How many years until we run out of original music?

[electromotive]

Never! Remember music before electric guitar? What about before the advent of "scratching records", hip hop and drum machines? Those are but a tiny example... we cannot imagine what is possible. Don't forget also the many permutations of keys and scales, rhythms, instrumentations, sound effects and word intimations and other vocal sounds that can be combined. I don't think we are anywhere close to being at the end of our creativity... although many songs today sound similar due to our limited cultural experiences and commercialism.

Exactly.

Take Christian "Flake" Lorenz from Rammstein for example... He plays keyboard, piano, organ, etc. but he's also credited with playing bass keyboard and bass guitar. When he was 16 or so, he was in a band with another fellow that later ended up in Rammstein as well. They wanted Christian to play bass guitar on occasion, but since he wasn't proficient at the instrument, he tuned his organ to produce the riffs they were looking for.

Now look at guys like Trent Reznor of Nine Inch Nails. One of his early projects, here in Ohio (Exotic Birds) was more or less absolute garbage, but it got him interested in tinkering more with a keyboard, and doing programming / synthesizing. He's now one of the most influential musicians on Earth, with NIN being a "one man band with hired help". He writes and composes basically everything himself, programs it, plays it all in the recording studio -- a virtuoso of sorts.

[T3sl4co1l]

Anyone who asks about "running out of" music, or written or spoken phrases, or computer programs, etc., need not worry.

It is physically possible to map out the simplest of all of these; most of them do nothing.  Gibberish and such.  A few extremely rare cases have relevance to us: meaning, feeling.

Anything with a length over a few hundred symbols will never be fully explored over the history of the universe.

Alternately, we can already say that all original media has been exhausted.  Consider:
Write a '1' and a '0' on two separate sheets of paper.  These will be our characters.  (This is arguably the language of computers as we know them today, a perfectly valid starting point.)
Create an interpretation of these characters, a language.  In that language, if you read '1', it translates to something; if you read '0', it translates to something else.  What that "something" is, is unique to the particular language used.
There are infinite languages, therefore there are infinite interpretations of '1', and of '0'.
We don't even need two characters, really; this can be done in unary, where reading the one character gives the presence of a statement, and not reading it gives the absence of a statement.

Or we might make a very simple language, and express all statements through reading a sequence of '1's and '0's.  We could even construct an infinite library organized by this system, mapping all original works, all translations, all covers and reinterpretations and critiques, and all gibberish, into a sequence of moves between buildings, floors, rooms, shelves and books.  We would have the Library of Babel.  I must warn you, though: it's a very boring place, where you'll wander your entire life having seen just a few complete words, never even a full sentence.

Such is the nature of large numbers, numbers of hundreds of digits.  These are numbers large enough to be dense in prime numbers, and yet difficult to factor (relevant to modern cryptography); large enough to encode meaningful phrases, or musical phrases; or computer programs, even (or especially!) recursive programs (a program that creates programs, whether as a quine or as a self-referential "strange loop", if you will).  We can encode mathematical proofs in these numbers, and we can equally well encode contradiction (Godel's Incompleteness Theorem).

Into the thousands, or millions of digits, we have quite complex numbers indeed, numbers that evolve -- genetic codes, codes that are also computer programs, which are running self-referential operations that will never resolve within our finite universe -- only propagate.

Tim

[rcarlton]

Never. People are infinitely creative.

[NiHaoMike]

Anyone who asks about "running out of" music, or written or spoken phrases, or computer programs, etc., need not worry.

It is physically possible to map out the simplest of all of these; most of them do nothing.  Gibberish and such.  A few extremely rare cases have relevance to us: meaning, feeling.

Anything with a length over a few hundred symbols will never be fully explored over the history of the universe.
...
Such is the nature of large numbers, numbers of hundreds of digits.  These are numbers large enough to be dense in prime numbers, and yet difficult to factor (relevant to modern cryptography); large enough to encode meaningful phrases, or musical phrases; or computer programs, even (or especially!) recursive programs (a program that creates programs, whether as a quine or as a self-referential "strange loop", if you will).  We can encode mathematical proofs in these numbers, and we can equally well encode contradiction (Godel's Incompleteness Theorem).

