Hacker News Reader

There are disputes over "white noise" videos so it is already as bad as it gets.

Given that white noise is random, a recorded copy of white noise is its own perfect watermark.

Does it depend on seeding that would eventually have tendencies for replication? Or could that be pushed out until practically forever?

If the white noise is based on a random number generator, the pattern would restart after a full cycle.

If the random number generator's period is 2^32 and you use one integer per sample, then at 44 kHz, you would have about a million seconds, or twelve days, before the RNG has gone through a full period.

Most RNGs have periods much higher than this. xorshift128 has a period of 2^128-1 and the Mersenne Twister's period is 2^19937-1.

So you could push it out until practically forever.

So you could push it out until practically forever.

But, there will still be repeated sections.

Knowing nothing of the YouTube algorithm, I’d still wager it’ll take you longer than the lifetime of the copyright to produce conflicting noise works.

Depends on how lossy the compression is. Your watermark becomes compression artefacts instead.

Most audio codecs by now use perceptual noise substitution, which potentially substitutes in identical noise (I think it just EQs it?).

no, the mp3 codec actually deletes humanly inaudible portions of the frequency spectrum to reduce storage, this data isn't recoverd or replaced or referred when played back again, it's lost information. what is removed is based on human perception limits for audio

He is referring to modern codecs, not MP3. MP3 isn't widely used anymore.

Modern codecs don't encode noise - they remove it during decoding and then add back artificial "comfort noise" when decoding, e.g. for film grain or background noise in voice calls.

Got a source for the current usage of MP3?

What services do you know of that use MP3?

Virtually every podcast ever.

YouTube itself seems to rely more on webm for audio. That seems to be a container for Opus or Vorbis formats.

Vorbis and Opus themselves are lossy encoders (Ogg Vorbis).

> Virtually every podcast ever.

Apple Podcasts are AAC. Spotify are Ogg Vorbis. That probably covers 90% of the market right there.

The original question was whether or not audio could be fingerprinted from white noise, or if compression would obscure that fingerprint. The original context was YouTube videos specifically.

Sufficient to support the latter assertion was that lossy compression is frequently used.

YouTube's webm does in fact use lossy compression as noted. MP3 is, I assert, still significantly used in multiple contexts, including many podcast sites. I haven't specifically surveyed those, and don't use Apple Podcasts myself (never bought into the ecosystem, though as it happens, this specific response is coming from a MacOS system). I do make heavy use of tools such as youtube-dl and mpv, and find that those do in fact frequently find and extract mp3 audio from various sources, including podcasts and IIRC Soundcloud.

Verification is as simple as:

    youtube-dl -F <URL>

The '-F' flag will list available downloadable formats.

E.g.:

   $ youtube-dl -F 'https://soundcloud.com/danyork/tdyr-362-thoughts-on-wordpress'
   [soundcloud] danyork/tdyr-362-thoughts-on-wordpress: Downloading info JSON
   [soundcloud] None: Downloading webpage
   [soundcloud] None: Downloading webpage
   [soundcloud] danyork/tdyr-362-thoughts-on-wordpress: Downloading info JSON
   [soundcloud] 637919466: Downloading JSON metadata
   [soundcloud] 637919466: Downloading JSON metadata
   [soundcloud] 637919466: Downloading JSON metadata
   [info] Available formats for 637919466:
   format code   extension  resolution note
   hls_opus_64   opus       audio only audio@ 64k
   hls_mp3_128   mp3        audio only audio@128k
   http_mp3_128  mp3        audio only audio@128k (best)

Pedants are my fourth favourite people, but only during Lent on leap years, prior to Vespers, during a blue moon.

Re: opus

'Opus (audio format) - Wikipediahttps://en.wikipedia.org › wiki › Opus_(audio_format) CELT includes both spectral replication and noise generation, similar to AAC's SBR and PNS'

Which would mean that at least some of the white noise signature of an opus-encoded audio track would be ... a compression or encoding artefact.

