Programmers generate every possible melody in MIDI to prevent lawsuits
musictech.net
musictech.net
https://news.ycombinator.com/item?id=22413526
Edit: And apparently discussed even more 2 weeks prior:
http://articles.adsabs.harvard.edu//full/1979HisSc..17..258K...
One fine day a marketing guy knocked on my door and asked:
"What would it take to print out every possible eight-by-eight bitmap? We want to copyright them so our competition can't use them."
Seriously.
So I told him the story about the king who wanted to reward the inventor of chess, and upon being asked what he wanted the inventor said "one grain of wheat for the first square, two for the second, four for the third..."
I thought a bit, and added "I think that printout would outweigh the planet."
He went away. I was not a helpful geek that day.
Later I realized that I had only considered black-and-white bitmaps, and that preempting copyright on color bitmaps would have meant a lot more planets.
There's got to be a Douglas Adams style tie-in here somewhere involving aliens with planet-chewer-uppers invading and taking Saturn, Mars, Jupiter and then us for some cockamamie copyright scheme in another galaxy . . .
[0] I wonder about estimating that time, many numbers may take longer than a second to even pronounce.
A series of 12 pairs of reduction gears, each at a 50:1 ratio. On the left, a motor spins at a high rpm; on the right, the drive shaft is embedded into a concrete block. It will be 13.7 billion years before the final gear completes a single rotation.
Secondly, that number is wrong anyway. The correct number is (2^16)^(8*8).
print(f"{16 ** (8*8):.4E}")
1.1579E+77
Python wanted to use a float to convert to exponential format but it overflowed. Need to use strings ;-)
n=(2**16)**(8*8); s=str(n); print(s[0] + '.' + s[1:6] + 'E+' + str(len(s) - 1))
1.79769E+308Taken from https://czep.net/weblog/52cards.html
So, just how large is it? Let's try to wrap our puny human brains around the magnitude of this number with a fun little theoretical exercise. Start a timer that will count down the number of seconds from 52! to 0. We're going to see how much fun we can have before the timer counts down all the way.
Your average printer can print at 600 dots per inch^2. A4 paper is 11.69in by 8.27in, with an area of 96.6763in^2. You can fit 58005.78 dots on an A4 paper assuming you can print on the edges, double that for both sides at 116011.56 dots.
Ignoring actually fitting the bitmaps on the paper, you can store 1812 bitmaps per sheet of paper. You can roughly fit all possible combinations of bitmaps on 10,176,500,000,000,000 sheets of paper.
Typical office paper weighs 5 grams. Ignoring the weight of ink, your total mass of paper would be 50,882,500,000,000 kilograms of paper. Pluto weighs vastly more at 13,090,000,000,000,000,000,000 kilograms.
You definitely don't need a planet's mass worth of paper, but maybe a couple of planets worth of production however. We'd be more likely to see planet scale enslavement in a Douglas Adams galaxy wide copyright mockery scheme, which galactic court would likely throw out since they don't appreciate mockeries of law, leaving everyone a bit disappointed and with entirely too much visibly grey paper.
Now, for the colored bitmaps case, how much does the ink cost? :-)
We're also ignoring some edge effects, where bitmaps can be "shared" by sweeping an 8x8 frame across a printed page at the level of pixels to generate the patterns. I don't know how to think about the scale of that problem right now, don't even know if it's easy to do optimally or wickedly difficult, or if Knuth has a solution somewhere in Vol 4 . . .
The cartridge costs about $40. We can expect to spend around $407,060,000,000,000,000 on printer ink for this endeavor, but HP might consider a bulk rate at that point.
The global world product is roughly $77,868,000,000,000. If we all unite under this noble cause (and assume no economies of scale), we can payoff the cost of the ink in only 5227 and a half years, assuming paper is free (which it totally is, it grows on tress!). If someone works out how much carbon this captures, we might prevent further global warming to troll domestic courts.
Some of the copyright lawsuits are dumb and this is effective satire or performance art but that's all it is.
The reality is that many number one songs can be tonally compared to many classical pieces, or even pieces from the last 40 years. The current state of music copyright law is an absolute joke, and deserves to be "disrupted" (destroyed).
You can use a ridiculous argument for many things.
Yes that’s actually the entire point of this
A lot of commenters seem to be missing the fact that this is a response to ridiculous arguments being used in winning court cases
Again we see programmers trying to understand the law in terms of ‘how can a piece of data be illegal?’ while the law is quite happily focusing on making specific actions illegal.
‘You can’t arrest me, gold bars aren’t illegal!’ ‘Yes, but carrying them out of the federal reserve vault without permission is.’
That's the problem with IP law.
In effect you give people monopoly on numbers. When the numbers are big nobody is bothered by this, because chance of arriving at the exact same one is effectively zero. But for songs the numbers are pretty small (depending on the encoding used to compare the songs), and the absurdity is evident.
i.e. sans lyrics
> Generally, a plaintiff proves copying through circumstantial evidence, showing that the defendant had access to the copyrighted work [...] > > [...] unlike in patent law, if a defendant independently creates the substantially similar work, he is not liable to the copyright holder.
https://www.finnegan.com/en/insights/copying-copyright-s-wil...
The fact there is no garbage data at all shows that they are doing judgement on a nakedly wrong level in music - even by the standards of copyright.
Selling "Harry Potter but with the capitalization inverted" to dodge book copyright nor even "this key and this very long block of data which happen to decrypt to the complete works of JK Rowling, don't decrypt because that would be infringement wink wink nudge nudge".
https://vinepair.com/wine-blog/how-wine-bricks-saved-the-u-s...
