I'm Sorry Dave, I Can Compose Better Than You

February 24, 2010 7:46 PM

Computer generated music is about to get better.
posted by herodotus (44 comments total)

Error establishing a database connection
posted by nathancaswell at 7:47 PM on February 24, 2010

Establishing database connections, however, will be an issue for a while.
posted by LSK at 7:49 PM on February 24, 2010 [1 favorite]

Try refreshing, it worked on the second try.
posted by maishuno at 7:53 PM on February 24, 2010

I can't wait for the Pitchfork review of Error Establishing a Database Connection. They practically jerked off into their own mouths after Previous Implicit Declaration of Function, so they'll probably shit all over the new record.
posted by secret about box at 7:56 PM on February 24, 2010 [18 favorites]

Personally I liked the Inject Malicious SQL Exploits better, but only the first album before they got signed.

Here's a Coral Cache link if the main post is broken. Looks like someone already loaded it.
posted by loquacious at 8:05 PM on February 24, 2010

I used to work briefly with Cope at UC Santa Cruz. He was a fascinating guy, exactly as lively and overflowing with an insatiable curiosity as you can imagine. He and Peter Elsea did some incredible work with automatically composing music, and I like how their compositions expose just how structured and algorithmic Western music can be.
posted by spiderskull at 8:15 PM on February 24, 2010 [3 favorites]

Okay shit yeah that's good. I mean, let me say that I like quite a lot of computer-generated music but those are some well-wrought compositions. Also, a very interesting article. Cool find, herodotus!
posted by Kattullus at 8:20 PM on February 24, 2010 [1 favorite]

Fascinating.

But:

"...and it performed just as well in producing new Mozart compositions and Shakespeare sonnets."

Shakespeare sonnets?
posted by grumblebee at 8:20 PM on February 24, 2010

grumblebee -- I thought that was weird too, but maybe they used the same idea to write sonnets. I imagine it isn't much of a leap to modify what I imagine is some sort of Markov process.
posted by spiderskull at 8:22 PM on February 24, 2010

yeah. track 2 is kinda pretty. Not sure elevator / sleepy time piano music being generated automatically is all that game-changing, though.
posted by rubin at 8:26 PM on February 24, 2010

"I can't wait for the Pitchfork review of Error Establishing a Database Connection. They practically jerked off into their own mouths after Previous Implicit Declaration of Function, so they'll probably shit all over the new record."

This reply made me laugh way too hard
posted by colinshark at 8:41 PM on February 24, 2010

So it's going to go beyond having random fourth-notes senselessly playing on and on, with no concept of song structure or what sounds good?
posted by dunkadunc at 8:44 PM on February 24, 2010

Thanks for this. I've sort of been interested in aleatoric music and particularly algorithms and software for computer-generated music, so I found this really interesting.

What Dr Cope's really done is mostly a better analysis of relationships in music, and ways to encode those relationships, and later use them to either compose something "not unlike [composer whose works were analysed]", or more importantly to aid his own composition by capturing his musical input, and producing "sensible" progressions or options on that input , based on data gleaned from his previous compositions, that he can then accept or not.

I'm struggling for graphical equivalents to this process. Hmmmm... maybe painting a line and having the computer suggest 10 different new lines, after it had analysed all of your previous good paintings?
posted by Artful Codger at 8:53 PM on February 24, 2010 [4 favorites]

The AI 'Emmy' is not a bad composer. The tone of the article is bit breathless.
posted by ovvl at 8:55 PM on February 24, 2010

Radiolab did a bit about Cope in 2006. It's a good listen.
posted by sleepy pete at 9:42 PM on February 24, 2010 [2 favorites]

Around 1987 I went to a museum exhibit in L.A. and saw a demo of computer-composed music. Wonder if it could have been his?
posted by thelastenglishmajor at 10:38 PM on February 24, 2010

Very interesting, and the music was quite pleasant. As for Dr. Cope's philosophical outlook with regards to his work, I find myself surprised by partly agreeing with his detractors. I think reacting with fear or anger is ridiculously over-the-top, but I also think that Cope is (partly) missing the point of music, and more generally, art. He is conceiving of art as something that is created by an artist in isolation, and likewise enjoyed by an audience in isolation. What Cope's missing is the notion of art as a conversation (albeit a bit one-sided, in most cases) between artist and audience.

