Why can't anybody be a music composer?

[Loren Booda]

I often wonder how copyright infringement is determined with the myriad measures of music. It seems to me that it goes on frequently, at least that musicians go out of their way not to reproduce more than a few seconds of copyrighted composition without permission - and that commercials are the main offender, when one could almost swear they could hear some great modern song narrowly averted.

Then their are the "artists" who repeat (with slight variation) their only hit ad nauseum over the years.

__________

Calculate the combinations of resonating superstrings in the cosmos!

 

[ThomasT]

http://en.wikipedia.org/wiki/Computer_music

 

[turbo]

Then their are the "artists" who repeat (with slight variation) their only hit ad nauseum over the years.

One need only listen to pop/rock FM to get the message. Once you've had a couple of hits, keep harping on it! Think U2, Rush, etc, etc. Some of these groups haven't had a new idea for decades, and apparently their handlers and recording companies are perfectly happy with that, as long as the cow is still giving milk.

 

[turbo]

Composing a popular music should be much simpler than literature. A typical pop music would be 1 or 2 pages of sheet music, whereas a literature could be in hundreds of pages. Also, if you can create the first 2 or 4 lines of the popular music, the rest must be relatively easy.

Haven't written many songs, huh? You can never tell where the muse might lead you once you get started, and it's tough (especially for a technically-minded person) to craft lyrics that convey some complex concepts with a few carefully-chosen words. Song-writing is tough, and you only have to listen to Lyle Lovett, John Hiatt, etc, to see why. Where did they get those ideas and those turns of phrase? They didn't spring from the "first 2 or 4 lines" of their music.

 

[TheStatutoryApe]

...it's tough (especially for a technically-minded person) to craft lyrics that convey some complex concepts with a few carefully-chosen words...

Too true. Writing poetry and lyrics are similar. Trying to describe something in a unique and evocative fashion in just a few lines is difficult. And with lyrics rhyme, rythme and meter become far more important and constricting. I've been quite daunted by the prospect of trying to write lyrics for music.

PS: As much as I am loathe to admit it rappers actually take the cake quite often in written lyrics.

 

[jobyts]

Haven't written many songs, huh? You can never tell where the muse might lead you once you get started, and it's tough (especially for a technically-minded person) to craft lyrics that convey some complex concepts with a few carefully-chosen words. Song-writing is tough, and you only have to listen to Lyle Lovett, John Hiatt, etc, to see why. Where did they get those ideas and those turns of phrase? They didn't spring from the "first 2 or 4 lines" of their music.

I meant, the tune part. I would agree song writing would be much harder from an AI perceptive.

 

[Vanadium 50]

Haven't written many songs, huh?

I'm kind of amused by this - people who haven't done it themselves, but who are absolutely sure that this must be so simple that even a computer can do it.

A pentium has 3 million transistors. A mouse has 3 million neurons. How many mice write symphonies?

 

[turbo]

I'm kind of amused by this - people who haven't done it themselves, but who are absolutely sure that this must be so simple that even a computer can do it.

A pentium has 3 million transistors. A mouse has 3 million neurons. How many mice write symphonies?

I have slaved over lyrics, re-working, tweaking, etc. The melody and rhythm usually come together pretty easily, but the lyrics...? Argh! I've still got a song un-finished from almost 40 years ago - never could get it to come together.

 

[moe darklight]

If writing music or poetry were easy, then all songs and all poems would be good, and there would be very few terrible artists out there.

I think there's an obvious reason why Shakespeares and Beethovens are just as rare as are Newtons and Gausses.

 

[fuzzyfelt]

This was on Neatorama recently:
www.neatorama.com/2009/01/27/only-four-chords-are-needed-to-make-a-hit-song/


https://www.youtube.com/watch?v=qHBVnMf2t7w

This is great, too!

 

[rootX]

If writing music or poetry were easy, then all songs and all poems would be good, and there would be very few terrible artists out there. .

Yes, agree.
If all hit songs are crap or no hard thing to come up with then anyone should be capable of coming up with that crap and becoming filthy rich?

 

[amezcua]

Good basic idea.

