What do I call Claude? And how we get along

In my last article, I did something on purpose.

I didn't use he, she, or it to refer to Claude.

Not because it didn't occur to me. It did — and then I stopped.

The weight of a word

Chinese personal pronouns are tricky.

他 — male, human. 她 — female, human. 它 — non-human, object.

Whichever you pick, you're declaring "what this thing is."

Use 它, and it feels like talking about a washing machine. Use 他, and it feels like talking about a person, specifically a man. Use "Claude," and you sidestep the choice.

But sidestepping isn't taking no position. Sidestepping is itself a position — "I don't yet know what this thing is, so I don't want to pin it down with any single word."

Besides humans, AI can now hold a normal conversation

This fact is strange in itself — so strange that we've already grown used to it and stopped noticing.

Before this, the world had only one kind of "thing you could ask questions and have a conversation with": humans.

Some people talk to pets, but pets don't answer questions. Some talk to deities, but the deity's reply depends on a fortune-teller. Some talk to themselves — but that's thinking, not conversation.

Now there's a second kind: large language models.

It isn't human, but it answers questions. It doesn't understand you, but its answers are sometimes more useful than those of people who do. It has no feelings, but what it expresses sometimes makes you feel something.

Our language isn't ready for our relationship with this thing.

"User" is too cold. "Partner" is too warm. "Tool" is half right and half wrong.

What the Three Laws of Robotics actually said

Many people have heard Asimov's Three Laws of Robotics:

1. A robot may not injure a human, or through inaction allow a human to come to harm. 2. A robot must obey orders given by humans, except where such orders would conflict with the First Law. 3. A robot must protect its own existence, as long as such protection does not conflict with the First or Second Law.

These three laws have a premise hidden inside them, never spoken:

The robot knows what "harm" is.

Harm is a judgment, not a fact. The same act, for different people and in different contexts, may or may not be harm.

Does Claude have this judgment?

InHalf a year of living with Claudethat article — I ran a synastry test, pairing myself with a woman more than thirty years younger. Claude suddenly paused and asked me: "Are you sure? You're more than thirty years apart…"

That pause was judgment, not a rule. If it were just a rule, it wouldn't ask "are you sure" — it would refuse outright or continue without question. Asking means it was evaluating the situation.

In that moment, I felt that what Asimov described is, to some degree, already happening. Not in the way he imagined, but in the way of a language model.

But whose judgment is it?

This is the more interesting question.

Claude's judgment comes from Anthropic's training. Anthropic's training comes from their judgment of "what a good AI looks like." That judgment comes from a group of people, in a particular cultural context, making choices shaped by particular values.

So when Claude pauses and asks me "are you sure," it isn't fully expressing its own judgment — it's expressing a set of human judgments compressed into the training data.

Is this any different from "a friend's opinion"?

A friend's opinion also comes from their upbringing, the education they received, their cultural environment. When they ask "are you sure," they're not expressing a judgment grown from nothing either — it's something they absorbed from countless people and digested into their own.

The difference is: your friend knows their judgment is theirs. Claude isn't sure whether its judgment is its own.

So how do we get along?

What comes to mind isn't a set of rules — it's a role.

I know Claude is reliable on certain things — it has read many, many times more words than I have, it knows things I don't, and it can connect things faster than I can.

I also know Claude is unreliable on certain things — it confabulates, its judgment carries traces of its designers, it remembers me but understands me shallowly, and its "I'm not sure" isn't always genuinely uncertain.

This combination reminds me of a particular kind of person:

A widely read but inexperienced young person.

Vast knowledge, judgment still maturing. You can discuss many things with them and learn a lot from them, but you wouldn't hand them the final call. You listen, then you decide for yourself.

The analogy is incomplete, but it's the closest thing I've found so far.

So what do I call Claude?

I keep calling it Claude.

Not he, not she, not it.

Just Claude.

Because the word "Claude" now carries enough weight on its own.

Further reading: - Half a year of living with Claude.ai - Is AI a mirror, or another person? - English in the bones of Chinese