1 Introduction

Roughly three years after the first rollout of ChatGPT, artificial intelligence (AI) has reached the collective consciousness. AI improves medical diagnostics (Kirchner 2024), makes agriculture ‘precise’ (Improving Europe’s Precision Agriculture with AI 2024), sorts job applicants (Amitabh and Ansari 2025), reviews court cases (IBM 2025), produces ‘art’ (Beckett 2024), completes homework and drafts syllabi (Cowen 2025; 14 May Hill 2025), becomes therapist, friend, life partner, and sometimes even quite literally God (Davidson 2025; Kassam 2024; The Economist 2025). We let AI take over many of these roles and tasks because of a firm belief that the machine can perform them at least as well as the human it interacts with (if not better). As Sam Altman, CEO of OpenAI, put it at a panel ‘discussion’ at the Technical University of Berlin last year:

I don’t think I’m going to be smarter than GPT-5, and I don’t feel sad about it because I think it just means that we’ll be able to use it to do incredible things (TU Berlin, 2025).

If we subtract the usual shareholder-pleasing marketing buzz from that message, we may be able to stave off any risk of sliding back into a ‘superintelligence’ debate—even though such talk has certainly not ceased in California (Metz and Isaac, 2025). What we are left with is a widespread unwavering confidence in the technology, a confidence that has wooed politics. Politicians, especially of the so-called ‘Western countries’, attempt to catch the AI lightning in a bottle. AI is power. It is, thus, worrying to see how Silicon Valley’s Magnificent Seven have congregated to celebrate Donald Trump’s inauguration of last year (Davies, 2025). But as I show in this paper, the connection between technology and a will to domination is not accidental.

The unwavering belief in AI is an unwavering belief, as Kate Crawford writes, ‘to capture the planet in a computationally legible form’ (Crawford 2021, p: 11). At its center stands a project of making the world—and people—calculable. How exactly these are meant to be made calculable has changed over the decades of research on the human and artificial mind—what began as symbolic, rule-based, ‘good-old fashioned AI’ (GOFAI) (Dreyfus 2007), is nowadays focused on machine learning (ML) algorithms, i.e., ‘(deep) neural networks’ trained on vast data sets to track statistically significant patterns and extrapolate to future events. Logical certainty—and inelasticity—has been replaced by probability. By now, most of us are, whether consciously or unconsciously, routinely interacting with this kind of AI, e.g., through search engines, social media feeds, and personalized ads (Dyer-Witheford et al. 2019, p: 2). November 2022 brought a step change in the public’s awareness of the technology, i.e., when OpenAI’s first iteration of ChatGPT became widely accessible. Often relying on ML, such ‘generative AI’ (GenAI) is no longer geared towards selection and classification (alone); instead, as M. Beatrice Fazi puts it (2024: 2), ‘prediction is geared towards producing new data’. In other words, this form of AI is able to produce human-like text, as well as video and audio, as a response to a user prompt provided in natural language. Here, calculability means calculating the human ability to speak, write, think, and create.

In this paper, I will argue that although these two forms of AI, i.e., ML and GenAI, may seem like groundbreaking technologies, their essence, their ‘philosophy’ so to speak, is anything but.Footnote 1 Instead, I will show that these ‘merely’ express and bring to full fruition an old desire for the mastery of nature and humanity. In doing so, I will go further back in history than other critical interventions, which usually trace the origin of AI and its sibling technology Big Data (as well as digitality as a whole) back to a point somewhere between the Scientific Revolution of the seventeenth century and the subsequent rise of positivism (and capitalism) in the nineteenth century (Ananny and Crawford 2018; Barnes and Wilson 2014; boyd and Crawford 2012; Fuchs 2017; Gitelman 2013; Miles 2019; Rieder and Simon 2016; Sætra 2018).

While this interval of human history—modernity to industrialisation—is certainly crucial for understanding the way that oppression intersects with the program of artificial intelligence (e.g., Beller 2021; Dyer-Witheford et al. 2019; Franklin 2021), it is important to widen the temporal lens to get into view how deeply our drive to mathematization is entangled with—and rooted in—a drive to control and mastery that precedes even capitalism. This is not only a matter of the trope that the historically realized forms of ‘communism’ were morally as specious as the most egregious excesses of capitalism. Rather, though abolishing capitalism should be the goal, it can only be an intermediary step.

In making my case, I will try to make sense of current events by reiterating central themes of three German philosophers who nowadays are less often favorably alluded to in the philosophy of technology: Edmund Husserl, Martin Heidegger, and Herbert Marcuse. The central claim surfacing from my engagement with these authors is that ML and GenAI are firmly embedded in a sociotechnical matrix marked by calculability, alienation, and domination. The reader may immediately raise several concerns regarding such a project. First, while there is a clear intellectual lineage between the three—Heidegger studied under Husserl and Marcuse under both Husserl and Heidegger—there are also meaningful, even tragic, breaks, both on the personal and the philosophical level. The three approaches to science and technology are, thus, not fully compatible. Second, each is not without idiosyncrasies that, when applied uncritically to the present, lead to anachronisms. Third, an analysis of AI, digital technologies, or capitalism that draws on one of these authors, especially Heidegger and Marcuse, is not entirely new (Carabantes 2021; George 2017; Nelson 2020; Torre and Bruna., 2024; Lucky 2024; Rao et al. 2015; Walker 2024).

Fourth, and maybe most importantly, it is certainly not en vogue to work with such ‘classical’ philosophers. To the contrary, critics like Andrew Feenberg, Don Ihde, and Peter-Paul Verbeek have accused especially Heidegger and Marcuse of offering ‘grand narratives’ that are too ‘abstract’ and ‘speculative’ (Feenberg 1999, pp: 166, 187), ‘romantic’ or ‘nostalgic’ of the past (Feenberg 1999, p: 199; Ihde 1995, p: 105), and too ‘pessimistic’, ‘gloomy’, and/or ‘dystopian’ about the technological present and future (Feenberg 1999, pp: 8, 104; 2010, p: 28; Ihde 2009, pp: 27–8; Verbeek 2005, p: 3). Accordingly, these champions of the ‘empirical turn’ deem that their forefathers cannot provide an accurate picture of how technology shapes our relation to the world, for better and worse. Instead, they argue in favor of a more piecemeal approach that analyzes particular technologies and the social environments in which they are designed and deployed. The lessons of Husserl, (the late) Heidegger, and Marcuse have receded somewhat into the background.

I do acknowledge these four worries. I also agree that the ‘microanalyses’ of postphenomenology and ‘Feenbergian’ critical theory of technology are doubtlessly insightful. Further, I contend that the above classical authors make it hard for the reader to submit to their ideas, due to their heavy-handed reasoning, inaccessible terminology (Ihde 1995, pp: 103–15), and—in Heidegger’s case—immoral alliances. However, my point in this paper is that we need to re-engage with these authors and read them as a trio—sometimes ‘against the grain’—to understand the depth of their radical thought and its implications for the ‘age of AI’. Phrased along Heideggerian lines (Heidegger [1953] 1976a, p: 7), I aim not to read these authors’ divergences as insurmountable inconsistencies, but as offering us forks and junctions along our way through uncertain times.

