The first question I would like to address is: ‘What was the network?’ With this we can also think about whether we are in a moment in which it is possible to historicize networks, and if so, why we would do that.

When I first read the question, ‘What was the network?’, I thought rather of the architectural topology that was not realized, the dream of the decentralized or even distributed architecture of the network, that didn’t come into being. The dream of a network that was in reality taken over by the more sovereign and mainstream infrastructures. And now there is this question that always takes me back to the expectations of the 90s, and the first platforms – IRC, Usenet. All these expectations were there – so what happened, what changed? The approach of Manuel Castells, for example, was all about how communication networks would bring change to society, politics, economy, and culture. And this change did happen, but not how it had been imagined. Now we also see the dark sides of the network. On a personal level, when I think back to the late 90s, early 00s, I still remember how important it was that connectivity had come along. I was in Athens at that time, working for a festival of art and new technologies, as we described them then, Medi@terra. At first the festival was Greek, then it expanded to include the Balkans, and then became international. The festival grew thanks to the networks that we built with other festivals and centers in the field, thanks to the research we could do online and to the interest that audiences and funding bodies showed in the emergence of digital networks. For me, the network was this potentiality. But now it doesn’t seem so possible to believe in any longer.

In media studies we love these kinds of past-tense questions. However, the current debate about digitalization seems to be completely ahistorical, as though the ‘digital’ had only just entered the stage. This historical oblivion is particularly true when it comes to networks and their implementation in digital media industries. Yet reflecting on the past doesn’t necessarily imply outdated historicism, in the sense of understanding a specific time in history that leads straight to the present. What I’m interested in is media genealogy, which is nonlinear and eclectic. Walter Benjamin calls this ‘historicity’, in contrast to historicism, or Jetztzeit (‘here-and-now’) – a term that perfectly fits the ‘eternal present’ of this year’s exhibition, ‘The Eternal Network’. What he means by this is that two widely disparate historical events may have more in common than two events close together in time. This historicity is ever-present, aligning the past with the here and now – and so also with the future. What Daphne said about the looming dream of the network and its potentiality for today is a good depiction of such a Jetztzeit.

When we talk about digitalization, we are of course talking about a decades- or even centuries-old development. But we don’t have to go through the whole history in order to reflect on it. Making the present intelligible through past events can be very episodic. This is also an interesting point about the network metaphor – that it has this untimeliness to it. It pops up in the 90s to make sense of quite different socioeconomic developments, such as a new worldwide communication infrastructure with the hope for democratic expression and the latest push toward a fully globalized capitalism. With this, the network becomes an all-encompassing, all-explaining concept, from food chains to supply industries to nervous systems. Patrick Jagoda has called this the ‘network sublime’: the network is everything and nothing1. And maybe that’s the best thing that could have happened to it – to become this weird and untimely concept. Like Jetztzeit, it can always actualize, it can connect to different types of pasts and futures.

And the fascinating thing in today’s context is that this ‘becoming a network’ or ‘network- becoming’ is also about becoming invisible. Most recent debates about digitalization tend to be dominated by debates about platforms. But it is still the network, at least from my perspective, that is the driving force – the motor – behind most of digital culture’s phenomena. Even though digital capitalism has solidified into platforms over the last decade, the inner working of these platforms, the way they produce value via data extraction and interpretation is still based on a network logic.

• 1. See, for instance, Patrick Jagoda, Network Aesthetics, Chicago: The University of Chicago Press, 2016.

Maybe this also has something to do with the inherently anti-narrative stream of thinking within new media and network theory, where linear representation is not an important issue. What was usually on its main agenda was how you acted or performed in a given project, rather than how you narrated it. The reluctance to write this history therefore also comes from the kind of anti-representational thinking inherent to working with and within networks, and the wish for forms immanent to the form itself.

I think that this also relates to the question of what we can actually learn from these histories of ‘Critical Internet Cultures’, and relatedly, what the blind spots of the contemporary moment are with respect to this question. Clemens for instance co-edited a book about ‘forgotten futures’1, pointing to the idea that we should perhaps also consider net cultures that never happened or were never heard about. This prompts another question as to the limits of networks.

• 1. Clemens Apprich and Felix Stalder (eds), Vergessene Zukunft. Radikale Netzkulturen in Europa, Bielefeld: transcript Verlag, 2012.

It depends how you see it. In the past there was a lot of discussion about networks being ‘walled gardens’. One could say that what lies beyond one’s network is difference, because networks are based on sameness. Other worlds, opinions, and realities are kept away from you. Networks are not porous. They are vulnerable, as Geert has discussed elsewhere, but they are not porous; you cannot easily break through them.

