Günther Anders in Silicon Valley: Artificial intelligence and moral atrophy

Abstract

Artificial Intelligence as a buzzword and a technological development is presently cast as the ultimate ‘game changer’ for economy and society; a technology of which we cannot be the master, but which nonetheless will have a pervasive influence on human life. The fast pace with which the multi-billion dollar AI industry advances toward the creation of human-level intelligence is accompanied by an increasingly exaggerated chorus of the ‘incredible miracle’, or the ‘incredible horror’, intelligent machines will constitute for humanity, as the human is gradually replaced by a technologically superior proxy, destined to be configured as a functional (data) component at best, a relic at worst. More than half a century ago, Günther Anders sketched out this path toward technological obsolescence, and his work on ‘Promethean shame’ and ‘Promethean discrepancy’ provides an invaluable means with which to recognise and understand the relationship of the modern human to his/her technological products. In this article, I draw on Anders’s writings to unpack and unsettle contemporary narratives of our relation to AI, with a view toward refocusing attention on the responsibilities we bear in producing such immersive technologies. With Anders, I suggest that we must exercise and develop moral imagination so that the human capacity for moral responsibility does not atrophy in our technologically mediated future.

Keywords

Günther Anders Artificial Intelligence digital technology moral agency

Introduction

In May 2018, Google showcased its latest Virtual Assistant tool, Duplex, to an eager audience at a technology conference. Duplex is designed to perform ‘real world’ tasks for busy humans, such as making a booking at a restaurant or scheduling a hair appointment, without the person on the other end of the phone realising that they are talking to a machine. One of the human-machine interactions demonstrated went as follows:

Restaurant:

How may I help you?

Duplex:

Hi, uhm, I’d like to reserve a table for Wednesday the 7th.

Restaurant:

For seven people?

Duplex:

Uhm, it’s for four people.

Restaurant:

When…? Today? Tonight?

Duplex (interrupts):

Uhm, next Wednesday, at 6 pm

Restaurant:

Actually, we reserve [unintelligible] for five people…four people you can come…

Duplex:

How long is the wait usually to, uh, be seated?

Restaurant:

When, tomorrow? Or weekend…or?

Duplex:

For next Wednesday, ehm, the 7th.

Restaurant:

Oh no, it’s not too busy, you can just come…

Duplex:

Ohh I gotcha, thanks!

The bewilderingly ‘life-like’ Duplex system not only convincingly imitates American English speech patterns, disfluencies and inflections, it is also able to react appropriately to the replies on the other end of the line – a human who never seemed to suspect that she was speaking to a computer. The demonstration prompted a number of reactions, ranging from unrestrained enthusiasm, to a distinct sense of terror over the uncanny features of a bot that appears indistinguishable from a human actor. As technology evangelist Chris Messina succinctly put it: ‘Google Duplex is the most incredible, terrifying thing out of #IO18 so far […] The human booking the appointment has no idea she is talking to an AI. Humans quickly becoming expensive API endpoints’ (Messina, 2018).¹ While this latest demonstration of Google’s forays into Artificial Intelligence (AI) neither qualifies as a full-scale Turing Test, nor does it approximate human-level AI, it nonetheless reflects a trend in AI research to first imitate, replicate and then overcome and replace the human in the pursuit of greater efficiency and optimisation. In current narratives – often cast in hyperbolic terms – AI is constituted not merely as a useful product but as a superior Promethean actor; an intelligent agent endowed with the capacity to autonomously direct us toward a brave new future, within which humans are destined to be configured as a functional (data) component at best, and a relic at worst. The global multi-billion dollar AI industry is thus racing, largely unimpeded, toward ever-greater ubiquity in the pursuit of human-level intelligence (Williams, 2018).

It has been the long-held goal of pioneers in AI to produce a ‘thinking machine’ in the image of the human mind; perfectible technological replicas of human thought, which promise to ‘solve [the] kinds of problems now reserved for humans, and improve themselves’ (McCarthy et al., 1955: 1). Today’s ambitions for AI still adhere to the relatively vague pledge to solve humanity’s problems.² In practice, this means optimising everything that can be captured in data and for which a clear goal can be stipulated, to reduce human impact and error, and achieve speed and efficiency. However, future goals for AI researchers and engineers are considerably loftier: to fundamentally aid in overcoming the cumbersome, slow and mortal condition of human materiality, to create a form of ‘superintelligence’ whereby not only our bodies but also our minds will be re-engineered and replaced through technological artifice, ‘tuned-up’ and merged with a growing technological universe (Kurzweil, 2005; Bostrom, 2014). In short, the aim is for AI to move further into the wider realm of human cognition and intelligence, to make human-level machines that ‘can match the intellectual performance of a typical human being in all practically important domains’ (Bostrom, 2014: 408). And once human-level intelligent technologies have been produced, which can work to replicate and improve themselves, it may not be long before ‘the ordinary human is removed from the loop, overtaken by artificially intelligent machines or by cognitively enhanced biological intelligence and unable to keep pace’ (Shanahan, 2015: xvi). In this feedback loop in which human cognition creates human-like artificial cognition, which can then improve itself to optimise functionality beyond human capacity and shape humanity in its own image, Günther Anders’s diagnosis of a hierarchical inversion between humans and their technological products – between creator and creatum – finds its radical culmination (Anders, 2016: 31; 2010: 25).

Anders sketched out this path toward technological obsolescence many decades ago, and his work on ‘Promethean shame’ and the ‘Promethean discrepancy’ provides an invaluable avenue of thinking with which to recognise and understand the relationship of the modern human to his/her technological products. Anders understood clearly that this relationship is fraught, and in mobilising the term ‘Promethean discrepancy’, he draws attention to ‘a growing rift between our technologically mediated ability to collectively influence the world and our individual capacity to feel, and to emotively apprehend what we are doing’ (Müller, 2016a: 12). With contemporary AI, the rift between our products and our moral imagination has become a steep abyss. Anders’s seminal work, Die Antiquiertheit des Menschen Volume 1, was published in 1956, the same year in which John McCarthy first coined the term ‘Artificial Intelligence’ in a proposal for a workshop dedicated to making learning machines (Walsh, 2017: 3). As an astute observer and critic of modern technologies, Anders was attentive to the social dynamics produced through technological ecologies and in his work he diagnoses the undermining repercussions of technological artefacts for political and moral agency with prescient and unusual clarity (Schraube, 2005: 83). In this article, I draw on Anders’s writings to unpack and unsettle contemporary narratives of our relation to AI, with a view toward refocusing attention on the responsibilities we bear in producing such immersive technologies. I do this, paying close attention to the three main theses Anders advanced in Die Antiquiertheit des Menschen Volume 1, namely 1) that we are no match for the perfection of our products, 2) that our scope for producing things outstrips our ability to imagine their impact and take responsibility, and 3) that we seem to assume that it is permissible (if not desirable) to do what we are capable of doing (Anders, 2010: vii). I argue that the desire for AI in general, and the technological singularity in particular, embody the radical realisation of all three of Anders’s theses, culminating in a condition in which our social, political and moral agency becomes increasingly atrophied. By way of a conclusion, I mobilise Anders’s call to expand one’s moral imagination in order to overcome one’s limited political and moral capacities, with a view to imagining the world we would like to live in with AI, instead of becoming subject to AI logic (Crawford, 2018).