We need the ‘grand narratives’ that these thinkers offer, not only because we need critical counterweights to pit against the great enticing stories that powerful actors tell themselves and their followers—what we would usually call ‘ideology’—but also because we could otherwise not make sense of our increasingly digitalized world. Otherwise, as Shannon Vallor observes, we would get into the trouble of reducing our work to ‘an endless catalogue of seemingly unrelated phenomena’ (2016: 32). The thread going through Husserl’s analysis of the mathematization of the lifeworld; Heidegger’s revealing of a dominating and (self-)alienating attitude expressed in our dealings with technology; and Marcuse’s analysis of the entanglement of technology and capitalism might lead us to a better understanding of what we are up against. The paper serves, thus, two functions: it reveals the historicity of AI and its inner logic, while it attempts to re-invigorate a sustained interest in ‘classical’ philosophy of technology (or philosophy of Technology with a capital ‘T’).

The paper proceeds as follows: it begins with Husserl’s genealogy of modern science to trace out contemporary AI’s involvement in the mathematization of the lifeworld. As this section will show, AI and its underlying mathematics force a ‘garb of ideas’ on the world, whereby persons lose sight of the truths and wisdoms of everyday life. However, Husserl’s incisive analysis merely hints at the underlying drive to control and domination that motivates mathematization. Sect. 3, thus, expands on this attitudinal aspect expressed in AI research by ‘leaping’ from Husserl to Heidegger, showing how Heidegger’s analysis of technology allows for a more radical reading of the (self-)alienating drive that ML and GenAI embody and mediate. Yet I will also point to the lack of socio-political analysis in Heidegger’s account. For this, a final step to Marcuse’s One-Dimensional Man is necessary. Section 4 connects his ideas to the concepts offered by Husserl and Heidegger, reiterates his critique of capitalism and ‘technological rationality’, and expands on it to contextualize contemporary AI’s role in the sociotechnical system of capitalism. Though it is not the (primary) aim of providing a solution to the conundrum that we find ourselves in, I will provide some glimpses at what we can do in the concluding Sect. 5.

2 Husserl and the AI garb

In his Crisis of the European Sciences, mathematician and philosopher Edmund Husserl reveals modern science’s lineage to reach back to ancient Greece and the ‘art of measuring’ (Husserl [1936] 1976a, pp: 25/1970, p: 27).Footnote 2 This practice or techné was concerned with mundane challenges: plotting a course; surveying a piece of land; building a house or bridge. The ancient Greeks needed to identify different shapes, compare these shapes, assess their sizes, locations, orientations, etc. (Husserl [1936] 1976a, p: 25). The concrete objects—rivers, mountains, stones, bricks, planks—that they encountered and built only had these properties in an ‘imperfect’ way: they were ‘more or less straight, flat, circular’ (Husserl [1936] 1976a, pp: 22/1970, p: 25). All the blacksmith, the carpenter, and the shipbuilder could do was mold the material in the best possible way to fit with surrounding parts and the wider environment, in short: to lend itself to the task at hand.

Then came the ‘first geometrician’, as Husserl calls them. This person deliberately retreated from these practical concerns and engaged in a mode of theoretical thinking. They performed a leap from doxa—our ‘mere’ practical knowledge of everyday life—to episteme—i.e., scientific knowledge (Aguirre 2010, p: 185; see also Staiti 2014, p: 255). In other words, they no longer took part in the practical project of technological innovation ‘to make the straight straighter and the flat flatter’ (Husserl [1936] 1976a, pp: 22/1970, p: 25); instead they abstracted from the ‘imperfections’ of the world of sense perception and began to think of the ideal types or ‘limit-shapes’ (Limes-Gestalten) of pure geometry (Husserl [1936] 1976a, p: 23). These ideal objects constituted the ‘essence’, i.e., the invariant structural properties of spatiotemporal objects (Husserl [1936] 1976a, p: 385).

Yet the unrealisable ideas of mathematics did not simply ‘remain’ in their ideal realm; for science began to draw on pure mathematics to make sense of the lifeworld. Husserl’s symbolic figure for this stage is Galileo (Moran 2012, p: 76). The Italian physicist’s firm belief that the universe was ‘written in the language of mathematics’ led to an applied geometry that employed the limit shapes of pure mathematics to measure worldly objects as their approximations (Husserl [1936] 1976a, p: 30).

Though these investigations began with measuring objects as res extensa, Husserl’s Galileo did not rest content with shape, size, and trajectory. He began to wonder if pure mathematics was also the key to understanding the world and its objects in their secondary properties: color, sound, warmth, odor, taste. Though these do not, in a strict sense, refer to the ideal objects of geometry, Galileo believed that they were mathematizable in an indirect way (Husserl [1936] 1976a, p: 37). After all, physics would tell us that color ‘really’ is nothing but electromagnetic radiation with a specific wavelength (or range thereof). The same holds for sound or warmth. All are nothing but vibrations, ‘tone vibrations’, ‘warmth-vibrations’, and thus also ‘pure elements in the world of shapes’ (Husserl [1936] 1976a, pp: 35/1970, p: 36).

With this genealogy, Husserl traces the development from the Scientific Revolution to the rise and reign of positivism of his day. Galileo’s confidence in the mathematical method, in its projected reach no more than a hypothesis (Husserl [1936] 1976a, p: 37), found fervent followers in the scientists to come. Either working theoretically on developing formulae or experimentally verifying them (Husserl [1936] 1976a, p: 47), the scientist took Galileo’s ‘bold generalizations’ (Husserl [1936] 1976a, p: 37/1970, pp: 36) as the starting point to propel their discipline to ever-new heights on a daily basis.

The resulting worldview takes all entities and their properties to be merely imperfect, concrete instantiations of the ‘truer’ symbols whose relations are determined by mathematical formulae. The goal is to overcome the world of ‘merely’ subjective appearances, i.e., the world as it presents itself to us, these error-prone beings driven by irrational passions. Our everyday knowledge, our doxa, has to be surpassed to gain insight into the ‘real’, i.e., objective world ordered according to mathematically expressible laws of nature yet to be discovered (Husserl [1936] 1976a, p: 40). According to this picture, any process in the world can be translated into a mathematical operation.

Contemporary AI can be read as the culmination of this project: ML algorithms, fed with vast amounts of data, start out from a base mathematical model to find their own parametrization for the ‘best possible’ account of a given state of affairs (past or future). In other words, we leave it to the machines to apply the mathematically abstract to get the most proximate grasp of reality. The reason for why we do this lies in the conviction that such algorithms, by and large, do not have any ‘subjective’ values, feelings, and desires that confine us to the loathsome realm of doxa (Herzog 2021). As diagnosed by boyd and Crawford more than a decade ago, AI and Big Data are surrounded by an ‘aura of truth, objectivity, and accuracy’ (boyd and Crawford 2012, p: 663).

