The decline of Latin as a universal language of scholarship and education is an incalculable tragedy. For example, despite all their national differences, the schoolmen could still speak to one another in one tongue. John Owen could read Calvin, and Calvin could read Aquinas, and Aquinas could read Anselm, and Anselm could read Augustine, despite there being almost 1,300 years of difference between them. The Reformers, despite being French, German, English, Czech, Swiss, or Dutch, could all read each other’s works. Today, untold volumes of the pre- and post-Reformation era are still largely inaccessible unless translated—and very few have been translated. For example, the Post Reformation Digital Library contains over 109,000 volumes from more than 6,000 authors. Most of these are still inaccessible to the majority of otherwise interested readers due to being in Latin.
It is not without reason that Douglas Adams called the Babel fish “probably the oddest thing in the Universe.” In The Hitchhiker’s Guide to the Galaxy, sticking the yellow leech-like fish in your ear would enable you to “instantly understand anything said to you in any form of language. The speech patterns you actually hear decode the brainwave matrix which has been fed into your mind by your Babel fish.”
I believe that with the advent of AI, retrieving otherwise inaccessible scholarship will begin to seem more like the science fiction of Adam’s Hitchhiker’s Guide. In the past eight months, I’ve updated into English and edited over 71,000 words from Jerome Zanchi that have been previously inaccessible, using ChatGPT 4 and the AI-powered optical character recognition service called Transkribus—all while on my lunch breaks.
Jerome (Girolamo) Zanchi (1516-1590) was an Italian Protestant exile and theologian. He was a close friend and colleague to Peter Martyr Vermigli. Zanchi represents the best of Reformed Thomism, and he did something that no Protestant had attempted before—to write a Reformed summa of theology. Richard Muller uses Zanchi as the prime example of the pre-Ramist method for treating loci on a massive expository scale. Zanchi’s work on the incarnation alone is nearly 1,000 pages.[1] However, don’t mistake Zanchi for a stuffy scholastic! His rigorous scholasticism served to enhance, rather than hinder, the practical devotion of his writings. In the words of Peter O’Banion, “Centuries after their first appearance, his works still impress. Their freshness, warmth, and rich insights command the attention of modern readers.” Some examples of Zanchi’s scholarly rigor and devotional warmth can be seen in the Works of God; the Incarnation; the Resurrection, Ascension, and Session of Jesus; an introduction to Ephesians; on the Opening of Schools; and also through his correspondence with Theodore Beza, Heinrich Bullinger, and John Calvin.
1. Richard Muller, Post-Reformation Reformed Dogmatics, I, 62.
However, almost all of Zanchi’s major works have never been translated into English. My long-term goal is to provide English translations from the nearly 8,000 untranslated pages of Zanchi. None of this would be possible at this speed without AI, and I believe tools like this will be significant to the study of post-Reformation scholastic theology.
In short, I use ChatGPT as a tool to do the heavy lifting of translation. What I create is a compound product: I use a customized ChatGPT 4 to assist in the first translation draft, then I check each line to the original and make changes. The three arguments I will make for adopting AI as a handmaiden to theological retrieval are access speed, division of labor, and richer resourcement.
1. Data Access Speed
Quick test: When you need to know the meaning of a word, whether English, Greek, or Hebrew, do you first pull out your phone or pull down a book from your shelf?
In biblical studies, we’ve generally accepted the benefits of using a program like Logos, Accordance, or Blue Letter Bible to speed up our study (to the chagrin of all our language professors). It’s all a question of the speed and efficiency of our time.
For example, last year, I taught a class at my church called Finding God’s Will. In preparation, I searched for every instance of “will” as a noun (thelēma/θέλημα) or as a verb (thelō/θέλω) in the New Testament and then placed every instance into categories. Of the 102 instances where God’s will was mentioned, not a single verse supported the Keswick/Henry Blackaby view (which states that God’s will for my life is discerned through internal impressions and subjective signs), but every instance applied to either God’s sovereign/secret will or God’s revealed/moral will. I was able to accomplish this feat in a few hours in an afternoon. Imagine how long that study would have taken if I didn’t have a digital search engine available. The same study could be done using physical Greek concordances and pen and paper, but conducting research more slowly when better tools are available doesn’t make much sense. Foresters have chosen to use chainsaws instead of handsaws for a reason.
