On Facebook and YouTube, classical musicians are getting blocked or muted
washingtonpost.com
washingtonpost.com
I realize this is machine learning and music sampling but I'm a tad surprised that the dataset cannot include a field such as "this recording is of public domain music"... and after a YouTube channel has to keep saying "No, I'm not playing someone else's CDs, we have a choir singing these live -- can't you hear the imperfections and that the soprano missed that eighth?" I suspect that at some point the channel will be cancelled because its too much of a copyright hazard.
Maybe YouTube just has to commit copyright seppuku to achieve enlightenment.
Their automation obsession has resulted in a firehose of erroneous takedowns that 100% aren't required by law, and they deserve all the ridicule and damage to their business that comes with that.
The fight doesn’t have to take place in the realm of law. At best copyright holders would have sued youtube into oblivion if they didn’t do proactive takedowns. At worst, well considering that Comcast owns both tons of copyrighted property (through Universal Studios) and major telecommunication infrastructure, I can’t imagine how dirty they can play if they had to.
So they will push music attached to dollars, according to the strictest legal terms of the loudest, most punishing complaints. Everything else is ignored, and thus sinks, or gets shoved off the top of the index by greed beasts that ignore courtesy, decency and rules.
Obnoxious greed beasts are sociopathic, and will game every tool Google arms them with. They want to be ranked highest for all keywords, as cheaply as possible.
So, how does one invert this payola scheme, as an eyeball used to justify advertising profits?
You have to tear the greed beasts limb from limb.
It may not be obvious, but the people that are filing spurious claims against you, are not good faith actors. They are the sock puppets of people who stand to make money from the keywords you stand to outshine all others on. You are likely crowding out low quality losers, who unfortunately have enough money to buy off-shore social media fleshbots that protest for them in surrogate.
If there's not a financial incentive to prevent situations like this, the likelihood that YouTube will correct it is approximately nil.
The sad truth is that the parties controlling the recordings that YouTube's algos erroneously protect are probably worth more to them financially than some church that broadcasts these performances.
It isn't an unreasonable critique but there are schools of capitalism that do not like the copyright system. That isn't an inherent part of capitalism. Peak capitalism tends to ignore copyright (eg, Silicon Valley in the early glory years, or the market for movies once BitTorrent was developed).
But, to a first approximation, in a capitalist system, if $THING doesn't make money, $THING isn't going to get done at any scale which matters. There are exceptions when philosophical principles and inherent human desires (such as curiosity and artistic expression) are involved. The corollary here, in true "the market adapts" fashion, is that if $THING is done and an entity with sufficient money dislikes $THING enough, then $THING will no longer be happening. I'm not sure whether the initial statement, or the corollary is fundamentally more distasteful.
That's precisely the issue. It is technologically impossible to solve the problem by analyzing the audio.
1. Two different performances could be audibly indistinguishable.
2. An infringing copy isn't necessarily indistinguishable. Often they have their own imperfections.
Are there other hosting options? Vimeo, one of the podcasting "networks", or even a terrestrial radio station which has online archives?
I didn't find out whether they consider religious content as controversial or otherwise unwanted on their platform, but from the overall view, it looks fine.
If 100B€ company is against me then all of us should expect a painful fine if we loose. For me that will be 10k€,for them 1B€ (or whatever).
They can afford to send lawyers whose job is to just do that: go to court. I have to take a day off.
If the fine is 10k it is insignificant to them and they can assume they will lose and this is no big deal. For me it is.
This unequality between a normal person and a company is painful. I belive that if companies were possibly in trouble when they loose then they would make more efforts to provide a good service (also out of court arrangements should be forbidden to force them to face the music)
What good are the courts if the law is on your side but justice is inaccessible?
The risk management calculation is completely different if you are an individual of limited means who cannot afford to lose as opposed to a deep-pocketed entity that can absorb the worst-case costs.
They already do this to get away with things like Hollywood accounting. There is a tiny little nominally independent corporation that owns the rights to a thing and is then completely unprofitable because whatever money it makes goes right back out as costs to whoever it is really making the profits.