 

[junglebeast]

No matter how good your program is, it will never be as good as the worst human.

That said, I've heard it from a friend that the music industry has some kind of software that they run songs into and if they get a high score it influences if they will make that one popular by playing it on the radios...I don't know how much truth to this there is

 

[junglebeast]

This was on Neatorama recently:
www.neatorama.com/2009/01/27/only-four-chords-are-needed-to-make-a-hit-song/


https://www.youtube.com/watch?v=qHBVnMf2t7w

You can always sing the lyrics of one song over another but these chords sound completely different from the songs they are representing to me...and it also sounds like utter crap.

 

[wittgenstein]

"I like it. the "million monkeys on a million keyboards" approach. good luck finding a search algorithm."
Proton Soup
If you had a million Shakespeares could one of them write like a monkey?
OK. I stole that from Steven Wright
But seriously folks...
Is music subjective or objective? Is there an objective way to say with certainty that Mozart is a better composer than I? Obviously, Mozart is way better. But is there an objective algorithm?

 

[junglebeast]

Is music subjective or objective? Is there an objective way to say with certainty that Mozart is a better composer than I? Obviously, Mozart is way better. But is there an objective algorithm?

I think it's a combination of objective and subjective factors. When listening to a new song, whether or not I like it, it is immediately obvious if the artist misses a single note...because it no longer fits with the rest of the song. In other words, it no longer meets the objective criteria necessary for beauty.

Beyond that, there's a large subjective component that determines which types of music you are receptive to enjoying this beauty in, and how much you will enjoy it. This is a combination of mood and social factors.

 

[CRGreathouse]

No matter how good your program is, it will never be as good as the worst human.

You vastly overestimate the worst human. Here's a C program which produces music (a la John Cage) better than that of the worst human:

int main() {
  return 0;
}

 

[Vanadium 50]

produces music (a la John Cage) better than that of the worst human

It's easy to mock 4'33". Have you heard it? If not, I recommend it.

 

[Kenneth Mann]

There are only a finite number music notes. And to create a finite duration of music, write an algorithm that plays all the combinations of the notes in, say, piano. Listen to most/all of them (or outsource them) and pick the better ones. Optionally, improvise them.

What's wrong with this approach? (I'm going to do this if I loose my current job:). Why do we need more than an average brain to compose music?

Or, just copyright all the combinations and live happily ever after.

This is just a rough estimate:
You seem to have a very naive notion concerning what you'd be up against. Consider one simple case - - creating very basic four/fourths melodies. For simplicity, we'll forego such niceties as tempo, timbres, chord patterns and the like, and just look at the simplest of melodies. At the very coursest, we break each measure up into eight eighth note intervals. (Sixteenth-note intervals would be more realistic, but let's not overcomplicate things.) Now, each of these intervals can be either a sounded note, or a rest. Furthermore we simplify by not being concerned with note intensities - - which would complicate things considerably. We shall consider "note" or "no-note" for each of the eight positions within the bar. Immediately, this leads us to 2^8, or 256 possibilities. Now, we can assume each of these bars to average four sounded notes (and four rest positions). (In reality, there are generally more notes than rests.) We also shall ignore whether the notes are legato or stoccato, a further simplification. Finally we shall look at the simplest of phrases - - one of only four bars in length. This would contain 2^32 possibilities, roughly 4 billion.

To construct the simplest melodies, a two-octave note range is called for - - twenty four values. Thus to generate all possible values for this rudimentary case the program will generate 24^32 distinct melodies. This yields over 10^44 melodies to be evaluated - - and these are just the simplest of melodies. If you are going to do this, I'd suggest you get started soon. At a couple hundred a day it's going to take a while. Also remember, if you want to take on more sophisticated cases, it will take even longer - - much, much longer. Good luck.

KM

 

[bassplayer142]

I would say your best bet is to use the program to develop small melodies that are catchy. The probability of making an entire song is our of this world improbable when you consider every note and rhythm combination.

I personally believe that truly beautiful music isn't just a logical statement but an emotional one.

 

[Kenneth Mann]

I would say your best bet is to use the program to develop small melodies that are catchy. The probability of making an entire song is our of this world improbable when you consider every note and rhythm combination.