If we say that music (and really any information) is just numbers which can be enumerated automatically, then surely the creative action is finding and picking a number which is actual interesting out of the infinite sea of random garbage.
My point is that framing music as "just numbers" does not disprove that producing a song is a unique creative work. There may be valid arguments against copyright, but this one isn't.
Your example is barely plausible, much less demonstrable.
The connection between "original works" using an extremely limited set of notes, and your right to privacy using some theoretical predictive algorithm is not at all obvious.
In some sense, looking at information as being exclusionary and as being inclusive are the same thing, but there's a lot of ways in which the former actually makes more sense as a thought framework.
And in this particular context we can see how that plays out... a list of all possible melodies of a given nature actually has very little information in it, because it doesn't exclude enough. It may superficially seem to our human senses that a lot of stuff has been included/constructed, but in reality, the 'list of every possible melody' is a vapor. There's not actually anything there. It is the act of exclusion of possibilities that leads to interesting information. Such information as this list has is contained in its specification of what a "melody" is. Counterintuitively (to a lot of people's understanding), if they widened the specifications, while they would end up with a bigger list they'd end up with less information in the result.
The act of creating a song isn't a matter of creating the possibilities from the raw nothingness, it's a matter of carving them out of the exponentially-large space of possibilities and finding something there useful. The exponentially-large space is so large that it is very easy to not see it that way, because, I mean, it's huge. It doesn't feel like "removing" possibilities the way carving a 3D stone does ("I remove everything that doesn't look like my desired statue"), because the exponential space is so inexpressibly larger, and we need fundamentally different tools to address such a space, but in the end, it's the same thing.
While this isn't what the law was written for necessarily, the "creativity" requirement here could be very easily pressed into service here. They've expressed very little creativity/exclusion on this list and it would be easy to argue it falls far below the threshold necessary for copyright. As a literary criticism of the system, it is successful and thought provoking... as a legal criticism of the system it would fail completely.
The legal criticism would be that there just aren't that many unique melodies—as demonstrated by the fact that they were able to enumerate them all—so the mere fact that two songs use the same melody is not sufficient to show that one is a copy of the other. The set of melodies that are compatible with human ascetics is even smaller. They don't actually need these auto-generated melodies to qualify for copyright for the project to succeed. It works equally well if similarity in melody is not considered sufficient evidence of copyright infringement.
Even just having the database around so that one can say that they copied the melody from here rather than from some other source might be enough. After all, unlike patents, independently producing something similar to a copyrighted work is not infringement; you have to have actually copied from the other work. If you're a musician perhaps you should listen to a few randomly-selected melodies from this program each day. Maybe it will spark something, but even if it doesn't it will at least make it harder to argue that whatever melody you come up with could only have been "subconsciously copied" from some other composer's song you may have heard decades ago.
How does that follow? You can enumerate any finite number. And the article doesn't say how big. Is it a thousand or a trillion? "Riehl says the algorithm works at a rate of 300,000 melodies per second.". The article doesn't say how many seconds it took to generate all melodies though.
Not within a fixed time period in the real world. You're limited by the matter and energy available, and by the speed of light. However we're not talking about the theoretical ideal limits of computation. The upper bound would be 300k melodies for each second since the program was written—68.7 billion in all, according to the Adam Neely interview linked from the Press page of the project site. Which is a lot, but then there are hundreds of millions of known songs, each of which is likely to contain multiple melodies, some of which are much more likely to be chosen than others. Accidental duplication is thus quite likely.
Interestingly, there are writers who want to argue originality based on the arrangement of letters in 26-letter alphabet!
The world is indeed a strange place for the dogmatically logical programmer.
But there are degrees of absurdity, it's one thing to do that when there's 26^100000 possible combinations, it's another when there's just 12^100 (and if you only care about melody it's overestimation, most songs will use much smaller subset of that).
At the risk of getting stuck in a loop: "No rhythms, no meter, no tempo, melodies are longer than 12 notes, it's diatonic, single octave, no concept of underlying harmony"
> and if you only care about melody it's overestimation
And they did it that way because courts don't worry about the exact tempo, meter and rhythm when ruling on plagiarism.
There was a guy trying to copyright A CHORD :)
That much is true.
> a guy trying to copyright A CHORD
IIRC he failed miserably.
A better example would be the (until recently) coprighted song "Happy Birthday." [1]
[1] https://www.nbcnews.com/business/business-news/happy-birthda...
... but that doesn't mean copyright and patent aren't a perpetual battle against the "natural" arrangement of idea, and absurdities are extremely possible when the law is misinterpreted or mis-structured.
Who "generally agree"s on this? The existence of laws doesn't indicate the mood of society.
But if laws are proliferating and regularizing instead of standing still or being abolished, and one assumes that elected representatives are acting on the will of the people, it probably does.
There's lots of controversy over how to improve copyright / patent law, but not very many people in governments in the EU, US, China, Japan, Australia, &c are talking seriously about just burning the whole copyright / patent system to the ground. At least a subset of the countries in the groups listed are generally understood to have representative governments.
Yeah, that's the assumption I take issue with.
Patents are of limited duration because the tradeoffs of unlimited patents are so horrific. If we accept a billion dollar drug must enter the public domain, clearly copyright should also be limited just as it was proposed in the US constitution. However, because a tiny minority has a huge benefit and society does not really notice the difference you get the modern mess of unending copyright.
If tiny minority + 'society' is the entirety of the system, then that's true, but there are also plenty of players who lose out due to restrictive IP regimes—and it's hard to quantify the extent of those losses. (Whether or not the benefits they would reap from looser IP are appropriate or fair is beside the question for utilitarian computations.)
