Deepfake Dances

Deepfake Dances for symphony orchestra.

12 minutes.

Standalone piece derived from Silicon, Movement II.

Instrumentation:

3.3.3.3/4.3.3.1/3perc/timp/synth/harp/strings

Note: Standard doublings apply in addition to 3rd clarinet doubling contrabass clarinet (can be adapted if instrument not available)

Programme Note:

Like much of my recent work, Deepfake Dances is music about the orchestra, about technology, and about what an orchestra might be in the age of rapidly advancing artificial intelligence (AI). It is also music that uses AI – both as an instrument embedded within the orchestra, and as inspiration.

Deepfake Dances is primarily concerned with authenticity and control. AI is used constantly to create fake, or ‘deepfake’, content online - I wanted to explore the idea of fake music, which I find both funny and deeply unsettling. There’s a prominent AI-powered instrument, called DDSP-VST, which was created for this piece in collaboration with Google as part of my PhD. This instrument morphs intelligently between the sounds of the human players – sometimes uncanny, sometimes convincing.

The music itself spins between three deepfake dance styles at the speed of a social media algorithm with a disappearing attention span, but the orchestra, like a marionette, is always beholden to the AI lurking, calculating, in the background. These styles are jazz-like, techno-like, and folk-like. For the folk-like style, I used AI to generate the melody which I then arranged, adapted, and orchestrated.

I wrote this piece as part of a larger orchestral project during my PhD, which examined the possibilities of artificial intelligence as a creative and disruptive tool during the compositional process.