Silicon Bach
Mon, 2 Feb 1998 14:09:58 -0500

This is a slightly dated story (3 months old) but its interesting and a bit
amusing. In a contest of sorts, two competitors, one a human music professor
and the other a computer program, were asked to write Bach-like musical
pieces. An audience then heard three pieces: one composed by the human
professor, one composed by the computer program, and one composed by Bach
himself. The audience concluded that the piece actually written by the human
professor was written by the computer program, the piece actually written by
the computer program was written by Bach, and that the Bach piece was written
by the computer program. An interesting state of affairs... (pardon the size
of the article)

date : Saturday 29 November 1997
source : New York Times
headline: Music by the numbers
reporter: George Johnson

In a low-key, musical version of the match between Garry Kasparov and
chess-playing machine called Deep Blue, a musician at the University of
Oregon competed last month with a computer to compose music in the style
of Johann Sebastian Bach. Dr. Steven Larson, who teaches music theory at
the university, listened anxiously while his wife, the pianist Winifred
Kerner, performed three entries in the contest -- one by Bach, one by
Larson and one by a computer program called EMI, or Experiments in

Larson was disappointed when the audience concluded that his piece - a
simple, engaging form called a two-part invention - was written by the
computer, but he felt somewhat mollified when the listeners decided that
the invention composed by EMI (pronounced "Emmy") was genuine Bach.

"Bach is absolutely one of my favourite composers," Larson said. "My
admiration for his music is deep and cosmic. That people could be duped
a computer program was very disconcerting."

The cybernetic pretender, which was invented by David Cope, a composer
the University of California at Santa Cruz, has been upsetting long-held
assumptions by people in the music world because of its ability to scan
pieces by famous composers, automatically distill some of their essence
and then churn out imitations of the work.

The cognitive scientist and author Dr. Douglas Hofstadter, who presided
over the contest at the University of Oregon in Eugene, calls EMI "the
most thought-provoking project in artificial intelligence that I have
come across."

"EMI forces us to look at great works of art and wonder where they came
from," he said. "Nothing that I have seen in artificial intelligence has
done this so well."

The artificial composer was the focus of a recent symposium, organized
Hofstadter at Stanford University, which included two live concerts of
EMI's music. In addition, a new CD of EMI's compositions, called
"Classical Music Composed by Computer" is being released by Centaur

The collection includes not only artificial Bach inventions but also
imitating a Beethoven sonata, a Chopin mazurka, a Rachmaninoff suite,
Mozart and Stravinsky and a Scott Joplin piano rag. Casual listeners
easily mistake many of the pieces for the real thing.

On a more ambitious level, EMI recently composed a full-scale Mozart
symphony and piano concerto, which were performed in April by the Santa
Cruz Baroque Festival on period instruments.

Linda Burman-Hall, who played the piano solo for the concerto, said, "It
felt a little different than playing a normal Mozart work, but it was
much like a work of the same period. It was certainly in the ball park."

All of this is rather disconcerting to music lovers who believe that the
unique style of a composer springs from a deep well of emotion and
experience, that the creator of a composition is speaking musically in a
very special way.

EMI is not trying to say anything. It has no passions, no memories and
knows nothing about life. Its entire inner world consists of rules for
to sift through past compositions for characteristic patterns of
harmonies and rhythms, and then to recombine them into plausible ( and
sometimes beautiful ) music.

The implication, which many composers and musicians emphatically reject,
is that musical style might just consist of a collection of simple
recipes, that the feelings and intent of the composer may not be so
important after all.

In his Pulitzer Prize-winning book "Godel, Escher, Bach," published in
1979, Hofstadter speculated on whether uplifting music would ever be
composed by an artificially intelligent machine.

A program that could produce music as mesmerizing as the great masters',
he concluded, would require more than simple routines for stringing
together notes. The machine would have to learn what it feels like to be
alive. It "would have to wander around the world on its own," he wrote,
"fighting its way through the maze of life and feeling every moment of
It would have to understand the joy and loneliness of a chilly night
the longing for a cherished hand."

Now he is not quite so sure. "I find myself baffled and troubled by
he said. "The only comfort I could take at this point comes from
that EMI does not generate style on its own. It depends on mimicking
composers, but that is still not all that much comfort. To what extent
music composed of `riffs', as jazz people say? If that is mostly the
then it would mean that, to my absolute devastation, music is much less
than previously thought."

Not all of EMI's illusions are equally beguiling. The longer the program
tries to sustain its masquerade, the more likely it is to stumble. "If I
turn on three seconds of EMI and ask myself, 'What was that?' I would
Bach," Hofstadter said, "but if I leave it on for twenty or thirty
seconds, it does not make sense. It is like listening to random lines
a Keats sonnet. You wonder what was happening to Keats that day. Was he
completely drunk?"

Burman-Hall said that EMI did its best renditions of composers like Bach
or Mozart, whose style is "more Apollinian, restrained, logical and

The program is not so successful, she said, with less predictable
composers like Beethoven or Bach's son Carl Philipp Emanuel Bach. "When
play C.P. Bach," she said, "I hear him having second thoughts, wandering
off and then feeling remorse and coming back to what he said before."
now, anyway, some of these more subtle motifs seem to elude EMI's
recognition routines.

Cope concedes that real Bach, younger or older, is still better than his
program's pseudo-Bach. "EMI produces beautiful music but maybe not
profound music," he said, but then how many people, he asks, can come as
close as EMI does to mimicking the greatest composers of all time?

As he continues to refine the program, he expects the imitations to
improve, incorporating more and more of the trademarks that make up a
composer's musical personality.

As a test of EMI's abilities, Cope likes to play its compositions to
people who do not know that they are hearing music that was written by a
computer. "When they assume that the music is human, they are obviously
moved and speak in the same terms as if it had been composed by Chopin,"
he said. They describe the emotions unleashed by the music and speculate
on what the composer was trying to say.

"When I tell them that there is nothing behind the music but cold hard
machinery doing addition and subtraction," Cope continued, "then they
not admit that they were moved." These experiments have led Cope to
believe that the meaning of a piece of music lies largely in the ear of
the beholder.