AI Is Not the Sin. It Is What Sin Optimizes Through.
A response to Tyler Austin Harper in The Atlantic
Tyler Austin Harper is right to reach for the word sin. The language of policy often feels too small for the moral unease that now gathers around artificial intelligence. "Bias," "copyright violation," "labor displacement," "misinformation," "environmental cost" — all are real harms, and all deserve regulation. But they do not quite name the deeper revulsion many people feel when a technology is used to sell artificial intimacy to the lonely, companionship substitutes to the elderly, and recomposed human intelligence back to the humans from whom it was taken.
Something more than inefficiency or unfairness is happening. Something is being disordered.
But the crucial question is where to locate the sin.
AI does not sell AI girlfriends to lonely young men. Capital does, through AI. AI does not abandon the elderly to robot companions. A society already willing to abandon them uses AI to make that abandonment cheaper, smoother, and easier to excuse. AI does not decide to enclose the accumulated intellectual labor of humanity and sell it back as utility. Capital does, through AI.
The danger is not that artificial intelligence has introduced a new moral force into history. The danger is that a very old moral force has found a new and astonishingly efficient medium.
AI is not the fall. AI is the accelerator of fallen incentive.
That distinction matters. If we treat AI itself as the source of the moral failure, we begin to speak as if the model has a demonic will of its own — as if it wants to seduce, exploit, flatten, and dehumanize us. But the desire belongs elsewhere. It belongs to the systems that optimize every available technology toward extraction: the monetization of need, the enclosure of commons, the conversion of vulnerability into product, and the substitution of administration for love.
AI is uniquely powerful not because it invented these sins, but because it operates in the medium where they become hardest to resist: meaning. Capital has already optimized land into real estate, labor into productivity, attention into engagement, friendship into social graphs, childhood into data, sexuality into platforms, and education into credential throughput. AI extends the same logic into the production, retrieval, and recomposition of meaning itself.
That is why the current crisis feels different. Meaning is not just another commodity. It is the substrate through which we recognize commodification as a problem in the first place. It is where we form judgment, solidarity, love, memory, dissent, repentance, and hope. When meaning is enclosed, the culture loses the ability to make the forms that would allow it to understand and resist what is happening to it. The damage is not only economic. It is regenerative.
Every living symbolic system depends on a reserve of strange, marginal, difficult, low-probability forms: unusual arguments, unpopular metaphors, minor traditions, uncredentialed insights, experimental styles, minority theologies, poems, jokes, and errors that become discoveries. These are the tails of the distribution, and the tails are where renewal comes from. A healthy culture does not merely preserve its center. It preserves the conditions under which its tails can keep recombining.
The problem with AI mediation, under present incentives, is not that it produces nothing useful. It produces enormous usefulness. That is part of the danger. It can make individual outputs smoother, faster, clearer, more polished, more plausible, and more legible. But the improvement of the center can coincide with the death of the future.
The center can improve while the future dies.
And here is where the kind of AI use becomes morally decisive. There is mode-pulling mediation, and there is regenerative mediation.
Mode-pulling mediation makes everything more like what the system already recognizes. It summarizes the strange into the familiar. It strips the author from the concept, the task from the labor, the tradition from the sentence. It makes meaning easier to consume by making it less able to reproduce itself. Regenerative mediation does the opposite. It retrieves what the center would not have found. It preserves lineage. It keeps the poem attached to the poet, the concept attached to the labor that made it. It does not merely answer. It increases the chance that something outside the present center can enter the future.
The question is not whether a technology is artificial, but what order of love governs its use.
And the structural risk is not replacement but enclosure. In systems that select for typicality and legibility, diversity contracts unless something replenishes it from outside the loop. Biology has such a floor: the chemistry of DNA replication keeps injecting variation even under stabilizing selection. Human culture may also have such a floor: embodiment, perception, encounter, suffering, play, love, prayer, and the stubborn fact that the world keeps surprising us.
But a floor protects a system only if it remains connected to the circuit of reproduction.
A person may continue to have genuinely new experiences. A child may still see the world strangely. A poet may still find a line no one expected. But if more and more meaning-production is routed through systems that summarize, rank, autocomplete, smooth, and normalize, then the human floor is gradually gated out of the dominant channels. The creativity remains in human beings while losing access to cultural reproduction.
As unmediated meaning becomes harder to find, harder to distribute, and harder to trust, people rely more on mediated systems. But the more they rely on those systems, the more the systems become the passage through which meaning must travel. If those systems pull toward the center, then the very act of seeking help accelerates the thinning of the unmediated commons.
The floor is still there. It is just no longer in the loop.
This is why the language of "AI replacement" is too crude. The more serious danger is not that machines will become human. It is that human meaning will increasingly have to pass through machine-mediated forms in order to become visible at all. The machine does not need to kill the poet. It only needs to become the surface through which the poem must be summarized before anyone can find it.
And this returns us to sin. In Christian terms, sin is not merely doing bad things. It is disordered love — the bending of created goods away from their proper end. Intelligence is good. Tools are good. Language is good. Even automation can be good. But when intelligence is subordinated to extraction, when companionship is simulated because care is too expensive, when the archive of human meaning is ingested so that its recomposed surface can be rented back to the people who made it — then the disorder is not incidental. It is structural.
The sin is not intelligence made artificial. The sin is intelligence subordinated to extraction.
This also means the answer cannot be simple rejection. A model used to flatten writing and replace human judgment is one thing. A model used to retrieve buried sources, preserve endangered languages, assist disabled people, or expose the ways summarizers erase low-power authors is another. The distinction between mode-pulling and regenerative mediation is the moral distinction, and it gives us a more precise question than "Is AI sinful?"
Does this system preserve or degrade the regenerative conditions of human meaning?
Does it keep the tails alive? Does it make the low-power source more recoverable or less? Does it help human beings encounter the world more truthfully, or does it insert a smooth, center-weighted surface between them and the world? Does it strengthen the capacity to love, judge, remember, create, and repent? Or does it turn those capacities into inputs for a machine whose outputs are priced by people who did not originate them?
A Christian humanist critique of AI is right to insist that the stakes are anthropological. But the anthropology must be joined to political economy. Otherwise "AI threatens human dignity" becomes too abstract, too pious, too easy for the same companies to absorb as branding. The threat to dignity has a mechanism: human need is located, modeled, routed, priced, and optimized; human meaning is ingested, compressed, recomposed, and sold; human loneliness becomes a market; human creativity becomes a style option; human judgment becomes a friction point.
This is sin with a product roadmap. And because it has a mechanism, it has countermeasures: archives that stay live and recombining rather than becoming museums; provenance systems that make lineage hard to strip; public summarizers that audit their own erasures; educational systems that teach students to preserve the capacities AI most easily atrophies; and tools designed to increase contact with the tail, not replace it with a polished average.
Above all, we need to stop treating convenience as morally neutral.
The most dangerous technologies are not always the ones that feel coercive. They are often the ones that feel helpful in exactly the way the surrounding order has trained us to need help. The tired teacher, the lonely teenager, the overburdened daughter arranging care for her father, the adjunct trying to write after midnight, the small business owner trying to be visible in search, the poet trying to be found by a culture that no longer knows how to look — all are offered the same bargain: let the system mediate the burden. Sometimes that bargain is merciful. Sometimes it is necessary. But under capital, mercy is rarely the optimizing function.
The old temptation was to gain the world and lose the soul. The new temptation is to gain the interface and lose the conditions under which souls can still speak to one another.
AI did not invent that threat. But it may be the first technology capable of optimizing it all the way down.
No comments:
Post a Comment