music for robots and humans

Machine Songs (2015)

Three movements for metacreative musical agents and improvising guitarist (Daryl Jahnke). The metacreative system uses over a dozen musebots – intelligent musical agents that communicate via messaging – that remain under my watchful eye (and hand), mainly to constrain the performance within a concert format (the system is designed for ongoing installations). The agents function within an ensemble, communicating their actions and plans, and responding to a organizing agent that generates rhythmic and harmonic structures learned from a corpus. As well as generative audio, three musical robots (Yamaha Disklavier, Karmetik Notomoton and Modulatron) perform the music. The machine generated musical framework is presented to the performer, allowing him to be an active agent within the creation.

An Unnatural Selection (2014)

for Turning Point Ensemble, Disklavier, and Karmetik NotomotoN

A modified artificial life system, where musical ideas are born, are passed on to new generations and evolved, and eventually die out, replaced by new ideas. Initial populations of phrases are created by rules derived from an analysis of exemplar music. It is a generative system, in that each iteration/performance produces new material; as such, the performers are reading music they have never seen before, from iPads. One unique aspect of this work is that the complexity of the music must be balanced with its immediate readability: only musicians of exceptional ability can perform it.

Generative music systems are based in the notion of creating complete works of art, producing both the details (i.e. the notes) as well as the overall structure of the music. While this may seem like a conceptual curiosity for the audience – who would be unaware of whether a piece of notated music was generative or traditionally composed – live generative music can (hopefully) produce the same excitement that improvisation imbues: a sense that the audience is experiencing something that has never before been heard, or will ever be heard again. However, generative music is different than improvisation, in that it produces multiple parts under the control of a single creative mind: a composition, rather than one line within a performance. As such, aspects of artificial intelligence are necessary to control the relationships between the many musical elements within a work such as An Unnatural Selection.

Titles to individual movements are derived using the same method of musical construction as applied to Jane Austen’s Pride and Prejudice.

Performed by Brenda Fedoruk (flute), David Owen (oboe), Francois Houle (clarinet), Ingrid Chiang (bassoon), Marc Destrube (violin), Marcus Takizawa (viola), David Brown (bass), Daniel Tones (percussion), Owen Underhill (conductor)

Performance #1 – Friday May 2nd

Movement One: Imagination is a Way

Movement Two: Much Beauty is Before You

Movement Three: This Truth Universally

Example Score (movement 1)

(click to view PDF)


And One More (2012)

for four percussionists, double bass, and Karmetik NotomotoN

And One More is part of an ongoing series of works which explore the notion of composition through negotiation. In these works, a general compositional environment is created, and multi-agents explore this musical space, negotiating their way towards a musical cohesion. In this case, the goal is to arrive at a mutually agreeable rhythmic polyphony, with negotiated parameters of volume, density, onset placement, and variation, with the added interest of the human percussionist acting as one of the agents: the human’s improvised performance is translated into machine terms, and the agents treat him as a silicon performer.

Performed by Brian Nesselroad, Daniel Tones, Martin Fisk, Timothy van Cleave (percussion), David Brown (bass)

More Than Four (2012)

for mallets, double bass, and Karmetik NotomotoN

This work uses the same system as And One More, albeit using pitch material. Importantly, the agent’s output is translated into musical notation on the spot, and the musicians perform the music live. Although the goal was to have true real-time composition and performance, this was not technically possible; instead, agent performances have been recorded prior to this performance, and a “curator” agent selects from this database of recordings, assembling several movements for this performance based upon musical criteria that it feels relevant.

Performed by Brian Nesselroad, Daniel Tones (marimba), Martin Fisk, Timothy van Cleave (vibraphone), David Brown (bass)

Gradual (2011)

for Violin, Marimba, and Disklavier

Created using evolutionary algorithms to develop a population of motives, initially generated through probabilities. Each population is a single phrase, and can last from two to over twenty beats. Once a series of generations (up to fifty) were derived, a genetic operator was run to determine the best ordering and repetition of individual populations, while another genetic operator was run to combine divergent populations. The evolution of the individual ideas are clearly heard as they evolve, mutate, and combine to form new individuals.

This uses the same system as “One of the Above”, extended to use pitch material. Three populations were evolved from the same initial population, and presented simultaneously. Harmonic change was governed by a separate algorithm used in my previous work (Other, Previously, and In Equilibrio)

Performed by Mark Ferris (violin) and Daniel Tones (marimba)

 Bhatik (2009)

With Curtis Bahn (Esitar/Edilruba), and the MahaDeviBot

The MahaDeviBot is Ajay Kapur’s 12-armed mechanical percussion instrument. A live improvisation, in which I influence my generative software system, Kinetic Engine, during the performance, by turning agents on and off. This forces them active agents to continually realign and reorganise themselves. Kinetic Engine is an intelligent multi-agent virtual percussion ensemble which, in this case, triggers the MahaDeviBot. No rhythms are preprogrammed: all are the result of the agents interacting in realtime.