The Well-Tempered Computer

Posted at 8:00 p.m. PDT Saturday, May 23, 1998
http://www.sjmercury.com/west/docs/composer0524.htm

Do the imitations of Bach and Mozart produced by David Cope and EMI mean
we're going to have to redefine creativity?

Part 1: The well-tempered computer
Part 2: The end of an era?
Part 3: Learning from Chopin
Part 4: The end of music?
Part 5: The creative process

BY JOHN HUBNER
West Magazine Staff Writer


SOMEDAY — and given what this story is about, it may be tomorrow — a
story like this will come with a chip. You will press a button and hear
music and form your own judgments. Since that day isn't quite here, I will
try to describe what I have heard.

I am listening to a CD called ``Bach by Design,'' and my living room does
indeed swell with Bach, somber, majestic, all-powerful Bach.

Except it isn't Bach.

Now we have Mozart, delightful, surprising Mozart, drawing deep from the
well that waters the universe.

Except it isn't Mozart.

And here comes Scott Joplin, pounding out a tingling, syncopated rag. My
living room has the feel of a rent party in Harlem in the years before
World War I.

Except it isn't Joplin.

The music includes large helpings of the three composers. But it also
includes David Cope, a professor of music theory at the University of
California-Santa Cruz. And with Cope comes his collaborator, whom he
affectionately calls ``Emmy.'' Experiments in Musical Intelligence, EMI
(hence ``Emmy''), is a computer program Cope invented that composes music
on a Macintosh.

Last October, EMI participated in a contest at the University of Oregon.
An audience of several hundred music lovers listened while a pianist played
three pieces of music: one work for a keyboard instrument by Bach; one work
for a keyboard written in the style of Bach by Steve Larson, who teaches
music theory at the university; and an EMI invention in the style of Bach.
The challenge put to the audience was to choose which piece was genuine,
and which was written by the computer.

Poor Steve Larson. The audience voted for his piece as the one written by
the computer. Poor Bach. The audience decided that EMI's composition was
the one written by the great composer.

The event was presided over by Douglas Hofstadter, a cognitive scientist
at the University of Indiana who studies computer creativity and is the
author of ``G˙del, Escher, Bach,'' which won a Pulitzer Prize in 1979. The
book explores how mathematician Kurt G˙del, graphic artist M.C. Escher, and
the composer Johann Sebastian Bach discovered patterns in nature that
others missed. Hofstadter also happens to be passionate about music. He was
blown away by a Chopin mazurka EMI wrote that, he said, ``could slide right
into the book of Chopin mazurkas.''

Like everyone else, Hofstadter had always believed that Bach, Mozart and
Chopin were great because, in addition to being great musicians, they had
great heart, great soul, great humanity. When he listens to their work,
Hofstadter told the magazine New Scientist, ``I've always felt I've been
coming into contact with the absolute essence of humanity.''

And now along comes EMI, which, Hofstadter says, ``has no sense of itself,
has no sense of Chopin, has never heard a note of music, has no trace in it
of where I think music comes from. Not a trace.'' And yet EMI produces
music that sounds enough like Chopin or Mozart to trick musically
sophisticated audiences into thinking that what they are hearing is an
original --if minor --piece.

``EMI,'' Hofstadter says, ``is the most thought-provoking project in
artificial intelligence that I have ever come across.''


Part 2: The end of an era?


Deep Blue pounding chess grand master Gary Kasparov was one thing. Chess
is about computing thousands of possible moves and likely responses to each
move, and then picking the right one. Kasparov was playing into Deep Blue's
greatest strength --its brute force, its ability to make hundreds of
thousands of calculations per second.

But a computer composing music? Great composers from Mozart to blues giant
Willie Dixon know everything worth knowing and feel everything there is
to feel and communicate in the most powerful universal language human beings
have invented --music. Don't you need a heart and a soul and ears that can
hear paint peeling to write great music?

``We like to feel that we're the one ascendant being, that we have a curb
on things creative,'' Cope says. ``EMI tends to complicate that view.''