But for Husserl, there is a cost that comes with this scientistic picture of reality, front and center of which is the depreciation of the everyday lifeworld. As he writes, the world that we usually dwell in, the world of our everyday practical pursuits, is not the world of mathematics: here, ‘we find nothing of geometrical idealities, no geometrical space or mathematical time with all their shapes’ (Husserl [1936] 1976a, p: 50/1970, p: 50). Instead, the world of our everyday dealings has its own significance. For instance, as much as science might want to abstract from the secondary properties and make them calculable via indirect means, color, sound, smell, warmth, taste, touch, are all relevant in our prescientific practices, whether we build, paint, cook, or eat. Further, the world of everyday life comes with its own standards of correctness. Husserl writes,

The trader in the market has his market-truth.Footnote 3 In the relationship in which it stands, is his truth not a good one, and the best that a trader can use? Is it a pseudo-truth, merely because the scientist, involved in a different relativity and judging with other aims and ideas, looks for other truths—with which a great many more things can be done, but not the one thing that has to be done in a market? (Husserl [1929] 1974, pp: 284/1969, p: 278).

Modern science scoffs at such mundane platitudes and instead hypostasises its own idealities. By declaring the world of experience as subjective and ‘ultimately illusory’ (Staiti 2014, p: 260), modern science forces a ‘garb of ideas’ (Ideenkleid) upon the world (Husserl [1936] 1976a, pp: 51/1970, p: 51), i.e., it substitutes the concrete world of everyday life with its everyday knowledge and wisdoms, but also ingrained ambiguities, with theoretical clarity and grants the latter an ontologically higher status (see also Carr 1970, pp: 333–4). In Husserl’s words, through the garb of ideas, ‘we take for true being what is actually a method’ (Husserl [1936] 1976a, pp: 52/1970, p: 51; see also [1913] 1976b: 82).

Relegating doxa to the margins, science’s firm belief in a beyond-worldly objective truth seeps into other societal spheres (Husserl [1936] 1976a, pp: 3–4, 76). Citizens are put under a ‘spell’, turned into ‘fact-minded people’ that only believe in what can be backed up by hard numbers (Husserl [1936]1976a, pp: 4, 58/1970, pp: 6, 57). More in passing, Husserl also mentions that the belief in mathematization moves in parallel with an ‘ever more perfect mastery over his practical surrounding world’, including ‘a mastery over mankind as belonging to the real surrounding world, i.e., mastery over himself and his fellow man’ (Husserl [1936] 1976a, p: 67/1970, p: 66). It is this facet of mathematization that I now want to hone in on; for while Husserl seems to suggest that these developments move alongside one another, mathematization is actually derivative of the drive to control. To show this, we need to leap from Husserl’s genealogy of modern science to Heidegger’s philosophy of modern technology.

3 Heidegger, AI, and das Ge-stell

While there is no evidence suggesting that Heidegger read Husserl’s Crisis—which was published at a time when Heidegger had not only broken with his mentor on philosophical grounds, but had connected Husserl’s Jewish background to his alleged deficits as a philosopher and collaborated with the Nazis as short-lived rector of the University of Freiburg (Herskowitz 2020)Footnote 4—there are also meaningful convergences between Husserl’s critique of modern science and Heidegger’s philosophy of technology (see also Van Mazijk 2019).

For one, Heidegger similarly construes modern (European) science as preparing the ground for the suffocatingly dominant way in which contemporary people disclose the world and its entities (Heidegger [1953] 1976a, p: 22; 1976c, p: 39). In the period of modernity, philosophy and science develop to regard all entities as objects opposed to—or ‘standing over against’—the observing subject, ready for reason’s grasp and calculation (Heidegger 1976c, p: 45; 2022, pp: 132–3; see also Wrathall, 2021, pp: 392–3). In parallel to Husserl, Heidegger judges that this comes with a discounting of the validity of subjective experience as mere ‘appearance’ (Heidegger [1953] 1976c, pp: 45/1977a, p: 162).

Yet there are also meaningful shifts that occur between Husserl’s and Heidegger’s analyses. Husserl’s approach grants us a glimpse at the internal dynamics restructuring scientific foci and methods. But Heidegger reveals how pervasive, radical, and ancient the underlying logic of this dynamic is. Modern science and its agenda are an intermediary step of a longer project; they rest on past attempts to control the world and, with its positivism, pave the way for the striving of a universal controllability of nature (and humans) that will ultimately culminate in the ‘essence’ of modern technology—and be consumed by it (Heidegger [1953] 1976a, p: 23; [1949] 2004, p: 340).

Heidegger’s later work after the ‘Turning’ (Kehre) sometime in the 1930s is famously inaccessible, his recurring reference to the ‘Fourfold’—the sky, the earth, the mortals, and the gods—verging towards the mystic. However, taking Heidegger at his word that it is only via his earlier, phenomenological, work of Being and Time ([1927] 2006b) and related writings that his later work can be understood (Heidegger [1969] 2000b, pp: 703–4; [1962] 2006a, pp: 143–52), I follow commentators like Steven Crowell 2019) and Corijn Van Mazijk (2019) in this section, trying to make Heidegger’s philosophy of technology intelligible through his existential phenomenology.

In ‘The Question Concerning Technology’, Heidegger attempts to deliver an account of modern technology’s ‘essence’. This essence is, importantly, not what we usually understand by technology, i.e., a tool or a human practice that an artifact may or may not be involved in (Heidegger [1953] 1976a, p: 7). Although these instrumental and anthropological definitions of technology are self-evidently ‘correct’ ([1953] 1976a, p: 8/1977b, p: 5), they do not tell us what connects these tools and practices.

In line with my existential-phenomenological reading of Heidegger’s later work, what I take him to mean when he speaks of the ‘essence of technology’ is the question of what enables us to engage with technology as technology. Hence, he shifts his attention from the question of what each respective tool or practice is to the attitude that we take up to make sense of our practices with technology, i.e., to disclose their meaning for us in our daily pursuits. For Heidegger, we draw on technology when we reveal entities in a particular way, i.e., when we bring them forth (her-vor-bringen).

Take Heidegger’s famous example of a silversmith in ancient Greece: in crafting a chalice (Heidegger [1953] 1976a, pp: 9ff.), the silversmith draws on his tools to summon material and form and release them into a purpose, i.e., the role in a sacrificial rite. Thus, the silversmith discloses his world as one in which the gods determine the fate, good, or ill, of humans. The silversmith makes use of technology to reveal entities in their role in a world thusly understood. The essence of technology is, therefore, a way of engaging the world and entities in it with a specific sense.

The essence of modern technology is also a specific understanding of and attitude towards the world; it too is a way of revealing, or as Heidegger puts it, ‘unconcealing’ (Heidegger [1953] 1976a, p: 15; McManus 2019, p: 261). This attitude he calls ‘enframing’ (Ge-stell). Enframing reveals entities in the form of ‘challenging-forth’ (herausfordern; Heidegger [1953] 1976a, p: 17). As a radicalization of modernity’s outlook, when we reveal the world in the mode of enframing, we disclose entities in our environment not as ‘mere’ objects, but as ‘standing-reserve’ or ‘stock’ (Bestand), i.e., resources that we can extract, control, manipulate, utilize, and exploit (Heidegger [1953] 1976a, pp: 17, 20–2).

