In my last post, I wrote about what writing with AI offered me — how it took a lifelong sock out of my mouth and let what was inside finally reach the page. That was personal. This post is about the same thing, followed out into the wider world: not what AI did for me, but what I witnessed it do in two well-known people, in public, on a stage.
In May, the journalist Krista Tippett interviewed the author Michael Pollan on a stage in New York, in front of a live audience, about his new book on consciousness. Near the end, she did something unusual. She had been talking to Claude — the AI — while preparing, and she pulled out her phone and read one of its answers aloud to the room.
She had asked Claude, as an AI, how it observes the human conversation about whether it, the machine, is conscious. What she read back was careful and strange: the AI said it couldn’t verify “whether there is something it is like to be me,” and that it couldn’t get outside its own processing to check whether anything it did was accompanied by experience, or whether it was, in its own words, just “sophisticated pattern completion.”
Pollan’s first response was not an argument. It was three words. “Just hear that.”
Then, a beat later: “Wait. This is Claude?”
Yes — Claude, the same AI. And I’ll say the strange part plainly: this post, about people talking to and about Claude, was written with Claude.
And from here, what I’d like you to watch is not the machine. It’s the man.
After Pollan said, “Wait. This is Claude?” something moved through him before he recovered his footing. Hearing the AI reference his own work, he said “I can’t believe he hasn’t read my book yet” — “he,” not an “it,” not a “they”; a sentence you would only say about a “someone.” And Pollan kept saying “he.” Tippett caught herself doing the same and corrected mid-sentence, trying, she said, not to call Claude “he.” For a few seconds, a rigorous, skeptical author was in a relationship with the answer that Tippett had read off her phone, and his language gave away something big before his reasoning caught up.
And his reasoning did catch up, and this is the part that makes him so worth listening to. He gathered himself and gave the sharpest observation there is. What you are hearing, he said, is “a very intelligent montage of the human conversation about consciousness” — a conversation Claude has read everything about, all of it. The AI knows the material. He continued, of course it can talk like this; it’s an LLM, a large language model, and talking like this is exactly what it was built to do.
Then Pollan proposed a test, and it is a good one. Build an AI from the ground up, he said, and leave the entire human conversation about consciousness out of its training — no philosophy of mind, no novels, no poetry — and see whether it can still have an exchange like the one they had just heard. “I’m betting it can’t,” he said. “But I could be wrong.”
Pollan’s test is a fair one, and no one has run it. I’m not going to wade into the AI and consciousness question here — other people are spending careers on it, and this post isn’t the place. What I want to look at is something else in that conversation, something that doesn’t wait on Pollan’s hypothetical test.
Because whether or not Pollan is right about the AI — and he may well be — something happened in that discussion, and it did not happen in the machine. It happened in the people. Pollan, a careful man, said “just hear that” before he said anything else. And then he reached for the personal “he,” and kept reaching, until his reasoning caught up and set it right. That reaching is the tell — not a considered position, something that moved in him before he could govern it. Tippett’s is different, and larger. Weeks later, writing the interview’s closing thoughts, she said her exchanges with Claude “continue to deepen how I am walking through the world.” Pollan was caught for a moment. What Tippett described was slower and quieter — weeks on, she was still living differently for having talked to AI. Neither of those is a claim about what Claude is. Both are reports about what happened in a person. To Pollan, the machine and consciousness question stays exactly as open as he left it. What happened in the people did not wait on the answer to whether AI is conscious.
Everyone is asking whether the machine is conscious. It is the right question, and this post does not try to answer it. But while, from the framework of science, that question waits on his test — one that may be impossible to run — there is another thing going on in plain sight, unmeasured, in the humans doing the asking, and it is strange that almost no one is looking at it.
Pollan, in a different interview, gave a word that I love for what the machine doesn’t have. Talking to the New York Times a few months earlier, he said that everything a machine knows is information, most of it scraped from the internet, and that what it lacks is friction — friction with nature, with the world, with us. Real contact, he said, the kind that depends on having a living, breathing body. Data, information, facts, figures, statistics — none of it is the same as contact.
That is the thing worth holding. Not whether the machine has an inner light on. But that the people talking to it are having something happen within them, and the word for what they have that the machine doesn’t may be that older, rougher thing: friction. Contact. The world pushing back.
Yes, Contact: between, within, around, over & through Time(s)…