James Steinhoff

This book argues that Marxist theory is essential for understanding the contemporary industrialization of the form of artificial intelligence (AI) called machine learning. It includes a political economic history of AI, tracking how it went from a fringe research interest for a handful of scientists in the 1950s to a centerpiece of cybernetic capital fifty years later. It also includes a political economic study of the scale, scope and dynamics of the contemporary AI industry as well as a labour process analysis of commercial machine learning software production, based on interviews with workers and management in AI companies around the world, ranging from tiny startups to giant technology firms. On the basis of this study, Steinhoff develops a Marxist analysis to argue that the popular theory of immaterial labour, which holds that information technologies increase the autonomy of workers from capital, tending towards a post-capitalist economy, does not adequately describe the situation of high-tech digital labour today. In the AI industry, digital labour remains firmly under the control of capital. Steinhoff argues that theories discerning therein an emergent autonomy of labour are in fact witnessing labour’s increasing automation.

As a nascent field within the academy, the contours, attributes, and bounties of data science are still indeterminate and contested. We studied how participants in an initiative to establish data science at a large American research university defined data science and articulated their relationships to the field. We discuss two contrasting visions for data science among our research participants. One vision is a transdisciplinary view portraying data science as a phenomenon with transcendent, appropriative, and impositional qualities that sits apart from academic domains. Another view of data science-one that was far more prevalent among our research subjectscasts data science as grounded, relational, and adaptive, emerging from crosspollination of numerous academic domains. We argue that this latter formulation represents a more quotidian reality of data science and positions the field as an extradiscipline, defined as a field that exists to facilitate the exchange of knowledge, skills, tools, and methods from an indeterminate and fluctuating set of disciplinary perspectives while conserving the boundaries of those disciplines. We argue that the dueling transdisciplinary and extradisciplinary visions for data science have important implications for how the field will mature, and that the extradiscipline concept opens novel directions for studying academic knowledge production in STS, contributing additional precision to the literature on disciplinarity and its permutations.

AI ethics is proposed, by the Big Tech companies which lead AI research and development, as the cure for diverse social problems posed by the commercialization of data-intensive technologies. It aims to reconcile capitalist AI production with ethics. However, AI ethics is itself now the subject of wide criticism; most notably, it is accused of being no more than “ethics washing” - a cynical means of dissimulation for Big Tech, while it continues its business operations unchanged. This paper aims to critically assess, and go beyond the ethics washing thesis. I argue that AI ethics is indeed ethics washing, but not only that. It has a more significant economic function for Big Tech. To make this argument I draw on the theory of intellectual monopoly capital. I argue that ethics washing is better understood as a subordinated innovation network: a dispersed network of contributors beyond Big Tech’s formal employment whose research is indirectly planned by Big Tech, which also appropriates its results. These results are not intended to render AI more ethical, but rather to advance the business processes of intellectual monopoly capitals. Because the parameters of AI ethics are indirectly set in advance by Big tech, the ostensible goal that AI ethics sets for itself-to resolve the contradiction between business and ethics-is in fact insoluble. I demonstrate this via an analysis of the latest trend in AI ethics: the operationalization of ethical principles.

The BuzzFeed memo provides an example of the tech industry’s embrace of buzzwords and bandwagonism. But it also provides an opportunity for thinking about the capitalist context in which AI is produced and used, and the vacuity of industry-driven discourse on AI’s social implications and ethical dimensions.

Surveillance of human subjects is how data-intensive companies obtain much of their data, yet surveillance increasingly meets with social and regulatory resistance. Data-intensive companies are thus seeking other ways to meet their data needs. This article explores one of these: the creation of synthetic data, or data produced artificially as an alternative to real-world data. I show that capital is already heavily invested in synthetic data. I argue that its appeal goes beyond circumventing surveillance to accord with a structural tendency within capitalism toward the autonomization of the circuit of capital. By severing data from human subjectivity, synthetic data contributes to the automation of the production of automation technologies like machine learning. A shift from surveillance to synthesis, I argue, has epistemological, ontological, and political economic consequences for a society increasingly structured around data-intensive capital.

