This work was commissioned specifically for Darius Jones to perform on alto saxophone with Seattle Modern Orchestra. How did writing with Jones in mind as a performer impact the creation of this piece?
I have been an avid follower of Darius Jones’ work since the premiere of his LawNOrder with Seattle Modern Orchestra back in 2019. Jones has the rare ability to transform musical material in a way that is analogous to transitions between different states of matter — just like a superheated gas shifts into plasma state, a simple melodic line in Jones’ music may become a visceral gesture under conditions of high intensity, or ossified structures might slowly melt into amorphous interplays of timbre. This project drew a great deal of inspiration from Jones’ music, but I particularly wanted to set up conditions that could resonate with the transformational aspect of Jones’ performance practice.
Where did the title “Inference Engines” come from?
An inference engine is a component of a neural network that infers new information based on what it has learned. There are three types of inference engines in this piece: a natural neural network (i.e. the musician’s mind) that creates new expressions based on the material at hand, the artificial neural network that generates new notation based on the written score, and another network which creates distortion by running inference on a face recognition network.
What other elements, either musical or non-musical, served as inspiration for this project?
Musical applications of deep learning have been a point of interest for me for several years. One of the first decisions I made about the piece was to use deep learning methods both for musical notation and audio processing. But at the same time, I wanted to make sure that the project would address a particular topic that is often buried under the hype: the complicated ethical issues that surround the field of artificial intelligence.
The audience will hear a distortion effect throughout the piece. It will randomly appear and disappear, revealing itself as an independent component, like sonic artifacts of a damaged equipment. Behind this distortion effect, there is a face recognition network that was trained on a set of photos which disproportionately features white individuals. As a result, the network cannot recognize faces of people of color. When another neural network that can process audio applies inference on the biased “insights” of the facial recognition network, it functions as a distortion component.
In 2020, a face recognition network with similar biases caused Robert Julian-Borchak to be arrested from his home and wrongfully imprisoned when the facial recognition software used by the law enforcement misidentified him. By exhibiting a similar tool in the form of an independently running audio effect, the piece attempts to separate it from the musical context, and seeks to emphasize its essential identity: a tool that carries considerable potential of inflicting harm.
How does the notation relate to the ideas and sound world of the piece?
The idea of inference is closely related to the interpretation of the notation. When one of the neural networks is fed materials from the score, it generates a fluid, constantly changing notation captured in short videos. These video-notations are meant to guide or inspire the performers during the performance. On the other hand, the musicians use the score as a basis for improvisation, rather than as a material to practice/learn in conventional sense. Thus the score serves as a model for both natural and artificial neural networks.
Can you explain what is going on in this gif? Is there a particular significance behind your use of color in the score?
This is from the video part for the drum kit, the colors emphasize different parts of the instrument. The notation stands somewhere between the conventional and graphic notation, and since it changes in time, the time-axis for the conventional notation becomes invalid. Percussionist Bonnie Whiting finds that this situation prompts the performer to catch glimpses of certain events and play them from the short-term memory. I am excited to explore the possibilities with this sort of musical communication, and I am very lucky to have started on this path with insights from Darius, Bonnie, Jordan, and Michael.
Is there anything else you would like to share with us?
I’d like to thank SMO co-artistic directors Julia Tai and Jérémy Jolley for their herculean efforts to keep the new music scene in Seattle – and in the Pacific Northwest at large – vibrant, relevant, and uncompromising.