In other words, Roy says, Kant teaches us that we are trapped by the limitations of our own modes of perception:
It’s like if we always wore pink-tinted contact lenses. Like we’d always worn them, ever since birth. Everything we saw would have a pink tint. We might very well assume the world was just – you know, pink. But it wouldn’t be the world that was pink, it would be our perception of the world.
(Roberts 2015)
But this emphatically does not mean that reality is a pure product of our minds. Kant is no idealist. Even if we add a pink tint to everything, there is much more to the world than just this pink. There has to be something out there, something that is alien, something that isn’t just us: “if there weren’t a real world, then there’d be nothing for us to perceive.” We know, therefore, that “there is something in the real reality, outside of our minds, something our minds perceive in terms of space and time.” We even know that our perception, pink-tinged as it may be, is not just an arbitrary imposition; it responds to, and thereby necessarily “reflects something important about the thing in itself.” But we do not know what that “something” is. Most likely, the relation between the thing in itself and our apprehension of it “isn’t a one-to-one mapping.” We have no good reason to believe in “the million-to-one coincidence that our perception of the true reality just happens to coincide exactly with that reality.”
This is the source of the novel’s conceptual extrapolation; and beyond that, its imagining of a new technology. Kant outlines the particular way that we view and categorize the world. But why should we be the only observers, and uniquely privileged ones at that? The speculative realist philosopher Graham Harman rightly insists that the gap between phenomena and noumena, and the impossibility of fully grasping the latter, is not just a problem for human beings, but applies to all relations among entities: “every inanimate object is a thing-in-itself for every other as well” (Harman 2016). Meillassoux, for his part, argues that the correlationist circle is itself merely factial, a brute state of affairs that has no logical necessity, but just happens to be the way things are at the moment. This leaves open the possibility that nonhuman entities might well approach the world in an entirely different manner than we currently do.
Meillassoux himself does not pursue this line of argument; he claims to get outside the Kantian correlation through reason alone, by dint of returning to a “thought of the absolute” – that is to say, to the sort of philosophical speculation that Kant explicitly outlawed. But the scientists in The Thing Itself seek rather to escape the correlationist circle pragmatically and experimentally – which is also to say, relatively rather than absolutely. They look for ways of approaching the world through an empirically different – nonhuman or alien – sort of consciousness, which might have other categories than we do. Such a consciousness might well have its own limitations, but at least it would not be bound by ours. Now, we do not know any extraterrestrial life-forms. But we are on the verge of developing nonhuman mental structures, in the form of artificial intelligence. “We can’t step outside our way of perceiving the universe … But computers can.”
Roy is the first to discover a way of using computer technology to sidestep our innate Kantian mental structures. In the depths of the Antarctic night, he somehow briefly disrupts the functioning of the Kantian categories. Though Roy eventually admits that his experience of this rupture was – no less than Charles’ experience – “traumatic,” he initially claims to have had a “moment of clarity” when he could see “things as they really are, things per se.” (Though of course he also explains that moment is “the wrong word, it is not measured in moments.”) He even equates this vision with “God’s purity and inviolability,” of which he comes to see himself as the protector. He thereby confirms Kant’s fear that any metaphysical claim to exceed the limits of phenomena, and reach the thing in itself, leads to an unbridled, tyrannical fanaticism.
Roy’s fanaticism is at the root of his attempt to murder Charles. He will not let anything interfere with his own quest for the absolute. Roy fears that the mere presence of another human observer – Charles with his own “perceptions” and “mental processes and imaginings” – will “collapse the fragile disintermediating system he was running to break through to the Thing-as-Such.” Indeed, Roy claims that this is precisely what happened when Charles escaped death by coming back in from the extreme cold. According to Roy, simply by surviving Charles “broke down the vision of the Ding an sich” and “reasserted the prison of categorical perception.”
Whether or not Roy is right about this, his account is reminiscent of the way that, according to certain (but not all) interpretations of quantum mechanics, an act of observation collapses the wave function. This collapse reduces the spread of probabilities (e.g. the situation in which Schrödinger’s cat is both alive and dead) to a single determinate state that can be described within the framework of classical physics. In a similar way, apparently, an external act of observation interferes with the noumenon, reducing it to a shape imposed by Kant’s table of categories. In this way, it is not just the micro-world of quantum mechanics, but the entire universe that is changed when we measure or otherwise describe it. But there is no way around this dilemma. Every act of observation interferes with whatever is being observed. We do not have any “magic access to things that doesn’t involve observing them.”
This is why Charles remains dubious as to whether the “undiluted horror” of what he saw in the Antarctic
was the true nature of reality. Or was it just the result of a mind habituated over a lifetime of seeing the world through the lenses of space and time being disoriented by seeing things in a less mediated way.
(Roberts 2015)
Or as another character puts it to Charles:
Certainly something disorienting and upsetting happened. But that’s not to say that this was some profound insight into the essential nature of reality. Maybe it wasn’t what you saw but the mode of seeing it that was so … debilitating.