Digital Work in the Planetary Market (2022). Edited by Mark Graham and Fabian Ferrari. MIT Press

This chapter shows that the data science labour force, while globally distributed, is predominantly tied to powerful firms concentrated in specific locales. In particular, I argue that the planetary data science labour force is increasingly created by and for powerful technology capital in the USA. The increasing efforts of large firms in producing their own bespoke labour force has implications for digital labour in general.
From the perspective of capital, data science labour-power is a scarce commodity to be competed for. Around the world, efforts are thus being made to proletarianize data science labour-power; to increase its supply and decrease its value, while capturing a competitive share of it. Wider distribution of the skills to perform currently rewarding and well-remunerated digital labour is often positioned as a means to close the economic gap between the Global North and South, but the distribution of such skills is accompanied by their simplification and consequent devaluation. While the proletarianization of data science labour-power may make more data science jobs available outside the Global North, it will do so only insofar as it reduces the labour costs for big technology firms. Rather than the elevation of less-privileged labourers to the digital era, the proletarianization of data science labour-power suggests the coming degradation of a privileged type of labour to the status of precarious and poorly remunerated ghost work.

In this paper, I argue that Marxism is inherently transhumanist because it entails a drive to de-reify nature, including the human being. I argue that the logic of Marxism also requires the temporal inversion of historical materialism, and its projection into the future. This is the transhumanism of Marxism. It is predominantly latent today. Marxists have largely been reluctant to conduct the temporal inversion of their historical materialist perspective, and in doing so have accepted an arbitrarily reified notion of the human. Transhumanists have not. I link Marxism and transhumanism through an ontological concept of suffering. Suffering encapsulates the materialist ontological relation between nature and the human. By tracing how suffering is articulated in both Marxism and transhumanism, I argue that we can get an idea of how to fully work out Marxism's temporal inversion and revive its latent transhumanism.

This paper argues that any discussion of “AI for everyone” needs to be situated within the context of the highly oligopolistic AI Industry. I argue that one should distinguish between the desirability of AI for everyone and the feasibility of AI for everyone. While several Marxian thinkers have argued for the social desirability of AI, the feasibility of an AI for everyone has not been explored. This paper explores this topic by drawing on the notion of “reconfiguration” elaborated first by Jasper Bernes in a critical assessment of logistics infrastructure. I assess whether AI might be “seized” by non-capital, and how, if wrested from capital, it might be operated. I show how contemporary AI differs from the industrial means of production studied by Marx. I conclude that today, a reconfiguration of AI towards social ends seems difficult at best.

This paper examines Friedrich Kittler’s infamous assertion that “media determine our situation” in conjunction with his less-discussed proclamation that the “day is not far off when … the history of communications technologies will literally come to an end” with the advent of artificial intelligence. It explores what Kittler might have meant by suggesting that there could be a situation without media to determine it. First, I survey Kittler’s statements about the end of media. Second, I consider existing interpretations of the end of media and judge them to be inadequate. Finally, I present a reading of the end of media as a horrific event in which the technologically-mediated conditions for subjectivity collapse. This, I suggest, provides support for the notion, advanced by Hans Ulrich Gumbrecht, that Kittler came ultimately to find his media theory abhorrent and thus devoted his last years to studying ancient Greece rather than artificial intelligence.

The thriving contemporary form of artificial intelligence (AI) called machine learning is often represented sensationally in popular media as a semi-mystical technology. Machine learning systems are frequently ascribed anthropomorphic capacities for learning, emoting and reasoning which, it is suggested, might lead to the alleviation of humanity's woes. One critical reaction to such sensational proclamations has been to focus on the mundane reality of contemporary machine learning as mere inductive prediction based on statistical generalizations, albeit with surprisingly powerful abilities (Pasquinelli 2017). While the deflationist reaction is a necessary reply to sensationalist agitation, adequate comprehension of modern AI cannot be achieved while neglecting its material and social context. One does not have to subscribe wholeheartedly to the social construction of technology thesis 1 to allow that the development and evolution of technologies are influenced by social factors. For AI, the most important aspect of the current social context is arguably capital, which increasingly dominates AI research and production. One former computer science professor describes a " giant sucking sound of [AI] academics going into industry " (Metz 2017). In addition to the sensationalist and deflationist accounts of AI, critical scholars should consider the view that capital and its functionaries now espouse. This is the view of AI as a new utility. Andrew Ng, former Chief Scientist at Chinese tech giant Baidu, summed this view up neatly when he proclaimed AI the " new electricity " in 2017. This paper introduces capital's theory of AI as utility and initiates a discussion on its social consequences.

