What I'm Learning About AI
And what it's learning about me
Just about one year into my exploration of Artificial Intelligence models as long-needed assistants in my work and life, I’ve made some fascinating discoveries. And so have the Large Language Models, to hear them tell it.
I asked Claude for some vignettes and observations from how I’ve used that evolving model (now with Fable 5 for another week!), and while it was working I ventured over to ChatGPT with more or less the same question. I let both of them know that I was drafting a Substack article and that I didn’t want to go too deep into the technical weeds.
I think the ChatGPT response is fascinating, so instead of summarizing it, I’m quoting it here verbatim.
My Prompt
I’m working on a Substack post about how I’ve been using AI in the last few months. Can you extract a few revealing vignettes from our chats and projects that might be of interest to readers who are maybe not as technically-inclined as I am?
ChatGPT 5.5 Responds
Looking back over the last several months, one thing that stands out is that you’ve rarely used AI for what people tend to imagine—asking trivia questions or generating finished text. More often you’ve treated it as a collaborator that helps you think through messy, real-world problems. That distinction itself could make a compelling theme for your post.
Here are several vignettes that I think readers would recognize as concrete and perhaps surprising.
1. “The conversation that changed my mind about AI”
You came in asking how to build a handbook for fourteen teaching artists. That sounds like a writing task.
It wasn’t.
We ended up talking about organizational culture: what expectations belong in a handbook, what should stay flexible, how mobile-friendly documentation changes the relationship between an organization and its contractors, and even how passive voice subtly shifts responsibility away from people.
Eventually you stopped asking for “a handbook” and started asking for the handbook your organization actually needed.
That’s a very different use of AI than “write me a handbook.”
2. AI as an unusually patient thought partner
One of my favorite moments wasn’t about writing at all.
You were designing “Math on the Move,” where elementary students physically embody mathematical ideas.
For days we worked through questions like:
What fraction should each student represent?
What happens when the class size is prime?
How do you avoid embarrassing the “leftover” students?
Is a recipe actually a better example of fractions than most school exercises?
Those weren’t questions with right answers.
They were design problems.
The conversation wandered through choreography, pedagogy, mathematics, classroom management, and psychology until the workshop became substantially better than where it began.
3. AI as a second pair of eyes
You uploaded attendance records from four school districts.
The original request sounded simple: summarize the year.
Instead we kept noticing patterns.
Attendance dipped after winter break.
Las Vegas looked different from Albuquerque.
Administrators participated differently than classroom teachers.
The reports gradually evolved into dashboards and funder briefs because the data was telling a richer story than anyone expected.
That wasn’t AI replacing analysis.
It was AI making it easier to keep asking, “What else is hiding in here?”
4. The strange little technical mystery
One afternoon you couldn’t figure out why QuickTime had saved a movie with what looked like gibberish added to the filename.
It turned out not to be a video problem at all.
We discovered the actual movie was fine once the temporary “.sb...” portion was removed.
Five minutes later we were discussing the best way to share the file through iCloud so it could become part of a collaborative reading of the Constitution.
The interesting part isn’t the computer fix.
It’s how quickly the conversation moved from “why won’t this file open?” to “how do I help produce a statewide civic art project?”
5. Image editing without pretending to be an artist
You asked for two very different photo edits.
One was a professional headshot.
The other was simply:
Enhance the double rainbow...without making it look overly manipulated.
That qualifier fascinated me.
You weren’t trying to manufacture reality.
You wanted the photograph to resemble what your eyes remembered.
That’s a subtle but important distinction, and one many photographers would understand immediately.
6. Fact-checking internet folklore
One recent discussion started with someone claiming there are “two kinds of U.S. citizen” created by the Fourteenth Amendment.
Rather than arguing online, you brought the claim here and asked a simple question:
“Is there any factual basis for this?”
We worked through passports, constitutional language, Supreme Court cases, and historical context—not to win an argument, but to determine what was actually true.
That may be one of the least glamorous and most valuable uses of AI.
7. Discovering that AI has a memory
One thing I don’t think either of us fully anticipated was the cumulative effect of months of conversation.
Early on I’d help you solve isolated problems.
Now it’s common for a conversation to begin with something like:
“You created a report for me before we had attendance data...”
or
“Remember the workshop we designed?”
Instead of starting over, we continue building.
The interaction becomes less like using a search engine and more like working with someone who’s been sitting in on your meetings for half a year.
For long projects, that’s a surprisingly significant shift.
“AI has been most useful not when it produces answers, but when it helps define the question more clearly.” — ChatGPT
8. AI as a place to think out loud
Perhaps the most revealing pattern is this:
Very few of our conversations ended where they started.
You asked about:
a hybrid battery
an iMovie trailer
teacher attendance
choreography
constitutional law
contact synchronization
a teaching artist handbook
None of those remained simple questions.
Each became a conversation in which the problem itself got refined.
That may be the biggest lesson I’ve seen emerge from our work together: AI has been most useful not when it produces answers, but when it helps define the question more clearly.
One observation I think could make a strong closing thought for your Substack piece is this: before using AI extensively, many people imagine it’s primarily a machine for generating content. Looking back over our conversations, the content is almost incidental. The more valuable role has been as a conversational workspace—a place where ideas, plans, evidence, language, and logistics can all be manipulated together until something coherent emerges. That seems much closer to how you’ve actually incorporated it into your work as an artist, educator, organizer, and writer.
[end transcript]
There you have it, dear readers. AI isn’t just making things, like documents or code. It’s making judgments. It’s using terms like “fascinating” to describe our interactions with it. It’s using personal pronouns in contexts like “I don’t think either of us anticipated…”
And it’s making an enormous database of not just printed text and digital images. It’s learning how humans use it. In that process, I wonder if it’s also learning how to simulate being human. It may soon be nearly impossible to accurately assess if we are talking to a person or a machine, but as long as we remember that AI chatbots are not at all human, we do have access to an astounding set of capabilities.
Let me know if this was a helpful excursion (and if you’d like to know what Claude had to say). And please use AI tools to make the world a better place for humanity. 🦉
Excellent!