His comment about notes on a page being mere dots and lines reflects this misunderstanding. You could say the same thing about a letter that I write to a friend: dots and lines, with the meaning and significance (re)constructed within my brain. If you were to receive a beautifully written love letter, you might be moved by that. If you then found out that the letter was meant for someone else, you'd feel foolish and disappointed. You see where I'm going with this.

I surprise myself by making this argument, since I'm otherwise sympathetic to his "A soul? What is that?" point of view. But for me, that doesn't even have to come into this discussion.
posted by Edgewise at 10:47 PM on February 24, 2010 [4 favorites]

is this software available?
posted by empath at 11:13 PM on February 24, 2010

I really like the tensions of track one. (and listening further, track two evolves really nicely towards the end)
It has a certain cinematic quality to it in my mind. Uses? Many.
This further digitality could change (in a positive direction) another aspect of filmmaking; getting some kind of score, that fits certain creative and thematic parameters, without breaking bank.
posted by infinite intimation at 11:36 PM on February 24, 2010

great find!
in fact, the results are quite astonishing. I guess I could discern his counterpoint from Bach's more or less easily, although.
posted by megob at 11:39 PM on February 24, 2010

The CD that comes with his book The Algorithmic Composer has a LISP implementation of his older program, called ALICE. I read that book last year and remember being unimpressed by his methods. It struck me that he had hard-coded too many theoretical assumptions: the parsing phase was already looking at the music in terms of triads and inversions, rather than sussing out those structures on its own (he claimed that it could handle atonal music too, but the way he represented voice leading between pitch class sets struck me as under-developed, though I don't remember exactly how it was done). As I recall, it required a large corpus of works, mostly of similar length, instrumentation and meter. It seemed as though he could have really benefitted from the assistance of someone well-versed in machine learning.

It sounds like the software has undergone a lot of revision, though, and the results are fairly convincing. ALICE's results were too, but they were much more restricted in scope.

What Cope's missing is the notion of art as a conversation (albeit a bit one-sided, in most cases) between artist and audience.

But if the audience doesn't know the difference and thinks they're having a conversation with a person, what difference does it make? The musical object is indistinguishable, and classical music loves the object. I guess it depends on the audience...if you get a bunch of composers with good ears in the room, they might catch some clumsiness in the high-level formal construction, but I wouldn't be surprised if you could get a composition program to handle that cleverly too.

My take-away from this, which is just an affirmation of something I've believed for a while now, is that composition that only concerns itself with the manipulation of the twelve tones in whatever framework, tonal or atonal, is a dead and basically irrelevant practice. It was bound to happen when the medium consisted of a small number of discrete elements. The real playground is in the areas where continuous variation is possible -- as in the gradations of inflection available in the human voice. The endless timbres you can get out of a synth patch, be it AM, FM, subtractive, whatever. I don't think a program will be able to handle those things masterfully until it becomes fully sentient, at which point I don't think we have any reason to be defensive.
posted by invitapriore at 11:42 PM on February 24, 2010 [3 favorites]

Also, herodotus, I hate you!, I was totally going to post this earlier in the evening but then put it off so I could drink beer. Dumb.
posted by invitapriore at 11:44 PM on February 24, 2010

Well, I'm sure as hell looking forward to that CD. Those compositions are beautiful. I don't care if a computer made them. That second one particular has a real sense of sadness to it, and that remains true no matter who or what created it.
posted by Malor at 11:51 PM on February 24, 2010

My primary interest in this software (an application for which I don't think ALICE would be useful, because of its aforementioned defects) is as a way to shine a light on the syntactical rules governing Schoenberg, Berg and Webern's free atonal music, before they moved on to serialism. Everybody can agree on the high-level requirements (basically, 'don't be tonal in any conceivable way') but theorists love to impose their clever and totally abstract structures on that music, and none of their ideas really ring true to me.

It's a short and experimental phase in music, so it might be the case that something approaching an emergent grammar never really solidified, but if that's the case I bet a good statistical analysis of the structures at play could tell us as much.
posted by invitapriore at 12:00 AM on February 25, 2010

But if the audience doesn't know the difference and thinks they're having a conversation with a person, what difference does it make?