Into the thousands, or millions of digits, we have quite complex numbers indeed, numbers that evolve -- genetic codes, codes that are also computer programs, which are running self-referential operations that will never resolve within our finite universe -- only propagate.

It doesn't even take that much complexity to get to the point where mapping out all possible combinations starts becoming impractical. Take a look at this done with just some discrete logic on a breadboard:

[apis]

There's definitely a limited amount of combinations since music is limited in time.

Four minutes CD audio (44.1kHz 16bit) would be 4*60*44.1e3*16=169Mbit and 2^169896000 combinations. That's a lot but most are noise and many would sound similar.

OTOH with 8 notes and a diatonic scale (7 notes) you get only 7^8 = 5 764 801 combinations, that's not that many, most of which would again be noise. If only one in thousand sounds good, then there would only be roughly 6000 such 8 note sequences that we would call music. Of course, you could then combine those into longer sequences, and so on.

I think collisions are unavoidable at some level, but the possibility for variations is clearly very large.

[ebastler]

Never. People are infinitely creative.

Well, but people are also lazy, and creating something original requires effort. (As witnessed by your not-so-original statement, which is all over the internet. )

So yes, there will be more original music, no doubt. But in parallel, there will also be an ever-growing stream of re-hashes, produced either out of lazyness or out of the expectation of easy monetary wins via safe bets.

[apis]

“Only two things are infinite, the universe and human stupidity, and I’m not sure about the former.”

[Kjelt]

A 19th century composer already stated that every possible chord was played.
However that was after the "wohltemperiertes klavier"where we in the western world are stuck with, if you look at other cultures that do not have relative positioned tones, there are many more possibilities to explore.

[apis]

It will still be some limited number of combinations though. There are certainly not infinite possibilities. If we knew the true number of possible variations (of a certain length, that sound good to most people) I suspect most people would be surprised how low it was. (Though more than enough to keep us entertained.)

[tszaboo]

The thing is, it doesn't need to be original. We can have a lot of fun with re-imagining music, playing it in different styles. Or if a few accord has been played somewhere else, who cares. Also, performing it mixes it with your personality. So if 400 years from now people will perform the Rick Astley - Never Gonna Give You Up with the current style... So be it.

[Kjelt]

It will still be some limited number of combinations though.

Sure since the spectrum is limited to 22kHz there is always a limit to the number of possibilities.
Luckily 90% of the possible combinations are not sounding nice.

[Beamin]

Assume that all songs use a Western musical scale and that such a scale contains twelve distinct intervals. Assume that a judge (not a musician but a judge) will distinguish three distinct note durations (which roughly correspond to eighth, quarter, and half notes, or through a trivial change in time signature, to quarter, half, and whole notes, or to sixteenth, eighth, and quarter notes). Thus, there are 36 possible distance vectors from one note to the next, and 36(n - 1) melodies of n notes.

Now, 36 to the third power equals 46,656 distinct melodies. No other melodies are possible in the Western musical scale.

source

Vsauce reads my posts! I should ask more questions 

[Alex Eisenhut]

Ever since I got into atonal music and all subsequent forms of ever more esoteric music, I'm not worried.



Steve Reich is pretty accessible I think:



Then you can make very compelling music with just percussion



You have the truly modern



Or the avant garde from half a century ago

https://youtu.be/VtMnDo7NItU

Or something really stunning

https://youtu.be/bU1hyzl8UJ4

Or just masterful

[T3sl4co1l]

I have board routing to do tomorrow... I know what I'm going to be listening to

Tim

[apis]

You mean this?

[Awoke]

Never. Just the other day I heard about country rap... that's apparently a thing. Someone is probably coming up with physics blues right now.

[MT]

[edy]

Someone earlier in the thread mentioned we could generate every possible sound and show that it is not infinite (obviously) but finite. However, we can calculate for all practical purposes it would exceed the lifetime of the universe.

For example, let's approximate that 2 to power 10 (1024) is about 10 to power 3 (1000).  So 2^10 ~ 10^3 roughly. If you continue this, 2^20 ~ 10^6 and so on, 2^30 ~ 10^9, etc... So whatever the power of 2 is, the ratio of the exponents to get to power of 10 is 10/3=20/6,30/9, etc...

