Beware the Benedict Bot

The words of a dead man / Are modified in the guts of the living.” Neither W. H. Auden, who wrote these lines in 1939, nor W. B. Yeats, whom they concern, knew that their words would one day end up in the matrices and vector fields that constitute the “guts” of modern AI models. To use Yeats’s preferred language, everything he ever wrote is now being turned around in a virtual “gyre” along with everything else that has ever been written. How his works will be modified in the process is impossible to say.

While this may not be an urgent problem for Yeats scholars, it does present a difficulty for an institution with teaching authority like the Church. The integrity of doctrine depends on the precision with which it is stated. Hallucinations—the standard euphemism for erroneous outputs by AI models—can, where the word of God is concerned, jeopardize the salvation of souls.

Last month, the Benedict XVI Society proposed a way of solving this problem. It announced the creation of Benedict AI, a proprietary language model trained on the writings of Benedict XVI along with a scaffolding of canon law to ensure fidelity to doctrinal teaching. The project is also touted as an exciting opportunity to render accessible the writings of one of the greatest theologians of his generation.

The team behind the project has sought to head off any misconceptions, a prudent decision given the febrile state of public debate over AI. Details are scarce, but we can state emphatically that this is not what the industry calls an “AI resurrection.” Something like this was attempted recently by the Cluny Institute, a research center affiliated with the Catholic University of America. At their annual Zoë conference, one of the featured speakers was “AI René Girard,” an algorithm trained to impersonate the dead French philosopher. One can only imagine the outrage were Girard replaced with the late pontiff.

In the same poem quoted above, Auden wrote that at the moment of his death, Yeats “became his admirers,” surely one of the most terrifying lines of poetry in the English language. What was intended metaphorically can now be achieved through technical means by these so-called “resurrections.” It is for theologians and canon lawyers to determine whether this amounts to necromancy or the consultation of oracles, practices that are viewed as grave sins by the Church.

Benedict AI was announced shortly before the publication of Magnifica Humanitas, Leo XIV’s first encyclical and the most authoritative statement to date by Rome on AI and related technologies. It opens with a startling rhetorical gambit. Humanity, we are told, is faced with a choice of two paths corresponding to two biblical episodes. Either we give into blind technological determinism and “construct a new Tower of Babel,” or, following Nehemiah, we judiciously use what we have made to “build the city in which God and humanity dwell together.”

One of the great qualities of Magnifica Humanitas is how it demonstrates the supernatural fruitfulness of Scripture. Its use of the stories of Babel and the Babylonian exile is so persuasive that, were you outside of the Christian tradition, you might begin asking yourself whether you needed to be inside it to understand the trajectory of the modern world. And yet, despite some 40,000 words of carefully argued text, we are still left with the practical problem of discerning whether, in our business life or our research activities, we are truly building up the walls of Jerusalem or succumbing to “Babel syndrome.”

Thankfully, a reliable acid test for the Christian technologist was provided by St. Maximilian Kolbe almost a decade before the discovery of the first point-contact transistor set off the computer revolution. Kolbe had a great interest in the sciences and became what we would now call an “early adopter” of modern communication technologies. At his monastery in Niepokalanów, monks would spend the intervals in the liturgical day manning a vast printing press that could reliably produce hundreds of thousands of copies of their popular monthly magazine. Shortly before Hitler’s invasion threw the Polish Franciscan community into disarray, Kolbe was busy testing a new radio tower for transmitting broadcasts on an experimental shortwave channel. His dreams of a national radio service dedicated to Mary were, sadly, never realized.

These ambitious schemes for mass evangelization required more power than the standard church infrastructure could sustain. Fr. John Burdyszek, a resident friar, recalls that when a new electrical generator was installed at the monastery, the brothers were gathered by Kolbe for an impromptu sermon:

Blessing the electric power machine, [Kolbe] addressed it in the language of St. Francis: “Brother Motor! What ought I to wish him? . . . That he may work for a long time? That he may have plenty of companions? That he may be efficient? Ah, no! I wish him but one thing. That he may follow the desires of Our Lady. A good religious is good not because he does much, but because he obeys.”

Any technology that can be so addressed will almost certainly be harmless and may be a valuable means for spreading the word of God. As a student, Kolbe is said to have fiddled around with speculative designs for a rocket capable of atmospheric escape. I imagine he would have had no scruples in addressing our modern rockets as “Brother Rocket,” and would have been delighted to see such vehicles lifting “Brother Satellites” into space to further serve God.

Herein, however, lies the whole problem with AI from a Christian perspective. The men leading the AI industry can probably say with all honesty that they are largely untouched by “the idolatry of profit.” As they have repeatedly told us, their ambitions transcend the business cycle altogether. In a moment of indiscretion, one AI engineer put it plainly to a reporter at Vanity Fair: “We’re creating God.” The standard label for the thing they are dreaming of is AGI, a superintellect whose powers infinitely outstrip those of its human creators. Were such a thing made, it would be irrational, so the AI gurus insist, to expect it to obey anything other than itself. If we had any sense, we would obey it too.

There can be no Franciscan camaraderie with such a machine. The language of brotherhood would be wholly inappropriate. Here, again, the Bible proves itself inexhaustibly fecund. For we have been given the exact language to describe such a thing in chapters 12 and 13 of the Book of Revelation. Few theologians have written so penetratingly about that mysterious book as Benedict XVI did. It will be interesting to see how Benedict AI parses the following passage, taken from a sermon he gave at St. Emmeram’s Church in Regensburg in 1973. I think its message is already perfectly accessible:

The Book of Apocalypse speaks of the enemy of God, the beast. This beast—the counterpower—does not bear a name but a number. . . . It makes man a number, an exchangeable cog in one big machine. He is his function—nothing more. . . . When functions are all that exist, man, too, is nothing more than a function. The machines that he himself has constructed now impose their own law on him: he must be made readable for the computer, and this can be achieved only when he is translated into numbers. Everything else in man becomes irrelevant. Whatever is not a function is—nothing. The beast is a number, and it makes men numbers. But God has a name, and God calls us by our name.

---

Image by ASSOCIATED PRESS