The strongest objection to what I am about to argue comes from the encyclical itself, so I will start there.
Magnifica Humanitas opens on the Tower of Babel, and it describes the tower precisely: “a single language, a single technology, a single direction” (§7). Its builders wanted to make a name for themselves, and the result is the one Genesis records: “When a city is built on pride and the claim to self-sufficiency, communication breaks down, languages are confused and people no longer understand each other” (§7). Three paragraphs later the Pope names the temptation in our own terms. We must avoid “the pretense that a single language — even a digital one — can translate everything, including the mystery of the person, into data and performance” (§10).
Read with AI in mind, that sentence is aimed squarely at the thing I want to praise. A translation system is a single digital language, owned by very few, offering to translate everything. The encyclical adds that these systems “may imitate language, behavior and analytical skills, or even simulate empathy and understanding, but they do not understand what they produce” (§99), and that the firms behind them “effectively set the conditions for access, determine the rules of visibility and shape the very possibilities for participation” (§95). If AI is anything biblical, the objection runs, it is the tower.
I take that objection seriously, and I concede most of it. Then I think it gets the story backwards.
Babel is not ahead of us
The encyclical treats Babel as a project we might still choose: “either to construct a new Tower of Babel or to build the city in which God and humanity dwell together” (§1). I read our situation differently. The confusion of tongues is not a risk on the horizon. It is the world as it has become.
We speak different languages, and that is the least of it. We speak different ideologies, different disciplines, different professional dialects, and we pursue different goals that rarely hear each other. An economist and a theologian can read the same encyclical and come away with two documents. A ministry publishes its rules in one language and the people they govern read another. A scholar who wants to know what a tradition actually says depends on whichever fragments someone chose to translate. Genesis describes the result exactly: God came down to “confuse their language so they will not understand each other” (Gen 11:7, NIV). That is not the danger. That is the address.
So the question I ask is not how we avoid building Babel. It is how we get out of it.
The answer the encyclical does not name
The Christian Scriptures already have an answer to Babel, and it is striking that a document built on Babel never names it. The encyclical’s counter-image is Nehemiah rebuilding the walls of Jerusalem (§7–§10), not the day of Pentecost in that same city. A full-text search of Magnifica Humanitas finds the word “Pentecost” only once, inside the Latin title of a sermon by Saint Augustine cited in a footnote; the body of the encyclical never mentions it.
For readers who do not know the story. The word comes from the Greek for “fiftieth” (day). It is the Greek name of the Jewish feast of Shavuot, the Feast of Weeks, which the Law counts as “seven full weeks”, “fifty days”, from the offering of the first sheaf at Passover time (Lev 23:15–16, NIV). According to the Acts of the Apostles, on that day the followers of Jesus “were all together in one place” in Jerusalem, a company the chapter before puts at about a hundred and twenty, the women and Mary the mother of Jesus among them (Acts 1:14–15). “There appeared to them tongues as of fire,” and “they were all filled with the Holy Spirit and began to speak in other tongues, as the Spirit gave them utterance” (Acts 2:1–4, RSV). Jerusalem was full of pilgrims “from every nation under heaven” (2:5), and they were bewildered, “because each one heard them speaking in his own language” (2:6). The Catechism of the Catholic Church says that on that day “the Church was made manifest to the world” (§1076).
Read the verses slowly, because the precise wording carries my argument. The crowd does not learn a new language. Nobody in Jerusalem is made to speak the Galileans’ language. The pilgrims ask: “how is it that we hear, each of us in his own native language?” (Acts 2:8, RSV), and they list where they come from, Parthians and Medes and Elamites and the rest, before saying: “we hear them telling in our own tongues the mighty works of God” (2:11). Babel scattered one language into many that could not understand each other. Pentecost does not undo that by collapsing the many back into one. It leaves every language standing and makes each of them understood.
That is exactly the distinction the encyclical draws when it praises Nehemiah’s Jerusalem, which “rediscovers a common language — not one of uniformity, but one of communion” (§8), and when it finds, “precisely from the plurality of voices and visions which, even though they sometimes remind us of the confusion caused by the diversity of spoken languages, a bright possibility emerges” (§10). Uniformity is the tower. Plurality made mutually audible is Pentecost. The encyclical describes the second without naming it.