So then you go to court with a "small business" with "less than ten employees" which qualifies for all the exemptions to whatever costs you're trying to pin onto major companies. This may in fact even make the problem worse, because very small entities they can shoehorn their legal troubles into may pay lower penalties than the current average.
The actual solution is to impose penalties that serve as a deterrent on everyone. Because why shouldn't a two person company that sends out millions of fraudulent takedowns end up with a billion dollar fine? The fine is proportional to the damage you do, not how much money you have. Otherwise what's to stop very small entities from causing rampant destruction and then getting a $25 fine?
The same proportion should apply to individuals and their incomes as to companies.
Problem solved.
Nevertheless, I do agree more should and could be done, especially for fines and penalties (less so for debts and obligations). It shouldn't be hard to prove those are shell companies, or fragments of a joint adventure.
Not even close.
There is no requirement for the company doing the dirty work to be a subsidiary. It can be "owned" by any random person drawing a lower middle class salary, because the entire premise is that that entity doesn't actually make any money. It does what you want because otherwise you don't do business with it, not because it's a subsidiary.
Class actions exist to solve the collective action problem of the harm not being large enough for a single person to pursue.
In other words the court should not be only there to fix harm, but also to teach a lesson.
Sorry what.. are you arguing against the concept of criminal law?
Because western states have happened to have chosen over the centuries to protect the general welfare by justifying its methods as "teaching people lessons" is neither explicitly permitted in any charter so far as I know (admittedly a small subset) nor necessarily desirable. I think it's not a n unfair argument that over the few centuries of modern history you can point to several incidents, sampled across many different nation-states, where particularly disgusting injustice has been promulgated under the cover of "teaching lessons", as if the general cover of "greater good" weren't bad enough.
As CS lewis quipped, "Of all tyrannies, a tyranny sincerely exercised for the good of its victims may be the most oppressive. It would be better to live under robber barons than under omnipotent moral busybodies. The robber baron's cruelty may sometimes sleep, his cupidity may at some point be satiated; but those who torment us for our own good will torment us without end for they do so with the approval of their own conscience."
That's not what they're intended for. If all harms are efficiently deterred and compensated, it's only our base animal instincts that suggest there's something more that needs to happen. Civilisation, of which the legal system is a subset, is there precisely to tame our baser instincts.
Just to head off a foreseen objection: the correct amount of deterrence isn't to prevent a harm from ever happening (trivial, just bankrupt everyone who does any harm), but to provide a negative incentive on the perpetrator equal to the harm they will cause by doing some act. If the chances of someone being caught are low, then it's proper to multiply the eventual fine by the inverse of the probability, which creates the required incentive. If someone showed malice, then that's evidence that the incentives were insufficient, and a reason to increase the fine. But none of that, from a legal theory perspective, is to punish them or bring fairness - it's to create the appropriate economic incentives.
Read up on https://en.wikipedia.org/wiki/Efficient_breach for an example of this kind of thinking.
Punitive damages are a longstanding feature of civil law.
This happens all the time, not just with copyright law, so doesn't that seem to imply that the penalty for the larger party is insufficient?
Not necessarily. With copyright law there's practically no penalty, and I think that should be changed as I said elsewhere in the thread. But the optimal penalty level should be the amount of damages, plus some amount of punitive to account for the probability of a case getting that far, if it's low. If a big company decides that it's worth paying that amount in expectation in a standard tort case, then it is, in fact, efficient to allow that to happen and to have them compensate the other side in the instances where it does happen. At least when we're talking civilly and not criminally, where different concerns apply.
A class action on behalf of content creators who are being abused combined with federal antitrust action might persuade it otherwise - especially if actions were brought simultaneously in the US and EU.
Someone tried suing Amazon over basically this theory and failed, same lawyer that I'm using in an antitrust case against tp-link for false infringement notices to Amazon.
See https://www.courtlistener.com/docket/6169804/mordys-applianc...
