Stacked Presencing
Tokenizing NSM primes involves turning the declarative sentence into or treating it as an output of the linguistic data stored as ostensive-imperative-ostensive circuits. I’ve taken the concept of an ostensive-imperative-ostensive circuit from Eric Gan’s example of the everyday working of ostensives and imperatives in The Origin of Language, which is the familiar “dialogue” between nurse and surgeon which has the surgeon requesting a “scalpel” with the nurse providing it while repeating the request: “scalpel.” My reading stretches this example, which only involves an imperative confirmed, in its fulfillment, by an ostensive, but nevertheless suggested to me a version of a kind of base reality (replacing talking of “percepts,” “sensations,” etc. and other concepts drawn from philosophy and psychology) which involves the world continually communicating with us by presenting us with things for our attention which entail imperatives (respond to, handle, this thing in some way, attend from this thing to some other thing, etc.) which we fulfill and confirm (or authenticate, consecrate, etc.) through some ritual repetition of the ostensive. Through this constant churning or mining of the tokenized world some imperatives prolong themselves into interrogatives which concern the availability of a particular ostensive but this is mediated through some obstacle to fulfilling the imperative. Think about how often questions have a trailing off tone, a kind of bridge between the hesitation caused by a problematic imperative and the “hope” for some restoration of the linguistic presence that is put at risk. The declarative is a kind of proposed bet that something we could recognize as the restored ostensive is out there, somewhere, and at least potentially available, if certain ostensive-imperative-ostensive conditions were to be fulfilled. The specific sentence lays out the conditions of that bet which, of course, may in turn raise any number of other questions, soliciting any number of subsequent sentences. The sentence, then, creates a field of possible ostensive-imperative-ostensive (from now on: OIO) sequences while weighing the likelihood of anyone of them playing out in such a way as to restore presence and/or instruct us in revising our expectations regarding what should “register” as an ‘authorized” presence. Clearly, I’m drawing upon the discourses of search engines, data search and machine learning to provide us with a vocabulary for extending originary grammar into the stack of scenes constituted by search engines, data search and machine learning.
What we are doing in language, through our idioms, is summon intentions embedded in the stack of scenes in the form of the trillions of OIO circuits deposited there. The initial imperative issued by or deriving from an ostensive is to intensify attention and since by “attention” I always mean joint attention, intensifying attention means multiplying its joints, spreading it out. What a thing, or, in full-fledged language, a scene centered in a particular way, wants (commands) you to do is join it, add it to the scene you’ve entered it from and carry it over to the next scene you’re going to. To be a carrier of the scene, which requires mining all the OIO circuits embedded in it, calling into motion all of your capabilities to do so. I’ll use the language of “intention” (which, more originarily, is just an outgrowth of the kind of attentional dynamic I’m describing here) because that’s the way questions of automated text generation tend to be taken up, as in the discussion in Critical Inquiry I discussed in a recent essay of mine which I’ve already linked to on several occasions. That online theory forum took as its starting point Steven Knapp and Walter Benn Michaels’s “Against Theory” from 1982—Knapp and Michaels there argued that the meaning of a text is its author’s intention, toward the identification of which interpretative efforts should therefore be directed, while the forum was launched to explore the possibility that AI, or automated text generation, had rendered this understand of text and authorship problematic (or, more precisely, called for a revisiting of the way in which poststructuralism had already rendered it at least theoretically problematic). My way of advancing what seemed to me, in the forum, a stalled discussion, was to take up Paul de Man’s understanding of “intentionality” which is essentially that the intention of a text is first of all to be read. This allows us to set aside the conventional approach to “intention,” which is usually to present a familiarized summary of the text (what Melville is trying to say in Moby Dick is….), and take up the question of what is involved in constructing, preparing and disseminating a text to be read. The author’s reliance upon genre conventions, specific histories, the language he uses, contemporary publishing conventions, etc., however much we want to attribute awareness of all this to the author, becomes part of our understanding of his intention because this all comes into producing a text to be read. Furthermore, intending a text to be read begs the question of “read by whom,” and, setting aside local inquiries regarding how a given author might have constructed his immediate readership, any author is aware, however vaguely, that there are many readers who will all make sense of the text differently and whom the author will want to serve as carriers of the text because wanting the text to be read means wanting it to be read by a steady readership, whether spatially or temporally and, even more, to be, if not the only text that is read (and reread…) the text through whose frame all other texts will be read—that is, if it’s a text worth reading—but, if the intention of producing the text is to have it read, part of that intention must be to make it “worth reading.” We have now passed through Harold Bloom and Leo Strauss territory so that we can get to the point that writing aims at generating an entire textual field within which that text will continue to operate pedagogically: it must produce its carriers, which is to say its promoters, defenders, imitators, even enemies, etc. The intention of the author is as much to do something as to say something. Such an understanding of intention takes us far away from the banality of Cliff Notes while still guaranteeing very close, sustained, multilayered study of a given text deemed worthy of such attention. For one thing, we want to learn to do, on a completely different terrain, whatever it did.