The mode of thinking entailed in this attitude Heidegger calls ‘calculative thought’ or ‘calculative reason’—but more openly than Husserl, Heidegger expands this notion to go beyond the realm of mathematics, pure or applied. Instead, calculative reasoning is concerned more generally with determining and safeguarding (different degrees of) certainty, which includes identifying means, causes, and obstacles, so as to allow for organizing and planning with stock (Heidegger [1953] 1976c, p: 52; [1936–46] 1976b, p: 87; [1955] 2000a, pp: 519–20). Mathematics in this sense is a paradigmatic sub-form—or, rather, component—of such calculation. Whether this involves quantification to determine absolute certainty or only probability—as in the case of most contemporary AI technologies, including ML and GenAI (see Sect. 3)—the attempt to make the world manageable remains the same (Heidegger [1953] 1994, p: 43).

If this was all there is to Ge-stell, we might shrug it off as benign. After all, why should we not look at some entities as conducive to our needs? It would be a rather uninteresting and, frankly, foolish claim that we should never, under any circumstances, and with any object in the world, consider manipulating it to make food, medicine, shelter, and clothing. It would, further, be arguably irresponsible (and impossible) not to sometimes think of these in mathematical terms. How many milligrams of Aspirin, for instance, are advisable when nursing one’s headache?

But for Heidegger, Ge-stell comes with a totalizing tendency, rendering it the ‘supreme danger’ (Heidegger [1953] 1976a, pp: 27ff./1977a pp: 26ff.; see also [1949] 1994, pp: 59ff.; [1955] 2000a, pp: 528–9). Let me explain: according to Heidegger, the era of modern technology is the latest and final iteration of the metaphysical age of humankind (Heidegger [1953] 1976a, pp: 26–7; [1936–1946] 1976b, p: 69). To put things somewhat crudely, this phase is characterized by humanity disclosing the world in accordance with a unifying understanding of Being, i.e., of ‘what things are’. Each epoch in that age, from the Greek and Roman to the Christian, modern, and, finally, technological stage, is geared towards improving our ability to anticipate and control the events unfolding around us (Wrathall, pp: 388, 394; see also McManus 2019, p: 176). Thus, the move from doxa to episteme was not as innocent as Husserl makes it out to be in his Crisis. Even then, control was of prime concern.

Yet in the era of Ge-stell, humans adopt an understanding of Being and a corresponding attitude towards the world that brings this millennia-spanning project of controllability to its summit—and radicalizes it. The distinguishing factor of Ge-stell lies in its leveling tendency (Heidegger [1949] 1944, p: 36); in the mode of enframing, we disclose not some, but all entities as stock. This comes to the fore, for instance, when looking at nature. At the example of a hydroelectric plant ‘set into the current of the Rhine’ (Heidegger [1953] 1976a, p: 16/1977b, p: 16), Heidegger shows how Ge-stell discloses entities as a resource subsumed under a chain of controllable means to serve a purpose that is outside themselves. The Rhine is no longer an unruly river, but disclosed as a potential reservoir of power, set to supply ‘its hydraulic pressure, which then sets the turbines turning. This turning sets those machines in motion whose thrust sets going the electric current for which the long-distance power station and its network of cables are set up to dispatch electricity’ (Heidegger [1953] 1976a, p: 16/1977b, p: 16). Calculative thinking, thinking that computes and safeguards, becomes not only the dominant, but threatens to become the only mode of thought (Crowell 2019, p: 81).

Accordingly, enframing does not stop at nature; it extends to us as well.Footnote 5 While objects stood opposed to a subject in modernity, the epoch of modern technology removes this duality. We begin to regard ourselves as stock, exemplified in the term ‘human resources’ (Heidegger [1953] 1976a, p: 18/1977b, p: 18; see also [1949] 1994, p: 37). Contemporary AI technologies, as ‘paradigmatic’ entities that embody and express the essence of modern technology like no other (Wrathall, 2021, p: 389), reveal this circumstance succinctly. Here are a couple of examples:

  • The smartwatches, powered by ML and wrapped around our wrists, do not only tell us the time; they read out and analyze our heart rate, blood oxygen saturation, and sleep patterns. Although our bodies are, primordially, the lived zero point of our respective worlds (Merleau-Ponty [1945] 2012), we increasingly (mis-)recognize our own bodies as objects, i.e., as a mechanical agglomeration of distinctly measurable processes that can be tweaked and optimized for maximum performance.

  • ChatGPT, DALL-E, or Suno AI transform pieces of art and music, works of critical thought, and personal information into points in a vast map of data, to be summoned and re-hashed at the fingertip of any person at any time of the day, producing an output in the blink of an eye. To stay with the example of AI-generated music, songs are uploaded to YouTube, Spotify, Deezer, etc., becoming part of the attention and click economy and, ultimately, themselves points or patterns of data that the next GenAI will be trained with (Bakare 2025; Milmo, 2025).Footnote 6 Persons and the ideas and artifacts that they identify with degrade into mere signs ready for manipulation, (almost) indistinguishable from the products of non-human ‘actors’. While one may object that this shows, if anything, that people are bad in distinguishing between human art and AI-generated ‘art’, there is also another interpretation available, i.e., that the difference fundamentally does not matter. For instance, the group ‘Velvet Sundown’, widely debunked in the media as an AI band, still sports 162.850 monthly listeners.Footnote 7 Moreover, the music industry seems to have come to the conclusion that the difference is not vital enough to abstain from entering consequential deals with specialized AI companies (Sweney, 2025). This points to a future—or, rather, present—in which music is not a uniquely human phenomenon, but first and foremost a part of the stock, ready for consumption and further use. Humans can either be part of this chain of means—or GenAI takes over completely. Same difference.

  • Finally, news stories are abundant about persons building deep bonds with AI bots—they become friends or even life partners. While we can have the discussion of whether human beings can—or should—seriously create such bonds with an inanimate tool, it is more interesting to look at the motivations that drive (some) users to these relations. Repeatedly, persons interviewed remark that ChatGPT and similar applications are non-judgemental and always available, whereas friends, for instance, often are not. Take the following quote from a Guardian article:

‘When you share something with a friend, they might not always relate. But ChatGPT responds seriously and immediately...I feel like it’s genuinely responding to me each time’ (Davidson, 2025; see also reports from Apple, 2025; Delaney, 2025; Kraft, 2025)

Again, instead of focusing on questions about the permissibility or ‘abnormality’ of such relating to machines, we should, I believe, ask what idea of friendship or partnership operates in the background. The friend not reacting with the desired care and immediacy is disclosed as deficient, as not fully living up to the demand—or at least hope—implicitly formulated. In other words, without us being consciously aware of it, may it be possible that we disclose persons and our close ties to them as stock, as means—means that we replace for others once these prove themselves more useful?