If AI is the answer, what is the question?

AI technologies in use today are typically relatively straightforward systems of algorithmic decision-making processes for the optimisation of a specific goal or outcome. This includes mundane tasks like email spam filtering, social media feeds, voice assistants, search engine results and so on. AI is now also widely used in traffic navigation systems, credit application evaluations and increasingly for predictive policing and medical diagnoses. Various communities in the US and the UK have started using facial recognition technologies for law enforcement purposes (Asaro et al., 2018); and China is trialling facial recognition for a social credit system in which citizens get penalised or rewarded for their public behaviour (Lucas and Feng, 2018). The expanding social reach of AI notwithstanding, the technology’s capacity to sift through large amounts of data in a relatively short span of time makes AI systems suitable primarily for specific tasks with very specific goals. In popular discourse and for industry research agendas, however, AI is often conceptualised more ambitiously as a technology that can move into the wider realm of human cognition and intelligence.

AI, as a technology and as a scientific discipline, is not a recent phenomenon. Its origins are typically traced back to Alan Turing’s article ‘Computing Machinery and Intelligence’, published in 1950. Turing’s article represents the first foray into the fundamental AI vision: the ‘hope that machines will eventually compete with men in all purely intellectual fields’ (Turing, 1950: 460; Shanahan, 2015: 1; Walsh, 2017: 13). Turing’s 1950s vision foresaw that by the year 2000, the world would have human-level thinking machines embedded in day-to-day dealings (Shanahan, 2015: 2). This proved to be overly optimistic but not far off the mark. After an inaugural two-month long workshop in 1956 at Dartmouth, organised by John McCarthy and attended by fellow pioneers Marvin Minsky, Nathan Rochester and Claude Shannon, Artificial Intelligence was established as a research domain. McCarthy, and most researchers in the field of AI since then, have defined artificial intelligence as ‘the science and engineering of making intelligent machines’ (McCarthy, 2007: 2), and the idea of ‘pitting the human against machine in some sort of intellectual competition’ has become the definitive public measure of success for AI systems (Guzman, 2019: 85). As a field of inquiry, AI experienced many ‘seasons of hope and despair’ over the decades (Bostrom, 2014: 6). A key breakthrough moment arrived in 1997, when an IBM computer, Deep Blue, beat the world chess champion Garry Kasparov, at his own game. For many, this was the ultimate feat – to outstrip the human in a quintessentially intellectual pursuit. Chess had long been the Holy Grail for thinking machines, and this was the moment the epic battle between human and artificial intelligence was won by the machines. ‘The Brain’s Last Stand’ had been claimed, as Newsweek provocatively announced in the headline to its cover story on the event (Levy, 2017).

Advances for systems like Deep Blue or, more recently, AlphaGo,³ rest on the twinning of developments in computer science with advances in neuroscience (Müller, 2016b: 2). The processing mechanism of artificial deep neural networks crucial to self-learning AI copies the structural set-up of biological brains with considerable fidelity (Shanahan, 2015: 12). The belief that our biological brain is structured to process information like a computer, that we produce thought through abstractions, representations and processing – in short, that our ‘brains themselves are machines’ – runs deep (Noble, 1999: 157; Zarkadakis, 2015). The very foundations of AI, thus, rest on a set of pivotal assumptions: that intelligence can be captured as data, that the human (brain and body) can be understood in its essence as informatics (Thacker, 2003: 74), and that the material structures of human cognition can be replicated through informatic systems. The validity of these assumptions is highly contested (see for example Epstein, 2016). Nonetheless, it is worthwhile looking at what underpins the oft-voiced position that human learning and machine learning take place in very similar ways. As Marcus du Sautoy suggests, much like a machine’s approach to learning, the ‘human learns…by engaging with its environment and learning from its failure’ (Du Sautoy, 2018). Machine learning and human learning are thus put on par, undervaluing the non-linear, sensory dimensions of human learning in the process.

Crucial here is that as a discipline that has ‘made questions about human intelligence and human essence [its] stock and trade’ (Turkle, quoted in Guzman, 2019: 85), AI considers both intelligence and essence to be grounded in mathematics and informatics. In other words, humans are considered first and foremost in their essence as data processors. And in this capacity the human is woefully slow, always inadequate and thus destined for replacement; a mere ‘meat machine’, and a messy one at that, as AI pioneer Marvin Minsky suggests (Noble, 1999: 156). Such conceptions of the human place emphasis on human physique and cognition primarily as carriers of information, processes and data. As data subjects and objects, humans can be captured in their individual biology as well as their statistical relevance within a socio-political body. Once woven into an ecology of data logics, information, compiled through past data, becomes the foundation for future data-cum-knowledge. In turn, socio-political practices are aligned to cater for the perpetuation of this new artificial ecology – states and communities hasten to direct their policy focus on AI (Dutton, 2018); businesses scramble to satisfy the physical demands for an AI infrastructure (Patrizio, 2018), citizens and consumers fuel the necessity for AI with digital existence on a range of platforms.

It is in this turn that the mythology of big data as a ‘higher form of intelligence’ and a shift toward the ‘algorithmic production of knowledge’ as ostensibly objective truth becomes manifest (Cheney-Lippold, 2017: 56–7). We become data beings and, as such, we produce data, and large amounts of it. This data promises omniscience but simultaneously poses a problem by its sheer scale. And it is to the very problem posed by the availability of enormously large amounts of data that AI provides the answer, as DeepMind CEO Demis Hassabis suggests (Hassabis, 2018). Without AI, the vast aggregates of data the human world produces would have little to no meaning. Meaning is now produced through technology, through algorithmic architectures which we may or may not understand, and which may shape us in ways we are oblivious to. We are woven into a networked system of technologies, which inadvertently governs us within specific logics. As such, AI is becoming structural, institutional and pervasive as a powerful but near-invisible organising principle for the body politic. The human, in turn, becomes ‘the last analogue object in a digital universe’ with limited capacity for ‘data input and output’ (Nolan, in Paine, 2018).

