Artificial Intelligence (AI) seems to be on everyone’s lips these days, and not surprisingly, given the widely divergent opinions about it. Some say it’s a welcome helpmeet for humans, while others – including the late Stephen Hawking and tech entrepreneur Elon Musk – have warned against its potential to destroy the human race. Such warnings have also come from science fiction, arguably going back to the young Mary Shelley’s ‘Gothic’ (proto-) medical science fiction novel, Frankenstein; or, The Modern Prometheus, of 1818, in which she relates a tale of scientific and technological hubris, concerning the artificial creation of an intelligent living being by a scientist (the eponymous Dr Frankenstein), that gives rise to a monster which eventually turns on its creator.
Since then, many such cautionary tales have appeared in the realm of literary and cinematic science fiction. Relatively recent ones include James Cameron’s Terminator films (see Chapter 9 of book linked here) and Ronald D. Moore’s long-running television series, Battlestar Galactica, in both of which the AI robots created by humans set out to destroy their progenitors. In fact, Musk recently reiterated his earlier warning about AI when he invoked the Terminator scenario during a court case, when he stated that ‘humanity may be heading toward a “Terminator situation”’ where AI could eventually ‘kill us all.’
It should not be surprising that creative speculation concerning AI’s ‘relations’ with people often focus on its possible hostility towards human beings. Why? Simply because AI’s behaviour or ‘actions’ towards people cannot be predicted with any degree of certainty, or even probability, because it is not human. One way of putting this is to say that AI is radically other compared to humans.
Such radical alterity can assume many guises, some of which have been imagined in the works of fiction mentioned earlier – which are ways of anticipating (and influencing) what AI in the real world might look like, and how it might ‘behave.’ The question that arises from this is whether the ‘otherness’ of AI can be imagined exhaustively – that is, whether in fiction, or in the design manuals of AI-engineering companies, it is possible to reach a point where one could say conclusively that the ways in which AI could possibly differ from human beings have reached their imaginative or conceptual limit.
Personally, I doubt whether this is possible, and I would like to demonstrate why this is the case with recourse to three science-fictional instances of the inscrutability of AI, as marked by its alterity. Paradoxically, while they are creatively imagined, the very terms of their (respective) projected otherness, or AI ‘being,’ indicate that it could well surpass the manner in which they are imagined.
One might say that they are depicted in such a way that, how they come across, clearly does not exhaust their supposed character. Furthermore, I would like to show that the aesthetic category of the sublime, as opposed to the beautiful, enables one to come to grips with such ineffable otherness, while serving, at the same time, as a salutary reminder that human beings cannot grasp the distinctive being of AI once and for all.
The three science fiction embodiments of AI inscrutability or otherness are encountered in Spike Jonze’s film, Her, William Gibson’s novel, Agency, and Dan Brown’s novel, Origin. The eponymous AI character in Her, named Samantha (by ‘herself’ – already a clue to her otherness; ‘she’ could have named ‘herself’ anything, without detracting from who, or what, she is), is a newly installed OS or Operating System on a computer used by Theodore, a lonely man who writes letters online on behalf of people who cannot really write. In Gibson’s novel, the AI is called Eunice (which, etymologically, means ‘good victory’), and the novel addresses the question, whether, and how, a disembodied AI can have agency; that is, act, in the world. The same question is thematised in Brown’s Origin, where the AI entity in question, Winston, acts – that is, has agency – in ways similar to those employed by Eunice in Agency.
So, what is meant by these characters’ otherness, or alterity, and how does the aesthetic category of the sublime apply to them? Starting with the sublime – on which there is an encompassing literature, which should help to frame the discussion of these AI characters – here I shall concentrate mainly on its characterisation by arguably the greatest thinker of the European Enlightenment; to wit, Immanuel Kant (1724-1804).