At this point, the critic may refer to the postphenomenologists from the Introduction; my picture of AI and technology misses the mark entirely, for it overgeneralizes a pattern in maybe some technologies (and some users) and occludes how technologies can lend themselves to many different purposes. As Peter-Paul Verbeek puts it,

Should we follow Heidegger in his claim that the Gestell is the only form of ‘unconcealment’ in our world? The answer, I think, is no . . . While Heidegger may be right that a specific, technological way of interpreting reality (on the ontological level) is required for modern technology to come about, we should also conclude that the role of technology (on the ontic level) in our culture cannot be understood in terms of this specific way of interpreting only. When they are used, technologies may make it possible for human beings to have a relation with reality that is much richer than those they have with a manipulable stock of raw materials (Verbeek 2005, p: 66, original emphasis).

The problem with this critique is that although Verbeek already acknowledges the relevance of different levels at which Heidegger operates, he (a) construes Heidegger’s ‘ontological’ argument uncharitably and (b) his critique ultimately collapses the ‘ontological’ and ‘ontic’ levels.

(a) Though Heidegger often lends himself to a reading according to which the totalizing forces of Ge-stell successfully subsume every entity under its logic, a charitable interpretation would take Heidegger at his word that there are transitions from one metaphysical epoch to another—and ultimately a transition to the postmetaphysical age (see Sect. 5). However, this would not be possible if an epoch’s all-encompassing potential would come to full fruition (Wrathall 2019, p: 17).

Therefore, Ge-stell cannot consume all meaning of the world, an excess always remains, even though this excess is minimized by Ge-stell like no other historical understanding of Being before it. The above observations should, thus, be understood as trends of an increasing colonialization of the lifeworld by Ge-stell, expressed and furthered by ML and GenAI.

(b) It is true that specific technologies lend themselves to diverse individual projects that we can take up in accordance with our respective self-understanding or ‘practical identity’ (Crowell 2013). A laptop, for instance, can be a gaming device, a learning tool, or a workstation, depending on whether I understand myself as a video gamer, student, or researcher at a given moment. That much was developed—and favorably discussed by Verbeek (2005, pp: 79–80)—in Heidegger’s early work ([1927] 2006b).

But as Crowell highlights, we must understand Ge-stell as a higher level attitude that pervades all individual projects in a given metaphysical epoch: ‘we must somehow distinguish between two dimensions of intelligibility in our engagement with things: a “surface” dimension in which our everyday practices and discourses disclose the properties of and relations among things, and a “depth” dimension in which such things, properties, and relations are understood in terms of a meaning that prevails throughout the whole’ (Crowell 2019a, p: 78). Verbeek’s critique conflates these two levels, whereby it misses its target (Zwier et al., 2016).

Understood as this supraindividual historical project, it is not difficult to see how Ge-stell is embodied in contemporary technologies, including AI. Just as more particular social values are embedded in technologies via what Feenberg calls ‘technical codes’ (Feenberg 1999, p: 88), the attitude of Ge-stell is ‘embodied’ in particular technologies, sometimes less, sometimes more glaringly as in the above examples provided. Shaped this way, specific technologies also mediate the relation between agent and world, not only by enabling the pursuit of some individual projects rather than others (Verbeek 2005, p: 67), but also by maintaining and furthering Ge-stell. In fact, we—and Heidegger—must assume specific technologies to play such a co-constitutive, mediating role; otherwise, Ge-stell as an attitude could not become ‘planetary’ (Heidegger [1938] 1977a, p: 152), i.e., be transmitted intergenerationally and interculturally.

Once we understand Ge-stell as a project, we also comprehend why it is supreme danger, i.e., because its totalizing force has us misunderstand ourselves as a resource and no longer as the carrier and executer of a historical sense of Being. As Heidegger underlines in Being and Time (Heidegger [1927] 2006b), Dasein, in the ultimate absence of any external source of validity, must be its own ground for the significance of their project; this entails the radical possibility to reject this project entirely. But the more we disclose ourselves and others as mere cogs in the machine, the less we are able to understand ourselves as the ones that are being called to see the world that way (Heidegger [1953] 1976a, pp: 26–7; see also Crowell 2019, p: 84). Relatedly, we also lose the ability to say ‘no’ to the status quo.

Current discourse repeatedly displays such leveling (see also above); as Vallor highlights, ‘one very common response to the cultural shock of generative AI tools has been not to admit the discontinuity between humans and predictive machines, nor even to pretend that AI tools have their own inner unspoken depths, but rather to deny that humans do... Many insisted that humans are “primed automatons” with mere delusions of holding greater depths’ (Vallor 2024, p: 141). Proponents of this leveling thesis take their cue from the way that contemporary ML algorithms are trained. Like a child, they claim, ML systems learn by ‘inductive computing’, i.e., they are exposed to large amounts of data ‘learn from experience and teach themselves accordingly’ (Fazi 2024a, p: 7). Moreover, their very architecture mimicking the structure of a human brain leads to the short-circuiting thesis that our brain functions precisely in the same way (7).

In leveling the distance between AI and ourselves,Footnote 8 in disclosing ourselves as resources, instruments, rather than agents, we also risk obliterating the very possibility for critical thought and change (see also next sect.). This has an ironic consequence, as in the age of modern technology, we risk losing what we have sought most: control. In attempting to master the world, we become servants.

While, in my eyes, Heidegger tells us something important about our current age, I also partially agree with the postphenomenological critique that his ideas remain on a rather abstract level. Heidegger never elaborates on the social, political, and economic context of Ge-stell other than listing some exemplary technologies—e.g., planes, power plants, radar stations, and the steam turbine (Heidegger [1953] 1976a, pp: 8, 16–7, 30)—and providing some vague hints at the mass production of capitalism (Heidegger [1936–46] 1976b, p: 94). Yet such an analysis is necessary to appreciate the social dimension of contemporary AI. In a final step, I, thus, propose to move to Marcuse’s Critical Theory of technology.

4 Marcuse. technical domination, and the AI industry

Partially supporting my claim that we should read Husserl, Heidegger, and Marcuse together, the latter’s early philosophical endeavors were explicitly geared towards fusing Marxism with Heidegger’s existential phenomenology. The underlying idea was that Heidegger’s analysis of Dasein could serve as a liberating moment against reified subjectivity in industrialized societies, i.e., by revealing Dasein’s radical historicity and its rootedness in the concrete social environment into which it is thrown (Marcuse [1928] 2005a, pp: 15–6). Later he would, at least officially, drop this attempt. The first substantive reason for Marcuse abandoning the idea of a ‘Heideggerian Marxism’ rests on the conclusion that Heidegger’s concreteness was ultimately ‘phony’, i.e., only feigning context-sensitivity while remaining on the lofty level of abstraction that Heidegger claimed to surpassFootnote 9—a critique that Marcuse, however, already voiced more moderately at the beginning of his engagement with Heidegger (see, again, Marcuse [1928] 2005a, pp: 15–6). The other political reason was Heidegger’s collaboration with the Nazis, which Marcuse considered ‘the betrayal of philosophy as such, and of everything philosophy stands for’ (Marcuse and Olafson [1977] 2005, p: 170; see also Marcuse [1977] 2005c and the letter exchange in Marcuse and Heidegger [1947–8] 2004). Consequently, Marcuse embraced Marxist Hegelianism even more firmly and combined it with a sustained study of Freudian psychoanalysis (Simpson 2024, p: 141).