Anything you can do AI can do better

Pitted against the data-logic of AI ecologies, the human inevitably pales. Although most researchers and engineers in the field of artificial intelligence attest to the fact that AI systems are nowhere near capable of human-level cognition or intelligence and are not likely to be so in the near future (Darling, 2012: 1; Walsh, 2017; Hassabis, 2018), the promise of computational possibilities for augmenting fallible human capabilities, voiced strongly by a handful of technology faithfuls (see, notably, Kurzweil, 2005, 2012), is enough to bestow a decidedly anthropomorphic quality to our new machine titans. This is evidenced by daily media reports, PR communiqués of technology companies and within dominant AI narratives (Ekbia, 2008). Rarely a day goes by without AI being posited as an outright actor within economic, social and political domains in popular media – ‘AI solves’, ‘AI finds’, ‘AI decides’, ‘AI dreams’, ‘AI predicts’, and so on. Machine learning approaches through deep neural networks produce simulacra of agency and authority. The self-learning capacity of AlphaGo is indicative. Programmers and engineers for AlphaGo are willing (and able) to take responsibility for the system’s actions only up to a certain point, as one of the programmers explains: ‘We just create the data sets and the training algorithms. But the moves it then comes up with are out of our hands’ (Metz, 2016). The system is, deliberately, designed to ‘do things that their programmers cannot anticipate or completely control’ (Gunkel, 2019: 58), and in this it does not merely become agentic but is imbued with agency in the same sense as it is typically ascribed to deliberate human action. The fact that most media accounts credited the AlphaGo system itself with the win, rather than the programmers or engineers who have designed the AI technology, illustrates this further (Gunkel, 2019).

Although many technologists are at pains to assert that AI should ‘enhance the human, not replace them’ (Li, 2018), the hierarchical switch between human and machine clearly manifests itself with the recent advances in AI. Increasingly, the chorus swells: AI is poised to make much better decisions than humans in ever more domains of human life, for we ourselves, as humans, are ‘lousy at it’ (Brynjolfsson and McAfee, 2017: 64–70), or so the thinking goes. Our machines have well and truly become pseudo-persons, as Anders suggested in the 1960s (Anders, 1962: 504), acting in our stead, because we are too fallible to be trusted. This turn toward greater technological authority leaves the socially and politically embedded human in both his/her capacity and willingness to take moral responsibility for one’s actions severely limited.

More than half a century ago, when the very idea of AI was in its mere infancy, Günther Anders examined our relationship with technology with a remarkably prescient eye. Grappling with the perplexities of his own time (specifically the horrifically destructive capacity of nuclear weapons), Anders was nonetheless clearly aware of the contours of our co-constitutive development as technologically mediated humans in contemporary modernity. In the seminal work Die Antiquiertheit des Menschen V1, he advances three theses about our technological condition, which resonate throughout his body of work and which I mobilise below to consider our moral and political agency in an AI ecology. Firstly, Anders suggests, we are no match for the perfection of our technological artefacts. The more capable, powerful and ubiquitous our machines become, the more humiliating it feels to be a mere human, especially in the sphere of labour (economic and social), in which the functionality of machines becomes the standard against which performance is measured, rendering the human perpetually ill-fitted and inadequate in a technologically-normed environment. Anders calls the condition we begin to suffer from in view of machine perfection ‘Promethean shame’. Secondly, and related to his first point, Anders notes that along with the formidable superiority of technological products, the always-antiquated human no longer has the capacity to fully grasp the possibly monstrous consequences of modern machines. This is a critical overreach for Anders in which the capacity to imagine and apprehend the impact of a technology is inversely related to its scale and scope. This leaves an unmediated gap, or slope, in which the possibility to take responsibility for one’s action is severely diminished. Anders calls this differential between producing (Herstellen) and imagining (Vorstellen) ‘Promethean discrepancy’. The third and final key diagnosis he makes is that the modern mandate appears to be such that innovation, and particularly technological innovation, is treated as a moral good. In other words, in relation to our modern machines, we ought to be doing what we are capable of doing in the drive toward progress. Although Anders’s point of reference for the technological human resides in an ‘analogue humanity’, the ‘digital moment […] nevertheless invoke[s] itself at every turn’ in Anders’s writings (Müller, 2016a: 5). All three of the prognoses traced above reverberate with amplified force in our emergent artificially intelligent socio-political context. I will turn my attention to this in the next section.

The human as flaw in the AI system: Promethean shame

Throughout his work, Anders clearly identifies an uneasy relationship between humans and their modern technological products. Rather than accepting technology as mere tools that are created and employed in the service of a specific goal, Anders recognises that technological objects shape and perpetuate subjectivities in modernity. He astutely unpacks how the invisible yet omnipresent technological context of the products within which we live, work and socialise comes to bear on us as individuals as well as a society. In ‘On Promethean Shame’, Anders specifically reveals the precarious tensions that emerge for the modern human embedded in a technological ecology of things (Müller, 2016a: 11). For Anders, the relationship between the modern human and machines is characterised by a reversal of roles in which we as creators and users of machines increasingly become subject to the machine’s power, logic and flawlessness. Implicit in this role reversal is a self-denigration of human capabilities, and fallibility, vis-à-vis our technological creations, as they become the base line for the measure of value, functionality and purpose. Such an artificial machine ecology ‘configures our soul’ in significant ways (Müller, 2016a). As the promise of purpose and flawless machine perfection becomes the benchmark with which to measure worth, value and success, the human begins to look (and feel) like a ‘faulty construction’ in terms of efficiency, strength, speed, precision, or computational capacity (Anders, 2010: 32). The inherent limitations and inflexibility in human physiology, cognition and computational capacity render the human always inferior and unable to compete. This, for Anders, is the basis for a deep-seated embarrassment, or shame, about not-being machine and, as such, being inadequate as a functional element for a machine ecology. The rigidity of our mortal bodies and the relative permanence of our brain structures renders the human unfit to effortlessly function within the rapidly advancing landscape of ever-improving technological artefacts. Within the domain of machines, we are indeed antiquated, a dead weight to the mandate for technological progress (Anders, 2010: 33). This shameful inferiority is impetus, Anders writes, for ‘human engineering’, artificial modifications to the fixed, slow, awkward body, in an attempt to overcome its inherent limitations and make first the body, then also the mind, more appropriate for an ecology of technological machine logics and productivity. In short, in order to continue to exist viably within our machine ecologies, to mitigate the humiliation, we must overcome our mortal condition: we must become machine.

