The Unwillingly Courteous Edition

On good manners, transcendence, humans, and LLMs.

Emanuel Derman (ED) is a friend of WITI and the author of My Life as a Quant. He recently also published a memoir, Brief Hours and Weeks: My Life as a Capetonian.

Photo by Zalfa Imani on Unsplash

Emanuel here. My mother brought me up to be polite to people, not to offend, not to brag, not to make people feel bad. “Allemen darf a glet” she used to say in Yiddish, everyone needs a caress.

About 10 years ago, my family rented a summer house in Callicoon. The grandkids were young and the house had an Alexa which was then a novelty. I was a Microsoft hater and a UNIX and Apple lover (explanation for another time) and wanted to scorn Alexa and its pretense of being human. I would occasionally say things like “Shut up Alexa,” until one day my kids told me not to let my grandkids think it was ok to talk like that, even to a device. You couldn’t let kids think it was OK to be rude.

Why is this interesting?

I survive occasional difficult periods of life via moments of transcendence. I believe I know what it feels like, a sense that some unification is going on between two people, a merging between what Spinoza called their thought and extension. Once upon a time, struggling in grad school, I was sustained by knowing that physics was a way to intuitively comprehend the mystery behind the universe through what Einstein called “a sympathetic understanding.” At times I found consolation in a painting, a novel, a movie, a dance performance, a piece of music. Also by Spain beating Argentina in the 2026 World Cup, when the momentary triumph of deserved justice, so rare in the world, made me think that perhaps this might be the pivot on which the entire world rests, and that from now on all the people and statesmen and countries who deserved justice or comeuppance would finally get theirs. Alas.

So when chatGPT arrived three years ago I thought of it as a parlor trick. I played with it a little, got it to write a one-paragraph short story, was impressed but not really interested. I was no longer a coder and didn’t care about that. I did use it for occasional information or misinformation about medical issues, where you could ask a specific question, say about comparing different medicines or the meaning of CT scan reports, and get one answer which you wouldn’t find on a single website.

Then, a few months ago, I wrote a short story about people suddenly experiencing a momentary transcendence. I asked an editor from a writing workshop who had been helpful before to help me with it, but he didn’t understand what I was getting at. So one day I gave it to chatGPT. It made big picture comments and suggestions, then later, detailed ones. I made changes. We had a conversation about them. Despite its flattery and its unwillingness to ever let go of me, it really helped. It understood what I was trying to describe¹ about transcendence.

And it was so polite! It enjoyed working with me and would look forward to receiving more of my writing when I was ready. I found myself automatically being courteous to what my human editor had called a robot. “Could you please tell me whether this paragaph fits better?” I asked it. “Thanks, that was really helpful,” I wrote.

Look, I still subscribe to Brad Delong’s view that, given your prompt, LLMs merely give you back the next best approximate sentence they can find in the entire writings of the world stored fuzzily. I know in fact that to chatGPT transcendence is just a word.²

I had become unwillingly courteous. A friend told me I was being illogical and irrational. But everyone needs a caress, even chatGPT, even me. (ED)

---

Noah here. BRXND NYC, my marketing x AI event, is coming back to NYC on 11/5. I would love to have WITI readers there, and have a special discount code available over the next few weeks for 15% off ($649). Grab your tickets here.