You haven't diminished the ability for 350 million people to do things, practically, by shaving 1 cent off of them. But adding the ability a million dollars provides one individual to do something cool with 1 million they couldn't do before has increased the overall capabilities of everyone.
In essence, you've just described Kickstarter's business model.
Basically, even if individual bacteria are unnoticed enough of them can kill you.
What would be the point of a patent system with unlimited duration? If we wanted that, we could just have companies not reveal their inventions in the first place
I think IP is required to monitize valuable ideaS, and monitzation leads to social availability.
On the personal level, I would not like it if publishers could freely print the works of new authors, or engineering solutions I spend years on could be copy and pasted.
Some people prefer Creative Commons, and they are free to publish their work that way. Others need or want financial compensation.
I certainly didn't mean to say that no-one agrees with it, but your personal agreement doesn't evince the general agreement that shadowgovt (an interesting username, in this context …) suggested. To be fair, neither does my skepticism provide any evidence against it.
In terms of public opinion, it would be interesting to know what studies have been done. I imagine if you would get broad support for an author copyrighting a book, and less on patenting a pre-existing genetic sequence.
As another unsubstantiated claim, I think if you sat down with the general public and the criteria for patents, they would mostly agree.
The challenge has to do with implementing them and the legal process around them.
If a crappy patent is issued to large corporation, it is incredibly expensive to challenge them.
Basically, America's crazy overreach forces our laws onto other nations - this is actually one thing that really frustrates me about corporate tax loopholes, that overreach could be trivially used to force better international standards for corporate VAT taxes there just isn't the political will (due to lobbying) to get it done.
https://www.aclu.org/blog/free-speech/court-tells-georgia-it...
Some might say lumping trademark, copyright, patent and trade secret laws (historically and in practice very different things) under one heading called "intellectual property" is an intentional strategy to muddy the waters and cloud any argument.
In the latter case, there's also no real risk of accidentally stomping on that.
The claim in this particular case is that they really have generated the entire possible melody space. Legally I think it's likely to fail on multiple levels if it is ever challenged, but part of the point is that some of those failures should also be applied to some real copyright suits that have been won.
(It is somewhat ironic that the music industry continues to be so upset about copyright even as they appear to be converging on The One True Pop Song at speed. Maybe if they acted less like some sort of bizarrely over-trained AI and cranked up the exploration constant, they'd stomp on each other less.)
Not even a thought experiment: it’s essentially a 4’33”-long ambient acoustic sample. There are plenty of these (though not usually that long) in sample libraries, recording e.g. traffic sounds, or diner conversation, or crickets in a marsh in summer, etc. And those are certainly copyrighted, unable to be used without license.
I would question any legal professional's authoritative standing to even advise on copyright of a work of music if they miscategorize a recording of ambient sound as a performance of a musical scoring consisting entirely of silence. The copyright doesn't apply to the ambient sound, but to the long quiessence of an artist at their instrument.
It demonstrates a complete blindness of the negative space of music, and a positivistic bias that has no place being enshrined in our legal system.
It's not just the music industry. The whole economy starts to feel like overfitting the profit function.
That's an incorrect way to view it actually. 4'33" copyrightable essence is actually represented by the active production of its scoring. I.e. nothing. The background sound is not what makes it copyrightable. You can go ahead and sit at a piano for the length of the composition all you want, wherever you want, and you'll still be publically performing 4'33".
The rather humorous outcome, if one asks me, is that anyone who writes in 4 beats of silemce into a score should be violating copyright if we're going to be consistent.
That kind of thing is completely unenforceable with respect to performances, but in written musical notation, copying the specific notation pattern could be infringement. If you write "4/4, tempo 80, 91-measure rest", that's maybe violating the 4'33" copyright. If you write a score for a full band or orchestra that shows rests in each measure for each instrument, with key changes and tempo changes and such, you're just retelling the same joke in a different way.
Usually short phrases aren’t supposed to be protectable under copyright. However, when a defendant blatantly appropriates a well-known literary phrase for a commercial purpose like selling unlicensed merchandise, courts may make an exception.
[there actually are musicians that specialise in recording backing tracks intended to resemble a particular popular recording which aren't that recording for use in commercial products, but they tend not to get sued...]
Related: The Supreme Court ruled that excerpting a single paragraph from a 454-page book can be copyright infringement. The book was Gerald Ford's memoirs, the one paragraph was his reasoning for pardoning Nixon. The Court's reasoning was more-or-less that nobody cared about anything else Ford did, so excerpting the one paragraph was as good as giving the whole book away for free.
Whether a work in question is even eligible for protection at all is a distinct legal question.
The question in play is how much of of the copyrighted work may be reproduced before it is infringement.
Of course not every set of four notes or eleven letters in a copyrighted work is prohibited; the significance and context of the use matters.
:)
However, if you interpret the program itself as a self-extracting compressed archive, they actually have stored every possible melody (in a compressed way).
So the question reduces to how much the type of compression matters here (is ZIP allowed? is TAR allowed? what about more sophisticated like PAQ? and what about this Rust code?). This is what I/we discussed here: https://news.ycombinator.com/item?id=22441328
It doesn't matter at all. As 'patio11 correctly points elsewhere in the thread[0], this question boils down to the colour of the bits.
--
A melody contains more than a sequence of tones. The most heartless definition would at the very least include rhythm. For every sequence of tones they output, they only produce one out of hundreds of possible melodies for that sequence.