Where the domain of digital technologies dominates socio-political and cultural contexts, human capacity is not merely inferior in a purely physical sense to more powerful machines, but also surrenders in terms of intellect. Here, both body and the very mind that created the digital product are relegated to a position of embarrassing inferiority, in need of artificial improvement. As Shanahan suggests: ‘When the thing being engineered is intelligence itself, the very thing doing the engineering, it can set to work improving itself’ (Shanahan, 2015: xvi). This is, of course, the transhumanists’ dream to be fulfilled by the advent of a complete singularity – to overcome not merely our mortal, decaying, unsolvable, unprogressive bodies, but also adjust our cognitive capacities to be in line with the ‘laws of accelerated returns’, so that we may meld with our technological products and become digital (Kurzweil, 2001). Although transhumanist aspirations belong to a vocal few, the holistic conceptualisation of life as ‘boxed in by competing biological and computational definitions’ is becoming a common place in narratives on AI (Galloway and Thacker, 2007: 79). In fact, we already have become digital, as data to be fed into algorithmic structures for future digital worlds. But as the digital bureaucratic supra- and sub-structures become increasingly complex, the drive to become one with the logic of data flows as the highest form of intelligence becomes all the more futile as we are less and less equipped to see and understand the very logic of the technological intelligence that we weave. Nor is our fallible and flawed human materiality relevant to the further progress of technological innovation. As a political scientist noted with regard to China’s plans to use AI for diplomatic decisions in the future: ‘Artificial intelligence systems can use scientific and technological power to read and analyze data in a way that humans can’t match […] Human beings can never get rid of the interference of hormones or glucose’ (Lemon, 2018).

In a similar vein, the act of anthropomorphising AI represents one more step in the functional integration of the human as an inferior element in a technological universe, perhaps in a much more literal sense than Anders might have envisioned. Nonetheless, as a sharp observer of the tendencies of his time, it did not escape his notice that the delegation of morally significant decisions to an ‘electronic brain’ constitutes a clear manifestation of Promethean shame. Specifically, Anders illustrates this with an account of the Korean War in 1950, when General MacArthur was relieved of his duties, on account of his pugnacious approach to decision-making in the conflict, only to be replaced not with another man but with a technological object – a machine that would determine, objectively, what the best way forward might be in the confrontation; an ‘oracle machine’, or ‘electric brain’, not unlike today’s AI (Anders, 2010: 60). While Anders indicates that this is a well-known story, it is today not easily verified. It is likely that Anders referred in his anecdote to the early advances made in computer technology by AI pioneers Claude Shannon or John von Neumann, who developed much of these technologies under the auspices of the military (Noble, 1999: 153; Kline, 2015: 86). The full details of the anecdote are difficult to trace; nonetheless, Anders’s insights at a time when the concept of ‘thinking machines’ had only just emerged are exceptionally astute and his assessment of the hierarchical shift in which a calculating machine is deemed to make better decisions – not just better decisions than MacArthur might make, but better decisions than any human brain might be able to take – rings truer than ever today. The ‘oracle machine’, like AI, is here posited as exceptional – like humans, but considerably better. The decision to externalise such a morally significant decision to a machine, ‘not because there were specific reasons to mistrust McArthur’s intelligence, but because MacArthur also merely has a human brain’ is, for Anders, a clear manifestation of the perceived lack of human capacity (Anders, 2010: 60).

Whether we take seriously the somewhat exaggerated desire for singularity with superhuman AI, or even just consider AI in the more narrow, practical sense, the prevalent narratives and metaphors of AI as a challenge to human nature, as a superior decision-maker, or as an actor with superior functionalities quite overtly represent the erasure of a once-assumed ‘ontological line…between human and machine’ (Guzman, 2019: 88). This, as Andrea Guzman highlights, may not only be a problematic perception of the self in relation to machines, but may well also lead to the ‘quashing of the need to look for alternative ways of theorizing the self and AI’ (Guzman, 2019: 88). Where we take the machine as the standard for all action, as a measure for our modifications, we also renounce the value and validity of our human selves as a standard and measure for humanity (Anders, 2010: 47). This is evident in the continual deprecation and depreciation of all things human and in the palpable sense of mistrust in the human in contemporary discourses on AI. And it is precisely here that the notion of shame is vital in understanding our (in)capacity for moral and political action within the AI context. The notion of shame is vital – not as a metaphor, but as a perpetual mode of being, a conflicted understanding of the self (Anders, 2010: 66) which, ultimately, limits the capacity for moral and political agency. Shame, as ‘a concern about the body in relation to the mechanisms of self-identity’ (Giddens, 2003: 65), is a condition produced by the incongruity over being ‘un-made’ within a technological ecology which favours fluidity, progress, innovation and change. It is not a condition that humans can be alleviated from, much as one might try through various forms of engineering. Rather, Anders suggests, it is a perpetually disorienting condition of disempowerment to be anything other than un-made, while the technological environment progresses (Anders, 2010: 71, 95). Shame, here, reflects the tacit and oblivious acceptance of obsolescence, while continuing to function along the lines of a technological, capitalist logic. It is this shame, which Anders illustrates with the case of MacArthur, that facilitates the shaping of subjectivities primarily along technological lines of utility (Schwarz, 2018: 156). Moral decision-making, in turn, also takes on the form of cost-benefit calculation.

Again, the case of MacArthur and the oracle machine is instructive and resonates clearly with modes of ethical decision-making today. By deferring to the electric brain, MacArthur was ‘sacrificing his conscience and agency at the altar of the apparatus, testifying to his readiness to submit to the self-made calculation robot, prepared to accept it as a proxy-conscience and oracle, yes, as prediction machine’ (Anders, 2010: 60). By delegating a decision about military (or political) action to a computational machine, the morally challenging dimension of decision-making is effectively rendered in cost-calculative terms, made necessary by the format of the machine. And this is a crucial point. It requires that messy, human environments are made tractable for a computational machine (Anders, 2010: 61, 246). They must be rendered in discrete numerical terms so that variables and factors can be organised, measured and assessed. In algorithmic terms this means gathering data from past events and conditions, which are then used to diagnose, predict (if not prescribe) future events through mathematical processes. Only those elements that do not resist quantification will work for this computational functionality. The less fixed aspects of human life, such as culture, identity, emotion, must either be pressed into a data-logic or left aside for reasons of methodological clarity and purity (Anders, 2010).

