Input ↔︎ Output | Part 3: Diagrams and Abstraction

This is the third blog post of a multi-part series making sense of THE VOID’s online video practices in the context of cybernetic participatory culture, the legacies of tactical media, stagnating platformization, encroaching AI, and the nascent Stream Art Studio Network. Read part one and part two. 

🟢 🟢 🟢

We share some other practices with our partner in crime Frank. One of them is diagram-making:

Rosenblatt’s diagram dissects the Mark 1 perceptron. It makes visible the notoriously evasive insides of the AI black box. But, to AI academics’ dismay, what we find inside is far from clarifying. No social or political values can be seen in it. Not even autonomy. In the same way that an artificial neural network doesn’t accurately represent how the brain works, the diagram does not show one-to-one how the machine thinks. Where among these lines and nodes can we say that discrimination happens?

Instead, this diagram is a way to highlight certain relations through an imaginary dimension reduction. It depicts neatly organized and abstract connections (not messy copper and plastic cables) linking sensory units to response units through adaptive units. This now canonical diagram even omits the feedback loop (oh so important for self-correction), only showing the one-way passage of information from input to output. In other words, it doesn’t show the output becoming input. For this, Rosenblatt sketched another diagram:

However imaginary, we also entertain ourselves with the task of depicting feedback loops (see diagram 1 again). Abstracting relations, subtracting the social and material dimensions from our technical setups to understand how input becomes output. While allowing us to picture and communicate to others the signal flows in our streaming setups, these diagrams omit the plasticky unruliness of cables, the shoeless feet, and the sometimes-tense-sometimes-fun relations among us and towards the public.

Like the perceptron to the mind, our diagram is not a representation, but “rather maps out possibilities prior to their appearance”[1] of our green-lined human-machine collaboration. It is more of a blueprint or a set of visualized instructions exiting somewhere in between the visible and the articulable (ibid.). The depicted lines refer to connections and relations meant to be carried out at some point in the future; the diagram is thus a picture that calls for action, guiding a dynamic operation between humans and machines. Following Pasquinelli, it is an abstraction that mobilizes social cooperation — a starting point for automation.

These blueprints have a complicated relationship to their hardware implementations. On the one side, they do not pretend to be a totalizing set of instructions encompassing the entire implementation process. That is, these are not a representation of but a trigger for implementation that might end up differing from the diagram. On the other hand keeping in mind Rosenblatt’s innovation of using performed processes as distributed memory, they are a way to paradoxically store and stabilize this practical know-how somewhere outside the set of emergent gestures. In a way, these diagrams are the comeback of memory-as-external-storage. Distributed enacted memory and centralized latent storage are then not at odds with each other. While the latter does freeze and abstract movements, diagrams might be useful and pretty but are not a network’s final form. They are not meant to be contemplated but are rather an instrument to reenact and, therefore, modify these operations in the future.

Like musical scores, instead of subsuming the socially and materially enacted dimensions of machines, these illustrations point them out. In other words, these diagrams are the leeway to automation not only because they abstract (separate) technological relations from social and material relations or because they externalize the task of memorization allowing for its further distribution. But counterintuitively, because they allow for the (re)production of these social contexts and relations. Social relations, diagrams, and automation do not exclude each other but are dialectically intertwined in a form of social and material abstraction. Diagramming, as a form of abstraction, is therefore not only the substraction of the messy social dimension.

Following Peter Damerow’s Hegelian understanding of the term, abstraction is not only a reduction of dimensions that leaves us with a partial, one-sided view of an object, but a precondition in the constructive process of concretization[2]. Or, as I like to called it on my last post, hardware implementation. Abstractions, such as numbers, have a social and material history; they are tied to material operations (i.e. counting), are mediated by tools and labor, and are directed towards specific goals. Abstractions are thus bound to external actions, give determinations to a social context, and as Jamila Mascat notes in her also Hegelian account of this notion: “abstraction constructs social bonds, builds up society and sustains the very structure of the body politic”[3]. Our diagrams, insomuch they are abstractions, stress the performative nature of our streaming setups/events/machines: they act as the driving motor for their reproduction by disposing[4] people and devices in such a way that transforms inputs into outputs and outputs into inputs.