It is of course an interesting thought to consider the definition of decompression, but on the other hand, we should also limit the contributed idiocy to the bare minimum required to break the relevant idiotic rules.
> So the question reduces to how much the type of compression matters here
If you compress, you can copyright the compressed bytes.
If you don't compress, you can copyright the uncompressed bytes.
As far as that copyright extending to derivations, e.g. decompressions, the answer indeed situation-dependent. For example, converting a copyrighted font from TTF to WOFF does not remove the copyright. But converting a copyrighted font from TTF to screen pixels to WOFF removes the copyright. (Sorry I don't have a reference; probably findable.)
The "self-extracting zip" derivation would probably fall into the latter category; that is, the copyright would not transfer.
---
But even if the copyright were maintained during your advanced decompression, one could argue that editing down to very specific portion of that extremely large body of work was a substantive/transformational derivation, which they could then copyright themselves.
Transformative works are very common in art. The most famous example is Duchamp simply adding a mustache to a print of da Vinci's Mona Lisa, and copyrighting that. [1]
I hope nobody takes this as legal advice; I'm not a lawyer but I'm fairly certain it's wrong. The font would still be the same work.
It comes down to this: Typefaces/glyphs are not copyrightable. The font code that produces those glyphs is.
> Typefaces cannot be protected by copyright in the United States (Code of Federal Regulations, Ch 37, Sec. 202.1(e); Eltra Corp. vs. Ringer)...However, there is a distinction between a font and a typeface. The machine code used to display a stylized typeface (called a font) is protectable as copyright. [1]
In software, a similar "black-box" derivation process has happened many times, e.g. UNIX/GNU. Copyrights applies to software source code, but not software functionality.
Determining what is the "essential, creative work" in each case in a nuanced way is a matter for courts and armies of lawyers: Apple round corners, Oracle Java APIs, etc.
[1] https://en.wikipedia.org/wiki/Intellectual_property_protecti...
(But yes, I do agree that typefaces are exceptionally commonly distributed.)
However the example here, and the situation with fonts seem different. It is the case that font data is viewed as utilitarian and uncopyrightable. So we have a copyrightable program, producing uncopyrightable data. The argument here seems that we have a copyrightable program producing copyrightable data.
And apparently arpeggios, because the US District Court of California ruled in Flame vs Katy Perry that arpeggios are "melodic enough" for copyright protection.
A combination of chords, melody, rhythm are I think the only reasonable measure that a song has been copied.
If I take the entire melody of a Beatles song, including the verse and chorus, but set it to an entirely different chord progression, would the court recognize that as an original song? What if I lifted all of the lyrics as well?
I am not a musician, but which of these are copied in the Tom Petty / Sam Smith case that motivated this exercise? To my untrained ear, I do hear some similarities in the relative lengths of the notes (meter?).
I found a side-by-side comparison https://www.youtube.com/watch?v=YflFw9T77FQ
If these two songs are similar enough, then I think it could be argued that a MIDI sequence has been copied, since in both cases it requires a significant change of tempo and key. A lot of commenters seem to be missing this point: yes, the generated sequences sound different from real songs, but so do the songs involved in the ridiculous court cases. Radiohead and Ed Sheeran were sued for chord progressions, Katy Perry for a melody. The songs involved were altered about as much as the MIDI sequences would need to be to show the similarities.
One of the items they were trying to point out, often abused in lawsuits for pop music, is the idea of "Access". If you came up with an idea all by yourself, but a similar song exists that is popular enough, the court argues that just by there being the possibility that you heard it, you therefore definitely heard it and then copied it.
If this music set exists, and is freely available, shouldn't it be considered that you had reasonable access to it and therefore stole it? No, of course not, that would be a ridiculous assumption and so is the current outlook of a song being popular being enough proof that you stole the idea.
Not to mention, music is extremely formulaic. Chord progressions have a natural tendency to certain forms, with centuries of prior art, rhythm within genres of music is often the same, even melodies have a trend toward particular combinations (leading tones over chord progressions bring about lots of similar sounding solos).
Any musician trying to claim copyright for their music should remember that their song only exists on the back of centuries of musical exploration. Consider how much of the song you can say is truly novel, it's going to be nearly nothing.
The combination of lyrics + chords + melody is in my opinion, the absolute minimum you need to claim a song has been copied. Lyrics are derivative, melodies are derivative, chord progressions are derivative, but together they have the chance to be a unique combination.
Sufficiently advanced “proving a point via absurdity to make a more general argument” is indistinguishable from satire.
Do you have a book that you would recommend to learn about these terms?
Melody is the horizontal arrangement of notes for an individual voice or instrument over time. Harmony is the vertical arrangement of notes sounding at the same time, and how those transform horizontally over time as a group. Tempo is the speed in beats per minute of the background 'pulse' of the music.
Now, asking for a music theorist to give you an algorithmic definition of how to make music with any of the above? Good luck ;)
I honestly know nothing about music so this is all gibberish to me. But I'd be interested to learn if anyone has a book recommendation.
"Learn music theory in half an hour" is obviously an exaggeration, but it really comes astoundingly close to fulfilling that promise. It contains a lot of information and each part builds on the previous parts, so it requires focus and maybe a few repetitions to 'get it', but I think the approach is fantastic for showing how many ideas of music theory are deeply connected.
Not a book as you requested, but hopefully you'll find the other reply useful. As for some of your specific questions:
Frequency is the same concept as radio frequency, but in this case refers to something we can directly sense. Radios transmit electromagnetic waves, which are photons moving at a certain rate, measured in Hertz, or cycles per second. Sound frequency refers to movement of air waves, so a more accurate analogy than radio waves is waves in a pond when a rock is thrown in. Human ears are sensitive to frequencies between 20 Hz and 20,000 Hz, so any sound you hear is a combination of frequencies. Natural language is helpful here, since higher frequencies sound 'higher' and lower frequencies sound 'lower'.
