Artificial Music
edited by Detlef Diederichsen and Arno Raffeiner (Music 781.76 Art)
The notion of artificial
intelligence (AI) and machine learning has been in the news a ton lately. First
we had all of those Deep Dream images a few years back, then the latest image
generators like DALL-E and Midjourney have made the news for the last year or
so. Most recently, it seems like practically every news outlet has multiple
articles about potential impacts of the new AI chatbot ChatGPT on society, from
education to employment to disinformation propaganda. Those involved with the
arts have their own concerns about AI, as evidenced by many visual artists
complaining about AI-generated art intentionally done in their unique personal
styles.
Of course music is already being
made with AI, too. There have been experiments with forms of computer-based
generative music back to the 1960s, although those predate our modern
conceptions of AI’s machine learning capabilities. Some artists are exploring
using AI in their own music, and there are already tons of AI-based music apps
out there to play with exhibiting various stages of sophistication. While we’re
not yet seeing pop songs generated entirely by AI hit the charts, the
possibility already exists. And there are some areas of the music market that
will likely become dominated by AI quickly: Billboard reported in March that
the “functional music” market (music that’s generally of an ambient nature and
marketed to folks as background music for things like concentration, relaxation
or sleep) is already seeing lots of AI-generated music. It’s easy to produce
with AI apps, and it’s a profitable market for copyright holders that could and
probably will be easily leveraged by some entity who uploads millions of
AI-generated ambient tracks.
All of this is developing faster
than the pace of most book publication schedules, but on our recent arrivals
shelf, you’ll find collection of essays called Artificial Music. This little publication is kind
of hard to categorize: it’s essentially a little book, but it’s part of what
will ultimately become a 25-volume series called “The New Alphabet,” published
by HKW in Berlin and Spector Books in Leipzig, Germany. All of the little books
in this series are focusing on the variety of new kinds of technologies and
philosophies affecting contemporary life, and likely to contribute to changes
we can barely even anticipate in the coming years. A few volumes relate to
music: we also have Volume 2, “Listen to Lists,” which discusses primarily how
new music streaming technologies are changing patterns of music consumption
worldwide. Each volume is made up of a series of essays from specialists
familiar with the topic at hand. In the case of this “Artificial Music” volume,
contributors include George Lewis, whose “Voyager” piece is one of the earliest
improvisational interactive pieces between live performers and software,
journalist Laura Aha, Professor of Cognitive Science at Indiana University
Douglas Hofstadter, and even a pair of AI apps that were used to supply some
images and text for the book!
The book starts with an
introduction and essay from editor Detlef Diederichsen. The introduction
defines the scope of AI activities in our era as those driven by “machine
learning and neural networks interacting with big data,” which is a pretty
succinct but accurate way of describing the technology as it stands. And the
relationship between this volume and the “Listen to Lists” volume is quickly
mentioned as well: “In the music industry, algorithms are already the norm, using
feedback functions to provide consumers with a range of music increasingly
tailored to their specific needs.” Indeed, AI is likely going to play a
significant role on the listeners’ side of music consumption as time goes on,
but for this volume, the focus will instead be on the creation of music, and
how the capabilities of AI may intersect with the activities of composers and
musicians.
The essay that follows acts as an
introduction to the work of composer David Cope, who started using
machine-learning concepts in the 1980s with his Emmy program, providing the
music of various composers to the software so that it could learn to compose in
the same style as those composers. Even in the 1990s, his system was already
robust enough that skilled listeners could only guess between works made by the
original composers or generated by Emmy between 40 and 60 percent of the time.