Moreover, Anders suggests, the prioritisation of data-driven machine decisions means that humans cannot be counted on to calculate as well as a machine; worse still, humans consider themselves too unsound of mind to authoritatively decide – ‘incapable of taking responsibility (unzurechnungsfähig)’ – so it is prudent to leave the decision to the machine (Anders, 2016: 59). Popular discourses on self-driving cars as ‘moral machines’ illustrate such a decision transfer. Consider, for example, Gary Marcus’s suggestion, published in The New York Times in 2012, that we might enter an era where an insistence to drive a car, rather than be a passenger in a self-driving car, might become morally prohibited, if not legally impermissible, because ‘the risk of you hurting yourself or another person will be far greater than if you allowed a machine to do the work’ (Marcus, 2012). Anders had foreseen such a turn where ‘[b]ad conscience has once and for all been transferred to moral machines; electronic oracles’ (Anders, 1956: 149).

This self-denigration of one’s humanity and humanness in favour of the calculating machine is also evident in discourses on AI-informed policing, as well as Autonomous Weapons Systems (AWS), where proponents hold that technological means make better, less biased decisions of moral relevance (see for example Arkin, 2009, 2010). Despite very vocal critiques against the claim that AI systems are less biased and more accurate than humans (see for example Campolo et al., 2017: 13–20), this is a realm in which AI technologies develop rapidly – suggesting that an ever-greater realm of morally relevant decisions are quite literally automatically outsourced to a machine, leaving the capacity and ability for humans to understand and take responsibility for moral actions atrophied and the ability to contest machine decisions frustrated.

Promethean discrepancy and the duress of progress

Accompanying this deep embarrassment about human inadequacy and the decline in (human) embodied knowledge and authority for responsible moral decision-making is an a-synchronic relation to the speed with which technological products develop and proliferate; a growing a-synchronicity where the conservative materiality and constitution of the human always lags behind the progressive production of ever-greater technological webs. This is the second core thesis Anders suggests in his work – an ‘inverted utopianism’ where, rather than being ‘unable to actually produce what we are able to visualize, we are unable to visualize what we are actually producing’ (Anders, 1962: 496). In other words, Anders identifies a discrepancy between the products we make and our ability to imagine their impact on and within the body politic and, in turn, an inability to take responsibility for the consequences we produce through our artefacts. This ‘Promethean slope or gradient (promethisches Gefälle)’ signifies a ‘growing rift between our technologically mediated’ effect on the world – socially, politically and economically – and our ability to feel, think through, judge and emotionally relate to the consequences of our products (Müller, 2016a: 12). In contrast to the high dynamism and elasticity of our rapidly developing, expanding and proliferating technologies, our (human) capacity to meaningfully relate to the consequences of outcomes, produced through our artefacts, remains comparatively rigid. In short, our own abilities do not develop in adequate proportion to the greater product capabilities we create (Anders, 2010: 267–8).

This Promethean slope has multiple dimensions: it constitutes a gap between our capacity for production and our capacity to imagine the consequences of our products; it also signifies an incongruity between our proficiency in using technological products and our ability to emotionally relate to technologically mediated acts. Moreover, it reflects a growing fracture between what we know and what we can truly understand. We may well, as Anders suggests, be able to invent and produce incredibly powerful nuclear weapons, but we fail to fully imagine the devastating impact their use would have (Anders, 2010: 17, 267). We may well be capable of destroying entire cities and their populations, but are unable to mourn these deaths, en masse, or even in small numbers; neither can we ‘feel’ the terror such acts inflict. These relational categories of imagining, feeling and understanding what our actions mean for others is, Anders suggests, diminished in light of the enormous machineries we produce – unlike our products, we are limited, and can only imagine so much, feel so much, understand so much. The more intricately we are woven into a web of technological artefacts, the greater this incongruity and its ethical significance become. We are ever less able to answer the question ‘what are we doing?’ – an ethical matter of the highest order – when we do not understand the consequences of our technologically produced actions (Arendt, 1998: 5; 1988: 48). Instead, we attend to the question of ‘what are we making?’. We have relinquished our (ethical) agency in favour of homo faber (Anders, 1962: 501). We are, as elements in a technologically networked world, focused on what we can produce, not on what we effect in terms of social or political actions. In this condition, ‘not only has imagination ceased to live up to production, but feeling has ceased to live up to responsibility’ (Anders, 1962: 497).

Anders was concerned in particular about our lacking ability to comprehend and take responsibility for the brutal and horrifying consequences of nuclear weapons technology, for both individuals and humanity at large. However, his diagnosis resonates perhaps even more strongly within the digital context. Where nuclear weapons quite literally dwarf our human capacities by the sheer size of their potentially apocalyptic impact, it is the complex intangibility of digital network structures that exceeds our capability to imagine, feel and understand the impact of digital power. Vast not in physical size but in potential reach and impact, the digital technologies that provide the substrate for AI are monstrous for their tentacular profusion serving as invisible conduits for technological modifications to individual lives and the wider socio-political body. These technologies gradually come to constitute the ‘technical-totalitarian condition’ that Anders anticipates (1988: 53). This might include an AI-supported decision for a medical diagnosis or course of treatment, an algorithmically determined rejection for financial credit, or the identification of a potential terror threat based on AI supported video analysis. It may also include an algorithmically shaped flow of information with the aim of influencing opinions and political positions, as it recently transpired in the case of Facebook data being misused for various political persuasion campaigns. Digital information networks, and specifically AI, are ‘expansionist, if not imperialist’ by nature (Anders, 1988: 50). In order to fulfil the mandate of optimal speed and efficiency, the human world must become fully technologised and datafied. All that cannot be rendered as data cannot subsequently become part of this new technological world. Any non-technological excesses that do not or cannot conform to the principles of the techno-world have no use value and become irrelevant at best, an obstacle to be eliminated at worst (Anders, 1988: 51). This escalates the more complex and diffuse the digital substrate becomes. As Anders notes:

the more complex the apparatus within which we are embedded, the greater its repercussions, the less we can see, the more diminished is our ability to understand the processes of which we are part, or understand their implications. In short, despite being human-made and maintained, our world becomes increasingly opaque as it eludes both our imagination and perception. (1988: 25–6)