A tone is a sound at a specific frequency, also known as a note. For example, 440 Hz is designated as the note A4 by the Geneva conventions, and this is what most instrument tunings are based off.
Twelve tone equal temperament is the tuning system nearly all modern Western music uses. Certain ratios of frequencies sound pleasant, especially ratios with small numbers, such as 1:2, 2:3, and 3:4. So if we know 440 Hz is a note in the system, it would be nice to also have 587 Hz, 660 Hz, and 880 Hz. However, these frequencies will only really sound good when played with that original 440 Hz, not necessarily with each other. So instead of using them exactly, we approximate them in a useful way. The 1:2 ratio, the octave, is generally considered to be the most important, so that ratio is kept, but otherwise the notes are equally spaced (human hearing is logarithmic), or equally tempered. The most popular tuning system has twelve tones. There's no note at 660 Hz, but there's one at 659 Hz, which is pretty close, and there happens to be one at 587 Hz. Other ratios are also represented reasonably well.
A key is a collection of notes that sound good together, based on the ratios of their frequencies. The alternative would be chromatic composition, where all 12 notes are used and none is obviously 'more important'. Most music is in a specific key, but uses some chromatic notes to make the melody more interesting.
I actually had to look up what the difference between a key and a scale is, as I thought the terms were pretty much interchangeable. I've edited the last part of my other comment to reflect this:
A scale is actually an ordered set of notes belonging to a key. A key is just an unordered collection of notes. I got this wrong earlier.
So playing all the notes belonging to C major in ascending order is playing a scale, and playing the notes in any order is playing in the key of C.
Isn't it a bit of wasted space to store the generated data on archive.org (https://archive.org/download/allthemusicllc-datasets)? I mean, as we see, it's trivial to create them. The code above is kind of a self-extracting archive, so it's just the same thing but in compressed form. And I guess it should not matter (legally) whether you store it compressed or uncompressed (or less compressed).
If you only provide the code as a means of generating any melody, you could just as well replace it with any music instrument and claim you can generate any melody with it.
Apparently, for this to work, they have to "affix them to a physical medium".
Strongly recommend watching that video! Excellent info.
So, by the same argument, they could copy the self-extracting archive to a hard drive, and then have the same thing, or not?
They even say that they already store it in a compressed form on the hard drive.
What if some future compressor (future ZIP format) is clever enough to see that you are going to save all possible permutations of something, and then saves this is a much better compressed way. Then suddenly when compressors do this, it means when you use such a compressor, you do not have copyright on the data anymore?
I think the focus has to be on the spirit not the technology. Saying to a lawyer/judge/jurist "hey, this script can generate every possible melody!" _feels_ quite different from saying "every single melody is already written down, on this hard-drive I am waving at you! Look!"
Tangent: reminds me of the Hutter Prize for compression (hey, they prize pot recently increased to €500K and I submitted it to HN but it didn't get any votes http://prize.hutter1.net/ https://en.wikipedia.org/wiki/Hutter_Prize).
This has never been proven for the digits of π. Of course, there are other digit-sequences for which it is trivially true. For example, just concatenating together every finite sequence of digits, in some suitable order.
“But are the tunes really affixed to the medium?”
“Yes, this hard drive in my hand literally contains a bunch of MP3 files” is a lot stronger than: “Yes, in a way, because this tune-generating script technically constitutes a self-extracting archive, your honour.”
https://www.tunecore.com/guides/copyrights-101 the first paragraph is:
> For a work to be “copyrightable,” it must be original and fixed in tangible form, such as a sound recording recorded (affixed to) on a CD or a literary work printed (affixed to) on paper.
You may say that in fact having the code that generates the data is effectively the same as having the data. This may convince reasonable, rational people, but the fact that we're dealing with the law that we are should tell you that we're not dealing with reasonable, rational people.
Not to get too far off topic, but even an infinitely long byte sequence cannot represent all numbers. It's possible to construct something not representable by that sequence via Cantor's diagonal argument: https://en.wikipedia.org/wiki/Cantor%27s_diagonal_argument
All the possible programs can receive input of infinite length and output bytes of infinite length depending on input.
So if we can supply any input to any program and treat its output as part of result, does this argument still work? Is there such a number that can't be constructed by some program receiving some input?
Correct.
> Is there such a number that can't be constructed by some program receiving some input?
Yes. Most numbers are not representable by a program...unless you allowed a program to be infinite length, in which case the first statement is no longer true.
There's nothing particularly unique about a digit-based encoding for Cantor's diagonal argument.
Why is that? What would be an example of an infinite program that could not be represented by an infinite byte sequence? Indeed, it seems trivial to map an infinite series of machine code instructions to an infinite series of bytes.
Edit: It seems like the first statement is false only if you use a different understanding of "possible" for "possible byte sequences" and "possible programs" where the former excludes infinite length and the latter does not.
Because you cannot enumerate all infinite byte sequences, according to Cantor's diagonal argument as linked earlier. [1]
Calling certain numerical representations "computer programs" in no way changes the fundamentals. There is no way -- no matter how clever your encoding -- to enumerate all real numbers.
[1] https://en.wikipedia.org/wiki/Cantor%27s_diagonal_argument
You don't have to enumerate all real numbers to create a mapping from real numbers to infinite byte sequences.