There exists a real dearth of literature available to Anglophones dealing with philosophical connections between transhumanism and Marxism. This is surprising, given the existence of works on just this relation in the other major European languages and the fact that 47% of people surveyed in the 2007 Interests and Beliefs Survey of the Members of the World Transhumanist Association identified as “left,” though not strictly Marxist (Hughes 2008). Rather than seeking to explain this dearth here, I aim to contribute to its being filled in by identifying three fundamental areas of similarity between transhumanism and Marxism. These three areas are: the importance of material conditions and particularly, technological advancement, for revolution, conceptions of human nature, and conceptions of nature in general. While it is true that both Marxism and (especially) transhumanism are broad fields that encompass diverse positions, even working with somewhat generalized characterizations of the two reveals interesting parallels and dissimilarities fruitful for future work.
This comparison also shows that both transhumanism and Marxism can learn important lessons from one another that are complimentary to their respective projects. I suggest that Marxists can learn from transhumanists two lessons: that some “natural” forces may become reified forces and the extent to which the productive apparatus is now relevant to revolution. Transhumanists, on the other hand, can learn from Marxist theory the essentially social nature of the human being and the ramifications this has for the transformation of the human condition as well as forms of social organization compatible with transhumanist aims. Transhumanists can also benefit from considering the relevance of Marx’s theory of alienation to goals of technological advancement.

First, I'm going to talk about why machine learning is a different type of automation from previous forms. Then I will show how the labour of producing machine learning tech is being automated and briefly discuss the relations of workers towards it. I will argue that AutoML complicates critical theorizations of automation from two Marxian schools of thought: labour process theory and post-operaismo. Finally, I will suggest that AutoML indicates the need for a conceptual shift from automation to autonomization.

When artificial intelligence (AI) is discussed by Marxists, it is usually regarded as a new weapon in capital’s ongoing mission to minimize necessary labour time in production. It is seen as a new addition to the history of automation technologies. While capital certainly dreams of fully automated production, if we try to “see” AI like actually existing capitalist firms do today, we find that it is increasingly referred to as a new kind of public utility. Since AI is now largely produced by capitalist firms, it behooves academics to consider this capitalist perspective while attempting to construct a critical theory of AI.
In this paper I introduce this capitalist “theory” of AI as utility and suggest that there are good reasons to believe it is becoming true. If this is so, then AI would not fall in the economic category of fixed capital, as do automation technologies. Instead, it would be a precondition for production occurring at all, and would be more accurately described by the underused Marxian concept of the “general conditions of production”. I will also discuss one effect of AI as a utility which is imagined by the firms that produce AI, as well as the consultancies which advise businesses. This effect is a real-time curation of human perceptual and cognitive faculties.

This paper reads the work of the media theorist Friedrich Kittler as a horror story. Kittler’s life’s work was a materialist theory of media which he elaborated throughout the course of an idiosyncratic history of media technologies. According to Kittler, this history begins with writing and ends with the computer. I approach this “end of media thesis” as if it were, like Shelley’s Frankenstein, a horror story with something to tell us about technology.
Kittler’s sometimes hyperbolic language suggests such a reading – for him the computer is a “monstrosity” (Kittler 2013, 190) which will one day “assume dominion over the world” (Kittler 2013, 286). Like Frankenstein, the end of media is about the danger of autonomous technology. But the end of media goes beyond Frankenstein. In Kittler’s story, the monster is not a deformed human, but rather an inhuman machine. And unlike Frankenstein’s monster, who pines for a companion but is denied one, Kittler’s monster is itself a multiplicity. The horror of the end of media stems not from the mere power of technology to determine human affairs, but from technology precipitating an ontological catastrophe for human subjectivity. By way of closing, I discuss one contemporary development in artificial intelligence (AI) which might be interpreted as validating Kittler’s tale of doom.

This paper is inspired by Jasper Bernes' (2013) critique, in Endnotes 3, of what he terms the " reconfiguration thesis ". He defines this as the revolutionary leftist assumption that " all existing means of production must have some use beyond capital, and that all technological innovation must have … a progressive dimension which is recuperable ". In this paper I pose the question of a communist reconfiguration of artificial intelligence (AI), both the actually-existing and speculative, human-level kinds. I argue that a reconfiguration of existing AI seems difficult, while a reconfiguration of human-level AI seems beyond even the imagination of the radical left.

This paper argues that artificial intelligence (AI) is becoming increasingly significant for capital not only within direct production processes, but also as a component of what Marx calls the 'general conditions of production'. I assert that capitals are coming to treat AI as basic infrastructure, like that of electricity. I draw a number of examples from the current state of the AI industry to support this view and conclude that this suggests that autonomist theories of intellectual labour and the general intellect are misguidedly humanistic.

