Archived Futures: Digging In The Crates Of Always

By Gary Charles

It wasn’t just about the stories themselves, though. It was about how memory was forced on the future. A better way to put it was this: history was also about the methods used to store it and the tools used to narrate it.

(Chude-Sokei n.pag.)

In a recent short story/essay, Anarchic Artificial Intelligence, Louis Chude-Sokei considers the role of memory and history in formulating conceptions of futurity. Through the lens of emerging artificial intelligence (AI), Chude-Sokei illustrates how representations of potential futures remain tethered to our interactions with the past, retaining the power to reproduce and reinforce existing inequalities and power structures. In previous work, Chude-Sokei has drawn on depictions of robots and automata in cultural works to illustrate the gendered and racialized understandings of artificial life within Western conceptions of modernity. Using Caribbean sound cultures as a central reference, he charts the complex inter-relationship between machinic innovation and human power relations, illuminating porous borders between the human and non/in-human. Cultural production, and more specifically the production of sound, is centred as a vital (but not exclusive) sign of both humanity and intelligence. “They communicated in codes so powerful that their masters heard something like intelligence in the music they made” (Chude-Sokei n.pag.).

Analyzing the contours of perceived time through cultural production provides rich ground for thinkers and writers. Theorist Mark Fisher draws on the work of Franco ‘Bifo’ Berardi in conjuring a landscape of flattening cultural time; a seemingly permanent present, along with an associated inability to produce seismic cultural shifts. Fisher draws on examples from literature and film, but it is within music cultures he finds the most compelling evidence. Even within the supposedly future facing world of electronic music, Fisher suggests we are locked in a loop of repetition, generating only minor incremental advancements over time, contending that “the very sense of future shock has disappeared” (Fisher 7). In this paper, I will consider how hyped technological advancements in AI serve to potentially intensify this repetition, reproducing not only cultural stasis but serving to entrench power relations and biases, leading us into what Robin James calls the “Age of Statistical Reproduction” (James 140). James equates repetition with Neoliberal hegemonic power, where probabilistic computation represents capital and its interests, eroding potentiality and devouring futurity.

Fisher’s sense of flattened time and diminished novelty has resonated beyond academia and critical thought. Popular manifestations of this understanding can be seen in memetic social media commentary reflecting on our perception of time through the lens of popular culture. Favourite examples include a wistful Will Smith contemplating the shifting notion of 30 years in youth culture, and a recent tweet that alludes to non-linear paths of progress, implying curbed evolution during recent years: “In the film Back to the Future Marty plays a song he considers an oldie: Chuck Berrys Johnny B. Goode. A similarly aged song today would be Smack My Bitch Up by The Prodigy” (@tametick). These instinctive reactions to the experienced passage of time lend empirical credence to Fisher’s contention that (in a cultural sense) time is stalling or thinning, that everything has been done before, with a “feeling of belatedness, of living after the goldrush” (Fisher 8). Music writer Simon Reynolds, whose book Retromania covers similar territory, charts the compulsive recycling of nostalgic material and textures by artists and musicians, and emerging retro fascinations of so-called hipster cultures. Reynolds quotes the novelist JG Ballard: “Everything happened in the 60’s, it was like a huge amusement park out of control. And I thought: well, there’s no point writing about the future, the future’s here” (410). For both Reynolds and Fisher, the dawn of the new millennium heralded less newness, an encrusting of categories, along with autonomic re-collaging of past musical forms and aesthetics. The first decades of the millennium are seen as exemplifying music culture stuck in a loop, repeatedly regurgitating the past into a never changing present. During the same period however, rapid, and far-reaching changes have occurred in the infrastructures that support music cultures; in technologies associated with distribution and consumption, and in financial and capital structures that reward artists and populate the industry. A prominent development is the transition to digital streaming, along with the datafication of curation and recommendation.

Today it is possible to instantly access entire archives of recorded music through streaming platforms. For the present-day listener, all eras appear equidistant and all temporalities, always available, immediately. Whether this unbridled access opens up new potentials for creativity, or further traps artists in recursive loops, is open to debate. Certainly, the quasi-monopoly held by the streaming platforms creates an imprint on shifting tastes, definitions of style, categories, and even methods of creation. Liz Pelly has worked extensively on uncovering biases and corporate interests embedded in both the recommendation and delivery mechanisms of Spotify. In her lecture “Music, Power and Platforms” at CTM 2019, Pelly highlights the way in which ‘Spotify For Artists’ purports to use proprietary data to ‘assist’ artists in making decisions about what to create and how to define themselves—in other words, how to modify artistic practice to best fit Spotify’s matrix. She goes on to discuss production houses commissioning work with the sole purpose of deconstructing Spotify’s curation algorithm to compose music with the highest probability of being ‘seen’ by the algorithm’s aural gaze. At this point she notes that “this is not music culture, it’s platform culture” (Pelly 31:16).

