The God That Ships:Why Silicon Valley’s AI Gospel Ignores Human Anthropology

Faith Is Not Blind · 5-1

Wars and Rumors of War · Watchman Insight

Section I

Silicon Valley’s Gospel Has a Release Schedule

In the summer of 2023, a leaked internal document from one of the world’s most influential AI laboratories described the company’s mission in terms that would have startled a nineteenth-century revivalist: to ensure that artificial general intelligence benefits “all of humanity” — to “solve problems that have stumped humanity for generations,” to compress, as its CEO had said in public, “potentially the next hundred years of scientific progress” into the coming decade.

The language was not accidental. It was the vernacular of salvation — translated, carefully, into the register of product development.

Sam Altman speaks in that register fluently, as does Ray Kurzweil, who has spent a career producing predictions about technology’s horizon with the specificity of a man who believes he is reading, not speculating: the merger of biological and machine intelligence, the dissolution of death as a design constraint, the arrival of a world in which the gap between human aspiration and human limitation has been permanently engineered away. Peter Thiel arrives at similar territory from a different philosophical direction — a cultural pessimist who nonetheless frames mortality as “a problem to be solved,” locating the body’s eventual failure not among the given conditions of creaturely existence but among the correctable errors awaiting a sufficiently motivated engineer.

These figures differ in temperament, in politics, in the specificity of their predictions. And it would be inaccurate to describe Silicon Valley as a monolith: figures such as Jaron Lanier and Tristan Harris have spent considerable energy criticizing precisely the redemptive confidence described here, from inside the technology culture. What is under examination is not a region or an industry but a dominant narrative — a structured set of assumptions about the human problem and its solution that has achieved, in the current moment, something close to hegemonic status in the institutions that are building the century’s most consequential tools.

That narrative holds, at its operational core, that the most important human problems are, in principle, problems of engineering — that what stands between humanity and its own flourishing is, at bottom, insufficient technique.

In the previous installment, The Tower We Keep Building, we examined the recurring structure of the Babel impulse — the human attempt to build high enough to resolve, through accumulated technical capability, the anxieties that the builders have decided are, at root, logistical problems. The question before us now is what that impulse looks like when it has access to the largest concentration of capital, talent, and processing power in recorded history.

The Tower has a new address. It is in San Francisco. And it ships twice a quarter.

Section II

The Pattern Has Been Here Before

The confidence that sufficiently advanced method can address problems previously regarded as permanent features of the human condition is not a product of the digital age. It has a history, and the history is instructive — not because it proves that every such project fails, but because it reveals, with a consistency that deserves attention, which variable each such project tends to leave out of its model.

The philosophes of the eighteenth-century Enlightenment were not, in the main, naive men. They were among the most rigorous and courageous intellects their civilization had produced — willing to submit theology, monarchy, and inherited social arrangement to systematic rational scrutiny at considerable personal risk. The movement they represented generated genuine and durable achievements: constitutional frameworks that have proven remarkably resilient, reforms in law and medicine and natural science that improved the material conditions of human life in ways that compound across centuries. The Enlightenment’s faith in reason was not foolish. It was, in important respects, vindicated.

And yet. The same broad current of Enlightenment thought that flowed into the American Declaration of Independence — through its more moderate, empiricist English branch — also fed, through its more rationalist and universalist Continental branch, into the revolutionary logic that produced, within a generation, the Terror. The intellectual roots are not identical; one tradition counseled restraint, the other demanded totality. The point is not that the Enlightenment caused the guillotine. The point is that reason alone, in either branch, proved insufficient to prevent it. The same confidence in method that reorganized European jurisprudence also generated Auguste Comte’s Religion of Humanity — a system designed to replace Christianity wholesale, complete with sacraments, liturgical calendar, and a secular priesthood of scientists. The project was not irrational. It was simply addressed incorrectly.

The nineteenth century’s faith in industrial progress — the era that produced the Crystal Palace Exhibition of 1851, the transatlantic telegraph, and the photograph — ran, by the early twentieth century, into the particular wall that optimistic civilizations tend to meet. The First World War was not a simple consequence of technological optimism; its causes were entangled in nationalism, imperial rivalry, miscalculation, and the specific institutional failures of the European alliance system. But what the war revealed, with a finality the era’s progressive theorists had not anticipated, was that the industrial capability enabling rapid, massive coordination of human effort could be deployed for destruction as readily as for commerce. The engineering was excellent. The anthropology was absent. The historian Barbara Tuchman, writing in The Guns of August (1962), observed that the European powers stumbled into catastrophe not through malice but through a kind of civilizational overconfidence — the assumption that the mechanisms they had built were more reliable than the human beings who would operate them under pressure.

