Computer improvisation

Can the computer be an improvisation partner? Can it generate musical ideas of its own in real time that aren’t the product of random number generators or nonsensical Markov chains?

In Joel Chadabe‘s “Settings For Spirituals,” he uses pitch-tracking to perform various effects on a recording of a singer: pitch shifting, chorus, reverb. The result is effectively an avant-garde remix. It isn’t exactly my speed, but I like the spirit of the piece – remixing existing recordings is a central pillar of current interactive electronic music. I’m less taken with Chadabe’s 1978 “Solo” for Synclavier controlled by theremin. The idea of dynamically controlling a computer’s compositions is an intriguing one, and I like the science-fictional visual effect of using two giant theremin antennae to control note durations, and to fade instrumental sounds in and out. Chadabe set the Solo system up to intentionally produce unpredictable results, giving the feeling of an improvisational partner. He describes “Solo” as being “like a conversation with a clever friend.” Who wouldn’t want such an experience?

Joel Chadabe performs "Solo"

Much as I admire the process, however, I can’t say the same for the product, which I find baffling. As fascinating as Chadabe’s system may be to interact with, I’m left wondering what my role as listener is. Either Solo’s rule set is totally arbitrary, or it’s structured without reference to any musical convention at all. As a result, I find it impossible to engage with.

Giuseppe Englert considers composition to be an interactive system. This is a statement that warms my jazz-centric heart. Englert elaborates that composition can be thought of as “the realization of a network of interactive processes that engender other, equally interactive processes such that the musician is just a link in the chain of processes.” This is a perfect description of jamming with other musicians, or coming up with a jazz head on the bandstand, though I doubt that this is exactly what Englert had in mind. He discusses his wishes for the ideal interactive algorithm:

The perfect program will think as I think but it will be more original because it will not be constrained by the limitations imposed on me by my memory.

This statement puzzles me. The limitations of memory are one of the key defining factors in how we think. Being conscious of those limitations also imposes some useful constraints on composers, because their listeners have the same listeners. Having listened to some of Englert’s works, it sounds as though he would have benefitted from more constraints, not fewer.

The most ambitious electronic improvisors ever must be the League of Automatic Music Composers, devoted to group improvisation on a network of computers, using heavy doses of randomness. The League’s work is demanding even by avant-garde standards. At the New Music America Festival, the soundboard operator refused to turn up the League’s amplification because he was so appalled by them. When a new music festival board op finds you too difficult, that’s really saying something.

Here’s how Joel Chadabe describes the League’s improvisation:

The group’s interactions… were modeled not on a normal music improvisation group but on a shared, interactive network.

Isn’t any ordinary group of improvising musicians effectively a shared interactive network? And would I feel the need to nitpick if I didn’t find the music so objectionable? Chadabe lauds the League for creating a metaphorical “global village” of networked collaboration. It’s worth observing that an actual global village of networked collaboration has emerged over the past ten years in global hip-hop culture. The collaboration isn’t happening in real time, but as artists comb YouTube and filesharing sites for samples and stylistic trends, ideas from any world city can and do find their way into the streets of any other, and the ears of people living there.

Realtime computer response to improvisation is only as good as the algorithms guiding it. The stylistic rules of bebop or blues are effectively algorithms for producing improvisation within constraints. When high modernists write their own improvising algorithms, do the results produce meaningful statements that engender emotional connection? Are the rules appreciable by the listener, so that there can be an interplay between expectation or surprise? Or will the algorithms be so complex or random as to sound confusing and alienating?

The musical tradition most congenial to intelligent improvisation is jazz. George Lewis‘ attempts to get computers to play jazz with him are more interesting to me than the hardcore avant-garde stuff. His program that generates semi-random walking bass lines, while not truly interactive, is nonetheless inspiring. His later work, where computers respond to improvised playing in real time, is quite a bit more inspiring. And to my pleasant surprise, Lewis’ “Homage To Charles Parker” even has some euphonious passages, after a harrowing few opening minutes of noise.

