“Intelligence is in an upward spiral,” asserted tech entrepreneur Marc Andreessen in a 2023 self-published manifesto, and “we are poised for an intelligence takeoff.” For Sam Altman, CEO of Open AI, the generative AI system his company developed—ChatGPT—marked the moment intelligence took flight. Two years after Andreessen’s manifesto, Altman announced the dawn of the “intelligence age.” The agrarian age passed into the industrial age, and now generative AI is here to crank the wheel of history. Dario Amodei, responsible for the generative AI model his company named Claude, predicts its culmination. “This is where humanity was already headed,” he writes about the peace, prosperity, and other materialized ideals that will characterize the intelligence age. AI just “gets us there more quickly.” The tone across these texts varies but the message does not: humankind is about to enter, or has just entered, a distinct next stage. We are at a beginning.
It is not always clear what intelligence refers to for the men announcing its impending reign. The term describes a faculty that is not exclusive to humans. It is actualized and increased in technologies, however, and technologies are nearly uniquely human. Through technology, intelligence will eventually realize humanity’s every hope. Striking a creedal note, Andreessen writes that “we believe in fulfilling our full potential, becoming fully human” through the development of generative AI in particular. “We believe in evolution” appears twice in his manifesto, in a strange echo of the popular yard signs telling passersby that this house believes “science is real.” The intelligence age elicits a prophetic tone. In the rapidly growing community of AI critics, the religion-like quality of these announcements has drawn suspicion. Philosopher Emile Torres attributes belief in the intelligence age to a “bunch of 20th-century atheists” whose lives “lacked the meaning […] provided by traditional religion.” Promises about intelligence amount to an insidious “secular religion” they offer the rest of us in its stead.
As the dream of the intelligence age grows, so does apprehension. Within the community that has developed around AI development, concern crystallizes around alignment. Derived from the French verb aligner, which in turn stems from ad lineam, the Latin phrase “in a straight line,” alignment describes a state in which things line up. With reference to generative AI, the things to line up are human values and the outputs of (yet-to-be-built) systems. Alignment, treated as an engineering objective and philosophical puzzle, ensures that intelligent systems do not act against our norms or interest when they take off, spiral, or take us somewhere very quickly. This elegant solution—only align—may wreak havoc in its own right, however. In an influential essay, Torres and computer scientist Timnit Gebru argue that alignment with the values held by Silicon Valley engineers and architects is not a safeguard for the intelligence age, but its greatest danger.
The problem is one of lineage. Gebru and Torres warn that the “bundle of ideologies” undergirding AI development is rooted in eugenics. The essay highlights institutional and biographical continuities between eugenic science and contemporary AI subcultures, and then extrapolates from these to suggest that the utopian idylls of the intelligence age will curdle into the horrors of the twentieth century. Such extrapolation overlooks the fact that early twentieth-century eugenics was not a perversion of science but its normal application. Precisely because it was ordinary, the association with eugenics is too pervasive to function as a simple taint. Beyond lines and lineage, the actual, live connection between the purveyors of the intelligence age and eugenics is secular religion.
***
What is a secular religion? A hundred years ago, eugenics certainly seemed a good candidate. In a 1904 paper, Francis Galton—who coined the term—explained that eugenics should become “an orthodox religious tenet of the future.” Many scientists working in his wake understood their commitment to eugenics to overlap with or reinforce their religious commitment to a particular kind of moral society, but that is not what Galton meant here. He meant that eugenics could and should now do the work of religion, serving as a resource for people in matters of statecraft and of sex. In the eugenic secular that Galton’s writing helped seed, scientific orthodoxies would be the engines of a better, more beautiful future. In this vision, secular religion is science with religion’s sway; science not as historically situated practice but as engine beneath history. This science, Galton and his peers believed, stood poised to take up its rightful role as the authorizing force underneath social and intimate choice alike, improving on religion by helping humanity go more quickly where it is already headed.
When Torres writes about AI development as in the grip of a secular religion, however, the term is meant to arouse suspicion. It is oxymoronic: the charge is counterfeit secularity. The community around generative AI appears to be undergirded by logic and guided by science alone, but it is in the thrall of something that acts like a religion. AI ideologues looking for purpose and hope commit themselves to a “normative worldview” that Torres sums up as “just like Christianity.” Particularly relevant to this assessment is Torres’s own drift: from a childhood steeped in Evangelical apocalypticism first to New Atheism—the mid-2000s movement invested in rescuing public life from religious “delusions”—and then to the pursuit of the intelligence age, because New Atheism turned out to be overly invested in rescuing Western civilization in particular. When the seemingly secular interest in the intelligence age began to resemble Christian eschatological preoccupations too closely, Torres turned into a self-described “ardent apostate.”