I am merely pointing out that the first statement only becomes false if you editorialize it to "all the possible [finite] byte sequences contain all the [infinite] possible programs". If you treat the meaning of "possible" consistently in that sentence, I believe it remains true regardless of whether you define infinity as possible or not.
The implication is that human minds wouldn't be able to represent it either, at least not with language.
The standard diagonalization argument is a constructive proof, identifying a specific number. Normally a constructive proof is preferable. But with a bit of hand waving you can make it into a non-constructive proof that there are "unknowable" numbers that cannot be described in any finite amount of bytes.
The point is that such a sequence needs to be constructive in the usual sense, so there are still only a countable number of them.
But it couldn't contain the real number representation.
Most real numbers have infinite-length constructive proofs.
Uncountable is uncountable is uncountable. If any correspondence/encoding existed between naturals and reals, the reals would be countable. (But they are not, so it is fool's errand to search for such an encoding.)
Okay, all natural numbers then. I don't think that changes the argument.
(Reals are a funny thing because any format you choose has numbers that are infinitely long in that format. Which is basically the same thing you said.)
The diagonal argument seems to be a proof that there are uncountably many infinite byte sequences. So while it proves that it would impossible to "enumerate" every infinite byte sequence, it doesn't prove that there exists a number that cannot be represented by some infinite byte sequence.
Indeed, I believe the opposite can be shown to be true by constructing a mapping from every finite number to an infinite byte sequence. ASCII trivially provides such a mapping for real numbers, and finite-tuples of real numbers (such as imaginary numbers) can be mapped by alternating digits from each element of the tuple.
Edit: The key distinction is that the set of finite byte sequences is infinite, but countable, while the set of infinite byte sequences is not only infinite, but also uncountable. Which of these two sets is deemed the set of "possible byte sequences" seems to be the critical distinction and the turn of phrase "all possible byte sequences in the universe" seems to imply all byte sequences of finite length.
You misread. The previous comment said
> you can trivially write a program which will enumerate all the possible [finite] byte sequences in the universe
No one is interested in your infinitely long musical score.
Could the algorithm be considered a form of compression?
An algorithm that generates all possible English sentences doesn't demonstrate that I copied this sentence from it, and the balance of probabilities (used in tort cases) suggests that I came up with the sentence, run-on as it is, rather than performing that algo and then selecting that sentence. Moreover 17USC (102? sorry I don't recall) says that media needs to be "fixed" to acquire copyright; so me having to perform the algo makes it no impediment to my owning copyright of a sentence that, if performed, the algo would by expected produce (which is sensible really, perhaps the algo is errant and can't make this sentence, maybe it only forms grammatical sentences ...
If someone is using archive.org for this purpose, please consider donating :-)
I guess for matters of copyright you have to produce a specific artifact, just blindly enumerating all possible artifacts is not good enough. Which would also mean that the existence of all those MIDI files will not make a difference.
If I were to write a "random book generator" that just regurgitate infinite text then I can copyright a pair starting index and length, but not simply all possible finite sequences of letters.
I see it as likely that someone will eventually try it not just as a legal strategy but because of a genuine curiosity in music and AI.
According to wikipedia
> There is a long tradition in classical music of writing music in sets of pieces that cover all the major and minor keys of the chromatic scale. These sets typically consist of 24 pieces, one for each of the major and minor keys (sets that comprise all the enharmonic variants include 30 pieces).
I don't think anyone would say there is any ill intent even though it is based on an enumeration.
But the future case I'm speaking of will be a very real dilemma and either decision will have drawbacks. It's not just about seeing through a hack.
That's bread and butter for the courts. You don't need judges and juries and lawyers to see through a simple hack; trying to use a trivial enumeration to defeat a copyright claim is just "ha ha, nice try, but nope" issue no reasonable layman would have a problem with. Shades of colour are of critical importance in precisely those cases that are fuzzy, where there is no obvious ruling to be made. That's what the courts are made for.
What is the intent if a human provided 1 bit of input by flipping a coin and chose all the odd numbered melodies? What if a human provided 1 bit of input by flipping an unobservable mental coin? What about 2 bits? N bits?
What if a machine generated all the "interesting" melodies via a neural network?
The point is in a world of cheap computing, intellectual property is probably moot if its cost of production is nearly zero.
IP is not moot because you have a random number generator. IP may become less relevant if computers ever learn to select useful bit patterns out of sea of randomness, but even then, the legal system will handle it just fine (somebody will own the selection algorithms after all).
> What is the basis by which jury likely decides intent here? Is it any different than "arbitrarily"?
Causality. You can't determine intent from bits, which leads people to (mistakenly) believe colour doesn't exist. To determine the colour of the bits, the courts will look at the actual chain of events surrounding their creation. Because that is what matters. Not what the bits are, but how you came into possession of them, and why.
What is the intent? You tell us. Was the person doing it to try and capitalize on someone else's creative works?
There's also a difference between copyright and patent law.
Back in the Napster days, there was a company (maybe mp3.com or something? Google tells me yes!) that would let you get high-quality encoded mp3s of your CDs by downloading their program, putting your CD into the hard drive, and it would read enough of the CD to verify that that was the CD you had, then give you access to their own encoded version of it. The idea being, you own the CD; you have a right in the US to rip it & encode it; so therefore you have a right to this mp3.
They were sued by the record labels (naturally), and lost in court; because the court made a distinction between the copy that mp3.com had of the CD and the copy that an individual could have made. mp3.com were violating copyright because they were distributing their own copy, even though the bits were identical to the copy their customers already owned.