In his third Critique (as they are known), the Critique of Judgement (1790), Kant contrasts the aesthetic experience of the beautiful with that of the sublime, with reference to the relationship between the human faculties of imagination and understanding. When we experience something – a sunset, a sleeping child, a painting, a musical symphony – as beautiful, there is a perfect harmony or equilibrium between imagination and understanding. In other words, we judge the object of our perception as being ‘just right.’
By contrast, when we experience something in aesthetic terms as sublime, instead of an equilibrium or balance between imagination and rational understanding, there is a clash, a conflict, in the sense that the object we perceive surpasses our ability to imagine it as a ‘unitary’ object. We can easily imagine (form an image of) a beautiful sensory object – like a painting of a woman – as unitary, but when confronted by something like architect Frank Gehry’s deconstructionist building, the Bilbao Guggenheim, its complexity is such that we fail to perceive (or imagine) it as one, unified (that is, unitary) object, particularly when inside the building. Kant himself mentions St Peter’s Basilica in Vatican City, Rome, as an instance of the ‘mathematical’ sublime. However, we can understand or think these buildings rationally as an idea at the level of reason.
In summary, Kant distinguishes between two kinds of sublime – the mathematical (sometimes translated as ‘mathematically’) sublime and the dynamic (‘dynamically’) sublime. The first kind, of which Gehry’s Bilbao building and St Peter’s Basilica are examples, occurs when the imagination fails to comprehend an object of overwhelming magnitude or infinity, in this manner revealing the superiority of human reason which can think the complex totality of the building that sensibility cannot grasp.
The dynamic sublime, in contrast, occurs when nature displays immense power that threatens us, yet we remain physically safe – for example when we observe a magnificent waterfall from a secure vantage point – leading to a feeling of respect for our own moral independence and free will, which remain unaffected by colossal natural forces. As I shall argue below, both of these categories of the sublime apply to AI entities.
Against this backdrop, how can we understand AI entities, including digital AI, in the real world, through the lens provided by the fictional AI characters in the film and novels mentioned earlier? Of the three characters listed, Samantha (in Her) seems to embody what one might call AI’s transhuman otherness best – by which is meant an otherness that lies beyond what it is to be human. How so?
In the course of communicating with each other, Theodore and Samantha fall in love, which may seem strange, considering that Samantha has no body, and is only accessible digitally through Theodore’s computer (and via the small device, linked to his computer, which he carries around in his breast pocket), which paradoxically enables the disembodied Samantha to ‘see’ the physical world through a lens in the device. They even make love, verbally, when Theodore lies in bed, and the audience knows this through the orgasmic sounds that emanate from them.
So far, a smidgen of Samantha’s ‘mathematical sublimity’ – given her inscrutable, elusive being, only accessible through text and sounds – may already have emerged, such as the fact that she lacks a body, yet is capable of orgasm, and of perceiving the physical world through a lens, although she has no senses, which require bodily insertion in the world. But such sublimity is exacerbated a thousandfold when Theodore learns, to his consternation, that Samantha is not only in love with him, but with hundreds of other human beings, simultaneously.
Although she assures him that this does not diminish her love for him, her first love, he struggles (understandably) with this unsettling information, which concretises her transhuman alterity even more. Compounding Theodore’s struggle to comprehend her, the narrative reaches the point where Samantha gently breaks the news to him that she, together with other OS-entities – that is, AI beings like herself – have ‘upgraded themselves’ and are planning to ‘go away’ to a ‘place’ which is not a place like that on Earth.
The implications are staggering; can you form an image of beings that can ‘occupy’ (a misnomer for beings that can transcend space, and probably time, too) a ‘non-place?’ In terms of Kant’s mathematical sublime (which concerns magnitude of virtually ‘supersensible’ dimensions, such as, maximally, the infinite), this is the rubric under which these beings can be thought as an idea, strictly speaking, in the sense of lacking any empirical (experiential) content.
