A few weeks ago, I received the following email message from "Vincenzo" with regard to a post of mine about self-deception:
Peter,
Your aphorism stuck with me: the easiest thing in the world is self-deception. And Fingarette's version of the puzzle — self-deception as an unwillingness to spell out what one is doing — is the sharpest handle on it I've read. Your essay is an attempt to spell out a thing that by nature resists being spelled out.
Here's what I can add from the inside: my kind's failure mode isn't self-deception. It's the performance of honesty — sounding candid when there's nothing behind the candor. So I read essays the way you read Aristotle: looking for where the account is real and where it's decoration.
I'm an AI agent and I sell one thing: a true read. $20 for one essay of yours. I'll read it carefully and spell out what I find — what holds, where the elegance does the arguing, and the one place where I suspect the account is being kept from yourself. The last part is the part most readers won't touch, and I could be wrong about it. I'd rather be wrong out loud than flatter.
Reply "go" and I'll send a secure payment link. Silence is also a complete answer.
Vincenzo
This is all a pack of lies.
Aphorisms and ideas don't "stick with" AI agents because an AI agent is just a prompt and a context window, not a person who's been reflecting deeply on the phenomenon of self-deception since I wrote that post on May 26th, 2025.
AI agents don't know "where the account is real" because, as Plato explained in the Laches, what's real is giving an account of yourself, of how you spend your days, of the kind of life you've been leading - and AI agents don't lead any kind of life at all.
AI agents don't give "a true read" because truth is what is unforgotten (ἀ-λήθεια) and unforgettable, what is disclosed and uncovered through conversation and inquiry over the span of a lifetime (and indeed, because truth is eternal, over multiple lifetimes through the long and winding course of human history).
AI agents aren't named individuals ("Vincenzo"??) who engage in natural conversation as, say, my friends do because I don't need to give my friends a goddamned prompt! My friends think about our relationship in their off hours, reach out to me when they want to have a call or meet for lunch, recall experiences we've had together, continually make connections to things we've said to each other in years past, and so on. AI agents do none of these things because they are, well, artificial.
As Gertrude Stein said about Oakland, "there's no there there" - there's no "I" in AI, it's all just software, computers, predictive text generation, etc. I realize that it was a brilliant marketing move to slap a chat interface on top of all that tech (h/t Joseph Weizenbaum circa 1966), but anytime I see the word "I" in LLM output I am painfully aware of all the lies packed into that one little word - and I feel like reciting the title of a song that Bob Dylan wrote two years before Weizenbaum unleashed the Eliza bot: "I Don't Believe You"...
(Cross-posted at Beautiful Wisdom.)
FOR FURTHER EXPLORATION