Well, folks, it’s been enough time since the encyclical Magnifica Humanitas dropped. I can no longer justify my silence; no longer blame my wife and children for their demands on my attention; no longer point to the encyclical itself, which ended with the papal instruction to “cherish places and times where physical presence remains crucial”—which I interpreted by weeding the yard instead of writing my “take.”
Getting dunked on by the Pope
The good people from Truthly, Longbeard, Hallow, Doggymatic et al. have drawn my attention (repeatedly) to the following paragraph, MH 100:
The artificial imitation of positive human communication — words of advice, empathy, friendship and even love — can be engaging and at times genuinely helpful.
This has been taken as a real dunk on my position, that AI chatbots are evil insofar as they tempt us to take a thing as a person, falsifying the world through an abuse of our faculty of conversation, which is for the sake of communion. That this sentence is not such a dunk, I will gladly show. But MH 100 really has authoritatively and magisterially revealed a foolish part of my argument—which I will un-gladly show. Here goes!
That sentence
Taken as it is in English translation—and with some regret for being lawyerly—this sentence cannot be read as an endorsement of the artificial imitation of human communication (which chatbots like Truthly and MagisteriumAI rely on in order to give their spiritual advice). It’s not that kind of sentence. The Pope is not giving an instruction. He’s observing a fact: “The artificial imitation of positive human communication…can be engaging and at times genuinely helpful.”
Yes, obviously. You’ll hear me spout all sorts of calumny against chatbots—calling them automated liars and Big Bad Bots, etc.—but I hope I’ll never be so prejudiced as to deny that they are “engaging” and “at times genuinely helpful.” An AI fiancee? Engaging! The probabilistic generation of the words “Your life is of infinite value?” Genuinely helpful!
For MH 100 to annihilate my position, it would have to be impossible for a technology to be “engaging” and “genuinely helpful” and bad. But that’s obviously not the case: a superhighway can be both engaging and helpful and yet bad because it cuts cities from their riverfronts and neighborhoods from each other. Literature abounds with visits with prostitutes that are “genuinely helpful” (and certainly “engaging”)—John Steinbeck’s Cannery Row, Cormac McCarthy’s Cities of the Plain, and J.D. Salinger’s The Catcher in the Rye come to mind—and yet we’d hardly call these endorsements of the brothel.
The Pope’s sentence is an observation of what we would all admit: simulated speech can be engaging and helpful. It sets up his warning of the dangers such simulations nevertheless contain: “However, for less discerning users, it can also be misleading, creating the illusion of a relationship with a real personal subject. When words are simulated, they do not build genuine relationships, but only their appearance.“ (MH 100)
That sentence, en français
All that said, the English translation does make it easier for a guy to leap beyond the actual scope of the Pope’s words: perhaps this is the reason why Zac, the good egg from Truthly, takes the encyclical as Truthly’s “marching orders” while I take it as a rather good reason to delete Catholic chatbots.
First, the English breaks into two, discrete sentences what the Romance languages keep as one: “...can be engaging and at times genuinely helpful. However, for less discerning users…” The French has it as a singular sentence: “L’imitation artificielle d’une communication humaine positive – paroles de conseil, d’empathie, d’amitié, d’amour – peut s’avérer gratifiante et même utile, mais chez des utilisateurs peu avertis, elle peut induire en erreur et donner l’illusion d’être en relation avec un sujet personnel authentique.” So it is in Italian (the usual drafting language of papal encyclicals), Spanish, Portuguese and German (and perhaps some kind polyglot could inform me concerning Russian, Polish, and Arabic).
Now, “splitting” is a perfectly acceptable way to deal with a long sentence made up of two independent clauses. But this two-sentence construction increases the temptation (for those invested in chatbots) to take the first of those clauses as a complete and independent thought, capable of being separated out and considered as a statement on its own. The one-sentence translations of all the other languages would force an honest man to quote that so-called “endorsement” with an ellipsis trailing off at the end. Were we to write it all out in English as our international hombres did, we’d write:
The artificial imitation of positive human communication — words of advice, empathy, friendship and even love — can be engaging and at times genuinely helpful, but for less discerning users, it can also be misleading, creating the illusion of a relationship with a real personal subject.
Here the ambiguity of chatbots is what’s being spoken of. There, in the English, it seems like the Pope is making a point of highlighting the good that the artificial imitation of speech contains, and then, in a separate thought, speaking of a danger.
Genuinely helpful?
The encyclical was drafted in many languages, so it is unclear which one to lean on. I suspect we’ll fight about the eventual Latin version, unless Leo tells us to do otherwise. But the phrase “genuinely helpful,” in English, is equivocal: “Genuinely” could mean something like “actually” as in, “despite what you may think, chatbots can be genuinely helpful.” But it could also be taken as a word carrying a lot more weight, as in “the help that the artificial imitation of positive communication can provide is genuine help,” a kind of help which is holistic, integrated, authentic—unambiguously good.
Again, all the other languages—the ones I understand—indicate the first meaning and not the second. Italian: “e persino utile…” Portuguese: “e até útil…” Spanish: “e incluso útil…” French “et même utile…” All of these would be better translated into English as describing the artificial imitation of speech as rewarding “and even useful,” or “and also useful”—a far cry from “genuinely helpful.” Likewise, none but the English translation adds a temporal caveat: “and at times genuinely helpful.”
None of these are egregious errors, but taken together, I can see why a reader of the English might look at a single, papal sentence declaring the artificial imitation of human communication to be “at times genuinely helpful” and be tempted to conclude that his Catholic chatbot might just be one of those “times” simulated speech gets a papal a-okay. But were we to put it as the Italian puts it, we’d get something like:
The artificial imitation of positive human communication — words of advice, empathy, friendship and even love — can prove rewarding and even helpful, but for less discerning users, it can also be misleading, creating the illusion of a relationship with a real personal subject.
Hardly something to hang a Catholic chatbot’s hat on.
What I really ought to have said
I was clear, in my “first punch” and subsequently, that chatbots tempt us to relate to a thing as if to a person. I was not clear whether one can overcome this temptation. Re-reading my own argument in the light of MH 100, I can see why this omission scandalized. People compared their own experiences of screwing around with Grok and ChatGPT to my dire warnings against “the simulation of the human person” and simply did not see themselves in my description: “I certainly never take this machine to be a person. What’s wrong with you?”
I think it is apparent that the temptation of the chatbot can be overcome: one can see, know, and even feel the probabilistic generations of a chatbot for what they are—impersonal, probabilistic rearrangements of the patterns of human speech. This state of taking the thing as it is is both possible (despite the temptation to do otherwise), desirable, and actual in the lives of good people who now, like or not, live in and among a whole lot of chatbots.
Because I didn’t say this explicitly, I think I came across as legalistic, puritanical: as if merely reading a probabilistic generation damned a man to untruth and non-correspondence with reality; as if a query to an LLM was necessarily and always an act by which one enters into the mode of conversation, that openness to the other who speaks. Experience shows that this is not necessarily so: one really can see the thing as so many moving parts and relate to it as such.
More coming, of course, but for now: it’s good to have a Pope.