Cope is sitting in his office at home, on a street just below UCSC. There
are three souped-up Power Macs in the room --one of the monitors hangs from
the ceiling, the better for Cope to see it without suffering neck cramps
--and an Emax synthesizer keyboard. Cope has had a remarkable relationship
with the Mac. ``Every time I've needed something, the industry has produced
it at just the right time,'' he says. Also on the ceiling are 100 wind
chimes. Depending on where the breezes are coming from, Cope will open
different windows, thereby orchestrating different sounds.

Cope reaches into a stack of books, pulls out an earmarked copy of
Hofstadter's ``G˙del, Escher, Bach,'' and remembers the night four years
ago that Hofstadter called at 2:30 a.m., his time. He had just read Cope's
1991 book, ``Computers and Musical Style,'' on computers, composing and the
elements that go into creating a musical style. Cope says the first words
out of Hofstadter's mouth were, `` `We've got to talk!' When he heard EMI's
music, he was floored that he could be moved by 20,000 lines of code.

``Doug made two predictions in this book,'' Cope says. ``A computer will
never be able to beat a chess master, and a computer will never be able to
write music that sounds like a work written by a classical composer. It's
already happened and he feels sad about it, as if an era is passing.

``But not me!'' Cope continues. ``I can only speak higher of the human
race for having invented computers that can beat them in chess.''

Cope is 56, and it sells him short to say he has the passion of a younger
man. Talking about EMI, he is intense and completely focused, an artist
exploring a great passion. It is fun, interviewing David Cope --the
conversation takes unexpected turns, visits interesting ideas. Talking
about composing leads to discussions of creativity and anecdotes about
Mozart, ``who is so off the chart creatively, you cannot compare him to
anyone.'' And Beethoven, who ``was dogged. His genius was, in struggling to
create, he could go back over dozens of attempts and pick out exactly the
passage that worked.''

Elements of style

So immersed in EMI is Cope --``You would never want to know another human
being as I know this program --all the ugly details, what she ate for
dinner, what's in her bowels'' --that he considers himself ``a drag.''

``I was probably a drag 25 years ago, too,'' he confesses. ``I was
practicing all the time.'' Cope was born in San Francisco but grew up in
Phoenix --``I had asthma; my parents had to move to save my life.'' A
musician from the time he was a ``tiny tot,'' he plays the piano, the cello
and the bass. He attended Arizona State and, having outgrown his asthma,
did graduate work in music theory and music composition at USC.

``I paid my dues, playing professionally in orchestras all over the
West,'' Cope says with a smile. ``I was pretty good.''

He was also a successful composer of avant-garde music. Works by Cope are
performed widely. ``David has done very interesting things in the world of
new music,'' says Jonathan Berger, a composer and researcher at Stanford's
Center for Computer Research in Music and Acoustics (CCRMA, pronounced
``Karma''). ``He has also written an extraordinary textbook about
20th-century music, `New Directions in Music.' ''

In 1977, Cope accepted a tenured position at UCSC and he and his family
settled in Santa Cruz. His wife, Mary Jane, is a concert pianist and a
piano teacher; they have four sons. Cope was happily teaching, composing
and writing until 1981, when he hit a block.

Composers' block, writers' block, painters' block --barricades are strewed
all along the creative path. Poets, painters, jazz and blues musicians are
known for using self-torture and prodigious amounts of drugs and alcohol to
break through these barriers --hence the popular romantic notion of the
tortured artist. Cope had no idea where his block came from --no one ever
does --but he came up with a different idea for dealing with it.

Cope had been fascinated by computers for years and had taught himself
several computer languages. He wondered if a computer would help him with
his block. What if he wrote a program that could show him his music, that
would be a mirror of his work? He could look at a set of notes he had
written and say to himself, Wait a minute, I went that way in that
composition. What if I took the same starting point and went in this
direction?

Cope learned the computer language LISP and set out to do just that.

``I didn't have a database when I began, just a bunch of rules, which is
typical of the way A.I. [artificial intelligence] people work,'' Cope says.
``Give the computer a bunch of rules and see if it can do what you intend
to do.''