And https://www.courtlistener.com/docket/16562550/thimes-solutio... for my tp-link case, I've mentioned it on HN several times. I have a list of similar cases, some of them use antitrust theories, but it's almost always against the brand/rights owner and not against the platform.
Personally I think the liability should be on the party submitting the complaint, not the platform. As long as the platform is permitted to reinstate the content after an appeal (which is what the DMCA counternotice safe harbor is about), that's enough. The DMCA framework works well [0], and I would like if it was extended to patents and trademark/counterfeit accusations as well.
[0] with the caveat that the standard for liability for false claims should be lowered, and they should be treated as per se antitrust violations.
Google has been trying to hide behind the "Oh dear our algorithm is experimental and fallible and sometimes it gets things wrong and that's very sad" defence, and it's obviously nonsense. You don't even need explicit evidence of collusion to understand why.
The "failure" of the algorithm has been happening for years now, and Google appears to have made no effort to correct the failure. But the real tell is that these "failures" disproportionately benefit large corporations which Google will have some relationship with - as opposed to small content creators who are disposable.
Ultimately, these actions are not mandated by law. So what's the legal - never mind moral - justification for them?
Nice theory, but in practice the process itself is often a punishment. Even winning a bogus case against you can come at great personal expense.
> The flip side is that anything you do to increase payouts also increases the number of suits brought
So if we flog lawyers for SLAPP suits, they'd do it more often? I have doubts about that; we should put that theory to the test.
You'd certainly have an increase in the number of frivolous anti-SLAPP motions.
Very related: I've been hit with a frivolous anti-SLAPP when suing a company that made a false accusation of counterfeiting against my Amazon account. https://www.courtlistener.com/docket/16562550/thimes-solutio... (can download the PDFs free at https://digitalcommons.law.scu.edu/historical/2153/ - feel free to contact me for all the case documents if you want.)
TP-Link is arguing that they have a 1st Amendment right to tell Amazon that someone is counterfeiting, even if untrue. They cite Noerr Pennington, which is a 1st Amendment doctrine. And this is a quote from their brief:
>Courts have found that the litigation privilege applies “regardless whether the communication was made with malice or the intent to harm.”
The anti-SLAPP was quickly denied. But the fact they brought it at all should show how this could get abused. Someone less confident than me might have dropped the case rather than risk having to pay legal fees if the motion was granted. And note that in a similar case against TP-Link, the anti-slapp motion was granted - that decision is up for reconsideration now, but the other side is asking for around 123k in fees. (See https://www.courtlistener.com/docket/14522683/tp-link-usa-co...)
It is my own personal, non-lawyer opinion that the anti-SLAPP motion in the careful shopper case was granted improperly. You can disagree; clearly a federal judge disagreed. So, in this context: do you think that lawsuit (or actually, it's the counterclaims that were subject to anti-SLAPP, so the counterclaims) should not have been brought? The brief facts of the case: TP-Link filed complaints to Amazon saying careful shopper was counterfeiting, got their account shut down, careful shopper sues for defamation and similar claims saying it was false and TP-Link knew it was false, TP-Link files anti-SLAPP saying even if this is true, they have free speech rights to say this to Amazon.
Steal stuff all the time. Sometimes I would get catched for whatever I was stealing. Proudly I will face all of the consequences: I just give back what I was stealing. All harm done is fixed. I keep all the other stuff that I stole where I was not catched. Why should I stop with that enterprise? The law certainly doesn't stop me.
Isn't that law way too gameable?
I think a system like this would only "work" if you catch _everything_ unlawful. You basically need a totalitarian police state dystopia for this.
I think the law should incentivize right behaviour and disincentivize wrong behaviour. Not just to fix harms done.
Also, certain torts, like stealing, are considered criminal.
Yes, youtube will give you a big scary warning about how repeated violations could mean your account is shut down. Every time this happened I went ahead and submitted counter-claim anyway. On all but one occasion the copyright claim disappeared immediately. The one exception required a bit of escalation but I still wound up on top (https://news.ycombinator.com/item?id=22488897 if you're curious).
