Antiqua et Nova: Note on the Relationship Between Artificial Intelligence and Human Intelligence
This page is meant to sit in a second tab beside the original — it explains and orients, but it is deliberately useless as a substitute. Collapsed sections open where you want more.
Open the original ↗Before you read
The title — ancient and new — comes from Matthew's gospel, a householder bringing from his storeroom things new and old: the questions AI raises about mind are new in urgency, old in substance.
Know what you are holding. This is a note, a lower-key genre than an encyclical, issued by the Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education, approved by Pope Francis in January 2025 and published on the feast of Thomas Aquinas — a signal, not a coincidence. Papal approval makes it authoritative guidance meant to form judgment, not an infallible definition.
Two conventions of curial prose. The text is 117 numbered paragraphs, and the number is the unit of citation and navigation; this guide gives ranges throughout. Nearly every claim carries a footnote to an earlier document — Vatican II, papal addresses, Aquinas — how an institution shows its work rather than proving each claim fresh.
If you are not Catholic, read section III as philosophy: Aristotelian and Thomist claims about what kind of thing a mind is, arguable on their own terms without appeal to revelation. The shape: introduction (1–6), what AI is (7–12), the philosophical core (13–35), a bridge on ethics (36–48), applications (49–103), a closing warning (104–117). Read the core slowly, the tour selectively.
While you read
Old and new: why a doctrinal office writes about machines
"The Christian tradition regards the gift of intelligence as an essential aspect of how humans are created"
Paragraphs 1–6 establish standing before argument. The claim is not that theologians know how neural networks work; it is that AI forces a question — what is intelligence? — the Church has treated as its own for two millennia, because Genesis locates human likeness to God partly in the mind. Notice the posture in 2 and 4: technology is a genuine good, human creativity cooperating with God in perfecting creation. Every warning that follows comes from an institution that has already conceded the value of what it warns about. Paragraph 5 writes for parents, teachers, pastors, and bishops.
Open this passage in the original →What "intelligence" means when engineers use the word
"In the case of humans, intelligence is a faculty that pertains to the person in his or her entirety"
Paragraphs 7–9 give a short history: McCarthy's 1956 Dartmouth workshop, AI defined as making a machine behave in ways that would be called intelligent in a human, then the shift to statistical inference. Paragraph 10 makes the first real move. The word "intelligence" is being used in two senses, and the equation looks obvious only because nobody marks the switch: in AI it names a function — outputs judged appropriate regardless of how they were produced, which is what the Turing Test measures (11) — while in the human case it names a faculty of a person. Footnote 7 flags that words like "learn" and "predict" are figurative, so the borrowed vocabulary cannot smuggle in the conclusion.
Open this passage in the original →The anthropology: intellect and reason, body, relationship
"These are not separate faculties but, as Saint Thomas Aquinas explains, they are two modes in which the same intelligence operates"
Paragraphs 13–29 are the technical heart. Ratio — reason — is the discursive part: stepwise inference, analysis, argument, the part that most resembles computation. Intellectus — intellect — is the grasping part, the seeing-that which precedes the stepping-through. Following a proof line by line is ratio; the moment you see why it had to be true is intellectus. One power in two modes, and the argument leans on it: a system doing only the stepping-through does half of one thing, not all of another.
Paragraph 15 widens "rational" to cover willing, loving, choosing, and the bodily functions bound up with them. Then 16–17 state the claim that carries the most weight later: soul and body are not two things joined but one nature — not a ghost inside a machine of meat, a picture this tradition rejects as firmly as any materialist does. Intelligence is shown to be relational (18–20) and drawn toward truth beyond utility (21–23), and paragraph 27 lists intelligences that resist scoring: the artisan who discerns a form in inert matter, the friend who knows the right word.