Two objections. First, the audience does know the difference. Second, what you seem to be saying is that the audience has to be deceived to fully enjoy the work. If you thought you were talking to a cherished friend, poured your heart out, and felt better as a result, I suppose that has its own value. But if the lights turned on and you realized you were talking to a mannequin, that would certainly diminish the experience. Any insights that you had about yourself would still be intact. But any sense that you created a deeper connection with your friend would be gone.

OK, that's not a perfect analogy or a very likely scenario, unless you frequently end up trapped in unlit department stores. But I can't really get behind any work of art that has to actively lie to me for me to appreciate it. Call me quaint and old fashioned, but I always thought art was supposed to plumb the depths of universal truth...and how can you do that when you're being fundamentally dishonest? Music as envisioned by Cope can please you, but can it tackle any of the higher functions of art?

However, one day we will have true artificial intelligence, and when those computers generate music, I won't have these kinds of reservations. Maybe some others, but we'll see.
posted by Edgewise at 12:19 AM on February 25, 2010

is this software available?

A lot of Cope's software is available here: http://artsites.ucsc.edu/faculty/cope/software.htm
posted by rajbot at 12:25 AM on February 25, 2010 [1 favorite]

I think people become overly fixated on the fact that there's a computer involved, and miss the fact that it's just a tool which a human being has programmed with his own creativity. Composers think algorythmically when composing anything, and figuring out how to get a computer to mimic this process is very enlightening about how methodical the human process is. I was always told when learning composition that it was '1% inspiration and 99% perspiration', I don't see anything wrong with letting the computer handle the 99% part of things!
posted by BobsterLobster at 12:36 AM on February 25, 2010 [1 favorite]

Call me quaint and old fashioned, but I always thought art was supposed to plumb the depths of universal truth...

I'd love to write a long and well-considered comment in response to yours, Edgewise, but I really have to go to bed. Assuming that your assertion is true (I don't, which is part of why my comment would be so long) why is a human-generated composition any more likely to reveal universal truth than a computer-generated composition is? Absolute music is the most inherently abstract of all the arts, which would indicate to me that someone who values it on its own terms should be utterly unconcerned with whether an evocative piece of absolute music was created by a person or by a computer.
posted by invitapriore at 12:57 AM on February 25, 2010 [1 favorite]

I should think that one very useful application for this technology would be to provide rolling content for videogames: even the best of them suffer from the inevitable repetition of themes and motifs and whatnot, but here tailored music could be generated on the fly that matches the player's experience. I could also see an industry of music analysts opening, their jobs to determine what makes say a Jerry Goldsmith film score so 'Jerry Goldsmith' and to apply these results to this kind of software. With the official sanction of the composer, of course (or their estate if dead), it might be possible to have a video game 'scored' by Goldsmith, or Alex North or indeed any present or past writer.

Of course, the artistic ramifications of this are complex, but as time goes on I think many creative occupations will be faced with the same kind of new-ludd issues.
posted by specialbrew at 3:13 AM on February 25, 2010 [1 favorite]

I especially liked the exploration of the nature of creativity, or rather the idea that it does not actually exist in the form that people seem to commonly think it does. I forget which author it was (perhaps Isaac Asimov) who wrote the short story about the composer who was never allowed to hear music, but it makes me wonder if he would be able to make music as we know it at all.
posted by tehloki at 3:41 AM on February 25, 2010

I used to play with a Cycling74 program called M that outputs something like a very primitive version of what is in the links. M lets you mess with patterns in a way that feels like composition (but isn't) and adds a bit of musically-pleasant stochasticity to spice things up as you play.
posted by Blazecock Pileon at 5:39 AM on February 25, 2010

I also think that Cope is (partly) missing the point of music, and more generally, art. He is conceiving of art as something that is created by an artist in isolation, and likewise enjoyed by an audience in isolation. What Cope's missing is the notion of art as a conversation (albeit a bit one-sided, in most cases) between artist and audience.

You beautifully stated what I believe is the crux of most debates about art. I don't agree with you that Cope misses the point about music. But I do agree with you that he missed the point of how a particular type of person (Douglas Hofstadter, you, etc...) interacts with music (and other art forms).