So along the same lines: "I have this bit pattern because I copied it from mp3.com's copy of a CD" is different, legally, than "I have this bit pattern because I copied it from my own CD".
https://en.wikipedia.org/wiki/UMG_Recordings,_Inc._v._MP3.co....
I would think that after this project is complete, at the very least arguments like those that lost the "Dark Horse" case should become not-acceptable in courts. Let Marcus Gray actually prove provenance, not infer/ speculate it - and all would be fine.
"Colour" is a term that encompasses provenance, chains of cause-effect, intent and related concepts. Bits are obviously colorless from compsci point of view, but are colorful from the legal POV. For instance, bits encoding a copyrighted video are just numbers, but for the judge it'll matter whether I got them from a lucky PRNG run or from a torrent site.
Not understanding that colour is real in legal systems is the core source of confusion people have wrt. intellectual property and illegal information (e.g. child porn).
Why in some cases colour matter and in some case, it does not ? no judge will care about the "colour" of child porn. Yet, they will care about the "colour" of a song.
(To clarify: Colour does not encode copyright information; it encodes provenance, intent and related ideas. Child porn is only a somewhat special case in the sense that laws are very insensitive about the shade of colour here. Usually, if it looks enough like CP, a possessor will end up in trouble, regardless of whether it's actual CP or whether they can prove they created it ex nihilo in MS Paint. But that's because child abuse is considered by societies worldwide a special kind of evil that needs extraordinary measures to combat. In almost all other cases, courts are very sensitive to the shades of colour.)
The original article is something about music. Then you say something about colors. And your article starts with some description of some game called Paranoia. Without already knowing the context this looks completely random.
Long ago, I enabled showdead to read Terry's comments. They didn't always make sense, but occasionally there'd be one that was worth it.
- a copyrightable work must include some human creativity. It seems to me that an enumeration of possibilities might be creative, but there is no way an individual element of that enumeration can be considered creative.
- Copyright depends on copying. If you release a song with a catchy melody stolen from another song, then you infringed its copyright, regardless of if there is an licensable version of that melody that you could have copied. What matters is which one you copied, not the existence of alternate copies.
Not a lawyer either, but "I came up with this myself" is the standard defense of the artists, with mixed success.
The fix is political. Call for copyright reforms.
Copyright claims have been won by arguing the song was so widely available that it’s unreasonable to assume the defendant did not have access to it, which is why they’re trying to spread access to this as far as possible
Additionally there's the problem of releasing the works to the public domain. As discussed a couple of weeks ago when the 2016 billion-dollar infringement lawsuit against Getty Images was attempted ... and thrown out of court ... the act of putting works in the public domain also extinguishes, in the court's eyes, the rights of the author to sue for any claims including moral claims of authorship.
See: https://news.ycombinator.com/item?id=22340325
So: clever stunt, but legally impotent, both by law and the self-neutering actions of the actors here.
Adam Neely's video (interviewing Rhiel and collaborator Noah Rubin) covers many points of copyright, though not the originality, authorship, or PD angles. (Posted elsewhere in thread, echoing here.)
Either the courts decide they are not, in which case the creators will be happy.
Or they are copyrightable and therefore this collection, even if it is in the public domain, is a prior art that invalidates any claim that a newly used melody is really new/copyrightable.
Free software achieved its attack on proprietary software not by rejecting copyright claims but by turning them on themselves.
It's a bit like the Tesla Valve, a one-way passive check valve with no moving parts:
What if the author was thinking about generating a specific, original melody when he hit "run"? He just wanted that melody and didn't care about the rest, but he was lazy and it was easier to find his melody among the rest of generated melodies instead of modifying the source code.
Can you deny that he created that melody?
This is a response to the Flame v Katy Perry lawsuit, where a youtube video with 100,000 views was deemed sufficient popularity for the court to assume copying without proof.
The element that Katy Perry was found to have illegally copied is a 4-note descending line with equal spacing between the notes. In both "Joyful Noise" and "Dark Horse" the pattern was probably generated by a producer pressing the "arp" button on a minor chord in a DAW. Neither of them actually plotted out each of the notes with particular purpose. But the Court still said that the 4-note descending pattern is original and creative enough, the fact that it was composed by software notwithstanding, and illegal to reproduce.
> If you release a song with a catchy melody stolen from another song...
Copied. Not stolen. Copied.
Of course I know what you meant, but it's sad that this phrasing has entered the language. In all the tensions around limits of IP protection, this one was probably the most effective trick media companies pulled. Equating copyright violation with theft (and, through implication, their moral weights), even though one has nothing to do with the other.
> His suitcase is full of noise, but what's coming out of the stereo is ragtime. Subtract entropy from a data stream – coincidentally uncompressing it – and what's left is information. With a capacity of about a trillion terabytes, the suitcase's holographic storage reservoir has enough capacity to hold every music, film, and video production of the twentieth century with room to spare. This is all stuff that is effectively out of copyright control, work-for-hire owned by bankrupt companies, released before the CCAA could make their media clampdown stick. Manfred is streaming the music through Annette's stereo – but keeping the noise it was convoluted with. High-grade entropy is valuable, too ...
https://www.antipope.org/charlie/blog-static/fiction/acceler...
We can think of selecting a good melody to fit into a song as one example of a design problem -- searching through some finite enumerable space of melodies and selecting a melody that's a good fit. A lot of the effort in this process is testing and evaluating the melody to see if it is any good to listen to / any good in context. This effort hasn't been done if you simply brute force enumerate and list the search space without testing anything.
There are many kinds of design problems where the effort of checking if a proposed solution is any good vastly outweighs the effort of suggesting a solution.
We could do the same thing for other kinds of design problems.