This concern is equally prescient when it comes to considering the much-hyped development of AI in music composition software. The protocols upon which the streaming platforms are based rely on categorization and labelling; a hidden topology, mapped taxonomies of what music is, calcifying styles, genres, affects and histories, treating music as settled and quantifiable. These logics driving curation are also embedded in models underpinning AI music creation. Neural Nets trawl archives, spewing infinite potential streams of music, probabilistically rendered against the vectors of an already established map. Data sets and protocols are presented as universal, apolitical, neutral representations of music, extending the idea of music as a Universal Language, with normative features and rules, applicable across cultures, localities and timespans. However, this notion negates situated understandings and side-steps cultural memory, illustrating the entrenchment of hegemonic Western-centred scientistic ideas of cultural understanding. In Whiteness and the Ontological Turn in Sound Studies, Marie Thompson looks at understandings of sound and music posited in new materialism and object-oriented ontologies. Thompson identifies and describes a modest ‘white aurality’ that underpins canonical works and pedagogy in Sound Studies and Western music academia – “a racialized perceptual standpoint that is both situated and universalizing” (Thompson 266).  However, the attraction to the idea of music’s universality is a powerful one, found in both public consciousness and works of art and literature. In The Dispossessed, Ursula Le Guin conjures two distinct worlds with divergent, unique sets of social structures, new understandings of gender relations, and even novel conceptions of physics and time. Yet, within these vividly imagined worlds, one of the primary commonalities that remains is the law of musical harmony. The protagonist, a physicist named Shevek, encounters a church while visiting the unfamiliar ‘m/other’ planet:

Shevek listened. Somebody was practicing the Numerical Harmonies on the Chapel harmonium. They were as familiar to Shevek as to any Urrasti. Odo had not tried to renew the basic relationship of music when she renewed the relationship of men. She had always respected the necessary. The settlers of Anarres had left the laws of man behind, but had brought the laws of harmony along.

(Le Guin 71)

While music does not play a central role in the story, this example demonstrates a common perception that music is inherently mathematical and universal. This idea of music as mathematical, with fixed statistical relations (Western classical harmony) acts as the foundation for the belief that music represents the most likely frontier for machine creativity. Behind Le Guin’s example lies the assumption that music is a solvable ‘problem’ with mathematical dimensions and calculable vectors. With a large enough corpus of example material, with sufficient data, these vectors can be simulated or re-aggregated, statistically reproduced. Problem solved.

This approach ignores social and political realities, negating Thompson’s situated cultures of reception. As Hito Steyerl points out in her discussion on algorithmic problem-solving, our historic deployment of technology tends to be highly effective when solving engineering challenges, yet less successful in complex social, cultural and political areas. In centring the commodity (music as noun) and ignoring the multiplicity of social and political activities (music as verb), a zombie machine is created, reproducing decontextualized music as hyperinflationary currency. This computational approach understands music as combinations of formal qualities that in aggregation can adequately explain the ‘object’, static and cast, removed from intentions, locality, and temporality. Music is reduced to notated, coded data, archived in proscribed statistical relations. In this domain, a string quartet notated in the 18th century is coded into the same language as ancient Inuit Throat Singing or a throbbing Gqom beat penned last week in a Kwazulu bedroom. To be machine readable, the entire archive is flattened and compressed into a retrievable symbolic, mathematical map. In digital media terms, this is a deeply lossy format. This reductive logic absents social and cultural memory, the interconnections, and underpinnings, of music as both verb and noun. It ignores the creative process itself, the multitudinous acts of creating and improvising in commune (or isolation), the learning, the interaction, the seamless flow and accidental collisions of conjuring sound, what Christopher Small calls “musicking” (1998). Taken to its logical end, if a comprehensive training set can describe music, then progression, evolution and novelty are curtailed. Interactions and movements negate models; multitudinous agents, acting both independently and socially, with ever-changing capabilities, desires, and intentions. These movements evade the algorithmic gaze, what Pasquinelli/Joler call The Undetection Of The New (2020).So, for a set of technologies so definitively hyped as futuristic, a primary function seems to involve reifying the complexity of the past—a data-driven hyperextension of a colonial instinct already present in Western Music Theory and Eurocentric music academia. 

