19 Learning in the New Dispersed Prime Time

Fuseki

Algorithms are dynamic objects that reveal their unprecedented effect on learning when the multiplicity of their temporalities is brought under examination (Deleuze 1994, 182–83). Drawing on critical theory, media theory, and examples pulled from algorithmic culture, this memorandum will analyze the transitional time in which contemporary modes of learning are taking place. I relate learning to a correlated heterogeneous series of social time, disciplinary time, televisual time, historicity, and rhythm. Through this approach, I hope to offer an open, nonessentializing and nontotalizing account. In fact, one of the most difficult aspects of writing on algorithmic time is to attend to its ontological character without ossifying and reifying contingent relational qualities—simply put, their temporality.

Philosophically, I push back at contemporary Platonic interpretations of process philosophy that find in algorithmic temporality an opening toward always-already present “worldly sensibility” or “vibratory continuum” that had been foreclosed in anthropocentric thinking (Hansen 2015, Murphie 2019) I favor more radically historicist thinking on the indeterminacy of Being or to be more precise with the essential temporality of Being wherein distinct moments do not partake in a common ontological substrate or schema that exist independently of Time (Heidegger 1977, 135; İşsevenler 2023a, 386). Thus, I argue that algorithms take part in irreducible novel events, with some poetic license I would term prime events, excessive to preexisting categories of understanding, however processual these might be, as these are demonstrably incomplete (Goldstein 2005) and inevitably limited and shaped by the technology and language of their time.1

The memorandum opens with the historical middle. The period wherein disciplinary time was declining in learning environments coincided with the advent of algorithmic social organization with its capacity to temporalize deterritorialized cultural bits. To further situate this transition in history, I conceptualize a move from televisual flow, as the preceding political economical model of managing production and circulation of information and value, to algorithmic timelines that learn and modify population behavior and preference through dispersed modes of customization without recourse to disciplinary structures.

To illustrate the arguments developed here, I offer a critique of the impasses of the narrative style of the AlphaGo documentary (2017) where the ambiguous historicity of algorithmic learning is compensated by an aesthetic recourse to national-time. I argue that the use of nationality as a master signifier to categorize both the algorithm and past Go champion is a compensatory attempt at creating symmetry between the players and continuity in time. This move demonstrates the difficulty in witnessing the alterity of algorithmic time on its own terms as it is more mundanely socialized into every day. What infinite “cultural” past goes under when AlphaGo internalizes lessons of thousands of Go games played across centuries? How to attend to the transcription of skills from one milieu to another where formal capabilities of the algorithm subsume vast existential domains out of which “the best” in Go has arisen in a historically specific moment? The memorandum concludes with notes on the political implications of a new temporality of learning at these critical historical junctures.

Social Time and Algorithmic Learning

Social time is a bodily, symbolically, and technologically kept regulative framework that allows the orchestrated reproduction of society (Durkheim 1995). Historical materialist theory places learning exactly at the moment of reproduction of labor through which normative behaviors are institutionalized and transmitted across generations. As a mechanism of reproduction, learning takes place not only in formally defined educational settings but across cultural life. Thus, social time is the historically specific punctuation of time through normative expectations and material designs that range from the routine rhythm of weekly or yearly schedules to sanctioned conditions of discontinuity such as performative sirens of emergency services that puncture the routine or politically declared states of exception that suspend normal distributions of rights in the name of national security (Foucault 2007, 343).

With the operationalization of algorithms in the temporalization of massive amounts of daily produced content on public-facing platforms (i.e., selection, distribution, sequencing), we witness the profound impact of algorithms on social time, undoing the process of reproduction that relies on the stability of social norms. New rhythms are added through computational protocols and contingent data sets that transform our senses of vitality and sociality. The conventional frames of disciplinary time that contain classical social institutions are constantly segregated and accelerated if not interrupted by self-activating algorithmic temporalizations that produce temporary “dissipative structures” (Chun 2021; Clough 2008) emerging in resonance and dissonance with the ruins of antecedent cultural forms at different scales such as metamorphoses of narrativity, family, and state by Instagram and TikTok’s Stories and Reels, Facebook and Twitter’s networks, and Amazon’s cloud services, respectively (İşsevenler 2023a).

From Disciplinary Time of Contained Duration to Overlaying/Interruptive Algorithmic Time

If technology classically functioned to help containment of social time through scheduling and programming of disciplinary bodily activity (e.g. exercise) in mutually exclusive enclosed spaces from school to work, from military to hospital (Foucault 1995, 141–69), the mode of algorithmic social time is interruptive in overlaying these previously separated domains of social life. Disciplinary time is a series of durations. “It is this disciplinary time that was gradually imposed on pedagogical practice specializing the time of training and detaching it from the adult time, from the time of mastery; arranging different stages, separated from one another by graded examinations; drawing up programmes, each of which must take place during a particular stage and which involve exercises of increasing difficulty; qualifying individuals according to the way in which they progress through these series” (Foucault 1995, 159). Algorithmic temporality manages turbulence and engages with productive chaos in the cultural vacuum created by the decline of the absolute authority of disciplinary knowledge systems (Grosz 2008). Yet, left to their own devices and freedom, subjects often seek the return of disciplinary measures (Deleuze 1992, 7) to render stable what appears to be new complex durations without stability or standardized measure, at best to make time more accessible/inhabitable and, often more problematically, to demand homogenization of rhythmic variation and to foreclose alternative creative arrangements of time; this tendency is becoming more apparent with algorithmic appropriation, reformulation, and juxtaposition of myriad cultural material.