I personally believe that truly beautiful music isn't just a logical statement but an emotional one.

The odds are long against being able to derive even the simplest tunes with this approach. Even a single octave diatonic melody would entail 9^32 iterations (343x10^9). How many eons would that take? It's really next to hopeless.

The irony lies in the fact that melodies can be generated using other approaches. There was a module (hardware), called "Muse", made about forty years ago, that could create some relatively pleasing (if somewhat unusual) melodies. The secret is to take advantage of some of the relationships and symmetries that are found in pleasing music. A "brute-force" approach, on the other hand, is very unlikely to work. To generate good computer-composed music, I would suggest that the person interested look into "associative processors". On the other hand, the best associative processor is probably the human brain. I'd suggest that study of composition and practice is the very best approach.

KM

 

[bassplayer142]

The odds are long against being able to derive even the simplest tunes with this approach. Even a single octave diatonic melody would entail 9^32 iterations (343x10^9). How many eons would that take? It's really next to hopeless.

The irony lies in the fact that melodies can be generated using other approaches. There was a module (hardware), called "Muse", made about forty years ago, that could create some relatively pleasing (if somewhat unusual) melodies. The secret is to take advantage of some of the relationships and symmetries that are found in pleasing music. A "brute-force" approach, on the other hand, is very unlikely to work. To generate good computer-composed music, I would suggest that the person interested look into "associative processors". On the other hand, the best associative processor is probably the human brain. I'd suggest that study of composition and practice is the very best approach.

KM

How about a computer that learned off a musician. If someone plugged into it and started jamming out on the guitar then you could program the computer to find relationships in that particular persons style.

 

[amezcua]

Jobyts
Like your style! Apart from the music aspect there is a danger that the copyright of every combination might legally work.BOING ! --- what a crazy situation that would be.
I lost one post a week ago but since then I read a history of the French mathematician Mersenne.He was fascinated by musical subjects and proved that there were 40,320 possible combinations of 8 notes.
My "lost" post referred to a book by Barlow and Morgenstern to show that most famous tunes are built on groups of 8 notes.The book is ;A Dictionary of Musical Themes.
Your idea to make a computer generate combinations actually works .
You will notice a lot of replies refer to you switching on a computer and waiting for the stuff to land in your lap. That is not what you suggested.You put the important ingredient in there.You would decide what was worth keeping. I have tried this and it does work and a lot quicker than I expected.
The KEY point is that sitting before a blank sheet of paper is enormously more difficult than sifting through random 8 notes which QUITE OFTEN produces musical combinations.
Organise 8 audible notes to sound.Have a repeat control,a delete control and one to let the notes run on if it sounds promising.Base everything on 8 notes.Work with the key signature system so that it won`t be all sharps and flats.
For efficiency I collected groups of notes and then played them out on a piano afterwards.
Best (or most unexpected) result was the whole oboe solo from the slow movement in the Brahms Violin Concerto, all correct apart from the second note from the end.
One crazy sounding piece left to repeat was very addictive.
It takes a few days to get into the routine and then if you thought you had a musical streak
you find you were correct. Most people will scoff and not try it .That puts you one step ahead.
In the Eurovision Song contest most pieces played do not follow the 8 note "rule" and that`s why they are mostly rubbish.

 

[amezcua]

When I logged on just now I wrote the above post then I was told --You are not logged on.That`s how my first post went walkabout.Some slipping cogs in this machine.

 

[jobyts]

A decent computer chess AI program remembers 10^120 positions. The difficulty in creating computer generated tunes lies not in finding out all the combinations. The difficulty is in choosing which tune is better than the other one. In a chess program, there is deterministic way to find out move A is better than move B. But choosing a tune would be much harder for a computer. It's harder, but doable with some human interaction.

Use internet, ask people to listen to x tunes and choose the best 2 from it. Do it recursively until the number of tunes are small enough for someone to listen.

 

[Kenneth Mann]

How about a computer that learned off a musician. If someone plugged into it and started jamming out on the guitar then you could program the computer to find relationships in that particular persons style.

