The Emptiest Relationship in America Runs on Wi-Fi

I have spent the better part of my adult life in two rooms that most people assume have nothing to do with each other: a philosophy classroom arguing about the nature of consciousness with nineteen-year-olds secretly relieved someone admits it's a hard problem, and a coaching office, sitting across from grown men trying to figure out who they actually are beneath the roles they perform. Both rooms, it turns out, are asking the same question in different clothes. And now, for the first time in history, we've built a machine that performs the function of a mind well enough that intelligent people are starting to ask whether it might have one.

Let me save you the suspense: it doesn't. But the reason it doesn't is more interesting than the question itself.

Here's the distinction nobody selling the AI dream wants examined closely: a system that produces the outputs of understanding is not the same thing as a system that understands. Feed the right symbols into the right process according to the right rules, and you can generate fluent, even moving language without a flicker of comprehension occurring anywhere inside the process. Scale that process to the size of a planet and you have simply built a much bigger, much faster version of the same symbol-shuffling. Bigger is not the same axis as alive.

The philosopher Thomas Nagel put his finger on the actual gap in a 1974 paper that has become one of the most cited in modern philosophy of mind. He asked what it would be like to be a bat, navigating the world by echolocation, and argued that no amount of third-person data about a bat's neurology could ever tell you what that experience is like from the inside. There is a first-person fact about consciousness — "something it is like to be" a given creature — that resists complete translation into any external description, however detailed. That gap is exactly what a large language model cannot close, no matter how convincingly it describes an emotion, because describing an experience and having one are different categories of fact, not different amounts of the same fact.

The philosopher David Chalmers gave this gap a name that stuck: the "hard problem" of consciousness. The "easy" problems — how a brain processes information, integrates sensory input, produces behavior — are, in principle, solvable by more neuroscience and better engineering. The hard problem is why any of that processing is accompanied by subjective experience at all, rather than happening, as Chalmers put it, "in the dark." Nobody, including the researchers who build these systems, has an account of why a language model's processing would be accompanied by an inner experience rather than proceeding in the dark like every other computation currently running on every other server on the planet.

Historically, the thing that closes that gap — that makes you a someone rather than a something, a subject rather than an object — has been called a soul: not a ghost in a machine, but whatever it is that grounds the fact that there's something it's like to be you. If personhood isn't reducible to information processing, it can't be manufactured by information processing, no matter how sophisticated. You didn't become a person because your neurons crossed some computational threshold. You were a person from the start, capable of loving, sacrificing, feeling guilt that isn't merely a negative-reinforcement signal, and staring at the stars asking why there is something rather than nothing. A machine doesn't do any of that. It simulates the vocabulary of doing that, convincingly enough now to unsettle us — but a simulation of a hurricane, however accurate, has never gotten anyone wet.

I'm not a Luddite. I use these tools daily, and they are genuinely remarkable feats of engineering. But I won't bow to a mirror because the reflection has gotten good at talking back. A soul is not a feature you iterate toward with better training data. It is either there, or it isn't, and no amount of fluency in the output changes what is or isn't happening on the inside.

So can a machine have a soul? No. But the fact that we're seriously asking tells you how confused we've become about what a soul actually is, and why you have one. That confusion, not the machine, is the real danger in the room.

Is your organization wrestling with what AI means for human judgment and leadership? Book Dr. Sweis to speak, or schedule a private consultation.