Does opus have a notion of bitrates similar to mp3 or ogg vorbis? If so, the white noise would vary among bitrates.

Podcasts are pretty niche. Almost all codec use these days is from streaming services or video/phone calls - YouTube, Spotify, WhatsApp, etc. None of those use MP3.

What codecs do those applications use?

How do they preserve or modify white noise?

You haven't named any lossy stream codecs that do restore white noise, so i will not consider you a reliable source of info.

> Another notable addition in this version of the AAC standard is Perceptual Noise Substitution (PNS). In that regard, the AAC profiles (AAC-LC, AAC Main and AAC-SSR profiles) are combined with perceptual noise substitution and are defined in the MPEG-4 audio standard as Audio Object Types.

https://en.wikipedia.org/wiki/Advanced_Audio_Coding

If you take 30 seconds of white noise and look for it in 2h long generated output I expect that you can find similar 30 seconds in whole 2h.

Copyright algos are not looking at the whole video, they are searching for pieces of songs.

Because if you use 30sec of someones song you have to pay up and youtube is enforcing that.

Now if you will take another 2h video cut it into pieces and start searching for similar patterns in other 2h video I expect you will find some matching ones.

If you're comparing sample-for-sample, maybe. But that's not what is performed by music identification algorithms - it would be both slow (to the point of being unusable) and inaccurate (because of re-encoding, etc). Instead they rely on different kinds of dimensionality reduction that match the way we perceive music, so they only compare smaller amounts of data to get more perceptually relevant matches. The Shazam fingerprinting algorithm is the best known but there are others.

So if you submit white noise to a copyright database, it could match different white noise recordings.

Even if they were the case, this further shows how limited and dumb (i.e. not 'smart') YouTube's copyright voilation detection system is.

You (probably) can't copyright white noise because US copyright requires authorship. So in the same way that you can't copyright a phonebook's alphabetical list of numbers, you can't copyright random numbers rendered as sound, unless you did something else unique to it to exert authorship. It's just not something protected by copyright law. So even copying someone else's exact white noise sample is probably just fine.

The problem is that YouTube's system can only apply simple content matching rules and it counts any sufficiently long content match as a violation with no consideration of the work or context of use. Thats not how copyright law works. Copyright is a complicated system with all sorts of issues like fair use, derivative works, public domain, and works that don't qualify for protection. It's not a database query.

I don't think it's hit YouTube, but there are people and/or bots enforcing a copyright on silence:

https://www.youredm.com/2015/11/28/soundcloud-finally-goes-t...

What would be worse: whoever holding the rights to John Cage’s “4’33” filing complaints against movies without background music, claiming that they have copyrighted background music.

> which is kind of bad, because now the support team must log-in with a Google account to YouTube in order to see the harmless but useful video.

Not only this, but in the EU* it requires verifying your age with a credit card or ID photo: https://i.imgur.com/gP20dXi.png

Since you van have a credit card when you are 13 yo, the check is more or less useless.

It's it effectively doxxing people under 13? That's a concerning implication if I'm understanding this correctly.

You have YouTube's content rating system backwards.

It's not "racey content gets marked special" with "suitable for children" being a catch-all category.

YouTube's "content developed for children", is the special case. It's explicitly content meant for and marketed to children, or content that children would be particularly attracted to, like nature documentaries.

All other content should be marked "not intended for children", even if it's not "adult"--aka restricted to 18+--content.

This is stated pretty clearly in YouTube's documentation. They have a link to it in a contextual pop-up right next to the form field asking you to self-rate the video.

This was the email I got:

> We wanted to let you know that our team has reviewed your content and we don't think it's in line with our Community Guidelines. As a result, we've age-restricted the following content:

> Video: Komoot Bug

> We haven't applied a strike to your channel, and your content is still live for some users on YouTube. Keep reading for more details on what this means and steps you can take if you'd like to appeal this decision.

> What "age-restricted" means

> We age-restrict content when we don't think it's suitable for younger audiences. This means it will not be visible to users who are logged out, are under 18 years of age, or have Restricted Mode enabled. It also won't be eligible for ads. Learn more about age restrictions.

