Christianity has always been a faith in translation. Its Scriptures were given in Hebrew, Aramaic and Greek, and its spread across the world has meant constant movement from one language into another — in sermons, songs, letters, papyrus and heavy printed tomes. The Gambian theologian Lamin Sanneh put it memorably: “Christianity is a translated religion without a revealed language. … Without translation there would be no Christianity. Translation is the church’s birthmark as well as its missionary benchmark.” That inheritance is now being tested in a new way.
This month Mere Orthodoxy launched The Faith Received, a large free online repository of theological texts. For the most part those texts are in English as the result of AI translation, and the project invites human readers to correct mistakes by comparing the rendering against the source. The online response was quickly polarised into three camps: those who see an obvious win, since obscure texts are now free to read; those who counsel cautious pragmatism, since AI has good and bad sides and accuracy must be checked; and those with deeper reservations, who see a project like this as more of a curse than a blessing. James Eglinton, writing for The Gospel Coalition, places himself — “mournfully” — in the third camp, while insisting he admires the people behind it and shares their desire to see Christians read the historic theological literature.
What troubles him is that the shift changes something significant in the historic nature of Christianity as a religion in translation. “The shift from Christianity as a human-translated religion to an AI-translated religion is far from seamless and shouldn’t be overlooked glibly,” he writes. Much depends on what translation is taken to mean. English, drawing on the Latin translatio, implies “to carry across”, a transfer from one language to another. The Tamil verb peyar, his colleague Hephzibah Israel has written, instead conveys “to dislocate, to uproot, or to unseat with force”. English invites its speakers to imagine translation as relocation. Tamil foregrounds the idea that translation is interpretation.
Eglinton offers a kitchen analogy. Monolingual English speakers, he finds, imagine translation as converting a recipe from cups to millilitres — the same dish in different units, with only superficial differences. To people who live in more than one language it is more like recreating a dish in another culture, where some ingredients cannot be sourced. Sometimes a substitution works and the dish stays recognisably itself: honey instead of sugar. Sometimes it approximates: lemon juice and sugar standing in for tamarind. And sometimes the ingredient has no equivalent at all. No blend of turmeric and paprika will reproduce saffron’s colour, floral aroma and earthy taste; you can represent what was there, but it is not the same. “As the chef/translator, to be truthful I should probably let you know that in this case, there wasn’t a direct way to move from one recipe to the other.”
He gives worked examples from Dutch, a language he learned over much of his adult life in order to make Herman and J. H. Bavinck available to English readers. The verb form ik lees can mean either “I read” or “I am reading”; English forces the translator to choose, because the ambiguity cannot be carried across. The word gezelligheid has no English equivalent. In the Calvinist poetry of Willem Bilderdijk, unusual vocabulary and rhyming couplets reinforce one another in a way English cannot reproduce: “I cannot simply carry over its stylistic composition and its conceptual meaning. Both are integral to the original. One of them must be sacrificed.” That raises an awkward question, he suggests, about what the result should even be called — Bilderdijk in English, or a new English poem prompted by reading Bilderdijk?
The project’s defenders answer that a reader can always check an AI rendering against the original. Eglinton’s reply is that this only works for readers who know the original language. To judge an AI translation of Bardesanes, a third-century Syriac theologian, “you need to know Syriac.” Everyone else must take the rendering at face value — and may pass on its errors, or mistake the AI’s translation biases for what the author actually wrote. He pictures a preacher quoting Bardesanes from an unverified AI translation: attribute it honestly and the moment deflates; leave the attribution out and the preacher is quietly being shaped by the practice, along with its assumptions about truth.
None of this is entirely new. Between the twelfth and fourteenth centuries, a medieval translation industry circulated a great quantity of Latin renderings of Greek and Arabic texts, often erroneous, received uncritically by readers who knew no Greek or Arabic. One fruit of it was a general laziness toward the study of languages among otherwise serious Christian thinkers — a norm that the English polymath Roger Bacon criticised, tracing its effects on biblical interpretation, scholarship and evangelism. AI translation amplifies that practice, Eglinton concludes, but did not create it. “In our day, the AI translation industry is our direct equivalent. Making the project of theological retrieval dependent on it is no small development in the long history of our ‘translated religion.’”
Source: The Gospel Coalition — thegospelcoalition.org