Ok, now assume a fairly crappy 22050 sample rate, 1 second worth of sound. We already figured that the number of possible combinations of sound would be 2^22050 (combinations of 0 and 1's in a 1 second sample). By way of power of 10, given the ratio of exponents if 10/3, 22050/x = 10/3 or 22050 *3/10 = 6615. So 2^22050 is roughly about 10^6615. Correct so far?

Now let us assume correctly so that the VAST MAJORITY of these combinations is nothing but noise to our ears. In fact.... Let us assume that the ratio of "good" sounds to the number of "bad" sounds is the same as the ratio of 1 to the number of atoms in the entire known observable universe!!!!! According to Google, on the high end the number of estimated ATOMS in the known universe is about 10^82.

By this analogy, we have 10^6615 / 10^82 and by the rule of exponents it becomes 10^(6615-82) = 10^6533. Ok, so we have now eliminated a HUGE number of possible 1 second samples that would sound like "noise". That still leaves a huge number left.

The age of the observable universe in seconds is about 10^24. Let's say you needed a group of people listening to the remaining 1 second samples every single second since the beginning of the universe. How many people would you need to listen, or how samples (in parallel) woudl you have to listen to every second to get through them all? You would need to play 10^6533 / 10^24 = 10^(6533-24) = 10^(6509) samples simultaneously since the beginning of the universe to listen to them.

OBVIOUSLY we cannot even appreciate the SCALE of these numbers. They are just TOO LARGE to fathom.

Let's try something even more simple. Let us take a sample that is not 22050 samples per second but 8000, which is around telephone quality (probably less than modern phone systems). We still encounter the same problems. 2^8000 possible combinations, in power of 10 we get 10^2400. The known number of atoms in the universe around 10^82. For fun, let us say we have 1 MILLION TRILLION universes... that's 10^16.... so total number of atoms in 1 MILLION TRILLION universes is 10^100... now if we even had only 1 possible atom out of all the atoms in 1 MILLION TRILLION universes worth of 1 second samples of 8000 sample rate, we would still have 10^2300 samples that *may* be music? And even if you listened to them all, without ever repeating one sound, since the beginning of the universe, you would have only listened to 10^24 samples because that's how many seconds there have been since the beginning of the universe... so you would still have about 10^2276 samples left to listen to.

I'm not sure if my math is flawed somewhere but even a rudimentary calculation like this shows that approaching music by creating all possible combinations of samples leads you to ridiculousness. Not only can you see that even with the crappiest sample qualities you still get gigantic numbers, but even selecting down by a huge number you still end up with too many options to be able to manage. Imagine using 44.1kHz audio? Remember it goes up exponentially for every extra sample in that second (you multiply again by 2).... so it is ridiculous. Not to mention that in the above examples we didn't even account for repeating 1 second samples... and music is very repetitive and structured, and the huge number of permutations of the ORDERS of those samples, to make up phrases, etc..

I rest my case... It is to all intents and purposes INFINITE because it vastly higher than the age of the entire universe by many many many orders of magnitude, even having eliminated most by many many many orders of magnitude as well.

[John B]

Someone earlier in the thread mentioned we could generate every possible sound and show that it is not infinite (obviously) but finite. However, ......<SNIP>...... purposes INFINITE because it vastly higher than the age of the entire universe by many many many orders of magnitude, even having eliminated most by many many many orders of magnitude as well.

The maths is all fine, but some of the axioms or presuppositions are wrong in that music isn't random or noise. I know, I know - there's tons of jokes to be made about whichever style. But music in general has intention and pattern, and patterns formed by constraints.

While randomly generated samples will eventually create something that sounds like music simply by the infinite monkey theorem, it would still lack the conscious awareness of what it is.

Humans are very good at exploring the boundaries within constraints, and by changing only a few, completely different creations emerge. The problem I think with "pop" music is that it is ultra constrained.

But even a few minor tweaks in tuning/temperament or rhythm could result in whole branches of new music.

Really there's nothing to worry about in terms of material. The sun will engulf us first.

[T3sl4co1l]

Yeah, it's important to note / understand that natural communication (whether language, music, social interaction, etc.) is very modest entropy.

It's peculiar I guess, that modern digital communications are -- quite intentionally and usefully -- high entropy.  These advantages probably apply to natural systems as well, but maybe on lower levels, neurological perhaps?  The apparent (high level) product -- language, music and so on -- is decidedly lower in entropy though.