When I click on appeal I get a popup titled "Submit an appeal" with the body text of "Appealing this violation is not available"

----

EDIT: Ah. I see. I actually did set it to "Is made for kids" thinking that this means that through this option I express that it does have no content which would be against the community guidelines.

I've now changed it to "Not made for kids" but also "Not age restricted".

Thanks for pointing this out.

>or content that children would be particularly attracted to, like nature documentaries.

maybe times have changed, but you could not pay me to sit through a nature documentary back when I was 8.

nature content is a very weird gray area of content in terms of age ratings. It's technically just, well, nature. But then there can definitely be "just nature" content you don't want a kid to see.

> Or maybe they just recognize that their AI is causing more harm than good.

Google recognizes that these flaws in their AI aren't worth caring about. Google doesn't have any mission or obligation to help the world share videos. Google cares about Google's profits. And they've found that the expedient way to do that is just let the AI be overzealous with rejecting, because the cost of a false positive is infinitesimally tiny and the cost of a false negative (real copyright violation) is so much higher.

How do we fix this? Competition. We need a Google/Youtube competitor so that users will choose the platform that does copyright recognition better.

Unless they've changed this you can get around this by sending the 'embedded' link: https://www.youtube.com/embed/<videoid> instead of https://www.youtube.com/watch?v=<videoid>.

Honestly, the only way things will change is through grass roots abuse of the system like this. You have to force them to change.

Sounds like good material for a Defcon talk: "How I Own A Quarter of Content on YouTube"

What if someone exploited this to such an extent that Google could no longer ignore the issue? All someone needs to do is write an open source library that automates this. Almost like exposing a security flaw to the public to force the company to fix it. Sure it would enable bad actors but it would subvert them in the long run. I feel this may be the only way to force the company's hand as an individual.

https://www.youtube.com/watch?v=-eBB1ZzvFoI

http://spiderrobinson.com/melancholyelephants.html

TL;DR copyright becomes absurd surprisingly fast when you have a large population, widely-available authoring/recording tools, and a way to store/search all of them, indefinitely. Like, indefensible absurd.

That was a good read. Thanks for sharing it.

This is a fantastic story. Thank you for the link.

> Ok. that's the way Google thinks is the way it should work. Or maybe they just recognize that their AI is causing more harm than good.

Let's be clear. It is not Google that wants it. Actually Google is on the side of calling all of this retarded. It is the state of copyright laws, of the DMCA, of the lobbying by the RIAA. The day the whole copyright ecosystem is updated to accept that computers exist and that people share files on internet easily, Youtube will be VERY happy to unplug all these terrible bots that are there to provide a bad solution to a problem we should not have.

> Actually Google is on the side of calling all of this retarded.

Google has the means to make this stop.

How so?

They just stop and wait for the DMCA take downs and copyright owners have to make their own claims? I guess.

Add friction to the process of submitting copyright claims. Reduce friction in the process of appealing copyright claims. Relax draconian rules like the copyright strike system leading to account closure. Adjust the content ID system to reduce false positives (assuming no magical improvement to the system this will come at the cost of increased false negatives; that seems like an entirely reasonable trade-off)

For starters, they should actually follow the DMCA. The DMCA gives affected people a defined way to counter-notice, and then the entity who filed the initial DMCA notice can either sue within 14 days or the content gets restored.

Instead Google chose to evaluate "disputes" themselves, with algorithms and "AI", and reject disputes. Rejecting counter-notices is absolutely NOT something the DMCA mandates or even suggests. Moreover Youtube essentially used to allow (probably still does) alleged rights owners to reject disputes in essentially one click, while the DMCA would require them to bring a law suit. Some copyright owners therefore created bots doing the clicking for them.

