How I Use ChatGPT & Latin
How am I translating Zanchi so quickly, you may ask. The short answer is that I use ChatGPT as a tool to do the heavy lifting of translation. Maybe you view that as an inappropriate tool. Does it make my translation illegitimate? That’s for you to decide. Let me walk you through my process, and you can judge if my use is proper.
Update: here is 3-part video of my process
1️⃣ I first divide the PDF into individual images via a pdfimages script in the console, and then I upload the files to Transkribus
2️⃣ I manually set the areas for word detection on the document and let Transkribus do its work of AI-powered OCR
3️⃣ I review every word, line by line, of the image to the suggested word. The model is usually 98% accurate. The AI can even detect shortened words and supply the full word. For example, neq3 is neque, and diē is diem, etc.
4️⃣ However, I still need to type in the Greek words and phrases manually. The current model can’t do it. Update: a new model I’m using can do Greek Ligatures and Contractions much more accurately. It requires a few manual corrections.
5️⃣ Once satisfied, I then download and format the plain text into a word processor. I manually add the marginal notes in brackets. I manually add the same paragraph breaks to reflect the original. I manually italicize the same words as the original
6️⃣ Then I feed the Latin into ChatGPT, 1-3 sentences at a time. I usually ask for 2-3 different types of translations: at least one translation that is “normal” and one that is woodenly word-for-word.
7️⃣ I compare the English translations to the Latin. Nine out of ten times, it does a good job on the first try. I sometimes have to ask ChatGPT to parse every word of the sentence and reprimand it when it drops out a word. The work session gets better with use.
8️⃣ At times, I will go thru Latin dictionaries to see if certain words are the best choice:
➡️ One example is with Isaiah 49:23, “And kings shall be thy nursing fathers, and their queens thy nursing mothers.” ChatGPT suggested “foster fathers and nursing mothers” from Zanchi’s “Et erunt (inquit) Reges nutritores tui, & Regina nutrices tuae.” That seemed to be an odd choice and one that I wasn’t comfortable with.
Yet, after all, I thought, kings do not nurse children, being men. Only women can do that. And, after all, “foster” comes from the Old English for “food,” which was similar to nutrix, which means to “nourish.”
But I also looked at Young’s Literal, Wycliffe, the 1599 Geneva, and the 1611: all translated it as “nursing father.” So I decided to stick with “nursing father and nursing mother” to respect the tradition and not be a distraction for the reader.
9️⃣ I also work as hard as possible to stick to word choices that would best mirror the Latin:
➡️ For example, Zanchi wrote, “Est haec Epistola ceu epitome totius doctrinae Christianae…”
ChatGPT made the bland suggestion of “This letter is like a summary of the entire Christian doctrine…”
But Zanchi used epitome as well as Epistola, so I thought using epitome and Epistle would serve us better. Thus: “This Epistle is like an epitome of the entirety of Christian doctrine…” It also preserves the alliteration.
🔟 I finally combine the suggested translations from ChatGPT, with my corrections, into one that I think best captures a reflection of the original’s arrangement of clauses and word choices. My goal is to produce a lively translation that would make me as invisible as possible while being as faithful as I can to the original.
🤖 How Do We Use Tools? 🤖
Why am I talking about all of this? I wanted to be as transparent as possible about how I’m producing these translations. Maybe it is also to ward off suspicion about how I was producing these translations so quickly. I guarantee you that I am no Latin phenom.
I also think that we’re in a watershed moment when it comes to AI and the tools we use. Compare the digital tools that were available to us in 2020 to 1980. That 40-year gap saw exponential and dramatic differences. It may not be an exaggeration to say those 40 years saw more changes than in the past 100 years. We’ve moved from kitchen phones to smartphones. We went from MS-DOS (1981) to Siri (2011).
Just look at how a program like Logos has changed how we interact with biblical languages. (Full disclosure: the last time I used Logos was in 2011 in seminary. I’m sure it’s gotten even better than the last time I used it).
How much time would it have taken 40 years ago to look up every word in a Greek sentence? Barring words you’ve already memorized, you would have needed to flip in your 1983 BDAG to read each word (technically, it would have been the 2nd ed. “BAGD” back then). In Logos, you just have to hover over with your cursor and see the definition and the full parsing of each word.
And now, in the last three years, the current leap in technology may prove to be just as dramatic as the move from kitchen phones to smartphones. We’re at the beginning of another exponential leap in technology with AI.
Today I can submit a 1,400-word essay I wrote to ChatGPT, ask it to review the content, and suggest 10 different article title ideas—and it can do it in less than two seconds.
❌ Objections ❌
⏺️ But doing translation work this way, David, will stunt your learning of Latin.
