Music composed by a software program

[composer]

Douglas Hofstadter writes about some music composing software that is able to write music in the style of different famous composers. He discusses the program here

The EMI Program (http://www.unc.edu/~mumukshu/gandhi/gandhi/hofstadter.htm)


In my lectures, I usually have a second musical interlude, this time involving mazurkas -- one by Chopin and one by EMI. One time, when I gave this lecture at the world-famous Eastman School of Music in Rochester, New York, nearly all the composition and music-theory faculty was fooled by the EMI mazurka, taking it for genuine Chopin (and the genuine Chopin piece, by contrast, for a computer-manufactured ditty). An Eastman music student, Kala Pierson, wrote me an email about this event in which she said, "I voted real-Chopin for the second piece, as did most of my friends. When you announced that the first was Chopin and the second was EMI, there was a collective gasp and an aftermath of what I can only describe as delighted horror. I've never seen so many theorists and composers shocked out of their smug complacency in one fell swoop [myself included]! It was truly a thing of beauty."

This demonstrates that composing music can be reduced to a series of steps and decisions–an “algorithm”.

Discuss…

[Ptah]

Like any other art from, some degree of formula is possible, if not behind (and giving rise to) certain "timeless" instances, if you ask me.

[composer]

How much is 'some degree', and how far can it be taken?

This program was able to fool experts.

[stuck]

My ground isn't shaken by this:

The methods used were 'recombinant', meaning that they took a bunch of works from one composer and scrambled them all together, having the computer work out some macro as well as some micro. However, the raw material, I'd imagine, is the thing doing the fooling.

[YHWH]

it's official, music is dead.

[stuck]

it's official, music is dead.

Don't despair. There will be much great artificial music, which will not be automatically devoid of pre-existing musical values. It's a meme- the music of computer intelligence, the sound of logic, the sound of mathematical operations. Humans will find their place in the singularity, and everywhere it will reflect us, our bias towards ordering the universe.

The ambient music you enjoyed of mine was composed by static logical and mathematic operations, by the way. I set the synths and hit record, while the circuits I set up worked themselves out. This arrangement allowed me to concentrate on distilling the feeling of the music, freed from any other technical concerns.

[composer]

The methods used were 'recombinant', meaning that they took a bunch of works from one composer and scrambled them all together, having the computer work out some macro as well as some micro. However, the raw material, I'd imagine, is the thing doing the fooling.

How is that different from how composers work? I understand that Mozart's 'formula' was 'decoded', in other words some musicologists figured out his likely algorithm for writing music. Explains how he was able to write so much so quickly, and why so much of the earlier work is so boring.

I worked with a <world famous composer> who said much the same. He thought the worship of composers was silly and counterproductive, he described it as having a craft. Sounds like an algorithm.

Alsop a small distinction, while I'm not familiar with the details of his approach, it doesn't seem that he 'took a bunch of works from a composer and scrambled them together'. What I read is the identification of recurrent structures of various sorts in a composer's output, and the reusing of those structures in new arrangements. Sounds like a design or architectural decomposition, which again is probably much the way people work.

[stuck]

How is that different from how composers work?

That's standard stuff, building something new out of parts of the old. My ground isn't shaken when composers simply do that, either.

"I'd be more impressed at someone working out the algorithm of Mozart that constructs itself from rules rather than Mozart samples + rules" is all I'm saying. This is what I concentrate on, because I feel I'd be able to then abstract the rule set and use that as a set of compositional tools to create something that doesn't sound like mozart at all.

[YHWH]

Don't despair. There will be much great artificial music, which will not be automatically devoid of pre-existing musical values. It's a meme- the music of computer intelligence, the sound of logic, the sound of mathematical operations. Humans will find their place in the singularity, and everywhere it will reflect us, our bias towards ordering the universe.

The ambient music you enjoyed of mine was composed by static logical and mathematic operations, by the way. I set the synths and hit record, while the circuits I set up worked themselves out. This arrangement allowed me to concentrate on distilling the feeling of the music, freed from any other technical concerns.

that's awesome. what I found hilarious was composer's conclusion. :facepalm:

composers have been working on algorithms for some while. many of my favourite composers and artists allowed chance, randomness, sampling, ready-mades to be part of their work or to define their work, that somehow didn't make art obsolete, quite the contrary.

[composer]

that's awesome. what I found hilarious was composer's conclusion. :facepalm:

YHWH, fuck off.

You knew that was coming.

[YHWH]

YHWH, fuck off.

You knew that was coming.

kid.

[asperger]

Actually we've already discussed EMI a bit:

http://forums.intpcentral.com/showthread.php?44554-Novelty-and-aesthetics&p=1495645&viewfull=1#post1495645

[asperger]

http://forums.intpcentral.com/showthread.php?44554-Novelty-and-aesthetics&p=1495645&viewfull=1#post1495645

I think I've listened to everything from Cope's programs that is on line and I'm not impressed. An interesting point:


One time, when I gave this lecture at the world-famous Eastman School of Music in Rochester, New York, nearly all the composition and music-theory faculty was fooled by the EMI mazurka, taking it for genuine Chopin (and the genuine Chopin piece, by contrast, for a computer-manufactured ditty).