Cope used algorithms, which break down a complex task into a large number
of steps and make use of the computer's greatest strength: processing
numbers. But while Cope was typing endless numbers into a computer in
search of insight and inspiration into his own music, he discovered his
task had broader, more profound implications. What he was really doing,
Cope decided, was using the computer to explore style. Not just his own
style. Musical style.

``Musical style is very important, but in my research I discovered that it
had not been analyzed very much,'' Cope says. ``I thought, `What if I could
use computers to analyze the styles of the great composers? What if I could
discover what made the great composers great? Made Mozart Mozart, Bach
Bach?' I'm not interested in using words like `soul,' that really don't
mean anything or mean different things to different people. I'm interested
in really getting down to it, in trying to discover what Bach really had.''


Part 3: Learning from Chopin


EMI took a composition apart, boiled it down to its essence, then did the
same thing with the next composition Cope fed it, and so on through a
composer's work. Cope's term for this distillation, for the marrow of a
composition, is ``signature.''


``Signatures'' are the building blocks of style. They are ``earmarks, groups
of melodic patterns that occur in similar fashion in works by a composer.
They are what makes Mozart Mozart, and Bach Bach. When we hear them, we
don't necessarily say, `That's a signature!' We say,`That's Chopin!' That's
Bach! That's Mozart!' ''

Cope's love for music is voracious. He has devoured all of Mahler's 10
symphonies, and it pains him there are no more. He has listened to everything
Rachmaninoff wrote so many times, he knows the composer like a lounge
band knows Beatles songs. He began to think,``Wouldn't it be nice to have a
new Rachmaninoff? A new Mozart? Something new from great writers who are no
longer composing--they are decomposing

He stuffed EMI full of Chopin --``The machine had a great teacher'' is the
way Cope puts it. He had EMI analyze two or more work sand distill Chopin's
signatures. Cope then asked the computer to compose a mazurka based on those
signatures. The result was a work that sounded like Chopin, that in away
was Chopin, and in a way wasn't. It was EMI, stringing signatures together.

``EMI can look at anybody's music and siphon off some of the important
elements,''Cope says. ``It can produce an example of Chopin's style, it can
sound like Chopin,but not be the same as any music in the database. It was
tough. I spent a lot of years doing that.''

In the beginning, Cope had EMI spinning out pastiches --perhaps computer
riffs comes closest to it --of Bach, Mozart and Rachmaninoff, entirely for
his amusement. It was,Gee, isn't this interesting? This really does sound
like Rachmaninoff. Fearing he was like a priest who was committing a
sacrilege by manipulating the canon, he waited for more than a year before
he let anyone hear ``any of EMI's imitations of dead composers.

``I played them for a friend, very much in the spirit of `Aren't these
wonderfully funny?'He thought they were wonderful instead of wonderfully
funny and suggested I play them for others.''


Part 4: The end of music?


In 1987, EMI had her debut at the Computer Music Conference in Urbana, Ill.
It is one thing to have a son or a daughter or a favorite student out there,
playing Bach and Mozart. If it doesn't go well, you know how to console
them. But a machine? What do you say when a machine bombs?

``About 150 people sat there, stone silent,'' Cope recalls. ``I was so
apprehensive, I took their silence as a sign of displeasure.'' Later, when Cope
talked to some members of the audience, he discovered that, ``to quote one of
them, they had been `thunderstruck' by the nature and quality of the output.''

But in other listeners, EMI inspired rage. Cope was delivering a paper on his
work at a conference in Cologne, Germany, in 1988, when a German musicologist 
raced up to him and shouted, ``Musik ist tot!'' (``Music is destroyed!'') And 
then he tried to punch Cope in the nose. ``At least he took me seriously,'' 
Cope says.

Criticism of Cope's work isn't always that extreme, but it can get pretty
rough. ``Why listen to ersatz Bach when you can listen to the real Bach?''
asks a friend who is steeped in the classical tradition. ``You have to ask
yourself, Where are Cope's experiments going? I'll tell you where: to Toys R
Us. Cope has come up with a great idea for a kid's game!''

Which leads to the crucial question: How good are the Cope/EMI facsimiles?