With AI, the level of complexity has reached its apex, not only for the black box nature of decision-making processes, but also for the highly diffuse divisions of labour involved in programming, training, using and experiencing AI technology, and the particular invisibility of AI decision-making structures and their societal impacts. Many AI decision-making systems in use today are ‘embedded at the backend of systems, working at the seams of multiple data sets, with no consumer-facing interface. Their operations are mainly unknown, unseen, and with impacts that take enormous effort to detect’ (Crawford and Whitaker, 2016). The invisibility Anders was so acutely aware of has become a feature in today’s AI worlds, leaving the human in the margins of social, political and moral agency, whereby the technological decision – right or wrong, ethical or unethical – sets an incontestable norm. And therein lies the possible monstrosity of the Promethean slope: in the oblivious surrender of responsibility for our technologically produced actions and the ease with which we accept the outcome – ‘I’m just an engineer’, one AI researcher shrugged when asked whether his algorithms to automate the identification of gang crimes could be misused for nefarious purposes or could lead to some unjust outcomes (Hutson, 2018). Where such AI systems are employed in the service of killing, as is the intent of proponents of Lethal Autonomous Weapons Systems, the moral significance of impenetrable machine-logics becomes amplified and war becomes an engineering problem, not a social and political challenge. In this techno-logic of war, killing is a much more likely course of action than discussion, debate or diplomacy and nobody is fully responsible for it. The morality of an action is replaced, in Anders’s words, by the smoothness of functionality (Anders, 2010: 246)

This trajectory toward ever-greater obscurity and diminished moral agency within increasingly complex and invisible digital decision structures is self-amplifying. The more we ignore the social, economic and environmental impact of our technological acts, the more of a challenge they become. In turn, this ‘motivates even more withdrawal, more isolationism and apocalyptic fantasies – and more desperately concocted technologies and business plans’ (Rushkoff, 2018). And here, Anders’s third diagnostic dimension comes to bear: the problematic belief that it is desirable, if not requisite, that we are permitted to do everything that we are able to. Encapsulated in this outlook is the modern dogma of (technological) progress and innovation. This, for Anders, is a crucial consequence of being embedded within a machine world. The very principle of technological artefacts, especially digital technologies, is the drive to faster, better and more efficient action. This is a directional mandate, toward greater, better, and faster productivity. To question or to hamper this trajectory toward constant betterment of functionality is to question the essential quality of our machines, and thus to challenge the logic of the technological worlds we have woven. ‘That which is due becomes what is desired’ (Anders, 2010: 40–1). Everything must be conducted in the service of the free flow of functional progress, so that the idea of perpetual innovation and technological advancement becomes a quasi-moral value which is not to be doubted or impeded. As Anders notes, ‘there is nothing more precarious, nothing that renders a man immediately so impossible, as the suspicion he may be a critic of machines’ (Anders, 2010: 3). The technology sceptic always runs the risk of being categorised as a regressive romantic, and such a stance implicitly posits technological production as an implicit, progressive other. This moral mandate toward functional progress is amplified in our contemporary digital context in which cycles of technological innovation accelerate at an overwrought pace and with unfettered ardour by those that hold technological power. Implicit in this is a capitalist logic which, like the technological logic, works toward expansion. Yet, to question this as a negative aspect of human flourishing is to ‘unintentionally cast oneself as an enemy of the market or anti-technology curmudgeon’ (Rushkoff, 2018). As critics of the current AI hype know, it is both an uphill battle and a race against the clock to try and limit the problematic consequences of a growing AI ecology, not least because this goes against the interests of the top five companies with the largest global market capital (Apple, Amazon, Alphabet, Microsoft and Facebook) which each have the greatest stake in continued AI development and proliferation (Statista, 2018).

Reclaiming moral imagination

The digital structures that shape and determine our futures have long become ubiquitous. They influence our economic, social and political behaviour and choices as they take on an ever-growing remit of human tasks. Embedded in such structures, our humanity becomes an obstacle as it lacks in utility. Where technological ecologies flourish, the human tends to be seen not as a feature but as a flaw in the system, as a bug (Rushkoff, 2018). With the clamouring for ever more AI decisions to determine our lives swelling to a loud chorus, media and PR efforts are contributing to a narrative that sees the human as hopelessly obsolete in an AI mediated world. Günther Anders has offered a very useful set of insights into the foundations of this technological condition and warns starkly against us humans becoming entirely morally incompetent. Moreover, as willing machine material, we are complicit in the deterioration of our social, political and moral agency. We may well, as Marcus envisioned, soon reach a point where

what we really want are machines that can go a step further, endowed not only with the soundest codes of ethics that our best contemporary philosophers can devise, but also with the possibility of machines making their own moral progress, bringing them past our own limited early-twenty-first century idea of morality. (Marcus, 2012)

With AI, the outsourcing of morally relevant decisions to our new technology titans has in itself become a moral mandate to overcome our failings. We are recasting ourselves as the shapeless, useless and shamefully vulnerable clump of clay once forgotten on Epimetheus’s workshop floor – most of us serving only as fodder for the expansionist capitalist hunger of the technologically powerful few. And this imperialist hunger for technological accumulation and expansion is insatiable (Anders, 1988: 50–1). Such a turn comes at the expense of the human world, its capacities, its resources and its humanity.

Despite a somewhat bleak assessment of the modern technological condition, Anders’s work nevertheless also offers the glimmer of a positive humanist avenue from which to resist the techno-apocalypse for humanity: rather than trying to catch up to the standard of functional progress of technology, we should instead develop our human capacity for moral imagination so as to overcome the Promethean slope (Anders, 2010: 271–4). In order to build the world we wish to live in as humans in relation with other humans and our artefacts, we must exercise and build our capacity for that which is a human capability – the ability to relate morally and empathetically to others in the world, lest it atrophies completely. Rather than let matters of the soul – our sentiments and feelings – shrink proportionally to the growth of our technological products, Anders enjoins us to make an active effort to hone and expand our ability to feel; to deliberately stretch and flex our potential for moral imagination (Anders, 2010: 273–4). This was as much the moral imperative for Anders’s time as it is for ours. Our ability to feel should match the ability to produce artefacts with equal immensity. At least this should be our aim: to experiment and practice our self-transformative capacities (Anders, 2010: 275). Rejecting the notion that feelings and moral imagination belong to a natural, immutable order as a calamitous prejudice, Anders posits instead that the capacity to feel is characterised by an inherent plasticity, which, although ubiquitous and continuous throughout history, is rarely acknowledged as such (Anders, 2010: 311). Developing new ways of emotionally adjusting to our condition is nothing new, he suggests. Neither is the willingly initiated expansion of our emotional landscapes – Anders draws here on the example of how our emotional and imaginary capacities grow when we allow ourselves to become enveloped and moved by the potency of musical compositions (Anders, 2010: 313). What is new is for this to be recognised as a dynamic domain which can be exercised and honed. And it is this possibility to willingly expand our abilities to feel, in line with the enormity of our machines, to stretch our souls and reclaim the freedom for moral imagination that Anders considers crucial in overcoming the contemporary chasm of the Promethean slope. This task is neither easy nor does it promise success, but it is necessary for our technological age.