I understand a good many people simply do not feel comfortable risking their youtube account being shut down from contested claims, but I suspect that for many the risk is a lot lower than they realize.
Sure, you can sue. Maybe even get your account back, months later... if you survive the emergency without communications of any kind.
Censorship-possible communications systems are an existential threat to a free society even if they are not used to censor 99.99% of the time. The fact that they can be wielded arbitrarily against people, outside of the law, or only with slow and expensive recourse (that fail deadly in the interim) should never be forgotten, especially during the times that they aren’t and everything seems fine and normal.
Do not use censorship/surveillance platforms, even if you are not being censored.
The "privilege" of being able to play your own original works on your own channel - generously granted by a random company that has no business relationship to you whatsoever...
Because I haven't heard of Facebook/YouTube doing so, though I'd love to know if they do.
And if it's Naxos, well that doesn't exactly solve the problem does it? Because the claim against your recording might be Naxos one day, but another label next week, and yet another label the week after that. Naxos is a big label, but it certainly doesn't have even the majority of classical music or anything.
We need to work on AI that can evaluate the musical talent in a recording. I mean yes, I was playing the same song as some famous musician, but anyone could tell I am an amateur and my recording was not professional.
Years of reading the arguments on HN (and other forums and aggregators before that) have convinced me that technology will not provide the answers to this problem. We need better laws from governments, and better policies from corporations.
YouTube’s company policy is essentially “ask us and we’ll take it down”, and has nothing to do with what the law requires.
---
This is exactly the problem. It's obvious that there is a "not-insignificant" amount of cases where the there isn't extremely high confidence in the automated model.
Instead of these cases defaulting to claims, it should default to human review.
The people who still dare to perform a classical concert somewhere in the hinterland are immediately punished and exempted by these collecting societies. In the age of the Internet, this is now even automatic. No wonder that there are fewer and fewer places that perform.
So today the shot is well and truly fired in the wrong diretion when composers and musicians assign their rights to such a society. Where there is hardly any market anyway, the promoters are penalised even more, although composers and musicians should be rather glad that someone is still listening to them.
I, for my part, am not affiliated with any such society anymore and have all my fellow musicians sign agreements whereby I pay them directly and no further claims of such societies are permissible.
Ironically, a musician working as a clerc for a collecting society earns at least an order of magnitude more than what the median musician can expect as annual royalties from the collecting society.
* As a YouTube channel owner you have the option of officially registering your channel with verified contact information.
* A video you have uploaded gets flagged by a bot.
* YouTube mutes the content and immediately texts you an alert.
* If you have opted to officially register with YouTube, you are given the ability to text back a countermand to the alert.
* If you countermand the copyright violation flag, your video is immediately unmuted, and the copyright holder receives instructions that if it still believes the content is an IP violation, commencing legal proceedings are the next option.
* If and only if the copyright holder affirms to Google that it will follow through with legal proceedings, it is given your contact information.
These companies should face criminal charges for this extortion. It's disgusting and outrageous.
> In Green v. AOL (2003), the court established:
> There is no real dispute that Green's fundamental tort claim is that AOL was negligent in promulgating harmful content and in failing to address certain harmful content on its network. Green thus attempts to hold AOL liable for decisions relating to the monitoring, screening, and deletion of content from its network — actions quintessentially related to a publisher's role. Section 230 "specifically proscribes liability" in such circumstances. Zeran, 129 F.3d at 332-33.
> There is immediately a question why this "quintessential relation" only works in one direction, i.e. why even when "monitoring, screening, and [deleting content]", a provider should not be considered a publisher, and thus ineligible for immunity under Section 230. The courts may not agree with that argument, but I would say that it highlights a serious deficiency, and the law needs to be changed in a similar manner as Josh Hawley's bill: https://www.congress.gov/bill/116th-congress/senate-bill/191...