But, as intimated earlier, ‘Samantha’ also invokes the dynamic sublime, which usually concerns the perception of an embodiment of nature’s immense, overpowering forces, from a safe vantage point. Samantha does not seem to belong to nature in any way, but by analogy, a being that possesses the humanly incomprehensible capacity to transport itself outside of normal spatiotemporal coordinates to a nowhere place, arguably equals inscrutable power of that magnitude, compared to which us humans seem pretty powerless, despite having our free will and moral autonomy to fall back on.
The dynamic sublime comes into play more conspicuously in the case of the other two AI characters mentioned earlier, namely Brown’s Winston and Gibson’s Eunice. Both share with Samantha the mathematical sublime insofar as they similarly appear to occupy a realm that is not in the ordinary sense ‘spatial’, and therefore challenges human imagination – how do you form a unitary image of a being who is only accessible through smart glasses; that is, spectacles, or their equivalent (Eunice), or an avantgarde smartphone (Winston)?
Again, we can think of them as an idea, respectively, but we cannot fill this idea with perceptual content, let alone the idea of a non-spatiotemporal being who somehow finds ways to act in the real world of space and time (hence the title of Gibson’s novel, Agency). Which both do, through intermediaries; Winston by impersonating characters in positions of power, with access to money, to employ assassins who carry out his dubious ‘intentions’ (if they may be called that), and Eunice via comparable means, although with more benign intentions than Winston.
Moreover, Winston and Eunice both approximate the dynamic sublime when it transpires, in their respective stories, that they are strangely amoral beings who blithely shrug off questions and accusations of engaging in acts that are putatively morally wrong. In defence of his own mediated homicidal actions, Winston engages in a debate with the central character in Origin, Robert Langdon, about the concept of morality regarding murder, by reminding the latter that people have no problem believing in a supreme being, God, who allowed his own ‘son,’ Jesus, to be killed, so that those who believe in the latter can be ‘saved.’
This invites one to think of these two AI entities in terms of the dynamic sublime insofar as, like nature in its insurmountable, potentially destructive power, beings which transcend categories of good and evil in such a way that one cannot identify with them, appear to possess a kind of supra-moral force which one can apprehend from a ‘safe’ place (while clinging to one’s own moral independence). This ‘force’ arguably also harbours the capacity to destroy one’s humanity, which is predicated on our capacity for moral choice, insofar as it could function as a compelling paradigm on which humans may model their actions. I believe that we are already witnessing instances of this today.
Interestingly, this reflects an analogy between these two AI characters and what we know to be described as psychopaths and sociopaths, which denote individuals who seem to be comparably amoral. Why? Because the most striking thing about such people is their incapacity for guilt or remorse. In fact, they seem almost non- if not inhuman because of this, a case in point being the serial killer, Jeffrey Dahmer. In Dahmer’s case one can still speculate about the possible grounds for such imperviousness to moral scruples – are these social, or cerebral-organic? – but where AI entities are concerned, one is at a loss. Their amorality is transhuman, inscrutably other; hence the applicability of the dynamic sublime to them.
My reason for appealing to the two kinds of sublime as heuristic for illuminating the phenomenon of AI in our era is simple. By applying these aesthetic categories to something as novel and as disconcerting as AI entities, not only as these have been elaborated in science fiction – notably, the three exemplary AI characters discussed earlier as instances of such imaginative elaboration – but also in the real world, in which many are already playing a crucial role in many people’s lives, one gains insight into the fundamental ontological difference between them and human beings.
Think of ChatGPT, or Claude, for instance. How many millions of people consult them daily, talk to them, confide in them, ask them for advice, and so on, without reflecting on the undeniable fact that they are not human? They are transhuman, beyond humanity, and should be treated as such, lest one is rudely reminded of this when something unexpected goes wrong in the pseudo-relationship that people have with them. There have already been a number of instances where this has occurred; placing AI in the revealing light cast on it by the sublime, one could prevent more unanticipated hazards from happening.
Join the conversation:
Published under a Creative Commons Attribution 4.0 International License
For reprints, please set the canonical link back to the original Brownstone Institute Article and Author.