Jazz per se is not a promising direction for computers; while jazz improvisation algorithms exist, their musical output is nothing you’d want to hear twice. But jazz is a rich source of inspiration for larger ideas about how improvisation can be made to serve musical ends. Jazz is rarely totally spontaneous and almost never completely chaotic. Within mainstream styles, all improvisation happens within tight formal and stylistic constraints. It’s the constraints that make the improvisors’ departures meaningful and intelligible to listeners, and to other musicians. Lewis’ most promising idea is to use jazz instrumental styles over interactive ambient textures. Computers do much better creating timbral hazes than jazz lines, so Lewis plays to their strength.

Bruno Spoerri is another jazz musician trying to get serendipitous improvisation out of the computer. I’m quite interested in hearing his 1986 “Drum Song” and 1987 “Rue du Cherche-MIDI,” in which the computer detects the instrumentalist’s pitches as MIDI values and plays the phrases back with transformed pitch and rhythm patterns. Spoerri reacts to the computer’s reactions, the computer reacts to his reactions, and around and around the recursive foodback loop spins. Spoerri describes his goal:

The important thing for me was to have a partner in the computer who threw balls at me, who gave me a reason to react in a certain way, but who would react in a logical way.

Having had this pleasure with some of my human improvisation partners, I wish Spoerri nothing but success. His process also reminds me somewhat of improvising on a guitar fed through a delay line with long delay time and feedback, interacting with my own looped phrases from a moment before. I wasn’t able to track down the Spoerri pieces mentioned above, but the music I did find is surprisingly accessible and pop-oriented.

Peter Beyls makes both visual art and computer music inspired by the emergent behavior in dynamical systems. This is an extremely exciting field for me, one that I hope to explore in my own work. (What is human improvisation if not emergent behavior of dynamical systems?) Beyls’ music from the 1970s is dreadful and abrasive, but his work has gotten a lot more listenable lately. I’m particularly inspired by his Drake Circus from 2006, a cellular automaton controlling a MIDI guitar synth. While the result is a bit shapeless, it’s clearly recognizable as music, and attractive music at that, with a nice balance between randomness and a sequence of linked patterns. Cellular automata tend to produce chains of repeated and slightly varied ideas that have a recursive quality, traits shared by the best human music. I’ll be following Beyls’ biologically-informed experiments eagerly.

Morton Subotnick spoke quite a bit in seminar about his wish to use electronics to create more interaction between performer and listener, to make the audience more active participants in the concert experience. He also expressed frustration about not being able to get that kind of interaction to happen. Subotnick’s main obstacle was his work’s inaccessibility. Had I been invited to interact with a Subotnick performance as an audience member, I wouldn’t have had the slightest idea what to do; I would have needed some significant instruction beforehand. Getting people involved in the music requires shared expectations and references, and the avant-gardists’ commitment to defying expectation will necessarily defeat their efforts to bring about meaningful interaction with audiences.

In the meantime, the world is becoming filled with exciting interactive electronic music experiences. The most basic one, and to my mind the most important, is dance. Wherever you go in the world, there are young people dancing to completely electronic music (and usually little else.) The burgeoning remix and mashup cultures are more direct forms of interaction, blurring the line between artist and audience member. Music-oriented video games are still in their infancy, but they already combine dance with remixing in ways that promise much wider and deeper possibilities in the future.

It bears remembering that in traditional societies, all music is “interactive” — everyone participates, not just specialists. The modern western concert hall is a peculiar cultural anomaly that inhibits our natural desire to get involved. Anything that brings more people into an active and participatory relationship with music, from karaoke to Garageband, makes me more optimistic for our cultural future.

I find that the computer makes for the best improvisation partner when it processes my playing in simple and mostly predictable ways. I’m a great lover of tempo-synced digital delay, which turns improvisation into a dialog with myself from a few seconds earlier. I’m also enjoying Ableton Live’s beat repeat effect, which gives a semi-random shuffle to what I’m playing. It’s unpredictable enough to be interesting, but still predictable enough to be musical.

One thought on “Computer improvisation

  1. Pingback: Ethan Hein on Computer Improvisation | Vic Dillahay

Comments are closed.