> In fact I would argue that his bill does not go far enough. Getting an FTC rubber-stamp certifying "no political bias" seems like a vague and squishy system that will be easy to game. Perhaps the law should state that any provider above a certain size which takes action to "monitor, screen, or delete content" that is not illegal, and does not respond to an appropriate claim by reversing this decision, should forfeit immunity under Section 230. This would be very controversial, because it effectively bans e.g. YouTube from having their own "code of conduct" independent of US law. But I think that is exactly what needs to happen - YouTube is far too big for a "community standard" to make sense, and any attempt to do this will be abused for censorship. These companies should be forced to engage with all of us in the making of laws about free speech, they should not effectively be able to declare what free speech is by fiat.
The part that I don't understand is: Shazam is really bad at classical music. Unless it's the exact recording of the piece as their database, it seems, Shazam won't recognize it at all. And a little noise or static is enough to throw it off.
How is it that FB/YT content detectors are so far ahead of the game that they can detect old music in completely new recordings? Or, alternatively, why can't FB/GOOG slap a simple UI on their services and make an app like Shazam for users but 50 times better?
We've got the worst of both worlds right now: bad detection of music when we want it, and overzealous matching of new recordings when we don't.
I don't think a lot of people know how truly bad Shazam is at classical music. Stuff that should be very obvious (NPR Live From Here, for example) routinely gets misidentified for me.
I feel classical music is especially tricky from a ML modeling perspective. How many unique recordings are there of Pachelbel's Canon in D?
This is probably similar in difficulty to identifying a specific bird in a flock of thousands, and then looking at another discontinuous (in time) photo of the flock taken from a slightly different perspective and being able to recognize the same exact bird again.
I was flattered and annoyed when one of my recordings was taken down. I contested it, it was restored. This was SoundCloud though
The article vaguely implies that there is also an issue of miscalibration of the matching due to emphasising commercial pop-music but I've never seen any evidence of that and wouldn't expect it from how these systems work. In particular, I've found various contentid services to be effective at identifying classical music-- so clearly their database contain plenty of it. I expect that this claim is pure speculation on the part of the article's author.
A lot of pop music "sounds the same" but the songs are actually different in meaningful ways note-wise. When you're dealing with classical music it's all the same notes! Just different people playing it.
So probably what's happening (some-how, some-way) is that the ID mechanism is trained to find songs that are "the same" and in pop music they're only "the same" if they're the _same song_. But in classical there are many songs that are different that are the _same song_ and so the model which was trained with one set of assumptions completely fails when a different set of assumptions are in play.
And because classical doesn't register in a big way to the people in charge this doesn't get fixed because it only takes the algorithm from 99.85% accurate to 99.72% accurate which as we all know, is meaningless!
Right. Pop music has a lot of variation in timbre even between 'covers' of the same composition, so if the notes match and the overall timbre matches it's a safe bet that you're looking at two instances of the exact same recording, or at most two versions by the same performer. In classical that just isn't true, the timbre of acoustic instruments is basically fixed and the notes are defined by the original composition. There's just too little variation for an auto-id system to work with.
This doesn't mean there won't be copyright claims, but the burden should be on the copyright owner not the user. If this platform is a "public streaming system", then no ads are shown and the limits of infringement will be constrained by the lack of ads (ie "fair use"). Having no financial incentives for people to make money off of other people's work makes it all much simpler.
A "Wikitube" would have its many challenges, certainly, but resolution would lean on the side of the public interest, something that has been wildly forgotten in the digital age.
> “We’re at the mercy of automation in order to uphold our obligations to our clients,” Hammons says in an email. Like other record companies, Naxos relies on Facebook’s and YouTube’s content identification systems to track potential illegal use.
Naxos, please explain who is forcing you to use software/algorithms implemented by other companies, that you implied you know is harming the classical music industry/community? It's clear to me who is at Naxios' mercy, but it's not entirely clear to me whose mercy Naxios is saying they are at.
> an original performance of an hour-long piece composed by Mozart in 1786
I'm surprised the bot recognized the piece when it was apparently played at about half speed. :)
Makes life a hell of a lot easier the ai recognised this is public domain and so no one can make a claim on it.