Our streaming setups, pop-up or fixed but always green, are a clearly demarcated circumstantial, provisional, and context-specific arrangement of things. This technically co-produced, delimited, and abstract social context is nothing new. It is just a fancy way to call a stage. I bring up this millennia-old technique from the performative arts in order to juxtapose a machine (the perceptron) with a performance (a play, an online stream, a hybrid event) and hopefully show the similarities between an AI engineer and online video streamers/producers. It also adds some theatrical and imaginative flare to the often-bland activity of cable connecting.

Although, if these similarities are not merely metaphorical, our diagram’s omission of social context is a woefully misleading blind spot. It is indeed an a-critical and almost unconscious decision to focus on the relations between devices as those that deserve to be conserved after the process of abstraction. As Damerow writes: “To understand abstraction essentially means understanding what has to be abstracted rather than merely knowing how it has to take place”[5]. We can argue, like Rosenblatt, that our diagram is a working but eminently partial model of our stage. We could even argue that this is precisely the purpose of technology: artifacts that afford in them certain uses and, therefore, ossify certain relations. It is then only logical to use them as fetishes that can successfully preserve and stand it for other relations around them. But I am not satisfied with these answers.

Some months ago, at a workshop we gave at the Royal Academy of Arts in The Hague, we presented students with the challenge of sketching a kind different diagram. We asked them to abstract the social from the technical. Sarah-Rose Antoun produced a compelling outcome:

For once, these diagrams show people (and even plants!). It also gives a sense of spatiality and distinguishes between figure and ground (something that artificial neural networks cannot do[6]. While our diagrams might help us to technically reproduce our studio in different contexts, they fail to give an idea to newcomer performers of, both, the space they’re going to be in and the things they’re going to do in it. We can show them a screenshot of our past streams, but their visual outcome is deliberately overloaded, difficult to read, and lack spatial clarity (oh, the pains of hybrid spaces!)[7].

Sarah-Rose’s diagram, on the other side, dissects the space that the green screens flatten. It distinguishes the different elements on it, acknowledging their physical dimension. Basically, the fact that they take up space. A space that is omitted in our diagram and the streams’ final video output, but that has implications on the things we’re able to do and the amount of people and props we’re able to fit on our hybrid stage. We cannot just place props and devices in whatever way best suits the input-output flows. Neither our diagram nor Rosenblatt’s can explain why cables quickly get messy. Nor will a reader interpret (after all, reading doesn’t exhaust the kind of activities someone can do with them) from them that, if they ever end up performing the blueprinted relations, they’ll spend most of their time disentangling cables and arranging sitting spots.

Another inspirational example for our diagraming practices are Paul Ryan’s ‘Klein Worms’. An assistant to Marshall McLuhan, Ryan regarded himself as an activist version of the media guru/theorist. Not only he wanted to analyze and critique media, but also transform it[8]. With this self-description, Ryan, along with Rosenblatt, is another of THE VOID’s awkward friends.

The worms were drawn by artist Claude Ponsot and get their name from Klein bottles, a non-orientable surface that has no inside nor outside.

Klein worms depict reversible relations between inside and outside. Every Klein worm contains itself. Consequently, the relation between contained and container is not a static one but one of constant becoming. Ryan introduced these diagrams in his text “Cybernetic Guerrilla Warfare”, published on the countercultural journal Radical Software in 1971. As the title of the article suggests, Ryan, himself a ‘radical’ video practitioner, was influenced by cybernetic ideas of self-regulating and self-organizing systems.

In his article, he presents these worms just after mentioning Warren McCulloch’s (yes, that notorious Cold War engineer, MIT person, and theoretical inventor of the perceptron) problem of emergen orders from multiplicity:

“the problem that I'm up against is the problem of organization of many components, each of which is a living thing, each of which in some sense, senses the world, each of which tells others what it has sensed, and somehow a couple million of these cells get themselves organized enough to commit the whole organism.”[9]

For Ryan these worms are an attempt to display abstraction without subsumption. The organization of multiple parts into a whole without reducing them to that totality. This is because the totality is not a self-perpetuating circle, an infinitely large grid, or a progressively ascending spiral, but an ugly worm. As Brian Holmes puts it:

“What one sees, in varying configurations each time, is a whole with emergent parts that split off and are then reintegrated, only to emerge again without ever being fully subsumed. This is quite different from a dialectic that constructs its higher unities through the suppression and sublation of opposites. The underlying notion appears to be that of cooperation without subordination, within a social whole whose differences go on differing”[10]

For the time being, I would just like to pin that cyberniticians (Cold War proto-AI engineers) as well as countercultural video makers were both invested in the same ideas, not only regarding emergent orders that keep on differing, but also cooperation without subordination, and abstraction without subsumption. A goal that doesn’t match with existing machine learning’s present obsession with automation, prediction, optimization, and identification.

