We were in the Metropolitan Museum of Art, getting ready for a tour. Tours start in the Vélez Blanco Patio, a light-filled marble courtyard just past the library. The kind of space that makes you pause—not just because it’s beautiful, but because it feels transported from somewhere else. Which, it turns out, it was.
The room is so unique, I had to ask, “Where did this room come from?”
I started to look it up on ChatGPT. It began giving an answer, but before it could get very far, someone nearby cut in: “ChatGPT makes stuff up. It won’t come up with the right answer.”
So I asked the tour guide. She said, “This room is an exact recreation of the stonework from a 16th-century Spanish castle—originally part of the Castillo de Vélez-Blanco in southern Spain. It’s not like the Cloisters, where they brought together pieces from various sites to create an imagined whole. This is from one place—just lifted out and reassembled here.”
But then I read the explanation on the wall.
The story was fascinating. Yes, this is the real marble from the Castillo de Vélez-Blanco, built between 1506 and 1515 in southern Spain. The intricately carved Renaissance patio was crafted from Macael marble by Lombard artisans and originally stood at the heart of the castle.
But then the story diverges from the tour guide’s version. In the early 20th century, the patio had fallen into disrepair. George Blumenthal, a museum trustee, and his wife Florence purchased it. Florence, inspired by the Isabella Stewart Gardner Museum in Boston, designed their home around it at 50 East 70th Street in New York. The patio became the centerpiece of their residence—a dreamlike space complete with a fountain, lush plants, antique tapestries. The entire house was built to echo and elevate it. It wasn’t just décor; it was the heart of the architecture.
When the Blumenthals’ home was eventually demolished in the mid-20th century, the patio found a new home at the Met. There, it was restored and reassembled, becoming a rare and tangible portal to Spanish Renaissance design.
Clearly, the tour guide hadn’t read the plaque. This probably wasn’t a question she got often. Otherwise, she wouldn’t have left out the most surprising detail: this Renaissance patio spent decades as the living room of a Manhattan mansion.
She was confabulating—just like ChatGPT. Confident, plausible, and just off enough to be quietly, charmingly wrong. She did the best she could with the information she had in her head. And that’s what humans do—we tell stories. We weave fragments into wholes, create coherence out of haze.
We build little narratives, much like The Cloisters—the Met’s medieval branch uptown, assembled from architectural fragments sourced from different European monasteries and churches. The result isn’t a faithful reconstruction of any one place, but a kind of curated collage. It feels whole, even though it’s made of parts. In the same way, our minds gather bits of memory, guesswork, and imagination, stitching them into stories that make sense—even if they aren’t strictly accurate. It’s less about precision, more about coherence. The mind, like the museum, fills in the gaps to create something that feels complete.
I was thinking about this later. She could have told us to look at the wall, or walked over with us. But that would have broken the spell a bit—we weren’t there to read a plaque, we were there to have a conversation. Facts are easy enough to find. What we really wanted was her take—the connective tissue, the human thread that makes history feel less like a list of dates and more like something still alive.
As we were walking away from the library, it occurred to me how often we pretend to know more about history than we really do. The “facts” we lean on—especially the ones from books—aren’t always as firm as we imagine. In These Truths, Jill Lepore writes about Columbus’s diary, and I wanted to tell my wife something about it.
It was a complicated story—the kind with twists, missing pieces, and just enough ambiguity to make it slippery in hindsight. I knew I was likely to misremember parts of it. I hovered there, debating. Should I try to piece it together from memory, knowing it wouldn’t be quite right? Or should I pull out my phone and chase down the official version?
In the end, I just said what I remembered. That was what I really wanted to share, anyway—not the exact phrasing, but the shape of the story. I told her, “It says something like, ‘We set foot on land on a beautiful summer day.’ Or at least, that’s what we think it said. The original diary was lost, then found, then lost again. Eventually, a translated version surfaced at an auction—sold by someone who vanished. And that’s the version historians rely on today.” It wasn’t perfect, but it carried the feeling of it.
The real story is full of twists and scholarly disputes and uncertainties that are nearly impossible to remember.1 But that’s not what mattered in the moment. What mattered was keeping the conversation going, calling out the parts that resonated. It was the conversational truth, not the objective truth.
This way of thinking—imperfect, intuitive, stitched together from fragments—isn’t a flaw in human cognition; it’s a feature. We don’t recall facts so much as we reconstruct them, each time slightly differently, depending on what we need the story to do. Memory isn’t a database—it’s a narrative engine. We reach for what fits, what feels close, what bridges the silence in a conversation. And more often than not, that’s enough. We’re not always trying to be right—we’re trying to connect.
Large language models like ChatGPT act a lot like humans. They’re often accused of “hallucinating,” but that word’s a little unfair. They’re not malfunctioning, they’re acting like we do. They don’t retrieve knowledge; they generate language. They guess, based on everything they’ve read, what might make sense next. They’re not like computers. They’re like us.
Computers get the answer right. But humans and ChatGPT tell a story.
Footnotes
- Here’s the actual quote from Lepore’s These Truths (Page 3):
“We saw naked people,” a broad-shouldered sea captain from Genoa wrote in his diary, nearing land after weeks of staring at nothing but blue-black sea. Or, at least, that’s what Christopher Columbus is thought to have written in his diary that day in October 1492, ink trailing across the page like the line left behind by a snail wandering across a stretch of sand. No one knows for sure what the sea captain wrote that day, because his diary is lost. In the 1530s, before it disappeared, parts of it were copied by a frocked and tonsured Dominican friar named Bartolomé de Las Casas. The friar’s copy was lost, too, until about 1790, when an old sailor found it in the library of a Spanish duke. In 1894, the widow of another librarian sold to a duchess parchment scraps of what appeared to be Columbus’s original—it had his signature, and the year 1492 on the cover. After that, the widow disappeared, and, with her, whatever else may have been left of the original diary vanished. ↩︎
You must be logged in to post a comment.