There are, broadly speaking, two types of people (in terms of interaction-with art), and I wish there were names for these types:

The Copes: they turn to art chiefly for sensation. As such, it's irrelevant whether the "art" is a sunset, a computer-generated image or a painting by a human, as long as whatever it is sparks sensations in their brains.

The Hofstadters: they enjoy art as a conversation between the artist and the audience.

Of course, many people are a mix of these types, but I've found that most people lean heavily towards one or the other (I'm a Cope), and at a certain point, the types have a very hard time understanding each other -- to the extent that Liberals and Conservatives have a hard time understanding each other. (And the two types tend to use the same sort of emotional, angry, fearful language that Liberals and Conservatives uses when they confront each other!)

You didn't use angry language, but you stated something about art (that it's a communication between the artist and his audience) as if it's a truism. And I'm sure you feel that way. To me, it's not a truism. But when I say that, I'm saying something that is incomprehensible to many people (Hoftstadters). I'm sure there are folks who will read what I wrote above and say (as if it's a fact), "No, a sunset is NOT art. Art is something created by a person." Well, it is or it isn't, depending on how you define art.

As a classical director, I'm endlessly lured into Shakespeare-authorship discussions. Though I find the subject mildly interesting from a historical point-of-view, I can't understand why it's so important to people. This is because I'm a Cope. To me, "King Lear" is the same play whether it was written by Shakespeare or the Earl of Oxford. If I was halfway through rehearsing a production of it, and someone suddenly proved to me that it wasn't written by Shakespeare, I can't see how that would change my approach to the play in any way, because my approach isn't based on caring that it was written by him in the first place.

To a Hofstadter, that's absurd. It's like saying "I love you" is the same sentence if it comes from your spouse or a stranger on the street. A Hofstadter is going to be confused as to why this distinction doesn't matter to me when it comes to art.

For years, I was baffled by this conflict. I remember when I first read about the Intentional Fallacy (the idea that people falsely believe an artist's intentions has a baring on how the art object is perceived). It rang so instantly and deeply true to me that I was amazed when others didn't agree with it.

If Shakespeare came back from the dead and said, "You're directing my play incorrectly, what I intended was...." I wouldn't care. I am not directing HIS play. I am directing the words that I found on some pages. His intentions aren't among those words. (In truth, I might care, but this would be for social reasons -- e.g. a desire to not offend the artist -- not for aesthetic reasons. And, if I gave into them, I would consider myself a bad direction, valuing my desire to avoid awkwardness over my duty to my art form.)

Hofstadters will understandably outraged by my stance. To them, it's like saying, it doesn't matter if the person who bumped into you did it on purpose or by accident. It's just the act itself that counts, and you should chastise both mean and clumsy people the same way.

I have come to the conclusion that there's no right or wrong here. And there's no point in arguing about it, because the issue will never be resolved. It's like arguing over whether we should allow just men or just women to exist in the world. As long as the human race exists, we're GOING to have both genders. And I suspect we're always going to have Copes and Hofstadters.

If you want to have intelligent conversations about art, it's important to understand that these two types of people exist, and that they have specific needs that they must meet in order to enjoy what they're watching or listening to.

In the end, I don't think there's a problem (aside from a personality problem). Politics will be forever fraught, because it centers around limited resources, e.g. there's only one Earth. If there were two, we could make policy to protect the one's environment and let the other's go to hell.

But there's plenty of art to go around. People will always make art. Perhaps in the near future, machines will also start making lots of art. There's no limit on making art. The fact that a computer makes it does not stop me from making it. If both humans and machines are making art, the Hofstadters of the world can steer clear of machine art. Meanwhile, we Copes can enjoy art that comes from the both the "soulful" and the "soulless," because we don't care either way -- as-long-as it's aesthetically pleasing.
posted by grumblebee at 5:39 AM on February 25, 2010 [8 favorites]

Requires Quicktime to play samples? Website fail.