Robin James, when attempting the tricky definition of difference between sound and music, contends that there are no absolute objective differences, and that embedded notions of difference are “institutional” in nature (James 37:30-40:44). She contends that the delineation and articulation of difference tells us more about institutions than the phenomena themselves. Western academic funding and pedagogic structures remain largely centered around a distinction between ‘art music’ and ‘popular music’. This distinction is echoed when assessing state-backed cultural spending and programme curation. This longstanding high versus low-culture debate resonates with racialized understandings of value within the arts. Western music academia (especially at the ‘prestigious’ end) has been slow in shifting from these distinctions. In New Music Decolonization in Eight Difficult Steps, George E Lewis demonstrates that contemporary art institutions and curators have been more forward looking than those in music, cultivating a “network in which New York, Lagos, London, Cape Town, and Basel were more or less equally important to a contemporary canon” (Lewis n.pag.). He laments slower progression in music institutions, the academy, and curatorial endeavours however, and sets out a plan for the decolonization of new music cultures. Revealingly, he calls for an abandonment of meritocracy (that there is such a thing as ‘a good composer’), investment in diverse practices, and cultivation of new consciousness that re-arranges the entrenched institutional definitions of music and musicking.

Yet, these same institutional conceptions are still utilised to populate the ontological register of AI training sets. These taxonomies are explicit in Google’s Audioset Ontology, a dataset of labelled/categorized audio events, including music divided up into genre, mood, concept and instrument. Underlying categories reflect a Western institutional conception of musical identity. However, we are discussing one of the largest corporate monopolies known to humanity, with unbridled global reach, already endowed with the power to shape definitions, consciousness, and the archeology of knowledge. Embedding classifications in the fundamental code from which future technologies may emanate represents a ‘hyper-flattening’, encrusting already problematic categories and conceptions into our understanding of cultural production. This represents a crystalline example of Chude-Sokei’s and James’s warning that machinic, automated ‘creativity’ results in the reproduction of historic power structures and inequalities—a hypercapitalised Autofac. Google’s high-profile removal of Timnit Gebru from its AI Ethics team is an example of how it has failed to challenge its approach to music. Gebru represented a critical voice within AI research, highlighting biases embedded in current practices and models. Gebru co-authored a paper looking at language models, On The Dangers of Stochastic Parrots (2021), demonstrating that greater volumes of data do not necessarily equate to diversity, and that finding patterns in static pools of data negates shifting social and political circumstance. These are similar concerns highlighted here within AI music processing and generation. And while music taxonomies may not contain the same potential for real world impacts as other algorithmic injustices, when considering the formulation of Audioset Ontology, and casting a glance at the authors and contributors to the project, it is hard not to agree with Gebru’s point.

Therein lies the paradox: by attempting to compile an objective database, a neutral meritocracy, what is highlighted is the situated understanding of the subject, identity. However, perhaps this points to a futurity that is less flat, more plural, where the outcrops manifest Lewis’s call for investment in multiple cultures and practices. In a recent article by Matt Bluemink following the tragic passing of SOPHIE, the artist’s work is deemed Anti-Hauntological, a rich example that challenges Fisher’s conception of lost futures. Bluemink posits that “regardless of our personal opinion of her music, it’s hard to imagine that she would fail to induce ‘future shock’ in listeners from 20 years ago” (Bluemink n.pag.). SOPHIE’s solo and collaborative work garners reception and occupation of multiple space: pop charts, the avant-garde, purveyor of noise, imploder of dancefloors. Fellow artist Lyra Pramuk captured SOPHIE’s essence: “Sophie’s sounds are a musical embodiment of transness, evading the rigidness of the 12-tone system, to express a fluid in-betweenness, continually exploring the infinities between zero and one” (@lyra-pramuk). So, while Bluemink’s assertion resonates, perhaps a more satisfactory classification would be Post-Retromania. SOPHIE sought to explore new possibilities in sound and thinking through sound. These were never statically informed, always moving, perpetual transition, never returning. This sense is best captured in an Arte interview, with SOPHIE lounging on a bed while espousing the merits of utilizing only the most high-fidelity synthesis. Returning to Chude-Sokei’s Anarchic AI: “But new life always announces itself through sound. That is where the artificial first becomes authentic.” (Chude-Sokei n.pag.)

