inherently unfair and suppresses small creators or to have no YouTube at all?
Personally, I think video hosting is the sort of thing that should be done by an NGO or non profit - one that has defined processes for fairness and transparency and which is required to be equitable to all instead of playing corporate favorites. Kind of like a PBS for the Internet.
Relatedly, Pewdiepie received claims on his own songs from someone "on behalf of Pewdiepie" (not commissioned by him). https://youtu.be/Ku1ykhGP764?t=388
32 comments
[ 5.0 ms ] story [ 72.2 ms ] threadhttps://news.ycombinator.com/item?id=27999813
This time it's another video, and another claimant, against a "song" that was released after the supposedly infringing video. It seems Content ID has a keyboard problem.
Google should be requiring people making copyright-infringement claims to provide payment information or a bond, and investigate enough of the disputes to make this scam unprofitable. Allowing this sort of thing to happen unpoliced makes YouTube videos de facto ephemeral; one of the important use cases is that people upload-and-forget things like lectures, and those have a tendency to disappear when no one's keeping an eye on them.
https://www.youtube.com/watch?v=s0ClqZHkSnc
It is not ripped from my video; it's just someone else typing on a keyboard. Content ID just seems prone to throwing false positives on non-musical noises, since it doesn't understand that the frequency spectrum over time of noise isn't copyrightable like the frequency spectrum over time of music is.
That said, yes, the system is broken in more weird ways. I have a video where I played back a B&W music video through a laser projector, and YouTube keeps offering me to make copyright claims against other videos of the same song/MV run through edge-detection algorithms and filters that look similar. The funny thing is that video is (correctly) Content ID copyright claimed/monetized by the song's owners, but I guess I have the right to place a sub-claim on copies of the video part? (even though none of the hits are actual rips of my video, just people doing similar stuff to the same source material, so they're all false positives anyway). These algorithms just don't understand derivative works; when two people use the same source material, it just thinks one ripped off the other.
I did get one actual wholesale match of a rip-and-reupload against an original video of mine though, claimed it, and the author did take it down, so it worked as intended in that one case. But the false positive rate is atrocious.
Edit: upon further inspection, the website of the distributor for the "song" in question says they specialize in Content ID and it's suspiciously generic: https://www.umag.io/. That makes me wonder if they're just a front for Content ID abuse. The "song" isn't stolen, but it might've been distributed as part of a collection of fluff to see if they get Content ID hits on noise. There are no hits for this "artist" other than the music stores themselves.
By automating their claim system and related customer service, Youtube achieves a high success rate for handling claims "properly" from the top few creators as organized by revenue. They probably give satisfactory service to the people who generate 90% of the money for the platform using their automated claim system, which is cheap to operate.
In other words, their system handles the people who make them money and screws over anyone who's not big enough or corporate connected. People who don't make them enough money don't matter.
I can hardly remember a time when Google's motto was "Don't be evil."
This is not Google’s fault. It is the fault of the DMCA as it has essentially no penalty for incorrect claims. Even a deliberately fraudulent claim has no direct penalty - as I understand google would have to bring perjury charges (costing money) and likely not receive any compensation.
The DMCA is trivially a useable, and it isn’t possible for a company to defend themselves from being used to enable such abuse, without opening themselves up to legal liability.
It could remove frames or other information from the uploaded video, or it could add some sort of invisible watermark.
Then it's possible that folks who download the video from google and try to submit it would have a less detailed video and could be easily seen as a copy.
> This is why I have a `sudo` password entry wrapper that mutes the mic (yes, really).
That’s a nifty concept for hardening security a bit. I imagine you might want to do the same for the camera (though that’s hopefully an even smaller risk than audio, unless you have a camera pointed at your hands or keyboard), blanking it briefly; I’m not intimately familiar with v4l2 or pipewire or whatever, but I don’t think video stuff has quite the same “mute” ability?
For v4l2, I imagine the simplest hacky way to do it would be to change the camera settings to result in a pure black or otherwise unusable image (e.g. zero exposure, zero brightness). That can be done with v4l2-ctl. But at that point you're probably using OBS for streaming anyway, so you might as well just use the OBS-websocket API to hide or censor the camera frame instead (and mute the audio that way while you're at it).
There has been talk in the past of inferring passwords based on the unique sounds of each key, time between strokes, etc... Moreover, the more frequently you do broadcast your password keystrokes, combined with all the other key stroking you do throughout the rest of your stream, the easier it becomes to come up with the password.
Now I'm not so sure that's what they meant?
Bc someone could decipher your passwords if you record the sound of your keystrokes while entering them?
Even the "legitimate" claims assume guilt on the part of the creator and disable monetization immediately until the claim is resolved. Then there are the false claimants, like third world channels copying popular videos from successful channels, then either appending a tiny "comment" section or flipping the video side to side and claiming it as their own. Heaven help the channels who then ask for credit or revenue from their own stolen content.. the thieves use the copyright claim system to abuse and torment the creators they stole from.
Until Youtube/Google are forced to disable their cheap AI based content claim system and implement something smarter that does the job right (and which must involve humans), small creators are going to suffer. Youtube holds a de facto monopoly on its niche, it needs to be forced to play fair, and either regulated or broken up.
Youtube, being an effective monopoly due to the network effect, has no incentive to lower its profits by paying the (large) amount of salary to the prospective claims checkers.
The whole Karen notion that "Big Tech should hire armies for my whims" is like saying Amazon should be exact solving Traveling Salesman because it has so much processing power. It fundamentally willfully ignores the scale.
That may mean that the present business model of YouTube isn't viable. If not enough revenue comes in to pay for managing copyrights, providing hosting, corporate management, marketing, etc then a new business model may be needed.
Or whether YouTube exists or not as a business may become a philosophical question - is it better to have a YouTube that is inherently unfair and suppresses small creators or to have no YouTube at all?
Personally, I think video hosting is the sort of thing that should be done by an NGO or non profit - one that has defined processes for fairness and transparency and which is required to be equitable to all instead of playing corporate favorites. Kind of like a PBS for the Internet.