Algorithmic circulation of information is increasingly felt in learning environments, most regularly challenging the authority of the teacher in controlling the attention during class time and the means available for students to complete a given assignment; we are witnessing the end of transmission of disciplinary knowledge through securely walled spaces of educational institutions. Just as probation takes precedence over the detention in the criminal justice system that thereby expands into civic life and just as medicine extends beyond the hospital to fragment care across unstandardized arrangements of continuous examination through bodywear, precarious forms of access to insurance and proliferation of myriad modes of self-care; education and learning leaves its compartmentalized disciplinary temporality to become a diffused, overlayed feedback process increasingly managed by artificial intelligence.2

A New Fractured Collective Rhythm

Deleuze and Guattari (1987) state that “action occurs in a milieu, whereas rhythm is located between two milieus” (314). As a result, collective simultaneity, that is, the shared pulse of culture performatively produced through mass media of industrial capitalism (e.g., newspaper, radio, television), that grounded formation of national cultures and their “shared” history, has been giving way to an increasingly more segmented, decentralized, and ungeneralizable nonsynchronous flow of time managed by algorithmic feeds. The disciplinary closure of social time (e.g., class time, worktime, examination time) that previously kept the interruption outside is now substituted with continual coexistence of learning, practicing, and evaluating through algorithmic infrastructures of tracking, analyzing, and immediate feedback. This simultaneity of subsequent steps is only part of the temporal puzzle of algorithmic governance. There is a paradoxical contraction of learning and unlearning in this new temporal logic. This radical deconstruction of learning demands cracking open correspondence of the arrow of time with growing mastery in the context of profound political and technical changes. To continue to internalize given corpus of a discipline might mean to collude with an erasure and unwillingness to experiment. To start writing the unwritten history might mean freezing of a particular disciplinary genre in time or alternatively, to use Spivak’s profound term, strategic essentialism giving measure to indeterminate potentiality of algorithmic processes beyond initial parameters and biases (İşsevenler 2023b, 69).

To limit the scope of analysis of algorithms’ impact on the temporality of learning, one might ask when and where teaching starts and ends. Yet contained spatio-temporalities (classroom i.e. transmission, library i.e. archive) become impossible to demarcate in that these instantiations of learning are diffused into the totality of social and environmental interactions. Media scholar Wolfgang Ernst (2021) argues that “different from the phenomenological time of lived experience, data processing is no continuous ‘stream’ at all, but nonlinear microstorage and transfer” (15). This noncontinuity arguably stems from “the dispersal of experience produced by contemporary media . . . which marginalizes not simply the operation but the very role and relevance of consciousness” (Hansen 2015, 252). With this dispersal, new continuities are synthesized at scales exterior to the boundaries of the phenomenological domain. Often unfelt everyday extraction and feedback of emotional, practical, and historical data offer an archival possibility and render any relation with artificial intelligence a potentially instructive moment—consciously registered or not. The ambulatory, interruptive, and distributed character of algorithmic learning undoes the hold of disciplinary boundaries—most commonly exemplified in the procedure known as “unsupervised learning.”3

This problem of what we might term phenomenological detectability of the instructive relation moves us to reconsider the temporality of the body as the center of creative indeterminacy between input and output. If the human body has been seen as a source of complexity and unpredictability in its internal processing of environmental input, machine-learning algorithms are increasingly tapping into the nondeterministic processes as well by their operation through probabilistic estimation. By getting closer to the social texture as an assisting technology through text-generators, algorithms modulate, then, existing arrangements of flow and disperse the locus of creativity.

Incompleteness from Televisual Flow to Algorithmic Flux

Encompassing individual programs and ads across different channels and networks, televisual flow is argued to be the temporal form of late-stage capitalism insofar as it minimizes the time of delay between investment and profit insofar as products need to circulate the market to realize their value through purchase by a customer (Dienst 1994). By staying with or zapping across content through the prosthetic technology of remote control that materializes subjective intent, audiences at home finalize what’s otherwise an unfinished product, that is, televisual time. By constantly drawing on the viewing as a work toward the resolution of a potentially infinite series of durations into a single finite track of audience time, televisual technology expresses a historically specific dynamic character of the political economy of time. In this resolution of circulating content into a single user time-track, there is neither technological determinism as in strict mechanization of time nor autonomy of constitution of temporality through the unpredictable aesthetics of subjective duration. Instead, an ongoing selective process unfolds between the activity of populations segmented into demographic and cultural categories and circulating massive amounts of cultural content (Chun 2021). Sorting algorithms in temporalizing social networks take over this dynamic relation as a function of contemporary political economy. The production of this immanent social rhythm that used to be finalized during domestic viewership is now deterritorialized through smartphones, contingent datasets, contingently bounded and segregated populations, and algorithmic feeds.