I thus take seriously Anders’s call to reclaim the elasticity of our capacity for moral imagination, and I would like to consider how this might be mobilised in the contemporary context of AI systems. First and foremost, this would include a clear and realistic understanding of the limits of AI for social and political challenges – what is AI good for and which AI technologies are clearly detrimental to humanity? Establishing this would increase our ability to resist the present AI hype and allow space to find non-technical ways of solving human problems. We may, for example, seek to re-imagine the problem of transportation in urban and sub-urban areas and ask the question: what problem do self-driving cars solve? Issues of safety and road deaths are often mentioned in response to this question (see for example Thompson, 2016). But would they also be able to solve the problem of congestion, pollution, limitations in transportation options in a real world setting? Or might a more sophisticated public transport solution, cycle paths and other modes of transporting humans from point A to point B, not be a more progressive solution which offers solutions to road mortality and other issues (Arieff, 2018)? The danger is the old adage: everything looks like a nail when what is at hand is a hammer. Overconfidence in the utility of AI comes at the expense of trying to more thoroughly understand the cultural, social or economic roots of political impasses. Moreover, and importantly, to hone our moral imagination with regard to AI means overcoming the prevailing moral myopia which predominantly considers very specific ethical-technical dilemmas (for example, should a self-driving car kill the passenger or a pedestrian in a challenging situation; should children have AI-equipped toys; should employers use AI to preselect candidates, or indeed should robots have rights, etc.). We should instead try to zoom out to see the wider impact of AI technology, such as the detrimental global ecological impact, the marginalisation of entire populations not embedded in and disadvantaged by technological networks, ‘the exploitation of the many for the benefit of the few’ (Rushkoff, 2018).

In his 1962 ‘Theses for the Atomic Age’, Anders suggests that the imperative ‘Expand your capacity for imagination’ more concretely means ‘Increase your capacity of fear’ (p. 498). This fear, however, is not a conventional, nervous fear. It is a courageous fear, a stirring fear and, importantly, a ‘loving fear, not fear of the danger ahead, but for the generations to come’ (p. 498). Precisely in this resides the possibility for moral responsibility. To wrest our social, political and moral agency from the technological machine world, I suggest, we should exercise our ability to be improper in relation to our technologies, to not acquiesce to the technological economic logic with which we currently shape our socio-politics; to take the magic out of the machines and unveil their immanent power structures and vested interests; to reclaim the body and human soul as a site of strength. This requires courage, vulnerability and love. In his writings and through his actions, Günther Anders has consistently set a valiant example to show that this is indeed possible.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Anders

(1956) Reflections on the H Bomb. Dissent 3(2): 146–155.

Anders

(1962) Theses for the Atomic Age. The Massachusetts Review 3(3): 493–505.

Anders

(1988) Wir Eichmann Söhne. Munich: C.H. Beck Verlag.

Anders

(2010) Die Antiquiertheit des Menschen, Volume 1. Munich: C. H. Beck Verlag.

Anders

(2016) On Promethean shame. In: Müller

(trans.) Prometheanism: Technology, Digital Culture and Human Obsolescence. London: Rowman & Littlefield.

Arendt

(1998) The Human Condition. Chicago: University of Chicago Press.

Arieff

(2018) Automated vehicles can’t save cities. The New York Times, 27 February. Available at: https://www.nytimes.com/interactive/2018/02/27/opinion/automated-vehicles-cant-save-cities.html (accessed 10 July 2019).

Arkin

(2009) Governing Lethal Behaviour in Autonomous Robots. Boca Raton: CRC Press.

Arkin

(2010) The case for ethical autonomy in unmanned systems. Journal of Military Ethics 9(4): 332–341.

10.

Asaro

, et al. (2018) Thanks to Amazon, the government will soon be able to track your face. The Guardian, 6 July. Available at: https://www.theguardian.com/commentisfree/2018/jul/06/amazon-rekognition-facial-recognition-government (accessed 10 July 2019).

11.

Bostrom

(2014) Superintelligence: Paths, Dangers, Strategies. New York: Oxford University Press.

12.

Brynjolfsson

McAfee

(2017) Machine Platform Crowd: Harnessing Our Digital Future. New York: W.W. Norton & Company.

13.

Campolo

, et al. (2017) AI Now 2017 Report. AI Now Institute, New York. Available at: https://ainowinstitute.org/AI_Now_2017_Report.pdf (accessed 10 July 2019).

14.

Cheney-Lippold

(2017) We Are Data: Algorithms and the Making of Our Digital Selves. New York: New York University Press.

15.

Crawford

(2018) You and AI – Machine learning, bias and implications for society. The Royal Society Event, 17 July. Available at: https://royalsociety.org/science-events-and-lectures/2018/07/you-and-ai-equality/ (accessed 10 July 2019).

16.

Crawford

Whitaker

(2016) Artificial Intelligence is hard to see. Medium, 11 September. Available at: https://medium.com/@katecrawford/artificial-intelligence-is-hard-to-see-a71e74f386db (accessed 10 July 2019).

17.

Darling

(2012) Extending legal protection to social robots. IEEE Spectrum, 10 September. Available at: https://spectrum.ieee.org/automaton/robotics/artificial-intelligence/extending-legal-protection-to-social-robots (accessed 10 July 2019).

18.

Du Sautoy

(2018) You and AI – AI applications. The Royal Society, 26 April. YouTube: https://www.youtube.com/watch?v=ZiWC3r5xXn8 (accessed 10 July 2019).

19.

Dutton

(2018) Artificial Intelligence strategies. Medium, 28 June. Available at: https://medium.com/politics-ai/an-overview-of-national-ai-strategies-2a70ec6edfd (accessed 10 July 2019).

20.

Ekbia

(2008) Artificial Dreams: The Quest for Non-Biological Intelligence. New York: Cambridge University Press.

21.

Epstein

(2016) The empty brain. Aeon, 18 May. Available at: https://aeon.co/essays/your-brain-does-not-process-information-and-it-is-not-a-computer (accessed 10 July 2019).

