Say what you will about ChatGPT and the humanities, but it’s been about 10 years since I first picked up Hermann Hesse in Berlin and I’ve never looked at the German text… until now.
I've been watching David Perell on the How Do You Use ChatGPT podcast, where he compares different translations of the Bible and discusses how the source language affects the English translation. I tried this out with one of my favorite quotes from Hesse’s Siddharta, this version from Stanley Applebaum.
“What is it that you've learned, what you're able to do?" "I can think. I can wait. I can fast." "That's everything?" "I believe, that's everything!" "And what's the use of that? For example, the fasting -- what is it good for?" "It is very good, sir. When a person has nothing to eat, fasting is the smartest thing he could do. When, for example, Siddhartha hadn't learned to fast, he would have to accept any kind of service before this day is up, whether it may be with you or wherever, because hunger would force him to do so. But like this, Siddhartha can wait calmly, he knows no impatience, he knows no emergency, for a long time he can allow hunger to besiege him and can laugh about it. This, sir, is what fasting is good for.”
Now in the original German as per Project Gutenberg.
"Was ist es, das du gelernt hast, das du kannst?" "Ich kann denken. Ich kann warten. Ich kann fasten." "Das ist alles?" "Ich glaube, es ist alles!" "Und wozu nützt es? Zum Beispiel das Fasten—wozu ist es gut?" "Es ist sehr gut, Herr. Wenn ein Mensch nichts zu essen hat, so ist Fasten das Allerklügste, was er tun kann. Wenn, zum Beispiel, Siddhartha nicht fasten gelernt hätte, so müßte er heute noch irgendeinen Dienst annehmen, sei es bei dir oder wo immer, denn der Hunger würde ihn dazu zwingen. So aber kann Siddhartha ruhig warten, er kennt keine Ungeduld, er kennt keine Notlage, lange kann er sich vom Hunger belagern lassen und kann dazu lachen. Dazu, Herr, ist Fasten gut."
ChatGPT came up with different translators altogether, here describing each translator’s style.
Hilda Rosner: Rosner's translation would likely focus on clarity and directness, maintaining the simplicity of the original German while making it accessible to English readers. Her translation would read smoothly, capturing the practical wisdom in the passage without overly embellishing the language.Joachim Neugroschel: Neugroschel's translation might delve deeper into the cultural and philosophical context, possibly adding footnotes or explanations to convey the full depth of the original text. His version would be rich and detailed, ensuring that the cultural and philosophical nuances are not lost in translation.Susan Bernofsky: Bernofsky's translation would focus on capturing the lyrical and meditative quality of Hesse's writing. She would aim to reflect the original tone and mood, ensuring the translation feels as poetic and introspective as the German text. Her version would resonate with the philosophical and spiritual elements of the passage.
While ChatGPT wasn’t able to find the actual translations, I got ChatGPT to translate from German to English in the style of each translator.
Hilda Rosner: "It is very good, sir. When a person has nothing to eat, fasting is the wisest thing he can do. For instance, if Siddhartha had not learned to fast, he would have had to take up some service today, whether with you or elsewhere, because hunger would have compelled him. But now, Siddhartha can calmly wait; he knows no impatience, no distress, he can let hunger besiege him for a long time and still smile about it. This, sir, is the value of fasting."Joachim Neugroschel: "It is exceedingly beneficial, sir. If a man has naught to eat, fasting becomes the most prudent action he could undertake. Take Siddhartha, for example; had he not mastered the art of fasting, he would be compelled to accept any form of servitude this very day, be it with you or elsewhere, driven by the coercion of hunger. However, Siddhartha can await events with serenity; he is a stranger to impatience, to dire straits, able to endure the siege of hunger for extended periods, and can regard it with a hint of amusement. Such, sir, is the essence and purpose of fasting."Susan Bernofsky: "It is profoundly good, sir. When one finds oneself without sustenance, fasting emerges as the most sagacious course one might embrace. Consider Siddhartha: had he not been adept in the discipline of fasting, he would find himself compelled to enter into servitude this very day, whether in your employ or elsewhere, as hunger's inexorable demand would dictate. Yet, in his way, Siddhartha can await with tranquility, untouched by the throes of impatience or desperation, able to let the pangs of hunger encircle him for lengthy spans, and meet it all with a trace of laughter. In this lies the true virtue of fasting, sir."
It looks plausible (I quite like Neugroschel's version) but it doesn’t hurt to be a little skeptical. I asked my friend Spencer Strub who teaches writing, and this is what he said:
This shows ChatGPT's strengths -- an impressively thorough digest of the approaches of each translator, and in nice prose, too! -- but that last paragraph also shows why students mess themselves up when they just turn in ChatGPT output: it's not going to tell us which would be best and why. That's a subjective judgment, and ChatGPT rightly doesn't offer it. It's a cool exercise, though...
Thanks for getting this far! In a nutshell, even if there’s room for improvement this is the furthest I’ve gotten into a closed reading of Siddharta despite reading it almost every year. All thanks to a word calculator.
I’ve been trying different things since the start of the year and this following week I’m experimenting with the following focus areas:
- Generative AI: Full Stack LLM bootcamp, build apps and deploy to the cloud (likely Vercel + Modal)
- People skills: Staff+ Engineer roundtable
- Communication skills: Ultraspeaking
On generative AI, I did a speed run of Andrej Karpathy’s Zero to Hero series at my most recent RC batch. I contemplated a re-run of that with exercises, or going through fast.ai courses. I met one of the instructors at Full Stack LLM at the Modal Labs meetup yesterday, took a closer look at the materials and thought this would be a good way to learn generative AI from a product-centric perspective.
On people skills, it’s a recent focus and now switching gears to a group setting. I’ve set up the Staff+ Engineer roundtable at Recurse Center for next week, and excited to discuss a few threads I’ve been pulling: professional coaching, networking, self discovery. In fact, I tried out the exercise from podcast above by pasting in the results of a personality test into ChatGPT and then asking who I work well vs less-well with. I was blown away by the less-well personas, perhaps something I can share at the roundtable if there’s interest.
On communication skills, I start the Ultraspeaking course next week! The Youtube channel has illustrative videos of what we’ll get into. I know myself well enough to know I need a class setting for consistency. Happy to share my findings as the course progresses.
OK so this post has been more about what I plan to do rather than what I did today. I shared David Perell’s podcast with more people, describing it as follows.
In this episode the writer David Perell talks about how deliberately he is at learning about himself and understand what he could be world-class at; in particular where ChatGPT fits in that process.
I replied to a few tweets: here, here and here. I signed up for more events: this one on applied LLMs and this one on sharing outage stories the following week.
The following quotes echo advice on building expertise in Python that I previously shared on this blog; this came up in an RC discussion. Here is David Beazley.
As an aside, sometimes people ask me "what can I do to improve my Python skills?" Much to their surprise, I often suggest doing a project in a completely different language or outside of their area of expertise. I think the main benefit of doing this is that you'll often see a completely different way of thinking about a problem that you can bring home to your own projects.
Next Peter Norvig.
Learn at least a half dozen programming languages. Include one language that emphasizes class abstractions (like Java or C++), one that emphasizes functional abstraction (like Lisp or ML or Haskell), one that supports syntactic abstraction (like Lisp), one that supports declarative specifications (like Prolog or C++ templates), and one that emphasizes parallelism (like Clojure or Go).
Finally Donald Knuth.
If you find that you’re spending almost all your time on theory, start turning some attention to practical things; it will improve your theories. If you find that you’re spending almost all your time on practice, start turning some attention to theoretical things; it will improve your practice.
Never gets old.
All the resources I shared