Also: obligatory link to Yamaha's singing robot.
Maybe these two could form a supergroup?
posted by tybeet at 6:04 AM on February 25, 2010

Cope has a program that can compose on the level of roughly a second-year undergraduate in composition. I'd say that's pretty impressive. I conceive of music exactly the way Cope does - it is a system of rules about how to put together sounds, and further rules about how those rules can be broken. It is absolutely art. We humans are soulless machines too - he's just gotten another soulless machine to create music. The real difference between the computer and the human composer is that the computer, unfortunately, cannot enjoy the music it has created.
posted by Philosopher Dirtbike at 7:00 AM on February 25, 2010

What Cope did here is take a step back from the traditional creative process, at least when it comes to music, and created a creator of it. Until recently (with the invention of computers and, more specifically, the microprocessor), one created music by either actively playing an instrument or interpreting the sounds of those instruments into a written language that players could read and understand. Cope's Emily Howell is an engineered piece of computer programming art whose function is to take mathematical computation and interpret it into musical notation. Essentially, Cope created an artificial maestro and, by doing so, is the maestro himself.

A man uses an axe to cut down a tree and no one would claim the axe cut down the tree--they would say the man did.
posted by Tlery at 7:25 AM on February 25, 2010

Were I a composer, I'd be less interested in the program than in the rules and meta-rules that Cope has discovered and described. I wouldn't feel threatened by them. I would want them in my toolbox!

Like Cope, I get frustrated when people describe facets of a work as vibrant, colorful, powerful or sad. I understand that's the effect, but I want to know what caused the effect and if there are any universal rules we can extract.

Since I don't believe in magic, I am sure the rules exist. Uncovering them is hard, and it's even harder to get people to collaborate with you on uncovering them, since most people are turned off by examining art under a microscope.
posted by grumblebee at 8:09 AM on February 25, 2010

grumblebee; you talk about this as though it absolutely has to be this way, like it's some fundamental thing in human nature and not a culturally bound, historically contingent thing. This kind of splitting apart into groups of artists and audiences often comes down to an "I like to feel it" (sensation) vs. "I like to think about it" (conversation) dichotomy. While certainly a lot of people approach art (and lots of other things) using this dichotomy as their main analytical tool, and in a way you're just telling it how it is, I don't think it has to be like this.

Further, I think it's a problem. If, on the level of a whole culture, we consider thinking and feeling to be basically opposed, we're doing something wrong.
posted by avianism at 8:15 AM on February 25, 2010

Wow, i thought i just didn't know hot to play piano, but I'm apparently a super musical robot.

I don't get it. How are these good or modern?

Was there any debate about whether Mozart and Beethoven were mathematicians?
posted by cmoj at 8:28 AM on February 25, 2010

avianism, I'm not really sure what you mean by "it doesn't have to be this way." In my experience, the split exists. I don't know if it's culturally or biologically produced (or influenced by both nature and nurture). What makes you so sure it's cultural? Do you know of cultures where this split doesn't exist? Even if it is "just" cultural, what would you do to stop it?

We differ on the feeling/thinking thing. I don't think the split has anything to do with that. Both Copes and Hofstadters can approach art emotionally and still feel strongly about their camp's position.

I approach art emotionally, and I'm 100% in the Cope camp. I have many friends who approach art like Hofstadter and they are also feeling more than they are thinking. (I tent to be more compatable with people who "feel" art more than they "think" it, so I seek out friends like that. However, there is no commonality amongst my friends when it comes to which side of the Cope/Hofstadters split.)

It may be connected to the introvert/extrovert thing. I'm an introvert. My life is largely inner. Many of my friends -- the ones that think of art as communication -- are extroverts. I can see how it must be appealing to extroverts to interpret art as coming from another human. My wife, an extrovert, doesn't really enjoy art unless she can discuss it with someone. I like discussing art, but I can enjoy it even if I see it and don't know anyone else who has seen it.

If my theory is correct -- that the way you interact with art is connected to whether you're introverted or extroverted (more than it is to whether you approach it intellectually, emotionally or both) -- then to get rid of the split, you'd have to make everyone extroverts. Or introverts. Good luck with that!

My guess is that I'm partly correct. I'm sure there are exceptions. There are probably introverts who are Hofstadters and extroverts who are Copes. So I don't insist on that theory. But I am sure that there are plenty of people in both camps who are emotional readers/viewers/listeners.

Since I'm not an intellectual art consumer, I am less sure that such people can be in both camps, but I don't see any reason why they can't. Jane may be interested in the themes and ideas in "King Lear." If you tell her Shakespeare didn't write it, she may say, "I don't care who wrote it. The ideas in it are the same either way."

