2
AI assemblage

Even technology makes the mistake of considering tools in isolation: tools exist only in relation to the interminglings they make possible or that make them possible.

Deleuze and Guattari, A Thousand Plateaus (1980, p. 90)

AI as social architecture

As argued in the previous chapter, this book does not approach AI based on any narrow technological definition of what 'AI' is. It has to be emphasized, however, that I am not engaging in any battle over what AI truly is, or what the correct definition may be. Definitions are tools and, depending on what we want to do with 'AI', different definitions will be useful. AI can be defined as 'intelligence displayed or simulated by code (algorithms) or machines' (Coeckelbergh, 2020, p. 64). This straightforward working definition, in turn, opens up the vast question of: What is intelligence? Luckily, this is not something to which we need to respond for the purposes of this book. It is sufficient to have a very general definition based in the spirit of the Dartmouth workshop in 1956 (see Chapter 1), where the idea was to simulate human intelligence in computers. Beyond such definitions, however, the battle over how AI is best defined is a recurring theme also within the community of computer scientists, engineers, and designers of AI systems (Ehsan and Riedl, 2022; Elliott, 2022, p. 5; Zerilli et al., 2020, p. 165).

AI is indeed a social reality. But the computer science definition of AI is not useful for analysing the social reality of AI. We don't need to know if AI is 'technologically driven forms of thought that make generalizations in a timely fashion based on limited data', or if it is rather 'the creation of machines or computer programs capable of activity that would be called intelligent if exhibited by human beings' (see Elliott, 2022, p. 5). What we need to know is how AI as a phenomenon in society takes shape, what it does to society, and what social, political, economic, and cultural processes of power and subordination it sets in motion.

This endeavour demands a different kind of definition of AI, which is focused on its broader underpinnings, tentacles, and practices. What is AI made of? How does it reach out and intertwine with people and their lives in society? What does AI do to social ideas and relations? Technology philosopher Shoshanna Zuboff (2019, p. 241) states that today's algorithmically driven digital capitalism - what she calls surveillance capitalism - operates through a 'ubiquitous apparatus' that renders 'human experience'. If we are to grasp AI in society, and its politics, the metaphor of a ubiquitous apparatus is quite fitting. Digital business researcher Ana Canhoto (2020) suggests that 'AI is a system, not a technology', and posits that AI can be seen as a collection of different components, not only the algorithm at the centre: 'when considering the benefits and pitfalls of adopting AI, we need to think of AI as an assemblage of technological components, rather than one technology'.

Aligning with this, computer philosopher Jaron Lanier and economist Glen Weyl (2020) contend that '"AI" is best understood as a political and social ideology rather than as a basket of algorithms', and as 'a suite of technologies'. AI, like other architectures in digital society are more than the mere '"back-end" of technical architecture', it is rather 'the relations among a number of elements' (see Burgess and Baym, 2020, p. 16). Kate Crawford (2021, p. 49), a leading scholar on AI, politics, and society, writes about the urgency in 'understanding the deep material and human roots of AI systems', because of the difficulties in seeing inside AI's 'complex assemblages'.

In sum, then, defining AI in sociopolitical terms has to do with understanding it as a ubiquitous apparatus entangled with human experience, and as a suite of technologies - a complex agglomeration of different components.

Making a monster

I suggest, for the purpose of sociopolitical critical analysis, that we conceive of AI as assemblage. This means sticking one's head into the conceptual complexity, unfinishedness, and contestedness that is the theorizing of philosopher Gilles Deleuze and psychoanalyst and activist Félix Guattari. They coined the notion of assemblage to provide a framework for thinking about multiplicities rather than things. As they put it in the quote at the very beginning of this chapter, 'tools exist only in relation to the interminglings they make possible or that make them possible' (Deleuze and Guattari, 1980, p. 90). Assemblage sees the social world not in terms of individual elements, but as abstract collectivities. I believe AI to be one such collectivity.

Drawing on the view of Deleuze and Guattari, we cannot see the individual computers or algorithms that drive AI as individual entities. Neither can we see the designers, programmers, and scientists as forces in their own, individual, right. The theory also stretches much further, to the point where we cannot see the people affected by AI as separate from AI either. This is because:

There are no individual statements, there never are. Every statement is the product of a machinic assemblage, in other words, of collective agents of enunciation. (Deleuze and Guattari, 1980, p. 37)

Drawing on Deleuze and Guattari's description of the nature of assemblages and applying it to the phenomenon of AI in society, we can contend that AI both consists of things, and does things. This relates to how 'an assemblage comprises two segments, one of content, the other of expression' (Deleuze and Guattari, 1980, p. 88). On the one hand, AI involves humans, computers, scientific practice, big business, imaginaries, visions, driving forces, and so on. It is a machinic assemblage of bodies, of actions and passions, an 'intermingling of bodies reacting to one another'.

AI also - if we now see it in terms of discourse and mythology - shapes social reality. This happens on several levels. On one such level it is where AI imaginaries impact on what kinds of future lives and future societies are constituted as thinkable. Media and communications scholar Amanda Lagerkvist (2020, p. 16) posits that: 'In the present age AI (artificial intelligence) emerges as both a medium to and message about (or even from) the future, eclipsing all other possible prospects'.

On a more material level, AI co-constitutes social and political reality through the ways in which it interpellates humans as subjects into relations of success, failure, precarity, inequality, and discrimination. Computer scientist and mathematician Cathy O'Neil (2016, p. 13) writes of how 'secret models wielding arbitrary punishments' increasingly 'affect people at critical life moments: going to college, borrowing money, getting sentenced to prison, or finding and holding a job'. AI's different elements function together in ways that produce very material consequences in people's lives. Or, once again in the words of Deleuze and Guattari (1980, p. 88): 'it is a collective assemblage of enunciation, of acts and statements, of incorporeal transformations attributed to bodies'.

Philosopher Manuel DeLanda has developed Deleuze and Guattari's original notion of assemblage into what he calls assemblage theory. While not popular with all Deleuzians (see, e.g., Buchanan, 2021, p. 2), this reading and adaptation of assemblage has an analytical edge that is lacking in the original sketches by Deleuze and Guattari. DeLanda (2016, p. 1) purports to strip away some of the 'additional conceptual machinery' to be able to provide a more 'coherent notion' of assemblage. DeLanda's definition of assemblage starts from Deleuze's (2007, p. 69) notion that '[a]n animal is defined less by its genus, its species, its organs, and its functions, than by the assemblages into which it enters'. Consequently, for our purposes, AI is not as much defined by its type, technological form, source code, and algorithms, as it is defined by the complex sociopolitical settings where it comes into play and co-functions. An assemblage, then, 'is a multiplicity which is made up of many heterogeneous terms and which establishes liaisons, relations between them' (Deleuze and Parnet, 2007, p. 69).

According to DeLanda (2016, p. 3), society consists of a potentially endless number of assemblages ('assemblages of assemblages'). Assemblages all have the same ontological status, meaning that while they may operate at different scales, they can also interact directly with each other. We can consider AI as one such assemblage, existing in society, interacting with a range of other assemblages - of economy, of culture, of social classes, of identities, of the natural environment, and so on. Furthermore, '[a]ssemblages are always composed of heterogenous components' (DeLanda, 2016, p. 20).

AI, in its broad meaning, consists of technological components ranging from the very hardware, through algorithms and software, to humans, ideas, governments, politics, cities, and so on. This is because, recalling the point from earlier, that AI - construed as assemblage...