I knew, writing “Delete Magisterium AI,” that I was asking for a difficult act from a great guy: to delete the Catholic chatbot. There is simply no way that Longbeard CEO Matthew Harvey Sanders is not the most likeable of nineteen schmucks, randomly selected. Differing viz probabilistically generated specifications of the Catholic faith, we would yet be grouped, by any disinterested bystander, as kin. I’m sure Sanders’s heart thrills to hear the St. Michael Prayer prayed by a large group of men; sure he likes vests; sure that he has the Salve Regina memorized when he sings it, but would find it hard to just say; sure that he is one of my people, all told.
But I did not know that my brother in Ecclesia was a genius, a veritable black-belt in argumentative jiu-jitsu. I was tense for a response, ready for what roundhouse kick would come flying for my idiot head. But—what a move! The kind you can really only appreciate from the floor, the wind knocked out of you, eyes crossed, bruises slowly forming over sixteen points of your broken body, the world dripping out of focus: beautiful.
For I made it clear that I thought it lame to talk to robots. I argued that to do so would be irrational, would be deliberately directing an act of speaking-with toward a thing that does not listen or speak. Sanders could have argued all sorts of things in response: that such natural law reasoning only applies to sex; that there is an essential difference between speaking-with and whatever it is that one does with a chatbot; that it was necessary to choose the lesser of two evils, etc. In any case, I was ready, bouncing on the balls of my feet. But what Sanders did was run my argument through an LLM, generate a response, edit it a bit, and rock it up on his website faster than a man can say “contra naturam.”
What a bind I am in! I made it clear that I shall not converse with machines and Sanders did the only thing that could stun me into silence: he “responded” with a machine. If I respond, I break what I understand to be a mitzvah. If I do not respond, I lose the argument. Woe is me, who must now choose between being a loser or a hypocrite!
It’s not just the vibes of Sanders’s thing. An entire class of professional, LinkedIn-types have begun to write and even speak like the word-generating machines they operate. They use end-of-sentence em-dashes; they habitually say that something is “not just this but more”; they slip into single-sentence emphatic paragraphs and bullet-pointed, numbered, ready-to-be-a-slide prose. Vibes alone do not tell: such things may well be said things. They may come from the man. They may be stood-behind and vouched-for and witnessed-to, so that the word remains a path back to the wordbearer and not to this new, lesser animal: the word-chooser and his robot companion.
Alone, such language would not have me keeled over and whimpering, bound between a rock and a hard place. But Sanders’s non-response included things I did not say, topics I did not address, quotation marks hugging words I never wrote. The whole thing is a probabilistic generation of what my argument might have been: close, but no cigar; correct genre, but wrong band; a Coldplay version of what I wrote, clearly signifying Sanders’s masterful stroke: to dare me to break my own rule and chat with a chatbot.
Sanders’s robot summarizes my argument in six points. Point 5 takes aim at my “Charge of ‘Gnosticism’ (Separating Truth from Body).”
Now, look: I did not write the word “gnosticism.” I did not indicate the topic of gnosticism. I neither wrote the word nor indicated the topic because I do not think Magisterium AI has anything to do with gnosticism. I did not write "separating truth from body.” I did not indicate the topic of “separating truth from body.” I do not know what “separating truth from body” means, and so I am incapable of accusing Magisterium AI of doing such a thing. The machine Sanders used generated the following as a quasi-explanation of my completely non-existent “charge of gnosticism”:
You argue that by turning the faith into a dataset, we risk a new form of Gnosticism—treating Catholicism as a collection of “secret knowledge” or “information” that can be extracted from the living Body of Christ and dispensed by a machine. You fear this disincarnates the faith, suggesting that “having the answers” is the same as “having the faith.”
Literally, none of it. My article is right there. It’s long, and I know anything longer than a finance-bro’s pitch for his niche cryptocurrency can cause seizures. But you can just ctrl+F the thing. I didn’t write “dataset.” The one time I even used the word “data” I did so, not pejoratively, but explanatorily: LLM technology, said I, was “somewhat helpfully described as ‘guessing the next word in a series’ on the basis of a numerical pattern superinscribed onto a large body of data.” I don’t think this, the modus operandi of LLM technology, is bad or even unseemly, much less “a new form of Gnosticism.”
Nor did I write what I am quoted as having written, to whit, “secret knowledge” (nope); “information” (nope, I somehow managed to write a whole article on without once using the word “information”) “having the answers” (nope) or “having the faith” (nope). But obviously, these are the sort of things that one of my boys, feeling squeamish about the synthesis of Catholicism and chatbots, might say, might even post, undoubtedly have posted (my boys can’t help it, but have mercy on them, Lord, they’re taking each day one at a time) and so the kind of thing that a probabilistic sentence-generating machine would dredge up in response to a query such as “summarize this argument against Magisterium AI.” Trawling an interconnected web of key-words, as is its wont, the bot netted some strange fish.