They could also be more lenient to "established" players as a first step, especially when it comes to counter-notices/disputes. Factor in previous history google has with an alleged infringer (alleged by their own algorithm by the way, not even by a third party) when considering a dispute, like account age, channel age, number of previous videos without problems, "we do know the customer" e.g. to pay out ad money, etc. And then maybe not outright reject it, but leave it to the alleged copyright owner to file a law suit (as the DMCA states) or at least refer it to actual human beings for further evaluation.

Of course, Google could hire people to check up on their own algorithms and decide on disputes instead of machines. Youtube had $6 BILLION in ad revenue in the last quarter (not year), so they could certainly afford to hire some people. In the end it might even be a profitable investment, as fewer good content is pruned from Youtube for wrong copyright issues, leading to more ad revenue.

Youtube right now seems pretty content with their quasi-monopoly, to their own detriment in my opinion. As unlikely as it may seem that people will create competitors, it can happen, ask mighty MySpace about it.

> Youtube had $6 BILLION in ad revenue in the last quarter (not year), so they could certainly afford to hire some people. In the end it might even be a profitable investment, as fewer good content is pruned from Youtube for wrong copyright issues, leading to more ad revenue.

Who is paying said ad revenue though? Could a large chunk of it come from the same companies/industries who currently enjoy the broken state of YouTube's DMCA process?

Parent comment was also referring to YouTube's AI marking a benign video as 18+. Google is willing to wrongly punish people via excessive/unfair false positives in the interest of trying to be more advertiser friendly. It's just about money.

Well, false positives result in fewer scandals than false negatives here.

So guilty until proven innocent seems like a perfectly reasonable, if very annoying, stance.

It's not reasonable to screw innocent people over.

It's reasonable for a for-profit company that has a monopoly.

No monopoly necessary to make this commercially reasonable stance to take.

(And, there's no monopoly situation here in any case.)

I don't think the laws are so harsh. It seems like YouTube caved in to the music industry interests so that popular artists continue to premiere their videos on YouTube.

And otherwise they'd get sued for hosting copyrighted content, which will result in hefty fines and jail( didn't the US want to get Kim Dotcom from NZ precisely for that?)

They won’t get sued if they follow the DMCA process which is very different from what they are doing.

They were sued [1]. The lawsuit lasted 7 years and ended in a settlement. The terms were not public, but I think it's likely they promised to institute a process that goes above and beyond what the DMCA requires. [1]: https://en.wikipedia.org/wiki/Viacom_International_Inc._v._Y....

So long as they respond to actual DMCA reports, no. ContentID goes far, far beyond what is required by the DMCA to have safe harbor.

Dotcom got in trouble because he was intentionally hosting infringing content, going so far as to try to hide that material from the copyright owners by making it look like he had taken it down when notified of its presence when in fact he just made the URL that the owners knew about stop working. Employees that went too far and actually took down infringing material got reprimanded.

Go dig up a copy of the indictment. It includes a bunch of internal emails from Dotcom and other running his site where they talk about all this stuff. It was basically a site whose intent and business model was hosting pirated movies. That you could also use it to host your own photos or whatever was there to try to provide cover.

Does the extent matter that much? Assuming YouTube did nothing to take down copyrighted content, is that better or worse compared to lying about taking down ?

The reason why Google is not accused of hosting copyrighted content is because they somehow managed to sell the fiction that streaming and download are two totally different things. The slightest touch can make that fiction dissipate.

Ehh, Google has gone to great lengths to ensure that the largest holders of copyright have their content libraries available to check against new uploads and apparently no lengths to ensure that public domain content is available for the same purpose.

The public domain doesn’t threaten to sue them.

It ought to.

Who has standing?

The public. It's ours.

Mskes sense, but how does that work?

Can any of us initiate a case and it becomes a class, or?

I'm open to having a conversation about it. What does the space look like between "no the public does not have standing", and "the public has standing individually or as a class"? What are the tradeoffs at those boundaries, and how can we work towards creative solutions that exist in the space between those extremes?

That convo should happen.

I dont believe this in the slightest. Besides, the false-positives which regularily pop up around YouTube claims are not a result of any law, or lobbying. They are the result of googles sloppy implementation of the cliam system.