It also introduces us to cognitive scientist Douglas R. Hofstadter, who
published a Pulitzer prize-winning book about limitations of artificial
intelligence in 1980. He had predicted that computers would never beat humans
at chess, which of course happened in 1997 with Deep Blue. He had also
predicted that computers wouldn’t be able to create emotionally-charged music,
because they simply don’t have the kinds of life experience that leads to
quality composing. But then he experienced some music composed in the style of
Chopin by Emmy, which he found emotionally moving. His response to this was
poetry, some of which is reproduced in this book under the title “Staring Emmy
Straight in the Eye—And Doing My Best Not to Flinch,” in which he somewhat
humorously acknowledges that experiencing music from Emmy has provoked him to
reconsider the very essence of what it might mean to compose beautiful,
meaningful music.
Laura Aha’s essay discusses the
nature of music, and how AI developments relate to it. It’s a fantastic brief
history of both issues that could serve as an introduction for anyone curious
about where we are today, and where we might go next with all of this. To
summarize, since music is relatively easy to boil down to mathematical
principles and basic rules of engagement, its circumstances create a pretty
optimal environment for machine learning to become very good at making music
that we like. She distinguishes between AI approaches that we’ve seen so far in
classical music and pop music circles: classical composers have mostly used AI
concepts to create self-generative works that sound unusual or surprising,
while pop music composers have focused on reproducing the conditions one finds
among most hit songs. But issues of ownership begin to arise here: Holly
Herndon’s work with creating an AI version of her own voice on her album Proto,
for example, point to us living in a time where machine learning can be
“trained” on a particular artist’s voice and then used to create new music that
sounds just like them. Who owns this music? Who made it? This reminds me of the
lawsuit that’s happening right now between stock photography company Getty Images
and Stability AI, a company that makes the Stable Diffusion app which produces
visual art after being trained on massive amounts of pre-existing images. In a
similar situation in February, the US Copyright Office declared that art used
in the comic book “Zarya of the Dawn” can’t be copyrighted, as it was produced
using the Midjourney image generator, and as such, the images aren’t made by a
particular human. We are certainly living in interesting times!
An essay by indigenous artist Tiara
Roxanne raises fascinating issues around colonization and AI. She thinks these
new technologies, which of course must in some way be extensions of the
dominant culture producing them, have the potential to create and further
sustain forms of “data colonialism” when they are initially programmed in such
a way that certain voices such as those of indigenous people are marginalized.
She raises a number of important points here that I haven’t heard in
discussions about AI before, and this is probably my favorite takeaway from
this book.
Composer George Lewis discusses his
work with using computers as improvisation participants, an art that he’s
worked on since the 1980s with his “Rainbow Family” and “Voyager” pieces. He
reflects on a number of his motivations for working toward having non-human
improvising partners that can generate ideas and adapt them in the moment, and
he further breaks down the act of improvisation into five aspects
(Indeterminacy, agency, analysis, judgement and choice) that can be broadly
applicable among humans improvising amongst ourselves as well. And Zola Jesus
ends the book with a powerful poem that questions the provenance and nature of
art, artificial or otherwise.
As a short book with a lot of
important ideas, I’d highly recommend Artificial Music for anyone who is pondering the
interesting moment we’re living through, and how it might affect our music.
(If you enjoy this, you may also
wish to try Listen to Lists edited by Lina Brion or Audio Culture: Readings in Modern Music edited by
Christopher Cox.)
( publisher’s official Artificial
Music web site )
Recommended
by Scott S.
Polley Music Library
Have you read or listened to this one? What did you think? Did you find this review helpful?
New reviews appear every month on the Staff Recommendations page
of the BookGuide website. You can visit that page to see them all, or
watch them appear here in the BookGuide Blog individually over the
course of the entire month. Click the tag for the reviewer's name to see
more of this reviewer’s recommendations!
Check out this, and all the other great music resources, at the
Polley Music Library,
located on the 2nd floor of the Bennett Martin Public Library at 14th
& "N" St. in downtown Lincoln. You'll find biographies of musicians,
books about music history, instructional books, sheet music, CDs,
music-related magazines, and much more. Also check out
Polley Music Library Picks, the Polley Music Library's e-mail newsletter, and
follow them on Facebook!
Posts