Point 4 is better disguised (probably in an edit) but Sanders still allows the robotic seams to show, binding my conscience as before. He summarizes my argument as one concerning “The Name: Usurping the Teacher?” As before, I wrote no such thing. (I am a teacher. To admit to a concern of usurpation would be weak, insecure, a bout of middle-class grasping for job security rather than that Charlemagnean confidence in one’s irreplaceable status that befits a man.) Sanders/robot explains my non-existent topic:
A critique I have heard often, and which seems to undergird your hesitation, is the name itself: “Magisterium AI.” It might appear that by applying this title to a machine, we are claiming the machine possesses the teaching authority of the Church, effectively creating a “Robotic Pope…I want to be clear: The AI is not the Magisterium.
Were this speech, and so representative of a speaker who can be said to “acknowledge,” this would acknowledge that the topic—Magisterium AI’s name—might not be in my article. But concern over Magisterium’s name is undoubtedly the sort of concern that gets posted—and so the sort of thing a language-trawling and guessing machine might guess was in my article.
But the problem of the name “Magisterium AI” does not “undergird my hesitation.” It’s simply not there at all. I never bring it up. I never mention “authority.” When I bring up the Pope it’s to crack jokes, not to fret over Robo-Pope.
Sanders would never use such an argument himself. It’s beneath him, beneath strawmanning, even, this schizophrenic swinging at non-existent vibes rather than the written argument, the one Sanders/robot said he/it “read…with great interest.”
What, could I say, “A critique of New Polity that I have heard often, and which seems to undergird Sanders’s hesitation, is that we’re a bunch of posers who don’t even lift. It might appear that, by not even lifting, we are not the based, absolute units of Catholic postliberalism we would like the youth to believe—effectively giving a false impression. I want to be clear: Denley and Boland lift enough for all of us.” This as a “response” to an argument that contains no mention of lifting? No, of course not. This is not Sanders: this is fake stuff.
Sandersbot’s point 3—“The Fear of ‘Impersonal Authority’”—is the same as the title of point 4, its explanation the same as the content of point 5, and all in the same form of AI response-generation: loosely approximating but ultimately unrelated to the reality at hand. I am said to “raise a valid concern” that “AI might flatten the living tradition of the Church into mere ‘data,’ detaching the teaching from the teacher.” I said no such things and have no such concerns. I never wrote the word “authority”—much less the quoted “impersonal authority.” I never said “fear,” because that would admit to weakness, see Charlemagne, etc.
Point 1 and 2 are somewhat related to an article I wrote—just not the article Sanders (cunningly) pretends to respond to. “You argue that conversation is teleologically ordered toward communion with another intelligence,” says Point 1. I do, and I did, but in a different article. The essay to which Sanders’s robot “responds” neither contains the word nor relies on the idea of “communion” for its punching power. This is the same with “non-mutuality”—wrong article. I am quoted as claiming that “using natural language to query a database” is a “moral lie”—I don’t know what a “moral lie” is, and so I never used it in any of my articles! (Heaven protect me from the sponginess of chatbots, how they soak up this and that and the other thing and squeeze it all out as if it were a whole!)
Against such maneuvers, I can only stand in reverence and awe; can only ask—beg, really—for Sanders to have mercy on me, a poor numbskull, and not make me into a hypocrite; not stand as middle-man to the very robots I have sworn off, to the same probabilistic word-generation machines that I have called irrational, undignified, and contra naturam to speak and listen to. The cleverness of Sanders’s move admitted, would he mercifully read my article and respond to it like a man?
For, up from the cyborgian murkiness produced between Sanders the man and Sanders the machine-operator, some arguments are struggling to raise their heads and bite at me.
I discern the argument that one is cleared of the charge of addressing and speaking with a non-intelligence insofar as one can be said to be only “consulting” one: but can I not consult a person? I appear to be assured that Magisterium AI “acts less like an author and more like a paralegal”—but is a paralegal not a man?
Against the argument that chatbots elicit a way of being and relating that is proper only toward persons, I am being told not to worry, because this is a Very Boring Person. I would like to argue (would that I could!) that there is the appearance of a person in the very appearance of the novel specifications of Church teaching to the particular query, regardless of whether or not that specification “cites its sources.” Men, too, cite sources, and never have I been invited out of the mode of reading what someone has written by virtue of a footnote: “This is really good stuff from Cardinal Ratzinger. Ah—nevermind. He just cited his source. This isn’t a person, after all. Just another dynamic card catalogue.”
I appear to be “told” that Magisterium AI is “focused on breaking the illusion of personhood” in AI chatbots—which would admit that there is an illusion to break. But is it broken by the fact “it does not have a name like ‘Father Justin’; it does not say ‘I feel’ or ‘I believe’”? Was the problem with “Father Justin” that anyone actually believed it was a person, such that a decreased quantity of anthropomorphism would solve it?
Magisterium AI does “say” “I function.” It does “say” “I can provide.” It is designed to refer to itself as an “ego,” a self, and frequently does so. Are we decreasing anthropomorphisms by adding “vocalization to responses?” “At present,” says Longbeard, “you have a choice between a male or female voice. We’ll be shipping many more options in future so you can pick a voice option that suits you.” Groovy! If I have this right, all this intense focus on breaking the illusion of personhood going on around here led Longbeard to develop/purchase voices to—what? Help us remember that we’re all just utilizing a dynamic version of the Summa?
But I cannot entertain all of these wonderful, possible arguments because I do not know whether doing so would be striving with men or striving after the wind. O tempora, O mores! O, release me from the bind I’m in, Sanders! (And take your time, no rush.)