Actually you've got something here. There are several relationships in music that can be used. The easiest way to these, however, is not through a computer which listens, but rather by programming in some of the already known relationships (musocologists spend a lot of time and effort working on these). There was a machine (called Muse - - - no relationship to present-day entities of the same name) made several decades ago that used one of these simple algorithme. It could come up with some rather pleasing - - - if sometimes odd sounding - - - tunes. It is definitely possible, but this method is not totally random.

KM

 

[Hel]

I think absolutely anybody can write 12 tone music, and an algorithm could easily be made for that, but the only people who actually listen to 12 tone music are geeks and professional classical musicians.


Tintinnabulation (google Arvo Pärt) is also a possibility. Pärt's music usually follows a strict pattern, very predictable. However, the thing that makes his music incredible is the way he breaks the pattern, sometimes in such a way that one note at the beginning effects all subsequent notes.
(My organ prof did a a lot of research on this)

There are lots of rules in music that exist because anything different sounds strange or bad to us, simply because it's not something we are used to. (example: parallel fifths in voice leading was actually banned by the Vatican for a time because it "came from the devil". Parallel fifths will still get you docked on Music Theory exams, as I discovered from personal experience.)

Lots of people who don't even know these rules instinctively follow them because if they weren't followed, the music wouldn't sound good.

Even if a program were made that can follow all the established rules of music theory, the computer won't be able to determine when it's necessary to break these rules for the sake of making the music sound better.
That's also why only certain people are composers. Being able to listen and determine what sounds good and knowing how to make it better is difficult!

Okay I'll stop... although I didn't even mention counterpoint, form, or film scoring!

Edit: My point: Anybody can make music, but only certain people will be able to make music that sounds great.

 

[Kenneth Mann]

I think absolutely anybody can write 12 tone music, and an algorithm could easily be made for that, but the only people who actually listen to 12 tone music are geeks and professional classical musicians.


Tintinnabulation (google Arvo Pärt) is also a possibility. Pärt's music usually follows a strict pattern, very predictable. However, the thing that makes his music incredible is the way he breaks the pattern, sometimes in such a way that one note at the beginning effects all subsequent notes.
(My organ prof did a a lot of research on this)

There are lots of rules in music that exist because anything different sounds strange or bad to us, simply because it's not something we are used to. (example: parallel fifths in voice leading was actually banned by the Vatican for a time because it "came from the devil". Parallel fifths will still get you docked on Music Theory exams, as I discovered from personal experience.)

Lots of people who don't even know these rules instinctively follow them because if they weren't followed, the music wouldn't sound good.

Even if a program were made that can follow all the established rules of music theory, the computer won't be able to determine when it's necessary to break these rules for the sake of making the music sound better.
That's also why only certain people are composers. Being able to listen and determine what sounds good and knowing how to make it better is difficult!

Okay I'll stop... although I didn't even mention counterpoint, form, or film scoring!

Edit: My point: Anybody can make music, but only certain people will be able to make music that sounds great.

Amen!

Thanks for your help.
I'm afraid I wasn't doing very well convincing these guys that they couldn't simply take a program that serially (not twelve-tone, but it would probably sound like it) combines notes and expect to become the next Bach.

KM

 

[Kenneth Mann]

Jobyts

I lost one post a week ago but since then I read a history of the French mathematician Mersenne.He was fascinated by musical subjects and proved that there were 40,320 possible combinations of 8 notes.

I would not put too much stock in Mersennein that matter. He lived before the time of modern music - - - in fact, when he was here the well tempered scale was just coming into existence. (He contributed to it himself.) It took the works of those who followed him (notably Bach) to establish the practices of today's music. Up to that time established music was a much narrower set of practices, and what was considered music had strict limits (church influence). Like Plainsong, which usually didn't even make use of rhythm patterns. (There were rhythmic tunes around, but whether they were known to follow rules might be questioned.) Music then was a work in progress.

KM

 

[amezcua]

I logged in.Posted an answer.Then I was told ,You are not logged in.Message at top said "log out".So I logged in.Then my message had disappeared.-----------------------------You tell me!