Thinking about the limits of networks and what lies beyond them, I am made to think of the system, which, somehow, was the first victim of the network. Before the 90s, the ‘system’ not the network was the dominant concept to describe society. However, with an increasingly globalized and networked world, the idea of social groups, institutions, and even the nation state as contained systems broke up. The system began to leak, and opened up into myriads of networks. For some this had a liberating effect, but it also created problems. Beyond a network is always another network. As Wendy Chun says, the network is such a compelling concept, because with it, or better within it, you are always searching and never finding1. You constantly zoom in and zoom out, switching from one network to another. The network gives you the opportunity or even the excuse not to make a decision, not to define an inside and outside, not to look for an exit. You are trapped within the network. Yet as Deleuze and Guattari demonstrated, every repetition has the potential of difference – for bifurcation. The things that seem to be especially repetitive are those that have the most potential to produce something new. What I like about this idea is that bifurcations happen all the time. This is what I’m trying to get at in my essay for this volume: that the network still has this potential; that it can connect different times and places. I want to argue against a reticular pessimism – that is, the idea that everything is trapped and captured within a network. You simply cannot capture everything.

• 1. Wendy Chun, Updating to Remain the Same, Cambridge, MA: MIT Press, 2016, p. 29ff.

The idea is precisely not to hide nor dissolve political categories, such as class, race, and gender, in some kind of network sublimity, but to make the edges visible and tangible, in order to enable bifurcations.

It is also important to consider what the dominant model of a network is for each era. Today, discussion has shifted to the area of artificial intelligence, with the dominant model being the artificial neural network. This brings us to topologies that are much more complicated, much more opaque, compared to the informational and social ones we have met up to now, even if all of these somehow intersect. I feel that this affects the discourse on networks, for example when we are talking about the Smart Home or the Smart City. Because these environments, the environments that we live in, are being adapted based on how these machines operate; how these machines see, read, and sense the world.

The invisibility of the network is also what made us stop referring to it in a way. That’s a bit like what Wendy Chun discusses in her book, Updating to Remain the Same: Habitual New Media. The less we see or pay attention to networks or technologies, the less we name them and reflect upon them. But that doesn’t mean that networks do not play a significant role. Actually, they play an even more significant role as we become the machines or the networks. They define our daily lives and habits.

In the late 90s network theory turned into network science, and then stopped. I am not saying that people have stopped thinking about networks but this specific trajectory stalled. Castells’s network society has not been widely adopted. Lately I’ve been in contact with some people in the European Commission in Brussels who are fierce promoters of network science. I challenged them to prove whether this science is alive and has any relevance. What has it produced lately? There’s a desire to bring scientists on board. The whole world of social networks has become so dark, fluffy, and messy to them that they felt they needed to bring scientists back on board to get rid of all the myths once and for all – the commercial interests and the hidden forces. In this view the network is a mysterious invisible power that produces fake news and then produces conspiracies.

In the social sciences more and more people say that we need to introduce technical solutions, because according to them, our understanding of society has completely failed. But we are already caught in a complex kind of technical, bureaucratic society: this is our reality. So this limiting of the horizon, it’s quite real. It does not open up discussions about alternatives at all. I wish there was another type of network theory that could now thrive. Then, the discussion around this table would be very different. What would have happened if decentralized networks would have been programmed to resist any form of centralization?

An interesting and somehow built-in ‘limit’ of the network in relation to AI and machine learning lies in the very beginning of cybernetics. As Orit Halpern has discussed, the cybernetic vision of Warren McCulloch and Walter Pitts, who theorized the possibility of an artificial neural network in 1943, led them to a computational rationality, which was no longer based on reason1. As a consequence, the network, in their view, turns psychotic; it leads to an overproduction of meaning, an unreasonable situation in which any form of symbolic closure is no longer relevant. This is the situation we find ourselves in today: artificial neural networks promote a hyper-inductive approach and, at the same time, dump the idea of symbolic reasoning. Just look into the data and the rest will follow. But it’s still people who build these models, and inevitably, they implement their very specific and biased understanding of what they want to do with the data. You can’t just dissolve this symbolic baggage in a supposedly flat ontology or hide it in network scientific discourse – as psychoanalysis has shown over and over again, every time you try to repress the symbolic, it reemerges somewhere else. So it comes as no surprise if artificial neural networks discriminate along the lines of a socially, that is symbolically unjust system. They are not so unreasonable after all, but follow the biases we produce as a society.

• 1. Orit Halpern, Beautiful Data: A History of Vision and Reason since 1945, Durham, NC: Duke University Press, 2015.

Yes, that’s exactly the point. I would just disagree on one point: I don’t think that these network models are out of reach; they are not black-boxed, as is often claimed to be the case. If you want to know more about neural networks or machine learning methods, you can, for example, download and use Google’s TensorFlow platform. Of course, you might object that this in itself is a technical framework, that for most people it is still out of reach. But for people in media studies, the arts, or activism, who want to engage with these debates, I don’t see why they shouldn’t take a crash course in machine learning offered by Google.

I think, Alexa, Siri, or Google Assistant are good examples for extending the notion of the black box. After all, when it was conceptualized in cybernetics, it didn’t mean that we shouldn’t touch it. On the contrary, the black box was introduced as a methodological tool in order to experiment with complex systems. So why not experiment with Alexa, Siri, and co.? This can happen on a technical, as well as on an artistic, theoretical, or even legal level. We should get our hands dirty if we want to formulate a critique of these systems.

There is a suggestion by Tiziana Terranova, quoted in your text, Geert, of shifting the idea of connectionism from our present model to quantum entanglement. It’s a very speculative proposal where she is saying that this could also produce ‘spooky’ results.