The Church has argued about this reading for a very long time. Gregory of Nazianzus, preaching on Pentecost in the fourth century, asked whether the miracle was in the speakers, who spoke foreign tongues, or in the hearers, so that “one sound was uttered, but many were heard.” He preferred the speakers. He also saw the Babel connection plainly: the old confusion broke up a wicked unity, while the new gift, “being poured from One Spirit upon many men, it brings them again into harmony” (Oration 41, in the English translation of 1894).
I take the reading Gregory set aside. I am not a theologian, and I do not need to settle what happened in Jerusalem. I need only notice that the text tells the story from the hearer’s side, and that the oldest commentators knew it could be read that way.
A digital Pentecost
Here is my thesis, stated as plainly as I can. AI does not let us all speak every language. It lets each of us hear every language in our own. It lets a person consult sources and documents in any language, any format, any field and any perspective, and receive them in the language they think in. That is the hearer’s miracle, made into an instrument.
I hold this as an analogy, and I want to be exact about its limits. The Holy Spirit is not a product, and a translation model is not grace. Nor does the machine understand what it carries; on this the encyclical is right (§99), and Acts agrees with it, because in Acts the understanding belonged to the hearers. The machine moves the words. The understanding is still ours to do. If anything, that makes the analogy more faithful, not less: Pentecost did not think for the crowd in Jerusalem either.
What changes is reach. The first Pentecost happened in one city, to a crowd of pilgrims who happened to be there. Peter, explaining it, quoted the prophet Joel: “I will pour out my Spirit upon all flesh” (Acts 2:17, RSV). The promise was already universal; the event was local. What we have now is an instrument that can reach a person one by one, wherever they are, in their own language. Not a gift for one room in Jerusalem and the crowd in the street outside, but one that could, in principle, be offered to the whole of humanity.
“In principle” is where the work is
Now the concessions, which are not small. The objection I started with is half right, and the right half is load-bearing.
These systems are concentrated, as §95 says. Their reach across the world’s languages is uneven, and the people who build them say so. The team behind a translation model covering 200 languages wrote in Nature in 2024 that the benefits of machine translation “are unevenly distributed”, having “mainly benefited high-resource languages, leaving many low-resource languages behind”; quality depends on large volumes of parallel text, and those are not equally available across languages. And a translation can flatten a language toward the one the system knows best, which is the Babel the Pope fears: one language passing itself off as all of them. That worry is mine, but it is not idle. A 2024 study of one family of open language models (Wendler and colleagues, on Llama-2) found that their internal “concept space” lies closer to English than to other languages, which, the authors wrote, “may have important consequences regarding the biases held by multilingual language models.”
None of that is a reason to refuse the gift. It is a description of what the gift still lacks. I have argued elsewhere in this series that innovation is not something to slow or to fence; my own words for it are: “Do not artificially truncate, limit, mutilate innovation. Be part of it, accompany it, and let’s see and be open to different directions when steering where it leads us.” Applied here, accompanying means steering toward the encyclical’s own program. “To speak of the universal destination of goods means finding ways of ensuring universal access to both technologies and the education needed to use them” (§109). The engineers reached the same place from the other side: the same Nature paper ends by saying that “policy interventions aimed at more fundamental issues surrounding education, internet access and digital literacy are imperative to eradicate the structural problem of language disparities.” That is the condition on which a digital Pentecost is a Pentecost and not a tower. A tool that reaches “all flesh” only after a subscription is not yet what I am describing. A tool that reaches the smallest language community, in its own language, with the education to use it well, is.
There is also a test the encyclical gives us, and I accept it: does the system restore technology “to the plurality of human cultures and ways of life” (§110), or does it reduce them to one? Translation that keeps each listener in his own language passes that test. Translation that quietly teaches everyone to think in one language fails it. The difference is not in the technology’s nature. It is in how we build it, who owns it, and whom we build it for.
The sharper ask
Magnifica Humanitas worries, rightly, that “most people are watching and waiting, observing from afar and merely hoping for the best” (§6). On this point I would ask for more than the encyclical does.
The Pope asks us not to build another tower. I would add the other half of the story. We are already standing in the ruins of the first one, unable to hear each other across languages, disciplines and convictions. We have an instrument that can let each person hear the others in their own tongue. The task is not only to avoid Babel. It is to finish the miracle that undid it, and this time to make sure it reaches everyone in the crowd, not only those who happened to be in the room.
— Carlos Miranda Levy Coordinator of Impact, Innovation and Intelligence