As Brian Roemelle points out, at the center of technology now is prompt engineering, which is essentially the designing of search terms that would elicit results from searches into textual databases (i.e., AI) that will most powerfully suggest the design of a new search term. As Henrik Karlsson (and, I must say, me) have said, prompt engineering or search term design goes well beyond database searches and reach into all social networks—your search results will situate you within the data, waiting there to be found via another’s search. This is arranging and treating your data as part of data exchanges, with others but ultimately the center, the repository of all data. Roemelle is also aware that this places language back at the center of technology, as even coding is coming to be generated through verbal search terms. But I think a better account of grammar and language as a mode of inquiry can be provided than Roemelle does, so I’d like to begin by presenting an explanation of grammar that I have used in my classes but can’t remember ever working through here or on the GABlog, and then move beyond that to a technologizing of language as, seemingly paradoxically, a mode of wisdom literature (which it seems to me now can incorporate the transdisciplinary discourse I have worked to turn GA into).
I present the declarative sentence as the unfolding of an anticipated dialogue embedded in the sentence. We begin the subject, something you could mention, name or point to; about this subject or topic we could expect one to ask, what about it? The answer to that question is the predicate: what about the horse? It’s running, or it’s brown, or it needs to be shoed, etc. We then bring in the modifiers, first of all the adjectives, which answer the possible question, which or what kind of [subject]? The brown horse, not the black one, the fast horse (not the slow one). (It seems to me that “which” has priority here, because singling out one object of interest out of several possible ones is logically prior to the more abstract categorization of types.) The adverbs, meanwhile, answer the possible question, how or in what way: the horse runs quickly, not slowly, it walks nobly, not nervously. I don’t fall into the trap of defining or explaining prepositions (although Anna Wierzbicka does think this can be done, and I’m sure she’s right, even though I don’t think it helps much in the analysis of sentences) for the simple reason that prepositions always either are or (much more often) go along with nouns to form adjectives or adverbs, which are already accounted for. (“I don’t fall into the trap…” which, or what kind of trap? The trap of defining.) But I have never bothered with conjunctions, or articulations of clauses, because it has seemed to me less important that students have a way of talking about them. But this is a very important issue for prompt engineering, and, grammatically, I think we can focus less on the paratactic “and” (interestingly not one of Wierzbicka’s primes) and more on the hypotactic “if,” ”but” and ‘because.”
I’m going to come back to something I haven’t discussed in quite a while (and which might turn out to be as important to originary grammar as I once expected), construction grammar, and with the help of Daniel Shore’s Cyberformalism which places the grammatical construction of formulas at the center of language, displacing the two poles of word and sentence that have dominated semiotics, and in particular for the purposes of designing and carrying out corpus searches across large databases. My plan is to target those constructions or formulas that do the work of “if, “but”” and “because” within phrases, or a group of words that together functions as a subject or (much more often, I assume) modifier. Let’s start with the following sentence, from the Cambridge University’s online dictionary definition of “to the extent that”: “It’s only worth planting flowers to the extent that it gives you pleasure.” This sentence could be rewritten as “if planting flowers give you pleasure, then it is worth doing it.” So, the causal/logical connection between planting flowers and receiving pleasure is made more explicit here, which is to say, the phrase to the extent that conceals or renders implicit the causal relation. At the same time, it makes the thought slightly subtler, by recognizing degrees of desiring to plant flowers which, presumably, up until a certain point or past a certain point would not give pleasure. And this suggests that making such causal/logical relations implicit will enrich meaning by (implicitly) adding questions of degree and other qualifications that make the thought more complex, simply by being translated out of these “prime” terms (I’ll note, though, that “but” is not a Natural Semantic Prime) into sentences including formulaic phrases or other means, such as the passive voice, nominalizations, turning verbs into adjectives within phrases, and so on. And if meaning is thereby enriched, the searches will presumably be further enriched, even if we would need the attentive interaction with the measured and mixed up database to see how this might be the case and to make it “richer” in the ways we want.
The search term as wisdom approach provides a powerful new application of the work I have done (included in Anthropomorphics) on Wierzbicka and David Olson’s theory of literacy, i.e., the metalanguage of literacy. If we start with the prime verbs focused on mental activity—think, say, want, know, feel, see and hear—and consider the way in which writing is (per Olson) in the first instance a record of reported speech which must indicate the attitude of the speaker whose speech (recorded in lieu of being able to imitate him) creates a whole field of verbs and subsequent nominalizations that in my hypothesis open up the disciplines. Words like “suggest,” “consider,” “understand,” “imagine,” “imply,” “indicate,” etc., are just ways of translating words like “know,” say,” “think” and so on into some stack of scenes build out of data including all kinds of other things that have been said, known, thought, said to have been known, etc. In making our way back and forth from the hypothetical scene we can draw from Wierzbicka’s primes (which finds some confirmation in a list Olson compiles in The World on Paper of pre-literate speech act verbs) upon which only the primes are used and the enormous range of far more complex discourses in which those words have been made to “trail” behind them whole series of scenes we can enter from any number of angles any discourse. I’m pursuing here my hypothesis (again, articulated in Anthropomorphics) that the disciplines, starting with philosophy, are nothing more than articulations of the metalanguage of literacy, picking up and translating along the way the more fundamental categories of ritual, debt (exchange with the center), and the juridical. The metalanguage of literacy produces a fabricated present contrived out of various match-ups between categories deposited in the mind and categories deposited in reality—the purpose of originary grammar is to deconstruct such a present into the stacked scenes presented with the problem of succession. The question of succession, the ongoing distribution of team members and teams across scenes, is the true present.