But EMI's pseudo Bach output does not fool such people. Personally I find EMI's output very unremarkable in general.

[gator]

My ground isn't shaken by this:

The methods used were 'recombinant', meaning that they took a bunch of works from one composer and scrambled them all together, having the computer work out some macro as well as some micro. However, the raw material, I'd imagine, is the thing doing the fooling.

I agree with this. While I think it's impressive that the program can distill the elements of a composer's style and do something new in the same vein, it's still sampling and recombining that narrow range of work. It's good at mimicking something that has already been created, rather than creating something new. The real challenge of composing or doing anything else creatively is being able to sample from things that influence you and then develop new styles and algorithms. Until they program something that does that I'm not going to be too concerned.

[YHWH]

even if a program could compose something original and seemingly creative, I don't see why any grounds would be shaken. machines are already 'composing'. only a couple days ago, I was at a factory and the sound/rhythm of a machine was awe-inspiring, I taped it, if a program composed something awesome, I'd enjoy it or sample it and reuse it. it's human perception and intent that make it art, not the object/sound-in-itself.

[asperger]

Nice post, thank you.

[stuck]

I'd feel the ground shake if someone were able to reverse engineer good composition techniques, purely because that's what I'm trying to do all the time, either codified/automated or informally. It'd represent a step forward for how music is taught, I think, and perhaps everything else, too.

[asperger]

Stuck, as usual I feel like I'm missing something here. Would you mind expanding on that? Thanks.

[stuck]

I have a hunch that some elements of style are universal- and could be most succinctly represented via algorithm. If we could have an intuitive grasp and computer tools to exploit this hidden wisdom, we could make better art, better data compression, better buildings, better everything.

[Skinart]

Something like a golden ratio for sound eh? But wouldn't that reveal more about our present preferences for musical sound than any intrinsic algorithm behind music?

There are definite algorithms clearly stated for many genres of music. Some of these algorithms are stated in the names--four to the floor comes to mind. When these algorithms are not in place, the music is seen as falling outside of that genre.

I think at best you could provide a taxonomy of music and say it was generated primarily using certain algorithmic forms--which is to some degree recognized when we identify a waltz vs. a reel. I don't think you could create a single algorithm to compose all of musicality, but you could probably devise one to identify and sort all of musicality.

Fringe genres, like noise and micro-sound are to be lauded for finding new ways to muck up musicality. In part because they do it without deliberately breaking algorithms--as jazz does--but by simply throwing them out and starting from scratch. They remind us that the algorithm of musicality is in the head listening, not the algorithm of the head composing.

[Chunes]

Humans are the EMIs of alien music. Time to stop kidding yourselves, music composers..

[composer]

First to say that I don't know the details about how this program was written except for secondhand through Douglas Hofstadter. However I think it's too easy to dismiss these efforts. Consider Watson, this was attempting so much more difficult problem which was to understand subtle distinctions in language, jokes and nuances. The problem of writing software to compose music is, I believe, much simpler. The combinatorial space is much smaller than the entire corpus of human knowledge and the English language. Even at that, the Watson problem was actually fairly pedestrian as far as AI problems go. It had never been solved before to this degree because it wasn't that exciting.

Additionally Watson was written by a funded team of IBM scientists, this was written by a single individual who, as far as I know, is not an expert in software or engineering. I believe if we were to put a team of top engineers on it they could write composing software that would achieve much greater heights. Why can't they write software that meets or exceeds the efforts of our best composers?

The reason I mention it is because I believe that we're going to have to grapple with this problem in the near future.

[rhinosaur]

I have a hunch that some elements of style are universal- and could be most succinctly represented via algorithm. If we could have an intuitive grasp and computer tools to exploit this hidden wisdom, we could make better art, better data compression, better buildings, better everything.

I think you would never escape the empirical approach.

You could poll a very large number of people, with wildly divergent musical tastes, and come up with some kind of master algorithm. Sit down with them and ask them to come up with four or five songs that they really really like, and identify specifically what section of the song they like the best. They don't even need to know why they like it. Apply David Cope's EMI algorithm to all of those song pieces, and combine it all. Wham! Classirockahopsijazz.

[asperger]

I believe if we were to put a team of top engineers on it they could write composing software that would achieve much greater heights. Why can't they write software that meets or exceeds the efforts of our best composers?

The reason I mention it is because I believe that we're going to have to grapple with this problem in the near future.

I believe that music becomes important to people when it makes a visceral/emotional connection. Thus it is my belief that “purely” computer generated music will remain bland and uninteresting until it can model the human “soul” and measure its work against that sounding board.

What interest me is the prospect of algorithms doing the grunt work while the composer handles the aesthetics.