Who Do You Think With?
How AI is changing the way we think — before we notice
Lately I’ve been noticing something strange: the more we study intelligent machines, the more they appear to resemble something disturbingly familiar: us.
In last week‘s letter, I referred to a long New Yorker article about Claude1, and what struck me most wasn’t just that Anthropic doesn’t fully understand what it’s built – it’s the conclusion the piece seemed to reach: that the more we study these systems, the more we discover that we’re really studying ourselves. P and I have noticed it in our own speech. We’ll say, “I was speaking with the bots...” or “I asked the bots...” without a second thought. And I couldn’t help but ask – are people forming emotional habits around tools like ChatGPT and Claude? Not romantic attachments or even emotional support, but something stranger and harder to name: cognitive companionship.2
There's a line in Joan Didion's 2005 memoir The Year of Magical Thinking where she describes her intellectual partnership with her husband, John Gregory Dunne: “I did not always think he was right nor did he always think I was right but we were each the person the other trusted”.
We have that relationship. But I couldn't help wondering – if we didn't have each other, would I too be running all my thoughts and ideas by a machine?