But none of this was really the “intention” of the essay—where I wanted to get from here is that if all of the supports and “affordances” of the text are part of the intention of the author, an intention that we, as readers, are prolonging, then the entire make-up of the world, all of the furnishings of all the scenes, are marked by the intentions of all the forms of authorship, from literary to administrative, that have gone into composing those scenes; even more, that in acting upon those scenes we are activating those intentions, or, again, prolonging them. No one, for example, can think seriously about politics today without asking where a particular political figure stands within the stack, in relation to the entire oscillation between debt and the intelligence agencies, as mediated by all the institutions and disciplines through which these oscillations are articulated. Fewer and fewer can take seriously the old commonplaces of political discourse about one’s “stand on the issues,” “getting your message out,” one’s “record,” etc. All of that stuff is merely symptomatic, far downstream. Politicians have intentions, and if they are sufficiently securely situated within the stack, we might indeed find threads from what they say they want to do and at least some of the results of their actions—but that just directs our attentions to their sufficiently secure situation within the stack—and, of course, securing that security is also part of their intention. But all of these features of the stack are themselves, and have been for some time, largely automated—as the academic Straussians like Harvey Mansfield showed, central to the argument for “popular government,” or the “extended republic,” or what was to become known as “liberal democracy,” was that the grounding of governance in the routines of the marketplace would make citizens predictable and controllable—which means that governance has for a very long time involved the pulling of levers, the oiling of machinery, rebalancing following disturbances, cybernetically taking in feedback from the environment and, now, “optimizing” the totality of social interactions.
Machine learning, then, merely further automates a largely automated process. It may have the added advantage of disabusing us of whatever liberal illusions of “public discourse” and “rational exchange” remain. But we also know that the machinery has never really worked as “intended”—indeed, much like current “AI hype,” “liberal democracy hype” has always been a series of propaganda campaigns that have performatively falsified its own idealisms. Governance remains a matter of “intentions”: the nomos must be expanded and attended to and the juridical needs to be preserved. We can, though bring declarative intentions in line with the tacit OIO background by thinking of governing intentions (and all intentions are governing intentions) as “stacked presencing,” that is the work of maintaining presence across the stack of scenes and through the reverberations of that stack of scenes. Processing the primes through originary grammar is the path toward making language programming and making it currency, against, I would guess, the discoverers of both these linguistic hypotheses (but, maybe, we can redeem Wierzbicka’s and Gans’s intentions by prolonging them through the stacks). Such idioms are to be as iterable as they are defamiliarizing and plastic; they are to attach terms from across the stack of scenes while being irreducible to instructions issued on any of them. They are to lend themselves to satire and parody and, even more, to deploy satire and parody as emissaries ensuring the spread of the idiom.
