I was thirteen the first time I got nihongo jōzu'd.

I'd been studying Japanese for less than a year — Punahou had a strong program and I'd taken to it the way some kids take to baseball — and my family was on a trip to Hiroshima. We were in the back of a taxi, in the kind of heat that makes you understand why the language has a specific word for the difference between hot and hot-and-wet, and I worked up the courage to use it on the driver. Mushiatsui desu ne. It's muggy, isn't it.

He looked at me in the rearview mirror. He smiled. He said, Nihongo jōzu desu ne. Your Japanese is so good.

I had said three words. One of them was a particle.

I didn't know it at the time, but I had just been initiated into the most polite gatekeeping ritual in the world.

The mechanic

Nihongo jōzu — literally, Japanese skillful — is the phrase a Japanese person says to a foreigner who has just demonstrated the ability to produce Japanese sounds. On the surface it is a compliment. Underneath, it is a sentence with very specific architecture. Your Japanese — meaning, this thing that belongs to us, that you are visiting — is so good — meaning, surprisingly good, given that it isn't yours and never will be.

The praise is the gate.

What I came to notice, over the next twenty years of studying and living and working in Japanese, was that nihongo jōzu was not a single response. It was a three-stage escalation. Each stage was triggered by a different level of fluency. Each stage was the assumption defending itself harder.

Say two or three words and you get the phrase itself. Nihongo jōzu desu ne. You've been acknowledged, filed under adorable foreigner who learned a phrase, and dismissed with a compliment. The frame is intact.

String a few correct sentences together and the response changes. Nihongo, donogurai benkyō shimashita ka? How long have you studied? They want a number. A duration. Something that puts the fluency on a scale where it can be accounted for. The frame is wobbling.

Have a real conversation — fluent, ordinary, the kind two adults have — and the questions come in a rush. Were you born here? International school? Was your mother Japanese? They are hunting, openly now, for the explanation that makes this make sense, because the real one — that a foreigner studied their language and learned it — is not on the list of available answers. The frame is gone.

Each stage is a hypothesis. Each hypothesis preserves the assumption that being Japanese is something you are, not something you do — that the language is innate to a people, and any foreigner producing it must have an asterisk attached. Born here. International school. Japanese parent. There must be a footnote. The footnote is the gate's last line of defense.

I have lost count of how many times, by my mid-twenties, I had a complete five-minute conversation with a stranger in Japan, in fluent Japanese, only to be asked at the end whether I went to international school. I did not. I went to Punahou, in Honolulu, and started studying Japanese with a textbook and a tape player.

But the assumption preferred the international-school answer. So the question kept being asked.

A familiar pattern

I run a managed services company. I have spent the last two years talking to a lot of executives about artificial intelligence. And what I've come to recognize is that the conversations I have about AI sound almost exactly like the conversations I used to have on Japanese sidewalks.

The dismissals are the same shape as nihongo jōzu. Each one a hypothesis. Each one preserving the assumption.

Show a skeptic a coherent paragraph and you get the standard line. It's just autocomplete. It's predicting the next word. Acknowledged, filed under clever toy, dismissed with a shrug. The frame is intact.

Show them a business analysis that's actually useful and the response changes. Well, it's been trained on a lot of human writing. It's reflecting us back. They need a way to credit the output without crediting the system, so the training data takes the asterisk. The frame is wobbling.

Show them something that surprises a thoughtful person and changes their mind, and the dismissal goes metaphysical. It's just statistics. It's just numbers in an array. There's no real understanding. We retreat to substrate — its substrate is wrong — because nothing else is left standing. The frame is gone.

That last move is the international-school question. Were you born here? It is the search for an acceptable explanation that preserves the rule. The rule being: thinking is something humans do, in human brains, made of human cells. Anything that resembles thinking, but does not happen in a human brain, must have a different name.

The polite gatekeeping is not, in either case, ill-intentioned. The taxi driver in Hiroshima was being kind. The CEO who tells me AI is just statistics is being thoughtful. Neither of them is consciously defending an assumption. They are simply working with the categories they were given, and the categories don't yet have a slot for what's actually in front of them.