Consider the average pop song: repetitive melody and structure throughout.  Some of that is driven by simple pattern recognition: a familiar pattern isn't necessarily a low-entropy one, but an obvious pattern one is very likely to be.  So I guess it stands to reason that even-numbered rhythms (4:4 time being most traditional), and stacks thereof -- roughly fractal structure with recurring sequences of the same length (4 beats, 4 bars, 4 phrases..), or what have you -- are likewise very common.

Very low entropy music is definitely a thing, take some slow meditation pieces for example.  It's decidedly..... not exciting, but when well executed, that's certainly beside the point.  High entropy music is there to some extent (look up "black MIDI" sorts of compositions, I guess?), but simply put: the more complex a piece becomes, the harder it is to understand, until it's eventually an onslaught of nearly-pitched white noise, and you can't tell what's going on at all.  (Or perhaps -- you can tell what's going on sometimes, and you're welcome to interpret that however you like, but that interpretation is likely to be as inconsistent, as noisy, as the composition it's supposedly derived from. )

So, in information-theoretic terms, we can rule out a large fraction of possible songs based on unintelligibly high entropy.  This isn't merely percentagewise for a given range of song lengths, this is a huge difference -- many orders of magnitude.  If we stipulate on repeating patterns as well, several orders of magnitude yet.  Still, what's left is a very big number indeed, one the universe is very unlikely to exhaust!

Tim

[metrologist]

Yeah, it's important to note / understand that natural communication (whether language, music, social interaction, etc.) is very modest entropy.

It's peculiar I guess, that modern digital communications are -- quite intentionally and usefully -- high entropy.  These advantages probably apply to natural systems as well, but maybe on lower levels, neurological perhaps?  The apparent (high level) product -- language, music and so on -- is decidedly lower in entropy though.

Consider the average pop song: repetitive melody and structure throughout.  Some of that is driven by simple pattern recognition: a familiar pattern isn't necessarily a low-entropy one, but an obvious pattern one is very likely to be.  So I guess it stands to reason that even-numbered rhythms (4:4 time being most traditional), and stacks thereof -- roughly fractal structure with recurring sequences of the same length (4 beats, 4 bars, 4 phrases..), or what have you -- are likewise very common.

Very low entropy music is definitely a thing, take some slow meditation pieces for example.  It's decidedly..... not exciting, but when well executed, that's certainly beside the point.  High entropy music is there to some extent (look up "black MIDI" sorts of compositions, I guess?), but simply put: the more complex a piece becomes, the harder it is to understand, until it's eventually an onslaught of nearly-pitched white noise, and you can't tell what's going on at all.  (Or perhaps -- you can tell what's going on sometimes, and you're welcome to interpret that however you like, but that interpretation is likely to be as inconsistent, as noisy, as the composition it's supposedly derived from. )

So, in information-theoretic terms, we can rule out a large fraction of possible songs based on unintelligibly high entropy.  This isn't merely percentagewise for a given range of song lengths, this is a huge difference -- many orders of magnitude.  If we stipulate on repeating patterns as well, several orders of magnitude yet.  Still, what's left is a very big number indeed, one the universe is very unlikely to exhaust!

Tim

Admire your posts, I do. If only I was an artist...

I found this creative and related.

Trained on Archspire with modified SampleRNN. Read more about our research into eliminating humans from metal: https://arxiv.org/abs/1811.06633

[T3sl4co1l]

Heh. Sounds rough, a bit undertrained I would guess (or reduced quality to make that more feasible to begin with), but seems to get the form right.

But will RNNs be able to come up with myriad new and obscure subsubgenres of metal automatically as well?

Tim

[KE5FX]

It will still be some limited number of combinations though.

Sure since the spectrum is limited to 22kHz there is always a limit to the number of possibilities.
Luckily 90% of the possible combinations are not sounding nice.

"Sounding nice" is all in the context.  For an extreme example, scrub this video to 2:30 and stop it at 2:50:



Now go back and listen to the whole thing from the beginning.  In this example, the duration of the "context" that differentiates horrible racket from the centerpiece of an interesting work is measured in minutes, not beats or seconds or samples.

Human attention is an exhaustible resource, so it doesn't really matter whether music or literature is.  Like Tim says, you'll drop dead of old age before you ever see a fraction of the Library.