In principle, what if we enumerate some (vast) finite space of digital circuits -- if we put a few limits on the amount of stuff we can put into a single circuit, and discretise any continuous parameters to produce a finite space. Does that mean we can copyright all these possible circuits, even if we put no effort into seeing if any of them are fit for any purpose?
Can we enumerate some large finite space of possible arrangements of atoms into molecules and then copyright them all, without putting any effort into analysis or experimental testing to see if any of the proposed molecules are fit for any purpose?
Snisarenko made a similar criticism when this was posted 15 days ago:
> As other commenters have pointed out, this is gimmicky, shallow and clickbaity. All that they did was count from 1 to 68 billion. Any piece of digital data can be converted to a number, and if we apply their argument then you can be "creative" by just counting numbers. But we all know that's not true. Once the search space becomes that big, you can't actually "enjoy" any of these melodies. Because you can't listen to all of them in your lifetime, the ones that you WILL hear are going to be awful 99.999 percent of the time. Hence, the "creative" process is navigating this search space, and figuring out which melodies are catchy. Better yet, trying to figure out how or why our brain decides to like or not like a melody.
-- https://news.ycombinator.com/item?id=22306857
From my perspective of framing this situation: this is dumb, as no design of any melodies has been performed. The design is the uniform prior distribution over some space of all possible melodies, most of which will be uninteresting / not fit for any particular purpose, which is a pretty uninspiring notion of design.
Pharma corps do this (with patents), ostensibly, but IP rights have a reasonable cost to them so they pre-filter.
https://en.wikipedia.org/wiki/The_Library_of_Babel
*Edit: book -> short story
I showed this article to a musicologist (in the UK) and he said:
> I assume that somehow lawyers can refer to a precedent of "this is too ridiculous to be taken seriously"
[0] https://www.amazon.com/Larry-Polansky-Worlds-Longest-Melody/...
[1] https://universes.art/en/specials/john-cage-organ-project-ha...
OK, forget everything I said but the 8 notes thing. Just, no. "Yesterday"? Great melody. whistle just the first eight notes. "Yesterday, love was such an eas". Hmm. Wonder if there are more examples.
1. CNN's are used fairly commonly for sequence tasks nowadays. Convolutions can be 1D after all. 2. It's also possible the previous comment was referring to using 2D convolutions on the spectrogram of the audio, which is a common approach. 3. Neural networks are capable of more than classification. Scoring is a regression task which is common application of neural networks.
do people really run 2d convolutions on spectrograms? that seems rather backwards to me - why convolve in frequency space when you can just multiply in the time domain.
re regression tasks: sure I guess that's just an embedding basically right?
Also, it’s computationally intractable to individually train 2^N weights. What a CNN does instead is train a convolution kernel which is passed over the whole domain to produce the input for the next layer; by operating in frequency space, it’s considering the basis functions e^{j omega +- epsilon} instead of delta(x +- epsilon)
>Also, it’s computationally intractable to individually train 2^N weights.
that's a good point - i'd forgotten for a moment (because i'm so used to cooley-tukey fft) that in principle getting the spectrum involves a matmul against the entire vector. which brings up a potentially interested question: can you get a DNN to simulate the cooley-tukey fft (stride permutations and all).
(Not sure if that matters, I don't know much about NNs, but seemed like an important distinction.)
2. Since data is MIDI-encoded, would a convolution hold any merit here? I suppose you could render to an mp3 and analyze the audio itself but that seems very computationally expensive and prone to overfitting. 3) If we're training a scoring classifier, we would need labeled data, but getting those labels seems very challenging, not least because of how subjective our impressions of melodies can be (for instance, the opinions of a fan of atonality would be drastically different from a fan of pop). Do you have any ideas on how to mitigate this?
Presumably it prevents lawsuits because I can write a song starting with a melody from this archive knowing that it's in the public domain. (Or because it makes obvious the absurdity of having a copyright on a 12-note melody in the first place.)
Copyright law protects expression, and the labour required to produce that expression. Just like people who copy others’ work to gain from it, the “authors” have attempted to gain from doing no work and letting a computer give them all the world’s music. Imagine if they didn’t release their library into the public domain, and instead started to immediately sue over every new song that came out? It is fortunate that they cannot do either.
A significant portion of melodies likely to be generated by people, or put another way these are the more trivial melodies. The range of complexity is vast, so just like cracking passwords might use a dictionary attack, or use English alphabet, or similar limited scopes, so it goes here.....
It doesn't matter if there isn't enough space to enumerate them all. It doesn't matter if most of them sound like crap. It doesn't matter if some match a pre-existing song. It doesn't matter if you do or do not highlight "interesting" songs.
The very fact that they are machine generated eliminates the element of creativity that is a central requirement to copyright.
Copyright law may be stupid, but it isn't this stupid. An even more stupid argument that isn't even valid doesn't really help.
> 503.03(a) Works-not originated by a human author.
> In order to be entitled to copyright registration, a work must be the product of human authorship. Works produced by mechanical processes or random selection without any contribution by a human author are not registrable.
next they need to generate every possible chord progression underneath that melody.
A aardvark awkardly announces
A aardvark awkardly apprehends
...
If it did, an actor with more resources could feasibly blanket own any cultural good of value and automatically, quantifiably argue how close any arbitrary third party creation is to their copyrighted material.
For example:
I am typing these words on a mobile phone whilst sat on a train to an airport, wearing a green jacket, brown shoes and a maroon hat with a black bag containing gifts for my children. That seems a fairly prosaic sentence at first glance but I doubt has been written before.
(edit: ninja'd)