Citation

Gary Charles, “Archived Futures: Digging In The Crates Of Always,” Alluvium, Vol. 9, No. 3 (2021): n. pag. Web 4 June 2021. DOI: https://doi.org/10.7766/alluvium.v9.3.04

About the Author

Gary Charles is an interdisciplinary researcher and artist, working across sound, moving image, installation, and conceptual practice. Gary also releases music under a number of monikers, including releases on High Strung Young and Flash Recordings as The Static Hand, and improvised electronics as part of improvisation collective, The Cosmic Asunder. His current research looks at the emergence of Artificial Intelligence approaches in cultural production, particularly in relation to creativity within contemporary art and music cultures. Through both research and practice, his focus is on uncovering the assumptions, misdirections and biases embedded in the models, as well as the protocols that underpin them. Gary is currently a PhD candidate at University of Birmingham, and teaches Synthesis, Audio and Cultural Theory at the British and Irish Modern Music Institute (BIMM).

Works Cited

Bender, Emily & Gebru, Timnit et al. “On The Dangers Of Stochastic Parrots: Can Language Models Be Too Big?” FAccT ’21, 2021, pp. 610-623. DOI: https://dl.acm.org/doi/10.1145/3442188.3445922

Bluemink, Matt. “Anti-Hauntology: Mark Fisher, SOPHIE and the Music of the Future”. Blue Labyrinths, https://bluelabyrinths.com/2021/02/02/anti-hauntology-mark-fisher-sophie-and-the-music-of-the-future/. Accessed 18 March 2021.

Chude-Sokei, Louis. Anarchic Artificial Intelligence. https://www.anarchic.ai/. Accessed 12 March 2021.

———. The Sound of Culture: Diaspora and Black Technopoetics. Wesleyan University Press, 2015.

Fisher, Mark. Ghosts Of My Life. Zero Books, 2017.

James, Robin. “Sound Is A Dimension Of Reality”. Clark Art, https://www.clarkart.edu/research-academic/podcast/season-2/robin-james. Accessed 12 March 2021.

James, Robin. “Neoliberal Noise: Attali, Foucault & the Biopolitics of Uncool”. Culture, Theory and Critique, Vol 55:2, 2014, pp. 138-158. DOI:https://www.tandfonline.com/doi/abs/10.1080/14735784.2014.899881.

Le Guin, Ursula. The Dispossessed. Orion Publishing, London, 1974

Lewis, George E. New Music Decolonization in Eight Difficult Steps. Outer National, https://www.van-outernational.com/lewis-en 2020 . Accessed 12 March 2021

@lyra-pramuk. “so much of Sophie’s sound design” Twitter. 5 Feb. 2012, 10.59am, https://twitter.com/lyra_pramuk/status/1357645085853294593

Pasquinelli, Mateo & Joler, Vlad. NOOSCOPE, https://nooscope.ai/. Accessed 21 November 2020.

Pelly, Liz. “Music, Power and Platforms”. CTM Festival, 2019. https://www.youtube.com/watch?v=6JuIlUMSiLk . Accessed 21 November 2020.

Reynolds, Simon. Retromania: Pop Music’s Addiction To It’s Own Past. Faber & Faber, 2011.

———.(No) Future Music. New Perspectives, Vol. 28(3), 2020, pp. 305-313.

Small, Christopher. Musicking: The meaning of Performing and Listening. Wesleyan University Press, 1998.

Steyerl, Hito. The Language Of Broken Glass. HKW, Berlin. https://www.hkw.de/de/app/mediathek/video/69577 2019

@tametick, “In the film “Back to the Future”” Twitter, 3 Apr. 2021, 8:52 a.m., https://twitter.com/tametick/status/1378253940479954945.

Thompson, Marie. “Whiteness and the Ontological Turn in Sound Studies”. Parallax, vol23:3, 2017, p266-282. DOI: https://www.tandfonline.com/doi/full/10.1080/13534645.2017.1339967

Feature Image: “Music has the right to children” by Benjamin Horn

One Reply to “Archived Futures: Digging In The Crates Of Always”

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.