Possibly. I’ll keep faithfully working through my Lingua Latin primers and my flash cards. Maybe one day I’ll be able to read Zanchi without so many helps. But even then, I imagine, I’ll still need to look up a lot of words in a Latin dictionary.
Using ChatGPT speeds up that process of discovery. I view it as the difference between looking up a word in a book versus using a digital search engine. It’s a question of the speed and efficiency of my time. We’ve generally accepted the benefits of using a program like Logos to speed up our study (to the chagrin of all our language professors).
I think there is a place for using a tool like ChatGPT to speed up our translation work further. What would be appropriate ways to use it? What would be inappropriate ways to use it? It’s worth the discussion.
⏺️ But no publisher will publish a book that used ChatGPT to help with the translation.
Possibly. Again, we’re going to have to have a discussion about the boundary lines of permissibility for translation work in this brave new world of AI that we’re living in.
⏺️ But why are you doing this translation work—if we can even call it “real” translation work—in the first place, David?
Because the field of Post-Reformation Scholastic Theology is the field of study that I want to be involved in, and I want to get into the work as quickly as possible. I would love to do post-graduate study in this area, if possible, and I want to dedicate whatever spare time I have to help promote the study of one particular scholastic, Girolamo Zanchi.
I think it’s a shame that there are thousands of untranslated pages of his work just sitting there untouched. To name a few:
On the 6 Days of Creation (1,153 pages)
On the Holy Scriptures (521 pages)
On Sin and the Law (1,193 pages)
On the Incarnation (943 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)
Commentary on Ephesians (546 pages)
Commentary on Hosea (779 pages)
Among many, many other works, including a book of correspondence (190+ letters) to and from leading figures in the Reformation, including Calvin, Beza, and Bullinger.
By delving into the work straightaway, I’m finding that the text is bringing loads of questions to my mind, and I’m learning along the way: Why did Zanchi teach at Neustadt? Who was Prince John Casimir? Why did the independent publication of the speech have a dedication to Chancellor von Ehem? How did the magisterial reformers view their patron princes?
If I may be blunt, my goal is not to be a Latin expert, but a Zanchi expert. I would be perfectly happy to have someone else translate his corpus. If all his works were in English, then I’d buy them all right now.
However, I felt I couldn’t in good conscience just sit by and wait decades for the work to be done, if it ever would get done at all. And in doing this work of translation, I have found that it has provided unique benefits to me as I have been forced to read and reread Zanchi’s words slowly.
My interest in tools, like drills, is to be able to use them, not to know exactly how the drill motor works. I’m not opposed to the languages. I’ve done my years of Greek and Hebrew. (I even took a course on Daniel in the original languages in seminary—so add a vague understanding of Aramaic to the list of the things I know). But it was enough time to show me that it wasn’t my cup of tea.
I still use the languages, but it’s not a romance. It’s a necessary step (for which I have a healthy amount of respect) in order to do what I love, which is theology. I love teaching about homoousia and hypostasis because of the significance it has for our understanding of the true divinity and true humanity of the Son. I love teaching about presbuteros, poimen, and episcopos, and every chance I get I’ll enthusiastically spell out the etymology on a whiteboard when teaching church members—but only because it is so significant for understanding the office of elder/pastor/overseeer. I love getting into the meaning of ekklesia and its Athenian city-state roots of the assembly—but only because I love the doctrine of the church and congregationalism.
I’m using tools to do a job, not because I aspire to become a tool teacher or a tool trade show salesman who can get into the differences between a Milwaukee and a Bosch drill. I’m sure that difference matters, but I can’t be bothered about it. I just need a tool to do the job.
Getting my Zanchi translations published is secondary to me—though, admittedly, I do have kids that still need to be put through college. And if my translations are not good enough for publishing standards, then so be it. Or just put a ChatGPT warning sticker on the insider.
The point is that I’m enjoying delving into Zanchi, and it’s exciting to see some of his works in English for the first time ever. It’s exciting to know I’m getting to enjoy a scholastic theologian’s works, over 440 years after he wrote them. I’m learning and loving the book of Ephesians more than I ever have.
(I’m especially excited to find out what he had to say about divination and astrology, as well as what he had to say about angels and the planets of the Primum Mobile in his 6 Days of Creation. Did any other Reformed Scholastics write on those subjects?)
I can’t wait to see how he summarized the teachings of the church Fathers and the great tradition on the incarnation, as filtered through the Reformation. (Did any other Reformed Scholastic write an entire volume on the incarnation?)
I can’t wait to see what he wrote to Calvin and what Calvin wrote to him. I can’t wait to see what he and Bullinger exchanged so many letters about.
I think it would make our Italian friend glad.