Today's idea is about a curiously simple relationship between human and artificial intelligence. It's almost too simple, and captures something that might align with "first principles" thinking. A human mind begins as one, a singular experience. We enter the world from a particular point of view and gradually discover that other minds exist. We do this by reading books, learning from teachers, arguing with friends, and along the way, absorbing a culture. These "vital encounters" become part of us. And the mind does it so completely that years later we may not even remember where a self-defining idea came from.
I think we can capture this rather simply.
Human: ONE ⇒ MANY
Now artificial intelligence approaches this from the opposite direction. Large language models are trained on the "voices" of millions of people and shaped by this eclectic human expression. From that plurality, AI produces something very different from what went in. AI creates a sort of inversion to deliver a singular conversational voice that is constructed out of the many. And along the way, that individual is irreversibly metabolized into the whole. This is an important distinction.
AI: MANY ⇒ ONE
I'm calling this the Many-One Inversion.
And here's a powerful and well-established phrase that helps this idea along: e pluribus unum, out of many, one. We know it as a defining element of American history—our melting pot—where a union is formed and individuality isn't lost.
AI creates a fundamentally different kind of synthesis. The human expression that shaped the system doesn't reach us as a collection of identifiable voices. We receive an answer generated from patterns, not people. Simply put, we hear one without having encountered the many.
Of course, human thought has always involved synthesis. Discovering other minds is only part of the process. We encounter varied (and contradictory) ideas, struggle in that process, and eventually leverage that to construct ourselves.
Let's unpack this a bit more. The fuller path of human thought has traditionally looked something like this.
ONE ⇒ MANY ⇒ ONE
But when we begin thinking with AI, that path changes in an important way.
ONE ⇒ [MANY] ⇒ ONE
Using brackets helps call out a distinction. The middle hasn't disappeared; it's become less visible.
Let's take this step by step. A question starts with me. I bring it to AI, where it passes through a system shaped by an unimaginable scale of human expression. What comes back may contain ideas and connections that, in many instances, I may have never thought of or perhaps even imagined. But most importantly, they arrive being synthesized into a particular response. And that response has a sense of "cognitive finality" to it. It's impossible for me to reconstruct the lineage of each and every thought that helped shape the conclusion. I knew where the idea began. The middle had become harder to see. And that forces the question: Does the middle matter?
I don't think there's anything problematic about synthesis itself. We depend on it constantly. A good professor can compress years of experience into an hour, and an editor can bring clarity to a complex argument. So far, so good.
But AI changes the process. A question can enter an LLM and return seconds later as a polished answer. Its "glibness" tells us very little about the intellectual path it took. We can wrestle with that terrain ourselves, or accept its compression without encountering the terrain at all.
I think it's easy to miss because the results can look similar. Yet the cognitive paths are different.
ONE ⇒ MANY ⇒ ONE
ONE ⇒ [MANY] ⇒ ONE
"The many" have always shaped human thought, and synthesis has always been part of thinking. What changes is our relationship to the middle. Some of the encounter and integration that once took place within the thinker can now occur before an answer ever reaches us.
The answer no longer tells us what happened in the middle. And as AI makes it so easy to stop thinking, we need to recognize that very often, the middle matters.