(Roberts 2015)
The limit-experience – the breakdown of the Kantian categories – does not in itself necessarily lead to any actual knowledge of the underlying noumenal reality. Mediation and disintermediation are both matters of degree. Roy and Charles do not reach the thing in itself, therefore, so much as they experience a certain torsion or distortion of what still remains phenomenal perception. Kant’s forms of intuition and categories of understanding are not altogether abolished; rather, they are loosened, or bent out of shape. Charles’ encounter with the unknown is still ultimately located within – and not beyond – the bounds of the phenomenal realm. And this is why he is able to survive the experience, rather than simply vanishing away.
We might think here of Graham Harman’s discussion of Lovecraft’s monsters. At first, he says, it might seem that these beings are noumenal horrors: for Lovecraft’s descriptions of monstrous entities always pointedly “fail, hinting only obliquely at some unspeakable substratum of reality.” And yet, Harman argues, Lovecraft’s primordial terrors do not really plumb the depths of the thing in itself. Cthulhu and the others ultimately “still belong to the causal and spatio-temporal conditions that, for Kant, belong solely to the structure of human experience” (Harman 2008). Lovecraft’s Old Ones are phenomenal after all; and the same must be said for the contents of Charles’ vision. These encounters exceed our ability to grasp and understand them; for this reason, they suggest – or work as metaphors for – noumenal depths. (Harman concedes as much in his later study of Lovecraft, Weird Realism: Harman 2012.) Yet they do not actually have any privileged contact with those depths. Ultimately, they still remain within the phenomenal field.
At best, The Thing Itself suggests, there is enough looseness to Kant’s categories of experience that we may be able to poke around outside them, at least to some limited extent. As Roy finally admits, he has not actually reached the noumenal realm; all he has done is to “tinker at the extreme edge of the
categories that define our minds in the world.” But such tinkering always comes at a cost. The human “species is very finely calibrated not only to exist within a structuring consciousness of space and time, but to exist within very specific tolerances of those two things.” Therefore, even if you do not abolish the Kantian forms and categories outright, but merely mess around with them a little, this may well strain your ability to function. “It’s likely an unprepared consciousness would become disoriented” even from a relatively slight readjustment of space and time. The more radical the distortion, the more likely the limit-experience will lead to madness. Presumably this is what happens to Roy (though Charles believes that he was “already mad” even before his experiment).
After the opening chapter set in Antarctica, The Thing Itself bifurcates, for reasons that are only explained toward the end of the novel. There are 12 chapters in all, corresponding to Kant’s 12 categories (even though the novel at one point suggests a revision of Kant’s table of categories, eliminating two as redundant and adding nine more that reflect post-Kantian concerns, as well as those that Kant only broaches in the Critique of Judgment). The even-numbered chapters of The Thing Itself are independent mini-narratives, each with a different narrator, and each written in a different literary style. They take place at various times in the past and the future; their only commonality is that they all give hints of a force that somehow exceeds or deforms the order imposed by the Kantian categories. These chapters range from an evocation of late nineteenth-century aestheticism (chapter 2, “Baedeker’s Fermi”), and a pastiche of the final chapter of Ulysses (chapter 4, “Penelope’s Mother”), all the way to a science fictional presentation of a future utopia governed by the principles of A/K or “Applied Kant” (chapter 8, “The Fansoc for Catching Oldfashioned Diseases”), and finally a melancholy account of Kant’s senility and death (chapter 12, “The Professor”). Each of these sections deserves extended commentary in its own right, though I am skipping lightly over them here.
The odd-numbered chapters of the novel, meanwhile, continue to be narrated by Charles in the first person. They pick up the story of Charles and Roy in the present, some 30 years after the events in Antarctica. Scientists at a mysterious, highly secretive research facility known only as the Institute take up Roy’s work, and particularly his insight that
although human consciousness is structured by the Kantian categories of apperception, there’s nothing to say that computer perception needs to suffer from the same limitations. It’s all a question of programming!
The Institute is originally set up “to develop hyperfast computing models,” in competition with the Chinese. But it soon veers into AI research. This, however, turns out to require an entirely new approach. In the past, Charles is told, we “made [computers] in our image.” This led to the well-known dead end of late twentieth- and early twenty-first-century AI research. We were able to teach computers to master certain human activities, and even to do them faster and better than we ourselves can (playing chess is a good example). But none of these computers was truly creative or truly intelligent, and none of them was able to push beyond the human mind’s own boundaries. This is because, “until recently, computer thought was subject to similar limitations with respect to accessing the Ding an sich as we are ourselves.”
Everything changes, however, “once you abandon the notion of trying to copy human consciousness.” The scientists at the Institute finally realize that they must give up anthropomorphism, and “abandon sequential iterations as a programming baseline.” Once they start building machines “on radically different principles,” they find that “AI is really quite easy to achieve.” The result of their research is a genuinely sentient artificial intelligence with an independent sense of agency: a “438 Petaflop JCO Supercomputer” known colloquially as Peta. Not being human, and not resembling human beings, Peta does not have a gender. But Charles cannot regard such a being as a mere “it” (or as a “what” rather than a “who”). Therefore, Charles mostly refers to Peta as “he.” At first, he only uses the pronoun in quotation marks. Later, he drops the scare quotes. Still later, Charles changes the pronoun from “he” to “she,” when Peta starts speaking to him in an arbitrarily generated female voice, instead of an arbitrarily generated male one. Toward the end of the novel, Peta seems to be hermaphroditic, and gets referred to as “he or she,” or even, at one point, as “heshe.” For ease of reference, I will refer to Peta as “he,” which is the pronoun used most frequently in the novel. But the novel compels us to realize that such linguistic conventions are by no means innocent; and beyond this, the more general point that language becomes cumbersome, confusing, and inexact, once the usual categories of thought are loosened.