As Roemelle points out, the search is conducted by giving the “AI” (let’s say, algorithmically saturated data such that the data elicits incipient reciprocal commensuration across reality) imperatives and asking it questions. The prolongation of the ostensive into the imperative, the imperative into the interrogative and the interrogative into the declarative studied by originary grammar will prove its potency here as well. The imperative is the command or demand that some object—which could be a person, a word, an arrangement—be made present. Using evolutionary language, we can say that those imperatives that can make the trek through to the declarative will be “selected for.” Those are the imperatives that can wait, survive changes in power relations and conditions on the ground, become requests or insistences, get differentiated in requests for information about the object, which in turn become requests for information about the possibility of obtaining information, and so on. In the end, an interrogative has to map onto a declarative—a good question is one that can be answered by some presence or a promise of presence; clearly the promise of presence is where things start to get interesting, because here is where various conditionalities enter into the equation and the whole array of formulas and constructions come into play. Think about a sequence like the following: “you’re not here… get over here… could you come here… why aren’t you here… when can you get here… is something preventing you from being here… could you tell me when you might be here… are you planning to come at all… how will I go on without you… I will be there with you as soon as I can, I am facing obstacles and opposition, the situation keeps changing, I won’t forget you, think of me always, I will always be there with you even if not bodily.” The questioner frames questions so as to “lock in” the interrogant, to provide language for a satisfactory answer, but also to prepare for the worst, and to provide language for an answer that might enable the interrogator to reconcile himself to disappointment with minimal resentment.
The data search is conducted in the assumption of and commitment to the programmability of all of reality, including the programmers; indeed, the always already programmed nature of all of reality. Programmability means all of reality is primed to engage mimetically with humans and with itself, part to part. We expect everything to become more responsive and in turn we become more responsive and anticipatory ourselves—we give data as we take it in. Roemelle recommends creating what are essentially fictional characters to interact with the algorithmically treated data—“you are a Harvard professor, specializing in…”—which is asking the program to imitate a particular kind of performance. But once we acknowledge the comprehensively and constitutively mimetic nature of human reality, and, especially, once we have recognized the rivalries that result from attempts at full bodied imitation, we can divert our thinking to another kind of imitation, one targeted to gestures, and more explicitly pedagogical. Instead of commanding (often tacitly) the student or initiate to “be like me,” it is possible to demystify that authority and say ‘do it like this.” Now, of course there can be no dangerous mimetic rivalry with the program, but there is a question of thinking here: if you ask for a result that might have come from a Harvard professor you bring along all the projections upon that figure that go into making up the Harvard professor. If, instead, we, say, formulate questions aimed at tracing a particular legal term through legendary or ritual material; or test out the boundaries between the ritual, the juridical and the disciplinary; design searches to identify a specified variety ways of translating “think” or “say” into disciplinary concepts; translate imperatives, interrogatives and declaratives into each other, and so on, we can work below the level of personalities to the basic moves and gestures that comprise the basic elements of a stack of scenes, which is to say, that fill in the various imperative gaps articulating center with periphery. We then hypothesize regarding the way the scenes are stacked, which is how we continue to stack the scenes we are on.
There remains the question of the programming language I’ve been working on in recent posts, that have applied the primes to the tripartite distinction of ritual, juridical and disciplinary. Before will have been the same as after; the part of all will have been the same; doing is the same as happening; and, let’s add, indicating the imperative threading through these domains, more happening can be because of doing. These serve as formulas, drawn from the primes, like the basic parameters of wisdom literature from which are derived narratives and commands. They are fleshed out with the metalanguage of literacy in ways that create stacks within the disciplines, preserving inscriptions of the ritual in the juridical, the juridical in the disciplinary and the disciplinary in the ritual. And disciplinary discourses can be reduced back into articulations and configurations of the formulas. Jaron Lanier in a recent article called for opening up the “black box” of the large language models, for the good cause of crediting the producers of the data used by those models—that is, to restore the juridical to the ritual and disciplinary explosion of technology. Lanier sees this as a way of compensating producers who contribute the data, but this goal seems unlikely; what is more likely is not to open the black box literally by tracing back all its data points but rather hypothetically, by designing prompts that expose possible inscriptions of the ritual, disciplinary and juridical in one another. Instead of paying people for past productions you would convert them into more effective producers now making contributions to institutions reduced to the problem of data security. Hypothetical trails would lead to ways of constructing and creating present trains of data; we provide not, primarily, information or knowledge, but language, which precedes and includes both. We would want to learn how to design rather surreal, or perhaps pataphysical prompts, asking, for example, to compose a story in a mock documentary style regarding a case that produces a differend (i.e., in which any decision produces an injustice) and that starts small but comes to involve powerful social actors. The program will produce something that touches awkwardly and perhaps unconvincingly, even bizarrely, on these commands—but then you begin to work backwards, engineering prompts, searching for court cases with language similar to the story, for mock documentary stories that might have served as models, for well known cases of a seemingly trivial case triggering wider social contagion. You shake loose the data in this way. We’d have new language for speaking about court cases, especially extremely difficult ones, for describing the ways powerful players get drawn in involuntarily while also exploiting such cases, for turning factual, presumably neutral language into satire at various levels of explicitness. The human is dissolved in the institutional and technological scenery while re-emerging by re-naming it all.