On their own, our diagrams also don’t fully attain these ideals. Ryan’s worms are perhaps a meta-diagram visually describing how ours are meant to be mobilized through hardware. That is, through its material and social concretization. Diagrams that become pop-up studios. Studios that become diagrams. Strategies that become tactics. Inside that becomes outsides. Outputs that become inputs.

🟢 🟢 🟢

[1] Zdebik, Deleuze and the Diagram, 1.

[2] Damerow, “The Material Culture of Calculation”; Damerow, Edelstein, and Lefèvre, Abstraction and Representation, 371–81.

[3] Mascat, “Hegel and the Advent of Modernity,” 42.

[4] See Panagia, “On the Political Ontology of the Dispositif.”

[5] Damerow, Edelstein, and Lefèvre, Abstraction and Representation, 371.

[6] See Pasquinelli, The Eye of the Master, 231.

[7] There is much to be said about the aesthetics of a VOID stream. Experimenting with a green screen as our main creative tool produces a specific visual output that flattens out the perspectival and spatialized image we have come to expect from camera footage. From rear-projection in classic Hollywood films, to digital cinema’s motion capture techniques and the layered compositions of animetism, picture-in-picture TV interfaces, Twitch streamer’s in-game overlay layouts, full- screen keying on YouTube video essays and dj sets, and the CoreCore trend on TikTok, it is possible to sketch an aesthetic genealogy of Stream Art’s flat and illegible hybrid spatiality. This is, however, a project that deserves at least an entire post.

[8] See Merrin, “Still Fighting ‘the Beast’”; Holmes, “Tactical Television. Movement Media in the Nineties | Regarding Spectatorship.”

[9] Ryan, "Cybernetic Guerrilla Warfare", 1. 

[10] Holmes, “Tactical Television. Movement Media in the Nineties | Regarding Spectatorship.”


References

Damerow, Peter. 2007. “The Material Culture of Calculation: A Theoretical Framework for a Historical Epistemology of the Concept of Number.” In Mathematisation and Demathematisation, 19–56. Brill. https://doi.org/10.1163/9789460911439_003.

Damerow, Peter, Wolfgang Edelstein, and Wolfgang Lefèvre. 2011. Abstraction and Representation: Essays on the Cultural Evolution of Thinking. Translated by Renate Hanauer. Digital reprod. of the original ed. Boston Studies in the Philosophy of Science, Volume 175. Dordrecht: Springer.

Holmes, Brian. 2015. “Tactical Television. Movement Media in the Nineties | Regarding Spectatorship.” 2015. https://web.archive.org/web/20240702152258/http://www.regardingspectatorship.net/tactical-television-movement-media-in-the-nineties/.

Mascat, Jamila M. H. 2018. “Hegel and the Advent of Modernity: A Social Ontology of Abstraction.” Radical Philosophy, no. 201, 29–46.

Merrin, William. 2012. “Still Fighting ‘the Beast’: Guerrilla Television and the Limits of YouTube.” Cultural Politics 8 (1): 97–119. https://doi.org/10.1215/17432197-1572012.

Panagia, Davide. 2019. “On the Political Ontology of the Dispositif.” Critical Inquiry 45 (3): 714–46. https://doi.org/10.1086/702613.

Pasquinelli, Matteo. 2023.The Eye of the Master: A Social History of Artificial Intelligence. London ; New York: Verso.

Ryan, Paul. 1971. "Cybernetic Guerrilla Warfare." Radical Software I, no. 03, 1-2. https://www.radicalsoftware.org/e/volume1nr3.html.

Zdebik, Jakub. 2012. Deleuze and the Diagram: Aesthetic Threads in Visual Organization. Continuum Studies in Continental Philosophy. London: Continuum. 

🟢 🟢 🟢

Part four coming soon…