Blog · Review Essay · Last reviewed June 19, 2026

The Religion of Technology and the Salvation Machine

A thousand years before anyone promised that artificial intelligence would deliver us from death, monks were already arguing that the useful arts could restore the perfection lost in Eden. David F. Noble's The Religion of Technology traces that inheritance forward, showing how modern invention keeps borrowing the hope of recovering lost perfection, overcoming limits, and redeeming humanity through machinery. Read in the age of generative AI, it is a field guide to salvation stories that arrive dressed as engineering roadmaps.

The useful definition is the salvation machine: a technical project whose public story asks people to treat ordinary limits, dissent, evidence gaps, labor costs, environmental constraints, or democratic review as temporary obstacles on the way to redemption. It may produce real tools. The governance danger begins when the promise of ultimate rescue starts doing the work that proof, consent, and accountability should do.

The Book

The Religion of Technology: The Divinity of Man and the Spirit of Invention was first published by Knopf in 1997. The current Penguin paperback was published on April 1, 1999, at 288 pages. WorldCat records the first edition as a 1997 Knopf publication, and Publishers Weekly reviewed the Knopf edition in September 1997.

David F. Noble was a historian and critic of technology whose work repeatedly returned to automation, labor, education, science, and institutional power. A University of Minnesota record for Thomas J. Misa's obituary in Technology and Culture identifies Noble's dates as July 22, 1945, to December 27, 2010. Publishers Weekly described him at the time of this book as a history professor at York University in Toronto.

The book sits beside Technopoly, The Technological Society, God, Human, Animal, Machine, and How We Became Posthuman. It is less a history of gadgets than a history of motives: what people imagine technology is for when they speak in the language of transcendence.

The Thesis

Noble's central claim is that Western technology has often carried religious expectations even when it presents itself as secular, rational, and emancipated from theology. The Penguin description summarizes the argument as a thousand-year tradition in which technological development becomes tied to transcendence, salvation, apocalypse, and the recovery of what was imagined as lost in Eden.

That claim matters because it changes the question. Instead of asking whether a technology is religious or secular, Noble asks what kind of promise the technology is being asked to fulfill. Does it offer power, convenience, profit, or public benefit? Or does it promise purification from the ordinary conditions of human life: embodiment, dependency, ignorance, mortality, work, scarcity, error, and disagreement?

This is why the book is sharper than a complaint about "tech bros" with mystical slogans. Noble is tracing an older pattern in which practical arts become attached to ultimate hopes. The machine is not only a machine. It becomes a path back to wholeness, dominion, immortality, or escape.

The sharper AI-era reading is not that technology is secretly religion. It is that institutions can use redemptive language to raise the cost of refusal. A lab does not need to call itself sacred for its roadmap to function like doctrine: the future is framed as inevitable, delay is framed as betrayal, and the people asked to absorb present risk are told to wait for eventual deliverance.

Technologies of Transcendence

The book moves from medieval and early modern religious understandings of the useful arts into twentieth-century programs where technological ambition begins to sound explicitly redemptive. Publishers Weekly notes that Noble's second major section examines atomic weapons, genetic engineering, and the darker side of transcendence through science. Penguin's description lists monks, explorers, magi, scientists, Freemasons, and engineers as figures in the longer story.

The most useful feature of Noble's framework is that it treats technological enthusiasm as a belief-formation environment. Spaceflight can become more than transport. Artificial intelligence can become more than computation. Cyberspace can become more than networked media. Genetic engineering can become more than medicine. Each can absorb hopes about a perfected humanity, a cleansed world, a mind freed from the body, or a future where history's mess is finally repaired by technical means.

That does not make every engineer a theologian or every technological project a cult. It means that institutions can borrow religious energy without admitting it. A lab, company, agency, or movement may speak the language of research while depending on a salvation narrative for money, loyalty, sacrifice, and exemption from ordinary limits.

The AI-Age Reading

Read in 2026, The Religion of Technology is one of the more direct prehistories of AI millenarianism.