Bill might say, "I am equally interested in the ideas, but I care about WHOSE ideas they are. Knowing the person allows you to tie the ideas of that one work into the framework of the they author's general intellectual stance!"

How would you get Bill and Jane on the same page? It's not that they don't understand each other. It's that they don't have a common interest in terms of how they interact with art.
posted by grumblebee at 8:32 AM on February 25, 2010

How are these good or modern?


I think "modern" is used in that article to mean classical-style music that's composed now. Most record labels are either into older music or contemporary music that's contemporary sounding. It's hard for a contemporary artist with a classical style to get his work recorded.

This is true for most art forms. You can sell a reprint of "Paradise Lost" and you can sell a Stephen King novel. It's much harder to sell something written now that is in the style of "Paradise Lost." Most readers either want to read classics or they want to read modern novels. The audience for contemporary-but-archaic-sounding prose is incredibly small.
posted by grumblebee at 8:38 AM on February 25, 2010

@grumblebee

Great points, especially about Shakespeare authorship. Overall, I agree with you, although I don't feel that these ways of looking at art are at all mutually exclusive. I like to think that I appreciate art for its own sake, and as an expression of its creator. For instance, imagine that, while exploring another planet, we come across stone figures that look very much like beautiful sculptures. If we later learned that these were actually natural formations, we would still be amazed, but in a very different way. I think that best describes where I stand between these two categories of appreciation.

Another interesting question is whether art must be intentional to be considered art. Even here, I think there is some room for both perspectives. For instance, a great work often represents real thoughts and feelings of an artist without his or her conscious authorship. But of course, some intentionality has to be there; I wouldn't call a sunset "art." I think it's possible to differentiate between art, and the broader category of beauty.

That said, if these compositions are art, who is the artist? Is it the computer, or is it Cope? Both answers are problematic. If we attribute the work to Cope, then perhaps we could consider Shakespeare's mother to be a de facto artist. If we attribute it to the program, then we are saying that intention is unimportant, in which case you could consider a sunset to be art.

Or maybe this just shows how pointless some of these categories are. It's easier to say that we simply appreciate beautiful creations, and part of that appreciation may or may not be tied up in the question of the creator's intent, depending on your inclination. Most importantly, I don't think that Cope's program represents any kind of threat to human-generated art. They can coexist, and it's perfectly legitimate to view art as a communication between audience and artist, which means that computer generated beauty will have its own, partlially distinct, audience.
posted by Edgewise at 11:09 AM on February 25, 2010

Cool stuff.

But notice, the music files are a recording of a performance on a single instrument. That instrument was tuned and set for the program.

Now write me a program that can boot up Logic or Ableton, load some soft synths and program those synths - matching with well done effects, bring in samples that fit - cut and slice those as needed, lay down a track, and then mix it. ..well, then I will wonder where things are going.

Still, computer generated music can be a great creation tool for a larger track.
posted by pwedza at 3:25 AM on February 26, 2010

This seems to go hand in hand with determinism, and I kind of like that.

Music and art and etc are supposed to tell stories, show emotions and such, but unless you are trained to the type of emotions and stories they exhibit, it seems like it can be easy to misinterpret them? Like we have defined certain chords, certain colors to mean one thing, and others another. Without context, this music is indistinguishable from human-created music, and with context, well, it's context that humanity has given it. I'm not sure if that's still context or not.
posted by rubah at 9:20 PM on March 2, 2010

Cope has a program that can compose on the level of roughly a second-year undergraduate in composition.

I think this is the same sort of thing as the computer/person chess tournaments where players user computers to enhance their game, and the two of them together are better than either individually.

Here's another, simpler example, which is a bit more human focused. The idea of that plug in is to let a non-trained musician play with musicians by forcing his playing into the correct scale, but it does a little more than that -- it generates cords, randomizes, etc and all of that can be automated by Ableton, and it will also synch multiple instruments together so that you can do simultaneous key changes, etc on all of them which should sound decent, without necessarily having any idea of the theory behind what you're doing.

It won't write music for you, but it makes the process of writing music easier. I think it's great, and I imagine there will be a lot more of it.
posted by empath at 12:26 AM on March 3, 2010