Composers who use crowdsourcing tools to generate musical material might be pushing technological boundaries, but this model of creative sharing is an illusion – and one that reinforces, rather than breaks down, hierarchies.
You don’t have to be Creative Commons founder Lawrence Lessig to argue that no technology is ideologically neutral. Just try listening to the output of composers who piggyback on social media and online task-outsourcing services to help write their work, and you’ll hear the implications that Silicon Valley’s current philosophy of disruptive innovation has – not just for selling and distributing music, but for creating it. Sharing and crowdsourcing services break down tasks to their smallest functions, which are then distributed over global networks to small-time, individual actors. These then fight for task crumbs, while service creators cream off profits, and maintain their control of how users interact with the service and with each other. When music is produced, what happens to ideas of community, or of place? And what happens to context, history, or the idea of a work as a whole?
Since his video for Lux Arumque went viral in 2011, Eric Whitacre has become one of the best known composers working with ‘virtual choirs’. Whitacre trawls the internet for amateur performers who record themselves singing into webcams. He then splices these recordings together with shots of himself making conductor gestures, despite not being present for any of the filming.
Calling Whitacre’s choirs virtual is a misnomer. One space choristers share is virtual – the mediated space of networked software platforms – but this conveniently elides the space in the physical world that each participant depends on in order to take part. People need space to eat and sleep and set up laptops with their microphones and broadband connections.
While Whitacre’s website boasts a list of countries from which people have submitted videos, his finished works remove the sense of shared place central to traditional choirs. These groups might have otherwise been brought together through a social interest – and there, in the rehearsal space, each singer would have accommodated the physical presence of the others. Online, they are unable to communicate, to respond to cues to harmonise or adjust their volume or phrasing. Whitacre controls all relationships within this network, and manipulates them to embellish his own brand.
An even more atomised take on crowdsourced composition uses task outsourcing, rather than sharing, as its model. Bicycle Built For 2000 (2009), by Daniel Massey and Aaron Koblin, used Amazon’s Mechanical Turk online service to recruit 2088 people to record the song “Daisy Bell”. Mechanical Turk is an unregulated marketplace that pits anyone with a network connection (and occasionally a microphone or webcam) in competition to complete small tasks online that computers can’t do yet – everything from answering surveys to reading numbers in blurry images. These Human Intelligence Tasks, or HITs, are paid for in pennies – one suspects their pay floor is linked to the capital cost barriers of an education in English as a second language, and network access in the cheapest parts of the world. It is built with a real contempt for its human users in mind: Amazon’s description of its Turkers as “artificial artificial intelligence” is telling, suggesting humans are only a stopgap until computational power is good enough to do everything, without pesky demands for payment.
Massey and Koblin sent several HIT requests to Turkers to listen to, and record themselves singing, notes from “Daisy Bell”. Workers were not given any information about the larger context of the project. These anonymised piecework requests, taken out of time and out of context, reduced to the level of note and phoneme, were compiled to create the final piece.
The song was chosen because it was used by IBM in 1962 to demonstrate for the first time a computer’s ability to sing using a speech synthesizer. But it is probably best known for its use in 2001: A Space Odyssey, as the song of the soon to be dismantled HAL. The murderous computer sings in an attempt to exploit astronaut Dave’s loneliness, empathy and tendency to anthropomorphise, in order to complete its mission. Bicycle Built For 2000 sounds like HAL in its woozy death throes, but more desperate because the situation is reversed – there is no empathy built into Amazon’s platform for the human Turkers to appeal to. The Turkers, probably desperate for their few pennies, don’t know what they’re doing or why, and it shows.
At least sample culture has the potential to call back ghosts from earlier songs. A well-deployed sample can stand in for the entirety of the song that originated it or for a cultural moment, a sound conjuring images from fashion or film. Samples can be used parodically or ironically; however, these are gestures that assume shared knowledge and agreed meaning between the sampler and the listener. They are on the same level. In a video about their project, Massey and Koblin explained their reasons for choosing “Daisy Bell” to Vimeo audiences. But if these audiences were worth telling, why then recruit people to create the piece without letting them in on the backstory?
Part of the answer might lie with a Silicon Valley attitude that differentiates between so-called creators and users: creators code social platforms and structure networks, while users provide the content, data and ad streams that make any of it valuable. Yet users are considered as a resource, or a pool of people who should be grateful to use these services. It’s old school industrialist thinking that dehumanises all but the top technologists and humanises the corporates that bankroll them – corporates are people, but workers and users are tools, and if they cost anything to maintain, it’s better to automate and bring on their robotic replacements as soon as possible. Beyond Mechanical Turk, think of Amazon’s exploration of delivery drones, or Google’s driverless cars. The creator/user dichotomy could also explain Whitacre’s visual attempt to hide his dependence on third party editing tools by inserting shots of himself emoting into a webcam with a conductor’s baton – as if to ask, who’s really the creator here?
Crowdband, described in a 2013 paper by Mary Pietrowicz et al and published in the proceedings of the First AAAI Conference on Human Computation and Crowdsourcing, is a project that references Massey and Koblin’s work with Mechanical Turk and develops it further. Crowdband systematically breaks down composition into discrete, manipulable modules that could be outsourced, produced in chunks as HITs on Mechanical Turk, then reassembled and aggregated for requestors, based on keywords describing sounds and duration. Meanwhile musicians could use the service to create chance events – using, say, the slippage between description of a HIT and the Turker’s lack of knowledge of the task to generate unexpected effects (like the confused slurring of “Daisy Bell”). Crucially, the people producing the sounds are referred to not as musicians but as workers, opening up the whole creator/user divide once more.
If a system like Crowdband ever becomes commercially successful, leaving aside the question of royalties and payment streams for contributors – either musician/composers or workers – what happens to the wider discourse between musicians? Or between listeners and musicians, and between musicians and history? Especially as so much of this discourse is not spoken or translated into keywords; it’s visceral, dependent on musicians’ physical responses to their instruments and to each other. And how much is lost by using workers as sound-creating tools, with no other value? Even composers working together across continents over Skype will bring their own lifetimes
of ideas and cultural memories to their practice,
along with responses to their working environments;
subtle things like angle of daylight or lag between time zones matter.
It’s unlikely that improvisation, or collaboration, or composition as we know them in all their spasmodic and messy glory are going anywhere soon. But the ceaseless drive to partition, modularise and automate music making– under the guise of profit and efficiency – destroys its basis in lived, shared experience that makes music fundamentally human.