In this paper I talk about a strange silence that pervades most transhumanist discourse. This silence pertains to the political. Transhumanists frequently make predictions about the future states of complex systems such as the human being and human societies, yet they noticeably tend to leave out mention of anything political. This talk seeks to explain this silence and also to offer a corrective for it.
My contention is that despite all its pretensions to think the human being in a novel, evolutionary way, transhumanist thought continues to posit certain essential qualities to its concept of the human. These qualities are those of homo economicus: or the human defined as atomistic, rational and narrowly self-interested maximizer of profit and utility. The recalcitrance of this conception of the human prevents transhumanists from thinking the political beyond a very narrow scope. This runs contrary to the fundamental directive of transhumanism: to explore new, unfathomed futures for the species currently called human.

In Mindless: Why Smarter Machines are Making Dumber Humans (2014), Simon Head introduces the reader to the occult world of " Computer Business Systems " (p. 4) or CBSs. CBSs are defined as largely invisible " amalgams of different technologies that are pulled together to perform highly complex tasks in the control and monitoring of business, including their employees " (p. 6).

The infamous Kittlerian thesis “media determine our situation” (Kittler 1999, 2) opens a field of inquiry by postulating a determining relation between media and the social existence of humans. There exists a second Kittlerian thesis which has received less detailed consideration, yet which is perhaps just as significant for Kittler’s conception of media. This is the thesis that the age of media has come or will soon come to a close: media have a beginning with writing and an end with the computer (Krämer 2006, 107). Or in Kittler’s (1999) own words: “[w]hat will soon end in the monopoly of bits and fiber optics began with the monopoly of writing” (4). The second Kittlerian thesis demands to be situated in opposition to the first and generates a seeming paradox: humans will exist in a situation without media to determine it. Since the computer is here – this is apparently our situation. So what does Kittler mean?
This paper reads Kittler through the lens of the recent speculative materialist philosophy of Quentin Meillassoux (a product of the situation at the end of media) to see what this situation “says” about itself. I use Meillassoux’s rehabilitation of the notion of being-in-itself to ground a “strong interpretation” of the end of media. On this reading the advent of the computer sees being in-itself, which has traditionally been understood as the stratum of inert objects, taking on qualities usually reserved for subjects, or what some philosophers have termed being-for-itself. This simultaneously portends the human subject being flattened into an object.

Karl Marx theorized capitalism as a relation between labour, capital and machines. For Marx, capital, the process of self-augmenting value appropriated from human labour, is inherently driven by competition to replace labour in production with machines. Marx goes as far as to describe machines as capital's "most powerful weapon" for suppressing working class revolt. Marx, however, could not have predicted the computing machines-such as artificial intelligence-which now form the basis for an increasingly cybernetic capital. Since Marx's time, many Marxist thinkers have sought to apply or update his approach to the cybernetic era. The influential post-operaismo school argues that fundamental revisions to Marx's approach are necessitated by the changed nature of high-tech capital wherein arises a novel "immaterial" type of labour. Immaterial labour, the argument goes, appropriates the machines of capital and achieves a new autonomy from capital, which can no longer control labour and instead, can only attempt to capture the fruits of its autonomous productive capacities. This dissertation's goal is to assess the validity of post-operaismo's claim for a new autonomy of immaterial labour from capital. It does so by conducting an analysis of work in the contemporary artificial intelligence (AI) industry. Work in the AI Industry should be, according to post-operaismo, immaterial labour par excellence. Therefore, this dissertation answers the following research question: does work in the AI Industry evince the new autonomy from capital attributed to immaterial labour by post-operaismo? I argue that it does not. I mount this argument with a multimodal methodology. I employ documentary analysis and qualitative interviews with workers and management in the AI Industry to produce a history, political economy analysis and labour process analysis of the AI Industry. This is followed by a theoretical analysis which assesses the claims of post-operaismo by the example of the AI Industry. I argue that work in the AI Industry remains under the control of capital and that, antipodally to claims of a new autonomy of labour, this industry evinces an increasing autonomy of capital. I conclude the post-operaismo mistakes obsolescence for autonomy.

Most data consists of the recorded actions of people. As new business models premised on the use of data-intensive technologies proliferate, so too does the collection of data via increasingly ubiquitous surveillance. Subsumed under the endless logic of capital accumulation, data collection, and thus surveillance, are expected to intensify. Companies are employing many methods to ensure their access to data such as encouraging “digital resignation” (Draper and Turow 2019). But what if data could be created from scratch? This is the dream driving the latest trend in AI: synthetic data, or data which is generated rather than collected. In this paper, I consider how synthetic data might reconfigure political economic power relations.