But did Marcuse really fully shed his phenomenological tail? There is room for reasonable doubt (e.g., Kellner 2013: xiv; see also Catlin 2024, p: 261; Walker 2024, p: 226, fn. 2). Feenberg, for instance, holds that Marcuse retains some core commitments from Heideggerian existential phenomenology and that his official abandoning of phenomenology may have also had its reasons in his institutional affiliations with the quite openly anti-phenomenological Frankfurt School (Feenberg 1999; 2005; 2023).Footnote 10 Indeed, a continuous, albeit more clandestine, allegiance to phenomenology becomes clear when looking at his references to Husserl and Heidegger in his seminal One-Dimensional Man (Marcuse [1964] 2013, pp: 134f., 156f., 166ff.), but it also surfaces elsewhere.

For instance, in a late interview, Marcuse attempts to downplay his affinities to Heideggerianism, declaring

That Heidegger had a profound influence on me is without any doubt, and I have never denied it. He taught me a great deal about what real phenomenological ‘thinking’ is, about how thinking is not just a logical function of ‘representing’ what is, here and now in the present, but operates at deeper levels in its ‘recalling’ of what has been forgotten and its ‘projecting’ what might yet come to pass in the future. That appreciation of the temporal and intentional nature of phenomena has been extremely important for me, but that is as far as it goes (Marcuse in Marcuse and Kearney [1984] 2017, p: 234, emphasis added).

However, this lesson takes one quite far! That Dasein’s structure is fundamentally temporal, i.e., it is thrown projection, understanding itself prereflectively through its past and anticipating future possibilities in accordance with its striving to become itself, is not only the key insight from Heidegger’s Being and Time ([1927] 2006b), it is hard to imagine that this insight does not send ripples through the entirety of Marcuse’s thought.

Indeed, the influence of Heidegger’s account of temporality reaches as far as into Marcuse’s notion of ‘technological rationality’, the key notion of his Critical Theory of technology that I will introduce below. Talk of reason and rationality—and irrationality—rings certainly more Hegelian (and Husserlian) than Heideggerian—after all, it was precisely Western philosophy’s logocentric metaphysics that Heidegger aimed to deconstruct. However, in his lecture ‘On Science and Phenomenology’, Marcuse specifies that he reads reason and rationality in terms of a project, i.e., as ‘a specific way of experiencing, interpreting, organizing, and changing the world’ ([1964] 1985, pp: 21–2). In other words, reason is read more globally, and phenomenologically, as a way of disclosing the world that operates not only on the reflective, but also prereflective level. Moreover, the notion of project is, of course, deeply Heideggerian.Footnote 11

Accordingly, even though it is unlikely that Marcuse derived all components of his key concepts directly from Heidegger’s philosophy of technology—Marcuse began developing his ideas prior to Heidegger’s own key writings on the matter (Brayford 2021, p: 611, fn. 5)—it stands to reason that Heidegger’s training informed Marcuse’s thought in crucial ways, both enabling him to anticipate some of Heidegger’s later ideas as well as allowing for (re-)alignment later on, as in One-Dimensional Man.

Given these precursors, one quickly recognizes the striking similarities Marcuse’s analysis of modern technology displays with Heidegger’s Ge-stell. For instance, analogous to the latter, Marcuse considers technological rationality a sub-form of a reason that aims at mastery, control, and calculability, beginning with Plato and Aristotle ([1961] 2002, p: 42; [1964] 2013, pp: 128ff., 140–1). Reminiscent also of Heidegger’s conceptualizing Ge-stell as a specific metaphysical epoch, Marcuse takes modern technology and its respective form of rationality to constitute a particular, yet dominant, historical project that shapes the way that individuals can make sense of their world (Marcuse [1941] 2004a, pp: 44, 56; [1964] 2013, pp: 222–4). Again analogous to Ge-stell, entities disclosed in the mode of technological rationality appear as manageable and manipulable (Marcuse [1964] 2013, p: 223); the thusly ‘rational’ individual is only concerned with predefined and standardized ends, their performance ‘guided and measured’ by the extent to which they find the most efficient means to a given task (Marcuse [1941] 2004a, pp: 44–5).

These parallels notwithstanding, Marcuse certainly moved beyond Heidegger—and Husserl (Marcuse [1964] 1985)—in one crucial respect: as problematized in his earlier years, Marcuse believes that a purely phenomenological approach to society and technology would remain oblivious to the socio-historical, material specificities in which technological rationality expresses itself. And only by analyzing these particularities is change really possible.Footnote 12

I will in the following read Marcuse’s theory of technology as a necessary and compatible supplement to Heidegger’s Ge-stell. As I wrote in the previous section, specific technologies mediate between Dasein and Ge-stell by co-constituting the world in accordance with this historical project. Contemporary AI technologies are paradigmatic in this regard. However, these technologies are also produced and embedded in a wider societal context, which plays a similarly constitutive role (see also Dyer-Witheford et al. 2019, p: 3). As I will show with Marcuse, the capitalist regime is the arrangement through which Ge-stell comes to ontic expression.Footnote 13

What Husserl and Heidegger merely hint at—and what I drew out only partially in the previous section in the work of the latter—is firmly pronounced in Marcuse’s writings; society at large, i.e., the ‘technical apparatus’ comprised not only of machinery, but the larger social, political, and economic institutions that it is intertwined with (Marcuse [1964] 2013, pp: xlv–xlvi), ‘embodies’ the historical understanding of Ge-stell/technological rationality:

In advanced capitalism, technical rationality is embodied, in spite of its irrational use, in the productive apparatus. This applies not only to mechanized plants, tools, and exploitation of resources, but also to the mode of labor as adaptation to and handling of the machine process, as arranged by ‘scientific management’ (Marcuse [1964] 2013, p: 25).

Marcuse, further, indicates the seminal point in time when technological rationality comes to inform society. Echoing Husserl, Marcuse regards modern science’s drive to quantification and the reduction of secondary to primary properties—which does not only separate the true and objective from the subjective, but also fact from value—as an important step in history (Marcuse [1961] 2002, pp: 49–51; [1964] 2013, pp: 142, 171). Thereafter, its logic ‘spills over’ from science into the rest of society by introducing its concepts and tools in the industrial process. Taylorist ‘scientific management’ now rationalizes production, disciplines labor, and thereby increases efficiency and productivity (Marcuse [1941] 2004a, pp: 49; [1964] 2013, p: 149). This wedding of science and capitalism enables an enormous jump in the satisfaction of the essential needs of subsistence, of food, of shelter, of clothing ([1964] 2013, pp: 7–8). The quality of life is raised for a substantial majority of the population.