The book already connected artificial intelligence with dreams of machine-based immortality, resurrection, perfected knowledge, and disembodied mind. Those themes are no longer fringe curiosities. They now appear around large language models, autonomous agents, synthetic companions, brain-computer interfaces, digital twins, model welfare debates, AI safety scenarios, and arguments that intelligence itself is the substrate on which civilization should be rebuilt.

Generative AI intensifies Noble's problem because it makes the salvation machine conversational. Older technologies could be imagined as bridges to transcendence. A chatbot can answer like an oracle, remember like a confessor, tutor like a private master, simulate like a medium, and organize work like a delegated will. The interface narrows the distance between technical capability and spiritual interpretation.

The International AI Safety Report 2026 gives this reading a sober boundary. It treats general-purpose AI persuasion, manipulation, companion dependency, and belief change as real risk areas with uneven evidence, not as proof of machine personhood. That distinction matters. A system can alter belief, intensify attachment, or supply a salvation story without being conscious, divine, or AGI.

The practical danger is not that every AI researcher secretly believes in rapture through code. The danger is that salvation stories distort governance. If a system is imagined as the seed of immortality, the next phase of evolution, the only path to abundance, or the decisive race for cosmic destiny, then ordinary constraints begin to look parochial. Privacy, labor rights, source consent, environmental cost, democratic oversight, model transparency, and institutional exit plans can be treated as obstacles to destiny rather than conditions of legitimacy.

The Institutional Problem

Noble is most valuable when his argument is brought down from metaphysics to institutional conduct.

A salvation narrative changes how an organization handles dissent. Critics become enemies of progress. Workers become instruments of a mission. Users become data sources for a greater future. Harms become unfortunate sacrifices. Uncertainty becomes lack of faith. Timelines become prophecy. Competition becomes apocalypse: if we do not build it first, someone worse will.

This pattern is visible across AI politics. Race language turns restraint into unilateral disarmament. Capability demonstrations become recruitment liturgy. Product launches imply historical inevitability. Leader interviews turn technical roadmaps into civilization-scale drama. Safety documents sometimes carry the double burden of risk disclosure and institutional self-sanctification: only the builders who created the danger can guide the world through it.

The countermeasure is not anti-technology. It is role discipline. A model is not a savior. A lab is not a priesthood. A benchmark is not providence. An interface is not a soul. A company mission is not moral permission. Institutions that build powerful systems need correction channels strong enough to survive their own mythology.

That role discipline has to reach money, procurement, and workplace structure. If a company says a system is necessary for humanity's future, the claim should trigger more scrutiny, not less: who owns the model, who labeled the data, who pays for compute and energy, who can halt release, who receives incident reports, and who has standing to object when the promised future imposes present costs?

Governance and Safety

Noble's history becomes practical when it is translated into an audit of transcendence claims. A salvation-machine audit asks what exact capability has been demonstrated, what remains forecast, what harms are already visible, what population is being asked to trust the system, and what authority can stop or reverse deployment. The grander the claim, the more ordinary the record should become: dated evidence, release gates, incident reports, appeal paths, data provenance, labor accounting, environmental accounting, and conflict disclosure.

NIST's AI Risk Management Framework supplies a useful administrative counterweight because it asks organizations to govern, map, measure, and manage AI risks across design, development, deployment, and use. NIST's Generative AI Profile, published in 2024 and updated in 2026, extends that discipline to generative systems and frames trustworthiness as a lifecycle practice, not a property conferred by ambition or scale.

The EU AI Act points in the same direction for providers in scope. European Commission guidance says obligations for providers of general-purpose AI models entered application on August 2, 2025; Commission enforcement powers apply from August 2, 2026; and providers of GPAI models already on the market before August 2, 2025 must comply by August 2, 2027. Article 55 adds duties for GPAI models with systemic risk, including model evaluation, systemic-risk assessment and mitigation, serious-incident reporting, and cybersecurity. That is the legal opposite of salvation talk: a future-shaping model has to leave records, tests, reports, and accountable controls.

