Replacement Theory
It’s been an AI kind of summer in the Boney household. I’ve been thinking and writing about the job and policy implications of AI’s increasing ability. My son, a computer science major, has been trying to figure out how to translate his increasing knowledge of AI into some kind of work after he graduates.
This week we decided to see if AI could replace what I am doing here on this website. The answer: yes; more easily than you might imagine, and a lot more efficiently.
We started by simply directing Chat GPT to the website: boneconnector.com. We asked it to read every post on the site (the 100+ posts I’ve put probably 1000+ hours into). Between 1 and 2 seconds later (yes, actually) it was done (and now, since the AI has read more of my posts than anyone else in the world, I’m granting it a nickname -- CG -- and awarding it probationary status as a human male. We can argue about those decisions another time).
Did “CG” think “he” could write a post in my style? Absolutely, he said (CG is relentlessly can-do). Even though my style was (CG said) “quite distinctive” (CG is nothing if not a flatterer), he thought he could replicate the “thoughtful, reflective voice that feels like he is talking to you – but with data and insight backing every point” (oh, CG, you say all the right things!), the “smart anecdotes” and “conversational, yet scholarly tone” (CG, you rascal; you are too much!), “occasional humor” (only occasional?), and the blog’s dogged unwillingness to end “in despair; it proposes solutions, moral frameworks, calls for participation.”
There was more, but you get the point: CG is a genial pleaser (I liked Meghan O’Rourke’s recent description of Chat GPT being like “an intern with the cheerful affect of a golden retriever and the speed of The Flash”). But, eye-rolling aside, I was stunned by what CG had picked up about me during his 2-second deep dive into my writing. It was time to put him to the test: could he write my next post for me? We gave him no ideas to work with. Maybe that would slow him down.
CG would be perfectly willing to deliver you from your writer’s block. See how happy he is? (Image generated by deepai.org)
After an interminable (three second) delay, he was done with the essay. It was pretty darned good.
In this post I want to tell you a little about the piece CG labored over for three seconds; next time I’ll share how I translated the ideas from CG’s post when I tried to write on the same subject.
The (+) side of CG’s post:
· CG decided to write something about the decline of libraries in the US and their potential to revive connection in communities. That kind of idea is perfectly in keeping with one of my core concerns – that we are losing “third places” (places outside of work and home where we can go to be with others). And even though the library is mostly where I go to write these posts, that thought hadn’t occurred to me. That seemed like a solid idea.
· CG found some research showing declining funding for libraries and, especially in rural areas, cutbacks in the number of hours they were open. I hadn’t read those studies; CG had.
· The style of CG’s post was similar to mine, and a reminder that I too often write in three’s. Here’s an excerpt that sounded uncomfortably familiar: “Too often, libraries are cast as quaint, outdated or passive. In reality they are frontline organizations: fighting disinformation, poverty and isolation – quietly, daily, effectively.” Ouch.
· CG’s piece was, I found, better edited than the ones I write and edit myself: once he made a key point, he didn’t go on and on like I sometimes do. He stopped. There’s something I can learn from.
The (–) side of the post:
· I found the writing a little stodgier than I like to think mine is, with a bit more jargon thrown in: “in a world of algorithmic content and misinformation,” CG wrote, “They (libraries) offer something more radical: context, curation and conversation.” I don’t really know what that means.
· And then there was the “hallucination” problem:
Since CG had observed I like to include personal anecdotes, it decided to make one up for me.
I asked deepai.org to make up a picture of the boy I never knew. Look how happy he is!
“I once knew a boy who walked to the library each day after school. Not because he loved to read – though he did. He walked there because the room was warm, and quiet and free.” CG (writing as me) went on from there, doubling down on the fake anecdote: “That boy wasn’t me. But I’ve met him in every county in North Carolina.” (note there are 100 counties in North Carolina; that’s a lot of boys for me to be meeting, in a lot of libraries). And at the conclusion of the piece CG brings the boy back again, allowing him to plead for keeping libraries alive so that that boy can “sit and read in silence, or ask for help without shame, or just be.” I might go to bat for that kid — if I knew him.
Those quibbles aside, I left the exercise impressed. CG came up with an idea I hadn’t thought of, found relevant research on it and put together a strong, structured argument. Is there a world where I could trust CG to write a post for me in the future? I think there is, and it is probably not far away.
But would I want to trust CG to write a post for me? I don’t think so. I kinda like the idea of CG serving as an occasional muse when I have writer’s block, or as an editor when I go on too long. But the act of writing is how I make sense of my particular conception of the world. I listen and I read and I bounce ideas off other people, and then I sit down (typically in the library) and try to make sense of them. Once I assemble the thoughts into some kind of order, they form my perspective; sometimes, the thinking that goes into writing shakes the snowglobe of my belief system.
What happens when a generation of students or grownups skips that step and allows CG and his buddies to do that synthesis for them? It’s so tempting to take short cuts, to avoid the hard work of making sense of the world. But it’s one of those things that makes us unique and fun and confounding and interesting. I worry what happens when we outsource our personalities and world views to CG and his AI brethren and sisteren.
So thanks for the idea on libraries, CG. Keep ‘em coming. But I want to do the writing.
Notes:
Meghan O’Rourke on Chat GPT: https://www.nytimes.com/2025/07/18/opinion/ai-chatgpt-school.html
Vauhini Vara writes on the challenges of “selfhood in the digital age”: https://www.thriftbooks.com/w/searches-essays_vauhini-vara/53367225/item/64940461/?utm_source=google&utm_medium=cpc&utm_campaign=us_shopping_zombies_lvs_21821394433&utm_adgroup=&utm_term=&utm_content=717525408665&gad_source=1&gad_campaignid=21821394433&gbraid=0AAAAADwY45hhicyfsSTj_n2Qc5gfWNhBf&gclid=CjwKCAjwv5zEBhBwEiwAOg2YKHkc8v7RNngZ0foY_IWAcxeAWwbpViP0Q0oOBYPAeirizW9Q7qSHexoCBegQAvD_BwE#idiq=64940461&edition=71609246