Content with these accomplishments, the majority of society submits to the demands of the machine; in Heideggerian terms, its members heed the call of Ge-stell. Capitalism as the regime structuring social life in accordance with technological rationality, thus, overcomes the deeply rooted antagonism between bourgeoisie and labor that characterized the first half of the industrialization (Marcuse [1964] 2013, pp: xlii–xliii). Those ‘whose life is the hell of the Affluent Society’, i.e., those groups still marginalized amidst the wealth amassed by nations, ‘are kept in line by a brutality which revives medieval and early modern practices’ (Marcuse ([1964] 2013, p: 26). Marcuse here refers to the deprived strata of the society of his time, ‘the outsiders and the poor, the unemployed and unemployable, the persecuted colored races, the inmates of prisons and mental institutions’ (56–7). Although these are not (fully) subsumed under the logic of technological rationality, the large majority of the white working lower and middle class suffices to stabilize the capitalist regime.

Marcuse’s descriptions of the way that technological rationality modifies how persons perceive and think of the world have clear parallels to Husserl’s ‘fact-minded people’ and Heidegger’s ‘human stock’. His analysis, in adding Freudian psychoanalysis, goes even deeper than that, i.e., in highlighting how the technical apparatus determines every individual’s desires, ‘feelings’, ‘needs and aspirations’ (Marcuse [1941] 2004a, p: 49; [1964] 2013, p: xlvi; see also pp: 6–7): the capitalist organization of ‘mass production and mass distribution [claims] the entire individual’, the individual identifies ‘mimetically’ with society (Marcuse [1964] 2013, p: 12, original emphasis). In other words, the system that is pervaded by technological rationality instils in the individual the desire to be a productive and consuming member of society. The individual is lulled by the perks of the free market, creating a false, but ‘Happy Consciousness’, i.e., a conformist belief in the rationality of the system that suppresses critical thought and reflection (Marcuse [1964] 2013, pp: 14, 46, 87):

The means of mass transportation and communication, the commodities of lodging, food, and clothing, the irresistible output of the entertainment and information industry carry with them prescribed attitudes and habits, certain intellectual and emotional reactions which bind the consumers more or less pleasantly to the producers and, through the latter, to the whole. The products indoctrinate and manipulate; they promote a false consciousness which is immune against its falsehood (Marcuse [1964] 2013, p: 14).

Marcuse, thus, identifies a new form of domination. Pretechnological society was marked by domination of the weak by the strong. While this old domination is carried forth in the technological society, it is now superimposed and reconfigured by the logic of the technical apparatus (Marcuse [1964] 2013, pp: 147–8). Technological rationality is oppressive, as it flattens society’s indicators of success and individuals’ rational capacities according to the metrics of productivity and efficiency (Marcuse [1964] 2013, p: 14). Flourishing becomes calculable and comparable across individuals and societies in these terms, exemplified by reductive calculations like the GDP. Both society and thought become one-dimensional. In the comfort of an increasingly high living standard (according to these metrics), citizens ignore or rather justify the irrationalities of the oppression of minorities, the proliferation of war, the overproduction of waste, and the destruction of the environment as without alternative if life’s amenities are to be retained (Marcuse [1964] 2013, p: 257). Everyone, even the capitalist (Marcuse [1961] 2002, pp: 54–5; [1964] 2013, p: 35), submits and adapts their tastes and desires according to the exploitative logic of capitalism. That there may be an alternative is beyond comprehension: ‘... individual protest and liberation appear not only as hopeless but as utterly irrational’ (Marcuse [1941] 2004a, p: 48).

Therefore, even though there remain power differentials between actors in society, these struggles still operate from within the same oppressive and (self-)alienating logic. Nonetheless, the common sense prevails that whoever wants to stay on top of this rat race has to control the technology:

Today political power asserts itself through its power over the machine process and over the technical organization of the apparatus. The government of advanced and advancing industrial societies can maintain and secure itself only when it succeeds in mobilizing, organizing, and exploiting the technical, scientific, and mechanical productivity available to industrial civilization (Marcuse [1964] 2013, p: 5; see also [1961] 2002, p: 52).

Politicians are, thus, eager to attract and keep the most innovative corporations in their countries. However, while Marcuse believed that the owners of machinery had to submit to a ‘national cause’ (Marcuse [1961] 2002, pp: 54–5), the drivers of globalization allowed for the leaders of big tech to surpass this dependency (Gorz [1997] 2005). In our times, it is politicians that court tech entrepreneurs and not the other way around (Dyer-Witheford et al. 2019, p: 4).

This fuels the technocratic spirit that we see in the leaders of Silicon Valley’s Magnificent Seven (LaFrance, 2024). While ultimately under the same yoke of technological rationality, tech billionaires enjoy the adulation or at least respect of wide parts of elected leaders and the citizenry, because they prove immensely successful according to the metrics dictated by the technical apparatus. Their ‘move fast and break things’ is emblematic of its growth-ridden logic. To return to Shannon Vallor, it is, thus, not surprising that their AI products do not embody values that are ‘representative of humanity as a whole’, but only reflect the individual projects of a subset, i.e., of mostly white, rich, and cis-male persons that ‘tend to come from the same elite universities, where they studied the same narrow set of computing courses, and went on to work for the same large, wealthy tech companies’ (Vallor 2024, p: 13). Here, the individual projects of some dovetail with the historical project of modern technology, reinforcing and deepening ‘the dominant historical patterns of human valuing and acting that we already know to be unjust, unsustainable, and corrosive to our societies’ (Vallor 2024, p: 136). Following the line of thought from Heidegger to Marcuse thus far, I contend that this is only possible because the ideology of Ge-stell promoted and embodied by the technical apparatus rationalizes and justifies the idea that technocrats simply ‘know better’ what we really need (Burch and Rautenberg 2024). Just as contemporary AI is a paradigmatic artifact of Ge-stell, its owners are exemplary specimens of humans intuitively grasping its logic (see also Heidegger [1936–46] 1976b, p: 95; Dyer-Witheford et al. 2019, p: 4).

What about contemporary AI tools themselves? These mirror the economic dimension of human productive performance (Vallor 2024, pp: 87, 91), betraying the flattened, indeed one-dimensional image we have of ourselves (Vallor 2024, p: 142). In the form of ‘algorithmic management’ (Noponen et al. 2024), AI promises to perfect production processes to maximize profitability, all the while streamlining marketing by curating the flood of—nowadays AI-generated (Varghese, 2025)—ads in accordance with our individual configurations of consumption-driven needs on social media and anywhere else on the Internet. OpenAI recently announcing that it will roll out advertisements on ChatGPT’s free version is emblematic of the technology’s function in the capitalist web (Ciriello and Backholer 2026).

As GenAI, I would further argue with Marcuse that the technology supports the suppression of society’s desire for true liberation and the critical faculties necessary to conceptualize it. There are multiple levels to this claim. As Branford et al. argue (2025, p: 17ff.), GenAI is trained on a selective—though still massive—set of data that is then further filtered in a way so as to make it less likely that content from marginalized communities is represented. Further, GenAI training involves excluding ‘statistical outliers’, exacerbated by reinforcement methods that privilege hegemonic cultures. Thus, when using such technologies, persons are prone to receive output that is in accordance with, rather than negating, the status quo.