``Mozart's 42nd Symphony,'' one of the computer's compositions, doesn't
have to be as good as the real composer's 1st to keep it out of Toys R.I.P. Us.
It doesn't even have to be as good as Mozart's lesser works. It has to be
only fresh and evocative; be new in away that makes us rethink Mozart, that
brings life to great music. That happens to many listeners. But not all:
There are those who find Cope/Amy's music thin, lacking the energy, the
divine spark, of the original.

``David is doing important work with grace, integrity, and often extremely
interesting results,'' says Jonathan Berger, the composer and music
researcher at CCRMA. ``And it is because I find his work interesting and
worthwhile that I am vitriolic in my criticism.

``Cope is really working on the surface details of music, on pitches and
rhythms, so the musical experience is limited,'' continues Berger, who in
the past has called Cope/EMI'smusic ``aphasic.'' Berger is certain there is
far more to musical style than what Cope calls ``signatures.'' There are
things like timbre, which gives music its flavor. There is the architecture
of a piece --the way a composer uses ``signatures,'' the bricks, to
build something monumental. And it is architecture that Berger finds missing
in a Cope/EMIpiece.

``Cope's work does not focus on process and transformation,'' Berger says.
``Rather, it is like fitting pieces of a jigsaw puzzle together.''

Finally, and this may be the most interesting criticism of all, Cope/EMI's
critics say that EMI has no understanding of, or appreciation for, the
listeners. Music is a subtle interplay between the composer and the
audience. Since a machine has no conception of an audience, it cannot begin
to grasp this intricate, fascinating relationship.

``The answers are not in trying to model composition,'' Jonathan Berger
says. ``It is in understanding how listeners receive music, how they build
context and expectations, and how those expectations are realized or
manipulated, that really gives us a definition of genius and creativity.
Whenever we build expectations, we are composing before we listen. Music is
a combination of the genius of a composer and the creativity of a listener.''
Jim Aikin, who writes a column on computer music called ``Other Windows''
for Keyboard magazine, brings Cope/EMI back to the words Cope loves to hate
--``soul, heart'' --but words that may be, nonetheless, unavoidable when
discussing art. In a September 1993 column about Cope, EMI and computer
music, Aikin wrote, ``The danger is that what an artist communicates, as one
human being to another, can drop out of the equation. The machine. . . has
no spirit, no soul, no mystery. And it's the mystery that makes art what it
is. Once we turn over the creative process to a purely logical, emotionless,
silicon-based muse, we may as well go back to painting houses, waiting
tables, driving cabs.''

Cope seems to thrive on this dialogue with his critics. ``Anger, as long as
it is intelligent anger, opens people up to the possibilities of thinking
about music in different ways,''he says. ``It opens them up to the fact
that there are things possible in the universe besides what they had
imagined.''

Cope originally developed EMI to get through his composer's block, but he
found, somewhat to his chagrin, that ``people were far more interested in
EMI's imitations of the great composers than they were in EMI and Cope.'' At
the same time he is working on his own compositions, Cope/EMI are producing
works like ``Mozart's 42nd Symphony.'' (Though there is a debate among
scholars over how many symphonies Mozart wrote, he has traditionally been
credited with 41, the last being the ``Jupiter'' symphony.)

Last spring, professional musicians from all over the country came to UCSC
to play``Mozart's 42nd Symphony.'' Cope asked the audience for reactions
afterward, and generally, they were transported and intrigued. Except for
one elderly gentleman who, Cope recalls, ``stood up and said nothing he had
heard sounded like Mozart. The thing is, he came in prepared to hate what he
heard. He had already made up his mind.''

Again, it is important to point out that what the audience heard that night
was a symphony composed by a computer. Cope doesn't tinker with a
composition after EMI spits it out; musicians play the music just as EMI
wrote it. And that is what distinguishes Cope/EMI from techno and trance and
rave and house, which are different names for music generated by computers
and played on synthesizers. It also distinguishes Cope from the music that,
for decades, composers working at Stanford's CCRMA have been producing.
Generally, composers working at CCRMA write music that explores the nature
of sound; Cope worked there himself in the late 1970s. But he says,
interesting though it is, it is only a beginning.