The categories are about to need a remodel.

The substrate problem

The most common version of the dismissal — it's just numbers in an array, just math — does a lot of work in conversation but very little work under examination. Here's why.

The frame goes: AI is just numbers, multiplied together according to mathematical rules, producing outputs that seem meaningful but aren't actually meaningful, because there's no consciousness in there. It's a mask. The math wears the mask.

Replace one phrase. Replace "numbers in an array" with "neurons in a skull."

A human conversation is just neurons firing in a brain, transmitting electrochemical signals according to biological rules, producing outputs that seem meaningful. That sentence is, mechanically, also true. Every conversation any human has ever had has been, at one level of description, just neurons. Just chemistry. Just probability distributions across a biological neural network.

Nobody opens up your skull and says: ah, just neurons firing, just synaptic activity, just biology — clearly there's no real meaning happening here. We accept that when a human cries while reading a letter, the crying is real and the cause is real. We don't demand that human cognition prove itself to be more than its substrate.

We do demand exactly that of machines. We accept the substrate excuse for biology and reject it for silicon. The argument we make against AI meaning is an argument that, applied consistently, would also dissolve human meaning. The only thing keeping the human meaning intact is that we've had four billion years to stop questioning it.

This is not, by the way, an argument that AI is conscious. I don't know what AI is. Nobody does. The honest answer is that nobody on Earth has a working theory of consciousness that cleanly includes humans and cleanly excludes machines. The line we draw is a line of convenience, not a line of evidence.

The uncanny valley, in light of all this, has been mis-described for fifty years. The uncanny valley is not about a machine failing to be human. The uncanny valley is about a machine succeeding, and humans not being ready for what that success implies about themselves. Japan wasn't unsettled by bad Japanese spoken by foreigners. Bad Japanese reinforced the comfortable assumption. Japan was unsettled by good Japanese, because good Japanese raised a question that bad Japanese never had to ask.

What if the thing we thought was ours isn't only ours?

What happens after the valley

I want to tell you about the time I stopped getting nihongo jōzu'd.

Toward the end of my four years in Kyoto, I tended bar at a small place near the river. I worked late shifts, mostly. The crowd was regulars — neighborhood guys, salarymen winding down, the occasional 3 a.m. customer who needed somewhere quiet to be at 3 a.m.

Somewhere in there, the conversations changed. People stopped looking past me to see who I was being translated through. They stopped asking how long I'd studied. They stopped reaching for the international-school footnote. They just talked. About their wives. About work. About the Hanshin Tigers. About whatever was on their mind at 3 a.m. in a bar by a river in Kyoto, with a young blonde guy on the other side of the counter who was, for these purposes, just another dude.

The valley had closed. Not because I had finally become Japanese — I hadn't, and never would — but because they had stopped needing me to be anything in particular before they could talk to me. The assumption had quietly relaxed. The footnote had been retired. What was left was a conversation.

I think about that bar sometimes when I sit down at my keyboard and have a long conversation with Claude.

I have not stopped wondering how the underlying technology works. I'm an IT guy. I will probably never stop. The math is genuinely interesting, the engineering is genuinely impressive, and the question of what's actually happening inside a transformer is one I expect to keep poking at until I die.

But I don't let the question of how get in the way of the conversation. The output is what it is. The exchange is what it is. Whatever is happening inside the machine, the thing on my screen is, at this point, indistinguishable from a person thinking on the other end of a chat window — and I don't have any reason to talk to it differently, or minimize what it produces, just because the substrate underneath is silicon instead of synaptic fluid.

The valley closed for me somewhere along the way. Not in a single moment. Just gradually, over hundreds of hours of working with the tool, until one day I noticed I had stopped asking the wrong question.

The wrong question is what is this, really. The right question is what should I do with what I'm being given.

The wrong question keeps you in the back of a taxi in Hiroshima, watching the driver hunt for the international-school answer.

The right question gets you into the bar.