Scholars of religion soon learn that commitment to the correction or even rescue of dupes, no matter how ardent, usually obscures more than it reveals. What are the footholds of the dream? The intelligence age finds purchase in ideas of human destiny and its fulfillment, in the vision of a world that is headed up. The pull that moved Torres from conservative Christianity to AI apostasy is of a piece with the pull that the salesmen of the intelligence age try to harness in their manifestos. That same pull animates the secularist succession narratives that men like Galton wrote, in which science and public reason take up the mantle of humanity’s civilizational progress now that Western Christianity has carried it as far as it could. Scholars in our field have problematized the neat handovers in such narratives of triumphant secularism, pointing out the profound continuities that stories of secular modern conquest skate over. “Secular religion” is not a gotcha but a guidepost, a signal indicating that this is a place where the better future is pressing its demands.
When I say that the intelligence age begins with secular religion, I mean that its launch requires a handover, a moment where our current age—industrial, late-capitalist—tips into the next. Altman writes that “society itself is a form of advanced intelligence.” He thanks our “parents and grandparents” for the “scaffolding for human progress” they built. Their society has become scaffolding for something new; their accumulated intelligence is loosening itself from the collective. Describing “the San Francisco consensus,” Google CEO Eric Schmidt predicts that AI systems “will soon be smarter than the sum of humans.” The intelligence of the past, in this vision, was the possession of human societies or masses, whereas the intelligence that delivers the future belongs to systems and models. This is a startlingly literal transfer of ownership, much more so than the transition from religious moral guidance to the scientific husbandry of biology that eugenicists envisioned. Only once generative AI models have taken in humanity’s shared visual and textual inheritance as data to train on can their intelligence—proprietary, available for subscription—succeed and supersede ours.
***
The purveyors of the intelligence age are sure that we have everything to gain and nothing to lose. The “techno-capital machine” will allow “natural selection to work for us in the realm of ideas,” preaches Andreessen’s manifesto. Its ideas will be so good, so “pro-human,” that society will reshape itself in their image—and we will love it. Even Amodei, generally the least exuberant in his predictions, agrees in his own manifesto. “Initially heavily bottlenecked” by “human factors” like bureaucracy and law, intelligence will eventually “route around” all but the most physically absolute of constraints. Yet it will take us where we were already heading. “If not inevitably, then at least statistically,” he writes, the intelligence age will bring “rule of law, democracy, and Enlightenment values.” He is right, of course, in the sense that AI models will produce outputs that resemble the inputs they trained on. There is no escaping inheritance in statistics, to the point that the discipline of statistics developed through and alongside eugenics, and yet inheritance is not the only way to imagine the future.
I first came across AI as a transformative technology while writing a history of online seduction advice, from 1980s pick-up lines to Andrew Tate. Twentieth-century researchers in AI worked to build machines that can think like us or better. In seduction advice, I saw humans seeking to better themselves by thinking more like an AI. The pick-up line, improved, is a conversational if/then algorithm; the boy, improved, is an alpha male possessed of “maxxed out” metrics. I called this computational creep: the language and logics associated with AI becoming a shorthand for efficacy even in those domains of life that seem unprogrammable or unpredictable. In the manosphere, the improvement of social and even biological intelligence requires its artificialization. Computational creep brings the intelligence age into an analytical frame that is more ambient and atmospheric than alignment. It attends to the seep and stickiness in the encounters we stage between intelligences, drawing out what humans lose and gain when they yoke themselves to system capabilities.
The intelligence age can begin only once the past has been made into its scaffolding. Secular religion calls for a handover that produces what it leaves behind, as scholars of secularism have theorized for an array of modern distinctions. When Galton announces that eugenic science is now poised to take over from religious orthodoxy, he reifies the power he ascribes to religion. The transition is also the continuity; religion’s obsolescence enacts its force. For civilization to enter its intelligence era, it needs to acquire a past in which intelligence already existed but could not yet properly be apprehended as such. Just as science needed a proto-science, needed to be nascent in religion in order to produce itself as religion’s successor, the society of our grandparents and everything our ancestors have written must become a proto-intelligence that a well-trained statistical model can improve upon. Computational creep illuminates how claustrophobic this is: computer programs, conversations, civilizational trajectories, and sexual desires all become a lock intelligence alone can open.
“Wonders [will] become routine,” Altman writes, “and then table stakes.” What begins as marvel first becomes habit and then price of admission—to a game we can no longer leave. He promises that the intelligence age can arrive gently, even though “the rate of new wonders being achieved will be immense.” This is one angle scholars of religion can bring to the alignment conversation: attention to the architecture of wonder, to the violent snips and spirals that produce a new age in real time, and to the forms that persist amid and alongside those cuts. We can offer AI “secular religion,” not as false advertising but as a sign that beckons us. It tells us where to look when humans assign miracles to a coming age.