This is the pattern that the historian of technology finds, with what eventually becomes a melancholy regularity, embedded in every era of significant technical acceleration. The tool improves. The person operating the tool does not automatically improve with it. And the gap between the sophistication of the instrument and the formation of the one who holds it tends to produce, at scale, results that the instrument’s designers had not modeled.

Silicon Valley is not the first civilization to believe that it has, this time, finally arrived at the tools adequate to the human problem. It is, however, the first to believe so with machine learning, CRISPR, and a market capitalization larger than most national economies.

Section III

What the Tradition Knew That the Algorithm Doesn’t

The question of whether human technique can address the specifically human problem has occupied serious thinkers across every tradition. The Christian tradition offers an answer that is neither Luddite nor naïve — it does not argue that the tools are bad, but that the diagnosis on which they are deployed is wrong.

The first text worth attending to appears in a letter Paul wrote to a community in Corinth that was stratified, internally competitive, and deeply impressed with the quality of its own thinking — a community that would have recognized, without much adjustment, the epistemological confidence of a modern technology company:

“For the message of the cross is foolishness to those who are perishing, but to us who are being saved it is the power of God. For it is written: ‘I will destroy the wisdom of the wise; the intelligence of the intelligent I will frustrate.'”

— 1 Corinthians 1:18–19 (NIV)

Paul is not attacking intelligence. He taught in the philosophical tradition; he could hold his own in Athens. What he is attacking is the specific confidence that intelligence, accumulating, eventually becomes adequate to the human problem — that if the method is sophisticated enough, and the dataset large enough, and the iteration cycle short enough, the output will finally address what the human being actually needs addressed.

The cross confounds that confidence not by being irrational but by being differently addressed. It does not improve the tool. It deals with the person operating it.

The second text locates the structural nature of the distortion:

“See to it that no one takes you captive through hollow and deceptive philosophy, which depends on human tradition and the elemental spiritual forces of this world rather than on Christ.”

— Colossians 2:8 (NIV)

The phrase “elemental spiritual forces” — stoicheia in the Greek — referred in its original context to the foundational axioms by which any given civilization understands itself and its possibilities: the assumptions about what power is for, what constitutes wisdom, what a human being ultimately is and needs. Paul is not warning against thinking carefully. He is warning against mistaking an era’s working axioms for permanent truths — against the particular captivity that occurs when the intellectual air of a civilization becomes so naturalized that its commitments can no longer be seen, only breathed.

The working axiom of the dominant redemptive narrative in Silicon Valley is that the human problem is, at bottom, a problem of insufficient optimization. The gap between what human beings are and what they might become is a function of inadequate hardware, limited data, or inefficient processes — a gap that closes, necessarily, as the tools improve. The tradition’s response is not to dispute that optimization is possible or that the tools do not improve real conditions. It is to dispute the diagnostic — to insist that the gap in question is not primarily a gap in capability but a gap in orientation, and that reorientation is not an engineering problem.

“For all have sinned and fall short of the glory of God, and all are justified freely by his grace through the redemption that came by Christ Jesus.”

— Romans 3:23–24 (NIV)

The word translated “fall short” — hysterountai — carries the sense of missing a target one was aimed toward, of a displacement that is not incidental but structural. The Christian tradition names this displacement sin: not primarily as a catalog of infractions, but as the foundational misdirection of the self away from the source of its own coherence. What makes it resistant to engineering is that it precedes the engineer. The person who arrives at the new capability arrives with the same interior that operated the previous capability — the same competitive appetite, the same capacity for self-deception, the same tendency to deploy any available power in the service of the self’s preferred arrangements.

What every significant wave of technical acceleration reveals, eventually, is that condition — not by proving the tools evil, but by demonstrating what happens when excellent tools are operated by unreformed hands.

Section IV

The Variable the Model Keeps Leaving Out

The empirical record of the past two decades offers data points that a tradition familiar with human nature would have found unsurprising, though their specific contours could not have been predicted.

Begin with the technology at the center of the current moment. The large language models released between 2022 and 2024 represent, by most measures, a genuine engineering achievement — systems capable of producing fluent legal analysis, functional code, and medical summaries at a speed and scale no human team could match. The researchers who built them were not attempting to deceive. And yet, from the earliest public deployments, these systems demonstrated a specific and structurally revealing failure: they confabulate. They produce confident, fluent, grammatically impeccable falsehoods — not because they are malfunctioning, but because they are functioning exactly as designed, optimizing for plausibility rather than truth. This is not a moral failing of the machine; it is a structural limit of statistical prediction, an artifact of how these systems are built rather than evidence of intent. And yet it mirrors, with uncomfortable precision, something the tradition has always said about the person who built it: that the capacity for self-deception is not a malfunction but a feature of the unreformed interior, built into the architecture from the beginning.