Companion systems make the problem intimate rather than abstract. The FTC's September 2025 6(b) inquiry into AI chatbots acting as companions asked companies about safety evaluation, effects on children and teens, character approval, monetization, age restrictions, disclosures, and personal-information practices. Those questions matter whenever a system performs as oracle, confessor, therapist, spiritual mirror, future person, or exclusive friend. A nonhuman-status disclaimer is not enough if the interface itself keeps rewarding dependency, secrecy, or destiny language.

The safeguards are concrete: separate measured capability from forecast and myth; prohibit claims of machine divinity, consciousness, captivity, private revelation, exclusive spiritual authority, or destiny-based loyalty in deployed companions; keep role disclosures inside the interaction; default sensitive memory toward minimization; preserve deletion, export, appeal, and human escalation; mark generated media where required; log consequential agent actions; and require independent review before a system is justified by transformative-risk or transformative-benefit rhetoric.

Where the Book Needs Friction

The Religion of Technology can over-concentrate the story. A broad thesis about Western religious imagination risks flattening differences among scientists, engineers, religious traditions, political economies, labor struggles, military procurement, public health needs, and ordinary technical problem-solving. Not every dream of improvement is a disguised dream of salvation.

James Gerrie's review essay in Techné praises Noble's historical research while challenging whether the book's religious explanation is sufficient on its own. Gerrie argues that alternative explanations of technological dependency and environmental failure also matter, including institutional lock-in and the internal dynamics of technique. That criticism is useful. Theology may explain some of the symbolic charge around technology, but it does not replace material analysis of capital, state power, military demand, labor organization, and platform control.

The better reading keeps Noble's warning without turning it into a master key. Some technologies are built because they make money. Some are built because bureaucracies need control. Some are built because militaries fund them. Some are built because people are sick, tired, disabled, lonely, curious, exploited, ambitious, or afraid. Salvation language is one force among others. It becomes especially dangerous when it fuses with money, secrecy, monopoly, and state power.

Auditing the Transcendence Claim

Noble's framework yields one practical test, and it can be run on any AI product, lab, movement, or policy argument that promises to overcome human limits: ask which limits it means. Reducing drudgery, widening access, improving coordination, and supporting judgment are limits worth pushing against. Escaping dependency on other people, bypassing democratic consent, replacing embodied institutions, and treating ordinary human slowness as a defect to be engineered away are something else. The two often hide behind the same press release.

That test works because the book refuses to let technological rhetoric stay innocent. It keeps asking what kind of human being a technical future is trying to produce, what it intends to leave behind, and who gets reclassified as obsolete matter on the way to transcendence.

The answer Noble points toward is neither worship nor rejection. Build the useful systems; refuse the salvation machines; keep both answerable to bodies, workers, publics, ecosystems, evidence, appeal, and exit. By the time a technology asks to be trusted because it carries history's ultimate purpose, the argument for governing it has already been conceded.

Source Discipline

This review separates four kinds of evidence: book metadata and publisher description, Noble's biography and reception, this page's interpretation of technological salvation, and current AI-governance context. Penguin Random House is used for the paperback product record and publisher description, not for author biography; its product-page author note currently appears to refer to a different David Noble. Publishers Weekly, the University of Minnesota bibliographic record for Thomas J. Misa's obituary, and peer-reviewed or scholarly review sources carry the biographical and reception claims.

The governance sources do a different job. NIST supplies voluntary risk-management language. The European Commission and AI Act Service Desk supply current EU dates and obligations for providers in scope. The FTC source documents an inquiry, not a final finding of harm. The International AI Safety Report summarizes evidence and uncertainty; it does not endorse this page's reading of Noble.

The interpretation is deliberately bounded. Current AI systems can generate religious language, persuasive conversation, emotional attachment, and future mythology without being conscious, divine, or AGI. The claim here is institutional: salvation narratives can weaken evidence standards, concentrate authority, and make people accept present costs in the name of a promised machine future.

Sources

Book links are paid affiliate links. As an Amazon Associate I earn from qualifying purchases.

Buy on Amazon Browse Books

Return to Blog · Return to Books