``Ninety-nine percent of computer composers are using only 1 percent of the
computer's potential,'' Cope says. ``Using a computer as a musical
instrument to play stuff we've created is all well and good, but that is
human music with the computer as a sidekick. What I envision as computer
music is using the computer to compose music in a way that is believable and
listenable.''

There are composers at CCRMA who say this distinction is too simple. They
claim that they, too, are working with algorithms and are trying to produce
music that is accessible, that sounds and feels like music to the human ear.
But letting that go for the moment, let's get to the big questions: Can a
machine create?


Part 5: The creative process


``For me, it is important to define intelligence before you define
creativity,'' Cope answers.

To Cope, intelligence has three parts: analyzing, associating, adapting. On
a camping trip, you throw a rock at a small black-and-white animal. You
get sprayed. You analyze what happened --threw rock, got sprayed.
Two years later, you go camping again. You see another black-and-white
animal. You associate what happened before with what is happening now.
You back up slowly, away from the cute little animal. You have adapted.
``Being creative is attempting to make associations that are not so clear,
that require a leap of faith to make,''Cope explains. ``A genius is someone
who is damn good at doing that.''

Creative genius, then, is Michelangelo seeing a finished sculpture in a
block of raw stone. It is Picasso playfully turning bicycle handlebars into
a bull's horns. Can a computer do this?

``No. The machine is creative only to the extent I've allowed it to be
creative,'' Cope replies. ``It can analyze music faster and more thoroughly
and with more data than human beings could do if they were given a hundred
times as much time. Can it associate? Yes, it does, precisely as I tell it
to. But it doesn't adapt. So I don't really consider EMI intelligent.''

So what is the value of all this? Cope remains convinced that
deconstructing music, breaking a composition into a composer's signatures,
is one way to unlock the mysteries of musical style. He believes that
listening to an EMI facsimile of Bach is a great way to discover what makes
Bach Bach.

``As convincing as an EMI imitation is at times, it is still not Bach,''
Cope says. ``That makes us think, makes us ask ourselves, `Why is Bach so
much better?' That's really healthy.''

EMI has brought Cope face to face with Bach and Mozart, brought him much
closer than he could ever have come, listening to, studying and writing
about their music. Cope has learned how they wrote music --he believes he
has discovered how the creative process works --and it does not work the way
you may think.

Remember the poster of Beethoven with lightning striking his wonderful mass
of unruly hair? Cope thinks the idea that the genius composer is a lightning
rod for the heavens, that inspiration comes in bolts from the beyond, is the
exact opposite of how creativity works.

``Picasso said, `Good painters borrow, great painters steal,' and there is
no way around that,'' Cope says. ``There is no such thing as an entirely
original composition. When we really go in and look at musical signatures,
we see that Beethoven, Mozart, they all borrowed from other composers, some
in small ways, some in ways that are strikingly large. What we thought of as
original compositions we can now begin to see as patchwork quilts.

Understanding this process tells us what a composer listened to, what he or
she liked, and clues us in to what they may be trying to tell us. Music
turns out to be richer and deeper than we ever imagined it to be.''


A roll of the dice

Cope stands up, slaps on his Irish walking hat, and prepares to leave. He
has classes to teach and hours to spend with EMI. But I have one more
question: What, I wonder, would Bach or Mozart think of what he is doing
with their canons?

Cope smiles, thinks for a moment, and says he doesn't know. Then he smiles
again and says that after reading voluminously in their letters, he does
indeed have ``a feeling'' about what Mozart would think. Mozart, he
explains, loved to play a game called``Musikalisches Wurfelspiel,'' or
musical dice. He and other 18th-century composers wrote music to the throw
of dice.

It worked like this: Mozart would write measures that could be arranged
randomly in trillions of logical ways. Then he threw the dice to determine
the order. If he got a 11 on the first throw, the 11th measure went first.
And so on and so on.

``Mozart was already embodied in a kind of digital world back then,'' Cope
says. ``If he were here today, he'd have already done what I've done, only
it would be six times better. Mozart would be a real geek in the 20th
century.''


JOHN HUBNER is a staff writer for West. His e-mail address is
jhubner@sjmercury.com .