I believe there is a place for using tools like ChatGPT to speed up our translation work. If we accept the usefulness of a digital search engine to look up the meaning of a word, then we should be willing to accept an AI service like ChatGPT or Claude, which can look up the meaning of dozens of words at the same time, place them in context, and provide an impressive translation.
2. Division of Labor
When using ChatGPT to translate, functionally, my work is akin to an editor reviewing the work of an assistant. I am able to delegate to AI the tasks that play to its greatest strength, namely, pattern recognition. AI algorithms, especially those based on machine learning, are excellent at recognizing and learning patterns in language. In Latin, with its structured grammar and inflections, I have seen how AI can quickly identify and apply grammatical rules, making it efficient for translating repetitive phrases. After reviewing tens of thousands of Zanchi’s words, I have been consistently impressed at ChatGPT 4’s ability to maintain a uniformity of style in translation as well as its ability to understand context, idiomatic expressions, syntax, abbreviations, and historical situations. This allows me to focus my time on higher-level translation questions while checking its work. One example comes to mind:
In a 1571 letter to Zanchi, Theodore Beza closed with an update on the affairs of Geneva, noting that their recent plague has been expiated “with the punishments of several most wicked witches” [aliquot sceleratissimarum veneficarum suppliciis]. This translation from ChatGPT surprised me, to say the least. I checked several Latin dictionaries for the range of meanings for each word to ensure I wasn’t missing any nuances.
Venefica is a female poisoner or potion maker when strictly observing the etymology. English translations of Cicero, Quintilian, and Seneca use the word poisoner, but generally in the context of a male tried in criminal cases. When the poisoner is female, the meaning takes on sorcerous undertones. Translations of Ovid use the word witch when describing Medea plying her art with magical herbs, and Horace uses it for witches or hags crafting vile philters of enchantment.
Were these women in Geneva punished for the crafting of poisons and potions, or were they punished for divination and magic? I deemed the safest choice was “witch” since the word has overtones of brewing poisons as well as the practice of sorcery.
The sentence still left questions in my mind: Did Beza truly believe that Geneva’s plague was a judgment of God for its local witches? Did Beza believe that the execution of witches removed the plague? I’ll leave those questions for the historians. But my time was well spent as an editor verifying the translation.
3. Richer Resourcement
I would be perfectly happy to have someone else translate Zanchi’s corpus. However, I felt I couldn’t in good conscience sit by and wait decades for the work to be done, if it ever would get done at all. I am persuaded that there are treasures of doctrine waiting to be discovered (or rediscovered) in the volumes of Zanchi. The use of AI could bring works of theology like his into the hands of grateful students to be studied again.
For instance, in his writings on the incarnation, Zanchi presents a model of theological retrieval. He refutes the Lutheran mistake of ubiquity by thoroughly examining the orthodox history of the person of Christ. In doing so, he shows his expertise by referencing 48 different early church fathers and sources, in addition to citing Peter Lombard, Bernard of Clairvaux, Boethius, and Thomas Aquinas. By retrieving the work of Zanchi, we will also retrieve an exemplar of how to use the network of theological resources throughout pious antiquity.
As it stands, thousands of pages of Zanchi’s work remain untouched. To name a few:
- On the Six Days of Creation (1,153 pages)
- On the Holy Scriptures (521 pages)
- On Sin and the Law (1,193 pages)
- On the Incarnation (932 pages)
- On the Three Elohim (891 pages; Davenant is currently working on a translation)
- On Divination and Astrology (469 pages)
- On the Marriage of Adam & Eve, and Marriage in General (542 pages)
- A Commentary on Ephesians (546 pages)
- A Commentary on Hosea (779 pages)
- Among many other works, including a book of correspondence (190+ letters) to and from leading figures in the Reformation, including Calvin, Beza, and Bullinger.
AI-enabled English translations of these works will only further our understanding of history while providing resources to address current theological questions.
Conclusion
If we have learned anything about retrieval for the sake of renewal or of the democracy of the dead, then we would be well-guided to recover the resources of the past instead of forging ahead without any guidance. In fact, the wise way forward is to go backward. Christ has given to His Church teachers, and those teachers include those who have since died. And the dead still speak—even if it is only in Latin. I hope that we will be able to hear them in our own language soon—perhaps with the help of an AI Babel fish.
