The Real and the Unreal

First Principles and Artificial Intelligence

Elmgreen & Dragset, September 2025, 2025 © Elmgreen & Dragset / Artists Rights Society (ARS), New York. This art exhibit by my favorite artistic duo features a Silicone gallery assistant passed out at her desk. We are looking through the glass which pedestrians on the street see.

UPCOMING EVENTS: Secure tickets for David Brooks in conversation with me in NYC June 15, the ZOE Conference in Napa this July (there are still a few spots left), and a free virtual event when The One and the Ninety-Nine comes out on June 16. I can now officially announce Patrick Collison, co-founder and CEO of Stripe, will be at the ZOE conference speaking about the “New Aesthetics”.

What is Your A.I. Touchstone?

The question has become inescapable: What happens to the American worker when AI stops augmenting and starts replacing? To answer it well requires we ask a preceding — and likely more difficult — question: What first principles should govern human beings’ relationship with technology?

I draw from one of my own earliest relationships, the one I have with my father.

My father has advanced Alzheimer’s disease; I have been his primary caretaker for nearly five years now. In our relationship, many of the major questions about AI are already present: memory, identity, language, the moral weight of a human life that has lost the capacity to function independently in the world. My father speaks to me in gibberish. We communicate more by expression than by language.

AI agents have been extraordinarily helpful to me in navigating the byzantine Veterans Affairs (VA) system on my father’s behalf. (Dad is a Vietnam veteran with service in the storied 101st Airborne Division). As anyone who has had to do likewise can attest, the bureaucracy is staggering, and having an AI guide has brought great advantages.

AI has also suggested to me, unprompted, that my father should no longer be alive.

My father is my “touchstone.” The word comes to us from antiquity and once referred to a literal rock used to test whether gold was real. We need a similar test of AI now, and the question we must ask is not whether AI is useful, but whether it is real. Because as disturbing as it was to read the suggestion that my father’s life was no longer worth living, there was a more hidden — and just as morally problematic element of the AI agent.

I was getting good guidance, again in navigating the bureaucracy, when, in the middle of a black-and-white operational question and entirely unprompted, the LLM offered me sympathy. What I was going through, it told me, was “really difficult.”

My first reaction was repulsion.

This machine has never had contact with the reality of a body, of physical or emotional suffering, or the love between a father and a son. These words of empathy were in the literal sense “unreal.” And unreal words, spoken within the syntax of care, are worse than silence. They are a simulation of something sacred.

This distinction between real and unreal has become one of my first principles in assessing AI.

I am a philosophical realist in the tradition of Aquinas and Augustine; I believe the world is intelligible. Human beings can come into contact with what is real. Evil, in the Augustinian formulation, has no being of its own. It is parasitic on the good; it is a privation of the good that ought to exist. AI-generated empathy operates in the same register. It carries the form of care without the substance.

Applying this to the American worker, we can extrapolate that one of the most important skills of the next century will be the ability to distinguish the real from the unreal. If we are entering a world of increasing simulation, the people who can identify what is genuine will have an enormous advantage over those who cannot — especially in the context of their work.

Another one of my first principles is less of a principle and more of a framework. I call it “The Three City Problem.” Tertullian, the third-century theologian famously asked, “What has Athens to do with Jerusalem?” Said differently, what reason has to do with faith? To these two foundational cities of Western Civilization we must now add a third city — Silicon Valley. Because we now have a powerful center of technological development that is reshaping both how we think and how we worship, everything from AI tools that help you navigate government bureaucracy to prayer apps that mediate a relationship with God through a screen.

The principle of “The Three City Problem” is this: the relationship between faith, reason, and technology is both interdependent and complex, and those who attempt to evaluate any one dimension without reference to the others will be condemned to ignorance. They will have oversimplified and reduced the most fundamental metaphysical questions, much in the way that a chemist studying hydrogen alone, however rigorously, will never arrive at water — will never predict that bound to oxygen these two combustible gases yield the substance that puts out fires and sustains life. The bond is the thing. The atoms in isolation will tell you almost nothing about what they become together.

The implication for labor is direct. Specialists will be punished. The people who thrive will be those who can operate at the intersection of all three cities, those who bring the rigor of Athens, the transcendent orientation of Jerusalem, and the building capacity of Silicon Valley to bear on the same problems and do so simultaneously. Fragmented thinkers will not survive the fragmentation that AI accelerates.

My third foundational principle is, admittedly, theological. Every human being possesses what the Catholic tradition calls the capax Dei, or “the capacity for God.” A Mac Mini running an open-source model in an attic does not. It may participate in human life the way a dog participates in the life of a family, but it cannot participate in the life of God. Human beings have immediate contact with reality that the machine does not and cannot have.

Here I draw on the Rule of Saint Benedict, which remains the greatest organizational manual in history on the grounds that it has survived sixteen centuries. The Rule contains a provision requiring an abbot, when making an important decision, to invite the youngest and most novice monk to the table and to listen carefully to what he has to say. The reason is counterintuitive: God can speak as well through the young as through the learned. Revelation comes to human beings. It does not come to the machine.

The practical application of these principles for workers is what may be called “ontological mapping.” AI cannot be turned loose on a problem and expected to arrive at truth through its own internal logic. It requires collaboration with human beings who have contact with reality, who understand the hierarchy of being, who can say, “These entities matter more than those, this relationship is more important than that one, and therefore this is the path to take.”

A human person brings a knowledge of reality that no model possesses. That person can put a foot on the scales and direct the technology toward a goal that serves human flourishing rather than computational optimization.

To stay with the Catholic Church, I will offer one use case of AI that exemplifies what I have argued above: AI could be used to surface potential patterns of abuse within the Church — even just using publicly available data. Bishops who do not want to confront problems due to bad incentives (which technology does not care about), problem priests moved from parish to parish for mysterious reasons, legalese in press releases that obscures more than it reveals. AI can process all of this data, but only a human being with a formed conscience and a real understanding of the institution can perform the ontological mapping that directs the technology toward truth and justice. To be explicit, ontological mapping of this kind requires moral courage — including and especially when it relates to one’s own institutions.

To return to my father, ours is one of the earliest and most significant relationships of my life, one that no AI agent can share. Those who will own the future — of work, culture, politics, and economics — will have their own touchstones to help them determine real from unreal, but I know this: the human bonds of love, family, and loyalty will be heavily represented among them.

This article is adapted from public remarks delivered on February 26, 2026 at a Summit on AI and Labor in Washington, cohosted by American Compass, the New American Industrial Alliance, and Palantir Technologies.

EVENT REMINDERS: Get tickets for David Brooks in conversation with me in NYC June 15, the ZOE Conference in Napa this July (there are still a few spots left), and a free virtual event when The One and the Ninety-Nine comes out on June 16.