The temporalities produced in this arrangement are complicated in that they prehend both the compositional creativity of the algorithms as well as the indeterminacy of human subjectivity. Deleuze’s notion of “sheets of past” is most suitable to account for decentralized temporal flows perpetually done and undone through algorithmic audiovisual networked feeds (Deleuze 1989, 105). In developing this concept as a thought-image to account for the novel cinematic strategies that creatively engage with the past and its relation to the present, Deleuze draws attention to the incomplete serial character of time whereby the past “potentiates” the present as a multiplicity. Through the compossibility of multiple series, the present is neither subordinated under a determinism of the past nor given pure autonomy.

Algorithms inhere in randomness that adds novelty to their process, that is, creative serialization of content. As alluded to in the introduction, there is a primeness to any algorithmic situation—insofar as the incompressibility of their logical processing (Parisi 2013) means there is an undecidability as to whether any given experience is a composite of other “basic units” (given hypothesis, code, parameter, variable or memory, affect, thought, social relation) or constitutes a veritable substantive novelty (Livingston 2012, 113–15). In this context of the complexification of temporality, the charges of presentism and spontaneity raised against the current “culture industry” managed by algorithms are problematic. As Deleuze suggested that, in editing, “irrational cuts” mark the emancipation of time from bodily movement, I argue that with some poetic license, the figure of prime numbers and ungeneralizable prime gaps that separate them allow us to mark and conceptualize the inexhaustible possibility of novelty and individuality of processes of serialized algorithmic learning.

Learning with Pure Strategy: Algorithmic Time Beyond National Historicity

Released in 2017 and published on YouTube in 2020, the AlphaGo documentary narrates with a significant dramatic arc the encounter between the top Go player Lee Sedol and Google DeepMind’s algorithm AlphaGo. Trained on thousands of games played by human players and improved through computer play, AlphaGo attests to the dynamic character of algorithmic objects. After its game with two-time European champion Fan Hui, the algorithm rapidly develops its skills through the impetus gained by Hui’s collaboration with computer scientists. Catching Lee Sedol by surprise by the speed of its improvement since its last public performance, DeepMind’s algorithm, AlphaGo, wins four of five games played under international televisual broadcasting and internet streaming. Interestingly, the setup of the competition places Lee Sedol under the South Korean flag and AlphaGo (and/or the person who becomes the prosthetic extension of the algorithm in placing its moves on the board) under the British flag (Figure 19.1). This marginal detail on the political aesthetics of the documentary can be taken as symptomatic of what I conceptualized as creative resolutions of ambiguity of algorithmic temporality (İşsevenler 2023a). As global corporations and transnational internet networks traverse the spatiotemporal boundedness of the political culture of nation-states, the latter is still vested with significant coercive, symbolic, and therefore affective power (Aretxaga 2003). Nationality allows the production of symmetry and historical continuity between ultra-modern technology of artificial intelligence and the ancient culture of Go covering over conceptually almost unsurpassable epistemic and ontological gaps across milieus. Perhaps this is the politics of the algorithmic time; how to perform heterogeneous transcription across different historically specific regimes, where the condition of transnationalism gives way to the emergence of a new transmedia condition. As Deleuze and Guattari (1987) speak of Go pieces as “elements of a nonsubjectified machine assemblage . . . A Go piece has only a milieu of exteriority, or extrinsic relations with nebulas or constellations, according to which it fulfills functions of insertion or situation, such as bordering, encircling, shattering . . . what is proper to Go is war without battle lines, with neither confrontation nor retreat, without battles even: pure strategy . . . maintaining the possibility of springing up at any point: the movement is not from one point to another, but becomes perpetual, without aim or destination, without departure or arrival” (353).

Thus, the alien time of the algorithm clothes itself with deterritorialized markers of the past and continues their serialization. Algorithms thereby disperse the prime time away from the evening, when in the modern tradition the ideological unit of the family gathers in front of the television to receive unifying national culture and scatters it into mobile screens moving as magic carpets. Then, identity markers derived from the culture (e.g., nationality), learned, refashioned, and redistributed by the algorithms across social networks, have only circumstantial function and do not express intrinsic temporal or spatial property.

Notes

1. 1. See Webb’s memo on “soft thought” and de Freitas’s memo on “delirium” in this volume for further discussion of algorithmic novelty.

2. 2. See Curinga in this volume for a discussion of potentialities of network mode of governance after the dissolution of what I approach through “disciplinary time.”

3. 3. See de Freitas’s memorandum on generative algorithms in this volume.

References