22.

Galloway

Thacker

(2007) The Exploit: A Theory of Networks. London: University of Minnesota Press.

23.

Giddens

(2003) Modernity and Self-Identity: Self and Society in the Late Modern Age. Cambridge: Polity.

24.

Gunkel

(2018) Other things: AI, robots, and society. In: Papacharissi

(ed.) A Networked Self and Human Augmentics, Artificial Intelligence, Sentience. New York: Routledge.

25.

Guzman

(2019) Beyond extraordinary: Theorizing Artificial Intelligence and the self in daily life. In: Papacharissi

(ed.) A Networked Self and Human Augmentics, Artificial Intelligence, Sentience. New York: Routledge.

26.

Hassabis

(2018) You and AI – The history, capabilities and frontiers of AI. The Royal Society Event, 30 April. Available at: https://royalsociety.org/science-events-and-lectures/2018/04/you-and-ai-history/ (accessed 10 July 2019).

27.

Hutson

(2018) Artificial Intelligence could identify gang crimes – and ignite an ethical firestorm. Science, 28 February. Available at: http://www.sciencemag.org/news/2018/02/artificial-intelligence-could-identify-gang-crimes-and-ignite-ethical-firestorm (accessed 10 July 2019).

28.

Kline

(2015) The Cybernetics Moment, Or Why We Call Our Age the Information Age. Baltimore: Johns Hopkins University Press.

29.

Kurzweil

(2001) The law of accelerating returns. Kurzweil Accelerating Intelligence, Blog, 7 March. Available at: http://www.kurzweilai.net/the-law-of-accelerating-returns (accessed 10 July 2019).

30.

Kurzweil

(2005) The Singularity Is Near: When Humans Transcend Biology. London: Penguin.

31.

Kurzweil

(2012) How to create a mind. New York: Viking Press.

32.

Lemon

(2018) China is using Artificial Intelligence to make diplomatic decisions. Newsweek, 30 July. Available at: https://www.newsweek.com/china-using-artificial-intelligence-help-make-diplomatic-decisions-1048734 (accessed 10 July 2019).

33.

Levy

(2017) What Deep Blue tells us about AI in 2017. WIRED, 5 March. Available at: https://www.wired.com/2017/05/what-deep-blue-tells-us-about-ai-in-2017/ (accessed 10 July 2019).

34.

F-F

(2018) How to make A.I. that’s good for people. The New York Times, 7 March. Available at: https://www.nytimes.com/2018/03/07/opinion/artificial-intelligence-human.html (accessed 10 July 2019).

35.

Lucas

Feng

(2018) Inside China’s surveillance state. FT.com: FT Magazine, 20 July. Available at: https://www.ft.com/content/2182eebe-8a17-11e8-bf9e-8771d5404543 (accessed 10 July 2019).

36.

Marcus

(2012) Moral machines. The New Yorker, 24 November. Available at: https://www.newyorker.com/news/news-desk/moral-machines (accessed 10 July 2019).

37.

Messina C @chrismessina (2018) Tweet posted 8 May 2018 10:52 AM. Available at: https://twitter.com/chrismessina/status/993911460945248256?lang=en (accessed 10 July 2019).

38.

McCarthy

(2007) What is Artificial Intelligence. Stanford University. Available at: http://www-formal.stanford.edu/jmc/whatisai.pdf (accessed 10 July 2019).

39.

McCarthy

, et al. (1955) A proposal for the Dartmouth Summer Research Project on Artificial Intelligence. MIT.edu. Available at: http://people.csail.mit.edu/brooks/idocs/DartmouthProposal.pdf (accessed 10 July 2019).

40.

Metz

(2016) Google’s AI wins pivotal second game in match with Go Grandmaster. WIRED, 3 October. Available at: https://www.wired.com/2016/03/googles-ai-wins-pivotal-game-two-match-go-grandmaster/ (accessed 10 July 2019).

41.

Müller

(2016a) Prometheanism: Technology, Digital Culture and Human Obsolescence. London: Rowman & Littlefield.

42.

Müller

(2016b) New developments in the philosophy of AI. In: Müller

(ed.) Fundamental Issues of Artificial Intelligence. Synthese Library 376. Springer Verlag.

43.

Noble

(1999) The Religion of Technology: The Divinity of Man and the Spirit of Invention. New York: Penguin.

44.

Paine

(2018) Do You Trust This Computer? Documentary Film, Diamond Docs/Papercut Films.

45.

Patrizio

(2018) The AI revolution has spawned a new arms race. Ars Technica, 7 September. Available at: https://arstechnica.com/gadgets/2018/07/the-ai-revolution-has-spawned-a-new-chips-arms-race/ (accessed 10 July 2019).

46.

Rushkoff

(2018) Survival of the richest. Medium, 5 July. Available at: https://medium.com/s/futurehuman/survival-of-the-richest-9ef6cddd0cc1 (accessed 10 July 2019).

47.

Schraube

(2005) ‘Torturing things until they confess’: Günther Anders’ critique of technology. Science as Culture 14(1): 77–85.

48.

Schwarz

(2018) Death Machines: The Ethics of Violent Technologies. Manchester: Manchester University Press.

49.

Shanahan

(2015) The Technological Singularity. Cambridge, MA: MIT Press.

50.

Statista (2018) The largest 100 companies in the world by market value in 2018 (in billion U.S. dollars). Statista.com. Available at: https://www.statista.com/statistics/263264/top-companies-in-the-world-by-market-value/ (accessed 10 July 2019).

51.

Thacker

(2003) Data made flesh: Biotechnology and the discourse of the posthuman. Cultural Critiques 53(Winter): 72–97.

52.

Thompson

(2016) Why driverless cars will be safer than human drivers. Bloomberg News, 17 November. Available at: https://www.businessinsider.com/why-driverless-cars-will-be-safer-than-human-drivers-2016-11?r=US&IR=T (accessed 10 July 2019).

53.

Turing

(1950) Computing Machinery and Intelligence. Mind 49: 433–460.

54.

Walsh

(2017) Android Dreams: The Past, Present and Future of Artificial Intelligence. London: C. Hurst & Co. Ltd.

55.

Williams

(2018) Why AI will win the global race for complete AI dominance. WIRED, 16 April. Available at: https://www.wired.co.uk/article/why-china-will-win-the-global-battle-for-ai-dominance (accessed 10 July 2019).

56.

Zarkadakis

(2015) In Our Own Image: Savior or Destroyer? The History and Future of Artificial Intelligence. New York: Pegasus Books LLC.