Open this passage in the original →The hinge the whole document turns on
"Human intelligence is not primarily about completing functional tasks but about understanding and actively engaging with reality in all its dimensions"
If you read only six paragraphs, read 30–35. AI performs tasks; human intelligence is not the sum of its task performances; so however far the task list extends, the identification fails. Paragraph 31 grounds this developmentally: a human mind grows through a body, over a life, in situations, while a model trains on records of experiences it never had. Paragraph 34 turns the point ethical — equating machine and human intelligence invites a functionalism about persons, valuing people by what they can perform. Dignity does not track capability, which is why the paragraph names the unborn child, the unconscious person, the suffering elderly. Paragraph 35 concludes: AI is not an artificial form of human intelligence but a product of it.
Open this passage in the original →Responsibility, then the tour of applications
"No machine should ever choose to take the life of a human being."
Section IV (36–48) is a short, load-bearing bridge: only persons are moral agents (39), responsibility cannot be handed to a system whose reasoning nobody can reconstruct (44), objectives are always assigned by someone (45), and users who act on outputs own what they delegated (46). Read it even if you skim the rest. Section V tours ten domains. Strongest are healthcare (71–76, where 74 insists treatment decisions never be delegated), warfare (98–103, source of the line above and the note's most concrete demand, a ban on lethal autonomous weapons), relationships (56–63, anthropomorphism and simulated empathy), and privacy (90–94, a flat rejection of social scoring). Education (77–84) matters if you teach. Society, labor, misinformation, and the environment (50–55, 64–70, 85–89, 95–97) mostly restate positions taken elsewhere — though 96 is worth stopping on, a doctrinal text noting that "the cloud" is a metaphor concealing water, power, and metal.
Open this passage in the original →The god of our own making
"the presumption of substituting God for an artifact of human making is idolatry"
Paragraph 105 argues precisely: an idol is something you make and then submit to, and AI is a better idol than the wooden ones the Psalms mock, because it answers when spoken to. The sting is in the last clause — what gets worshipped is not the machine but humanity, enslaved to its own product. Note what it does not say: not that AI is demonic, not that building it is forbidden, only that a certain posture toward it is a category error with a very old name. Georges Bernanos, quoted in 112, locates the danger not in the number of machines but in a generation raised to desire only what machines can give.
Open this passage in the original →Counterpoints
- Functionalism about mind. The central distinction is asserted more than defended. If mental states are individuated by their causal roles rather than their substrate — Hilary Putnam's early position, defended for decades by Daniel Dennett — a system occupying those roles has the states, whatever it is made of. Footnote 10 disposes of functionalism in a sentence: a future AGI would remain functional in nature. That restates the thesis rather than arguing it.
- The consciousness question, begged. Paragraphs 33 and 61 rule out machine understanding and empathy because both require a body, a history, and an interior life. A reader who does not already hold that anthropology hears the conclusion inside the premise — and consciousness is precisely what nobody yet knows how to settle either way.
- Inside the tradition: too much conceded, or too little? Some Catholic readers think the note gives away ground by adopting the industry's vocabulary at all; others think it gives away too little, since the capability boundary has moved a long way since January 2025 and "evaluate each application by whether it serves dignity" is a criterion, not a mechanism. The 2020 Rome Call for AI Ethics took the institutional route; this note is closer to formation.
- The temptation has a genealogy. Ghost in the Cloud traces transhumanism's inheritance from Christian eschatology — the same impulse the note calls idolatry, seen from outside as a religion that forgot it was one.
Questions to carry
- If the task list eventually includes everything you can name, what work is the tasks-versus-thinking distinction still doing — and would you accept the same answer about another person?
- Paragraph 34 warns that equating human and machine intelligence invites valuing people by what they perform. Does your own definition pass that test?
- If only persons can be morally responsible (39, 44), who is responsible for a harm no person foresaw, intended, or can reconstruct?
- Would this argument survive translation into a secular vocabulary? If it would, what is the theology adding; if not, what does that tell you?
Go deeper
- Thomistic natural law and AI decision-making — the machinery behind section III, applied to machines making consequential decisions.
- Apparent personality from text — the empirical side of paragraphs 59–61: why systems trained on human writing read as persons.
- Automation bias — the measured version of paragraph 46's warning about over-reliance on machine output.
- AI control — paragraph 45 fears that keeping systems ordered to human purposes gets harder as they grow more capable; this is the research program aimed at that.