I think at this point I’ve done enough idiom mining that the sequence ostensive-imperative-interrogative-declarative, on the one hand, and ritual (originary distribution, nomos)-juridical-disciplinary can serve as allegories of each other for generative purposes. Of course, ritual is not purely, ostensive, the juridical is not purely imperative, and the disciplinary is not purely interrogative/declarative—we’re speaking of weights here: declaratives issued on a ritual scene aim at guaranteeing the presence or effectivity of the ritual object, ostensives on the juridical scene flow more directly into imperatives while declaratives are more closely tied to the justification of imperatives; in disciplinary spaces, meanwhile, OIO circuits need not be guaranteed—their hypothetical or virtual character can be posited. I’ll remind you of, or mention for more recent readers, Peirce’s elaborate semiotic mappings where he fractalized his icon-index-symbol trichotomy to the nth degree (implicitly unlimitedly). The continual generation of language out of and its return to the primes is the medium through which these allegories are performed. Think, know, do, and so on stretch out as words like “faith,” “assume,” “hypothesis,” “consider,” “perform” and so on as words are treated and prepared to operate on scenes that refer to, include, nestle within, form an infrastructure for, other scenes. A “species” of the “family” “think” has emerged upon the semantic field held by the primes because of all the ways it has been deployed to maintain presence under specific threats to that presence—insofar as we take the primes as our point of departure, a threat to presence might be a threat to the boundaries upon which portions of the semantic field cleave (i.e., separate and connect). A threat to presence, then would be a threat to the boundary articulating “think” and ‘know,” or “do” and “happen,” or “same” and “other,” or any articulation of these provisional opposites. If we set up a kind of Laputa-style laboratory for the study of language we might construct hypothetical models showing how a single prime might mutate into all of the words that would rely upon it for its definition or explication simply by working through all the meanings that word would accrue as it was used in various forms of address to various audiences, as an ostensive or exclamation, across the imperative continuum, as a question growing in complexity by being prolonged through a series of answers and, finally, as a declarative deriving an imperative from the field of reality so as to issue an implicit command to suspend demands until a space in which the desires indexed by those demands can be created. That’s the work of stacked presencing—to perform the stack of scenes as an ever problematic, never guaranteed prolongation of the originary scene. A unit of currency is a miniaturized replication of the originary scene, made available as a possible ostensive horizon of a declarative.
Wierzbicka’s NSM primes provide us with elementary building blocks out of which language is composed—any word or statement or discourse or, for that matter, gesture, can be explicated using the primes. Out of elementary building blocks we then build secondary building blocks and tertiary building blocks, and so on. This is what it means to construct an idiom: as much as possible you draw upon or, in fact, let it be explicit that you build upon, “pre-fab” phrases and sentences. While literacy was more directly based on orality, in particular oratory, it was a given that the production of texts involved working with inherited commonplaces to be mastered by the writer or speaker. With the unmooring of literacy from orality, which perhaps can be conveniently located in Romanticism, the way is opened for writing to become programming and language to become performative in this new way. Idioms now need to be manufactured rather than inherited, and a good way of thinking about how to do this is to work on inventing sentences that have something paradoxical and imperative about them and that are readily iterable and take on more paradoxicality and imperativity the more they are repeated; moreover, these little idioms maintain consistency when one pre-fab phrase is switched out for another, taking on a kind of Ship of Theseus character, wherein it remains the same throughout, even because of, changes. Generating collisions at the boundaries of the primes across the semantic field through words further downstream up the ostensive>imperative>interrogative>declarative circuit is the method for doing so I’m proposing here. Why rely upon philosophy or any of the disciplines derived from it when we can endlessly explore the relations between the primes manifest in the language we already have: where “think,” “feel,” “know” come up against each other and generate new combinations which, in turn, come up against the boundary between “do” and “happen,” and then between “same” and ‘other” and so on. Wierzbicka provides comprehensive and convincing explications of individual words but once her method is brought into collaboration with originary grammar and stacked presencing explications will become much more complex and debatable: try to explicate an entire sentence into the primes and you will end up with something far more elaborate and questionable than what Wierzbicka provides us with. Such a discipline will remain highly grounded in certain “fixed” terms that in turn lend themselves to infinite and orderly variations. The implications of this for training LLMs is to be explored.