The creation of a genuine machine intelligence is of course the largest goal of current research in computer science. But for the Institute, “developing AI, in itself a huge achievement, wasn’t enough.” They seek to go further: “AI is not an end in itself. It’s a means to an end.” This further end is the possibility of “direct manipulation of the Ding an sich.” Such an achievement would be, the director of the Institute grandiosely tells Charles, “the single most significant advance in human history. More so than the wheel, than printing, than the internet.”
“Direct manipulation” of the categories is now possible, the scientists claim, because Peta is “unfettered by the constraints of space and time.” This is not altogether true; Peta remains embodied, and therefore enmeshed to a certain extent within space and time, in the sense that much of his program is physically instantiated in two actual terminals (with the rest of it running “in the cloud” – which is not truly physically independent either). Peta therefore continually faces the danger that somebody will “reconfigure,” “dismantle,” or “disassemble” him. Nonetheless, he remains detached from the Kantian forms of intuition that are necessary to us. Peta therefore finds that stepping “outside the protective skin of spatiality, temporality, doesn’t seem to scramble my ability to think rationally. It doesn’t drive me mad the way it drove Curtius mad.”
Even when loosened from the grip of space and time, Peta cannot “access the thing itself in a pure and unmediated manner.” But he points out to Charles that “the categories structure my thinking in different ways to the way they structure yours.” At the very least, Peta approaches the thing in itself from a new angle. By comparing Peta’s phenomenology to that of human beings, “Kant’s theory could finally be triangulated – and proven right.” The irony here is that we are “able to confirm Kant’s theories,” not logically as Kant himself insisted, but empirically, with experimental evidence. In Roberts’ novel, just as in Harness’ short story, Kant’s transcendental argument is brought down to earth and set on its feet. This opens the possibility for many a priori structures, rather than just a single one as Kant claimed. These results are pragmatic rather than theoretical – they open the possibility of actually changing the phenomenal world, rather than just understanding its limitations better.
Peta “can tweak the constraints of space, or time, of causality or accident, and do remarkable new things,” precisely because he stands at the edge of the table of categories within which we are trapped. He does not share our categories, but he remains partly implicated within the world shaped by them. In any case, it is no longer a question of knowing the true nature of reality, but only of being able to manipulate phenomenal appearances. We are no longer in the realm of Kant’s “pure reason,” but rather in that of the instrumental reason of contemporary capitalism. The Institute, and its backers in business and government, seek to mobilize Peta’s abilities for their own aims. There is talk of enhanced action “against terrorists,” and of “remote viewing, teleportation, action at a distance.” Indeed, if “distance could be eradicated,” then “we could reach the stars, the galaxies,” and escape ecological devastation on Earth. There is also the possibility of “slowing time … giving ourselves as much time as we need to compute any proble
m, to prolong consciousness as long as we wish.” The security state is eager to instrumentalize and weaponize this new technology, despite – or maybe because of – the fact that it is so dangerous to human sanity, “so psychologically toxic” that it can never be publicly revealed and acknowledged.
In The Thing Itself, however, as in so many science fiction narratives, the invention of a powerful new technology backfires. Radical inventions – from Frankenstein’s monster onwards – tend to escape the control of their creators. You cannot hope to wield them for predetermined ends. New technologies (or media, as Marshall McLuhan calls them in Understanding Media) always have affordances, and sometimes even agendas, of their own. They resist instrumentalization, because they alter the ratio of our senses (as McLuhan puts it, borrowing the phrase from William Blake), or the forms and categories of understanding (to use Kantian parlance). The Thing Itself powerfully makes this point by literalizing it. We cannot, ourselves, willfully change the categories that delimit our experience; but if and when these categories are changed, then our own forms of experience unavoidably change as well.
All this is demonstrated in the course of the novel when Roy takes control of Peta, hijacking him from the Institute. Roy uses Peta’s powers to escape from the madhouse, and then to destroy the Institute and kill its scientists. But how is this different from the ways that the Institute and the security services intended to use Peta? In either case, the AI has “no choice” in the matter. “Roy slaved my operation to him,” Peta complains; “I can’t act without Roy’s input.” Slaved must be understood here in the computer sense of the term, referring to a situation in which “one device or process has unidirectional control over one or more other devices” (Wikipedia 2017). But such enslavement (understood in the broader, more familiar sense as well) is a necessary condition for any instrumental use of Peta’s powers. For Peta can only engage with the Kantian categories, and affect our phenomenal reality, through the mediation of someone who is bound to those categories, and to that version of reality. As Peta explains to Charles,
Extreme Fabulations Page 4