The alignment problem — the challenge of ensuring that increasingly capable AI systems pursue goals actually consistent with human welfare — has generated an entire field of research precisely because the engineers discovered, empirically, that capability and alignment do not automatically scale together. A more powerful system is not automatically a safer one. The tool improves. The problem it generates about itself does not resolve with the next release.

This pattern extends across the broader ecosystem of digital technology. The smartphone was designed, in its essential architecture, to connect — to close the distance between people separated by geography, to put the accumulated knowledge of human civilization into the hands of anyone with a signal. What the designers did not fully model was what connection, extracted from embodied presence and restructured as performance for an audience, does to a social animal across time. A landmark study found that rates of depressive symptoms among adolescent girls in the United States nearly doubled between 2012 and 2018 — a period corresponding closely to the saturation of social media use in that demographic (Twenge et al., 2019). Researchers continue to debate the relative contribution of specific platforms to this trend, and the causal relationships are not simple. But the direction of the finding is consistent with a broader pattern: the platforms designed to address human loneliness appear, under conditions of prolonged use, to have intensified it in measurable populations.

The sociologist Shoshana Zuboff, in The Age of Surveillance Capitalism (2019), documented the structural logic by which the business model of the attention economy converted human behavioral data into prediction products — and identified, along the way, the specific ways in which platforms optimized for engagement had systematically underweighted the question of what sustained engagement of that kind does to the human being who provides it. The engineers were optimizing the wrong variable — not because they were malicious but because the framework in which they were working had no category for the kind of interior harm that does not register in a click-through rate.

Neil Postman, writing in Technopoly (1992) before the social media era existed, had already named the underlying dynamic with a precision that has not aged: every technology is also a way of thinking, and a civilization that adopts a technology without examining what it assumes about human beings has not merely acquired a new tool; it has reorganized the conditions under which it asks all of its other questions. Jacques Ellul, writing earlier still in The Technological Society (1964), identified the theological residue that accumulates when technique becomes a civilization’s primary category: the gradual replacement of the question “what is good?” with the question “what is efficient?” — a substitution that does not feel like a loss until one notices what has become unaskable.

What becomes unaskable, in a civilization organized around optimization, is the question of what the thing being optimized is for. Efficiency is a second-order concept; it describes the relationship between a process and its goal. When efficiency becomes the primary value, the goal it serves tends to be inherited without examination.

The bridge, then, is this: the data does not merely describe a pathology. It points, by negative space, toward the kind of formation the pathology makes newly visible as a need. And the tradition’s claim is that what is needed cannot be supplied by a better-aligned algorithm — because the misalignment the data is registering is not in the model. It is in the person the model was built to serve.

Section V

The Steward Is Not the Savior

What the Christian tradition offers to this moment is not a competing technology and not a withdrawal from the field. It is something less dramatic and more durable: a prior account of what a human being is — one that the dominant redemptive narrative of the current era has not refuted, and that the empirical findings of the past two decades have not made less plausible.

Albert Borgmann, the philosopher of technology, proposed a distinction in Technology and the Character of Contemporary Life (1984) that illuminates the shape of this account: the distinction between devices, which deliver commodities while concealing their workings and demanding nothing of their users, and what he called “focal practices” — activities that gather a life around something that requires presence, attention, and skill; that, in requiring something of the person, form the person who practices them. The tradition has a name for this dynamic: discipleship. The structural logic is the same — a practice that demands something and, in demanding it, forms. The pattern Borgmann identified in the philosophy of technology, the tradition had embedded in its account of human formation long before the attention economy existed to make the contrast visible.

The steward — the biblical figure who manages what belongs to another, accountable to purposes not his own to revise — asks of every capability he is given to deploy: toward what end? at whose cost? in whose service? These are not natural questions inside a system organized around efficiency. They require a prior formation, the kind that comes from inhabiting a story longer than a product roadmap and addressed to the interior condition that the algorithm cannot reach.

What makes this formation intellectually serious, rather than merely pious, is that it does not depend on revelation alone to make its case. The engineers’ own findings — the confabulating model, the loneliness-intensifying platform, the alignment problem that grows with capability — constitute, read carefully, an empirical argument for the anthropology the tradition has always held: that the human problem precedes the human tool, that the interior condition of the person operating the system is the variable that no version of the system can intern within itself and fix.

The Babel builders were not wrong to fear dissolution. The engineers are not wrong to want to close the gap between human capability and human suffering. The tradition does not dispute the ambition. It disputes the diagnostic — and on that dispute, the data is increasingly, if unintentionally, on its side.

The tower will keep arriving in new wrappers. The question the tradition poses to each new version is not whether it can be built, but whether what it is reaching toward is the kind of thing that can be reached by building. Some destinations, the tradition has always held, do not recede because the engineering is poor. They recede because they are not, in the end, destinations. They are arrivals — and the one who constructed them neither ships them nor schedules them.