One of my earliest assignments as a young interpreter was to provide simultaneous interpretation for the proceedings of an ecumenical council that brought together all Christian denominations. As my homework, I dutifully read scripture, the gospels, papal encyclicals and the conclusion of the first council of Nicaea.

There was, however, one thing I had not foreseen. Mass was held not in the conference hall, but in the church itself, where there were no booths and the interpreter was required to stand discreetly on the altar. Here, translation alone would not suffice – the interpreter had to perform the part of the priest, with his unmistakable clerical timbre, the arms outstretched then folded in prayer, the gaze repeatedly lifted towards heaven.

My childhood experience as an altar boy helped, as did that innate instinct for the theatrical that seems always to come naturally to Italians. My performance was so flawless that when a telegram arrived from Pope John Paul II wishing the council well, I was entrusted with translating his Latin. The temptation to give it a Polish accent was strong, but I restrained myself.

Whether the latest developments in artificial intelligence and voice-to-voice interpretation will include a “priestly voice” setting and the whimsical option of a specific accent, I cannot say. Should they do so, future participants in ecumenical councils will be spared a most curious spectacle – and, I venture to think, deprived of a certain charm.

Live voice-to-voice interpretation, which the Cologne-based AI translation company DeepL unveiled earlier this month, marks the crossing of a frontier in artificial intelligence and in the realm of language from which there will be no turning back. The age of the interpreter is over: that ambiguous figure poised between the shrewd mediator who averts conflict and the scapegoat, who made communication possible not only between speakers of different tongues, but between different worlds and different ways of apprehending reality.

The machine will perform this task far better – cleanly, without siding with one party or another – and the economic savings will undoubtedly be considerable. The transformation in human communication will be profound. But are we certain it will be progress? Will the crossing of this frontier truly enhance communication and mutual understanding among people of different cultures and languages?

The first effect of the AI translation revolution will be to render the study and learning of languages superfluous for individuals. It will be enough to turn to our phones to understand whoever speaks to us and to translate our own speech into any language. Eventually, we shall be able to read information in every language, to write texts that can be read from one end of the world to the other. Yet knowledge – true understanding of others, of their cultures and customs, of the cast of mind of another country – will not thereby become ours. This body of knowledge will reside in AI systems, not in us.

If no one studies other languages and cultures any longer, we shall know nothing about the person to whom we are speaking. Until now, to study a language was also to enter its culture. And to learn language and culture one must love them, become passionate about them, feel a kind of infatuation with that country and its world. One always learns something because one loves it; only thus does one truly learn it. With AI, this process of conquest through knowledge will be lost. The passion for knowing and discovering another people will disappear. Languages will become for us mere codes to be deciphered, and we risk knowing nothing at all about the people who speak them.

Nor is it certain that AI systems will prove infallible in translation. However thoroughly they may be supplied with every possible piece of information about a country and its culture, they will always lack the capacity to judge the situation – the moment in which an encounter takes place and translation becomes necessary.

After my brilliant debut at the ecumenical council, my career as an interpreter continued with more prosaic assignments. I was once hired to provide simultaneous interpretation of lectures delivered by a group of Neapolitan engineers to a group of technicians from several French-speaking Arab countries, on site at a production facility in southern Italy. But my work did not end in the classroom. It continued in the evenings, over dinner and in conversations between the engineers on both sides.

The Neapolitan engineers were very curious to know how many wives their Arab colleagues had. Clearly, their knowledge of the Arab world went no further than a distorted vision drawn from the Arabian Nights, tinged with a rather backward attitude towards women that was then still common among southern Italian men. It was obvious that I could not ask such a question. So instead I asked the north African technicians how many children they had. Out came figures that satisfied the Neapolitans: at least two, but often three, even five. The Neapolitans widened their eyes, offered congratulations, slapped their colleagues on the back; the north Africans, in turn, basked in what they took to be praise of their reproductive prowess. Everyone was pleased, and my false translation served the good cause of understanding and conviviality.

It may be that the AI of the future will learn to master the particular fixations of future Neapolitan engineers. But there is a poetry, and even a certain nobility, in attempting to speak – however imperfectly – another language, even at the cost of provoking laughter through an error or a misunderstanding. It is, in the end, a form of courtesy to try to learn another’s language, a sign of interest and regard, a tribute to their culture. With AI translation, the humanity, the sense of wonder, and the emotional reshaping that comes with discovering people different from ourselves risk being lost for ever.