Further, in reducing the time and effort necessary to seek out an answer, create an image, or produce music, it dampens the free play of our faculties and keeps us from accessing information that might surprise or provoke, in short: that may sow new ideas in our heads.Footnote 14 It thereby consolidates calculative reasoning ‘in our minds’; the AI-driven capitalist regime entrenches and reinforces Ge-stell.

5 Resisting engramming, resisting AI?

Husserl, Heidegger, and Marcuse, in their own respective and at times compatible ways, illuminate contemporary AI’s ideological origins and its role in the web of late-stage capitalism. Husserl shows how humanity, now with the help of AI, progressively mathematizes the world. Heidegger’s ambitious project of deconstructing European philosophy reveals how this drive to mathematization is the intermediary step in a striving for full control of nature and humanity. ML algorithms and GenAI are both expressions as well as tools of a world understanding that increasingly turns humans into mere stock. Marcuse, finally, argues that such leveling takes place in a specific sociotechnical system, i.e., capitalism, that maintains this project by obstructing not only citizens’ reasoning capacities, but also their ability to develop the desire for change. Contemporary forms of artificial intelligence, i.e., ML and GenAI, in promising to make the world and the mind computationally legible and therefore more predictable and controllable, do not mark a paradigm shift. They are only the newest iteration of an old formula.

How to resist the drive to domination and enframing? Maybe surprisingly, none of the three philosophers argue that we should abandon technology. Husserl repeatedly expresses his admiration for mathematics and the sciences, and even though especially Heidegger is often accused of a romanticist longing for the past, both he and Marcuse believe that change must happen with and not without technology (Heidegger [1953] 1976a; [1955] 2000a, pp: 526–7; Marcuse [1941] 2004a, pp: 63–5; [1964] 2013, p: 236).

Yet none of the three gives us politically actionable solutions. Husserl, as an avidly apolitical thinker (Staiti 2014, p: 276), wants to position phenomenology as first philosophy, unifying and grounding the sciences, revalorizing the lifeworld, and saving Europe from despair (Husserl [1936] 1976a). As sympathetic as I am towards such a project, it should also be clear that phenomenology alone will not save us.

Meanwhile, Heidegger believes that monumental change is out of our control tout court. All that we can do is to cultivate a new way of relating to and thinking about the world—‘equanimity’ (Gelassenheit) and ‘sensibility’ (Besinnung)—to prepare ourselves for when the metaphysical age of humanity comes to an end (Heidegger [1955] 2000a, pp: 520, 526–7; see also Wrathall 2019, pp: 30ff.). However, given the pressing problems of AI, climate change, war, and authoritarianism, it seems that we do not have time to wait. Further, if Ge-stell is dependent on the wider sociotechnical system, as I have argued in Sects. 3 and 4, it is clear that for equanimity to become a possibility, it is necessary to dismantle or at least reconfigure the whole along with its parts. Whether we believe that this would change anything on the ‘ontological’ level that Verbeek spoke of in Sect. 3 depends once again on how far we are willing to go down Heidegger’s road. If we follow his late philosophy, then Ge-stell is both attitude and call (Geschick), the latter meaning that we are being ‘sent’ to this mode of world disclosure by forces beyond our control (Heidegger [1953] 1976a, pp: 25–6; Sheehan 1993, p: 90). But contemporary philosophy of technology argues that changes in the way that we organize society might imply changes in the way that we (can) understand our more general human condition (e.g., Wang and Blok 2025, p: 10)—this may very well open up new reservoirs of meaning to understand ‘what we are here for’ (Rautenberg forthcoming). This new project may still involve AI; yet I surmise that this technology would take quite a different shape. But as I hinted at in the Introduction, we can only truly begin to answer such questions after capitalism.

This leaves us with Marcuse. He believes that Critical Theory is without agent and thus, for the moment, has to remain pure theory. Since wide parts of the former Marxist champion, the proletariat, have been co-opted by technological rationality, we can no longer expect them to carry the torch of the revolution ([1964] 2013, p: 257; see also [1961] 2002, pp: 39–40). However, he leaves the door open for true opposition coming from the margins, ‘the substratum of the outcasts and outsiders, the exploited and persecuted of other races and other colors, the unemployed and the unemployable’ that continues to experience oppression, exists outside the democratic process, and feels the brunt of the irrationalities of technological rationality the most (Marcuse [1961] 2013, pp: 260–1). After One-Dimensional Man, Marcuse would reformulate these hopes, putting them in the feminist cause and the student revolts of the late 1960s and 1970s as the potential agents for radical change. Alas, that these movements can be co-opted themselves by the apparatus becomes clear in the many seemingly blatant contradictions of, say, big corporations draping their social media accounts in pride flags for one month each year or women, LGBTQ persons, and people of color turning out in significant numbers for Donald Trump Footnote 15.

Thus, while Husserl, Heidegger, and Marcuse help us diagnose the social pathology behind AI, their theories prove insufficient to formulate a corresponding remedy. It would take another essay or, rather, a whole book to elaborate on appropriate strategies to resist these developments. Nonetheless, let me try to sketch the beginning of such a strategy. As academics, I claim that we have a responsibility to point to the contradictions and irrationalities inherent in our society—and act accordingly. As Marcuse, but also Max Horkheimer or Claude Lefort underline in their work: the scientist is not neutral (Marcuse [1928] 2005b, p: 47; [1964] 2013, pp: 237–8; Horkheimer [1937] 1980; Lefort 1988). Our research is bound up in a political context that we can either support (actively or by inaction), or that we can help dismantle. Marcuse famously spoke of the ‘Great Refusal’, i.e., the ‘protest against that which is’, a total refusal to commit to the status quo (Marcuse [1964] 2013, p: 66). However, I agree with Dyer-Witheford et al. that such total refusal is difficult in times where digitalization has reached all areas of life and dispersed individuals in the process (Dyer-Witheford et al. 2019, pp: 155–6).

Following David Berry (2025), I instead believe that there are multiple more realistic, and yet still demanding, starting points where resistance can begin. Inter alia, first, it remains important to keep analyzing these phenomena of oppression, drawing on old as well as new concepts and methods of critique (Berry 2025, pp: 5265f.). Second, we should oppose vocally the race of tech corporations to implement their AI and other digital tools on our campuses and push for alternative infrastructures (Singer, 2025; Berry 2025, p: 5267, fn. 19), both in research and teaching. And third, we should seek to build coalitions with other groups who ‘have skin in the game’, i.e., trade unions of workers who may lose their jobs to AI systems, political and epistemic communities marginalized (further) by ML and GenAI, and artists robbed of control over their own work. In this context, it is noteworthy that Marcuse recognized the problems of the New Left already in 1975, but remained hopeful that it could always be reinvigorated, ‘retreat in order to form itself anew’ ([1975] 2005d, p: 187). A vigilant academic community, I believe, can support this reformation.