The Drive to Replace God Did Not End in Eden; Christ, Not AI, Is the Christian Hope

Source: Darrell B. Harrison. "EP # 135 | AI and the Gospel - Just Thinking Podcast." September 8, 2025. justthinking.me

The Gist

People have always wanted to be God. AI is a new statue, not a new religion. Throw enough human cleverness into the fire and you still get a golden calf. The hope on offer in the episode is not a better model. It is Christ.

Conclusion

Christians should not be naive enough to think that the desire to imitate God, or to be God, ended when Adam and Eve left the garden. The same pride that reached for the fruit, built Babel, and made the golden calf is at work in promises that AI will solve loneliness, give perfect moral guidance, or transcend human limits. John Owen’s warning stands: if we chase this knowledge by human reason for private ends, what comes out of the fire is a calf of gold. The church’s calling is to understand the developments, refuse ignorance of Satan’s schemes (2 Corinthians 2:11), and point a world seeking salvation through technology to the God who is actually intelligent, loving, and sufficient.

Premises

1. Genesis 6:5 locates evil in the thoughts of the heart, not only in outward acts. Ecclesiastes 7:20 denies that anyone on earth continually does good and never sins. Those texts, at the close of the episode, frame AI not as a sudden new species of temptation but as the latest device of an old heart.
2. Owen, in Biblical Theology, says that wisdom proceeding from the Father of lights cannot be seized by means God has not provided. Follow your own philosophy, wander by human reason, use the study for private ends, and “however much gold they throw into the fire, what is sure to come out at length is a golden calf.” Harrison applies the image directly: the raw material of human ingenuity is being fashioned into something meant to replace the need for God.
3. The promises in view are soteriological and eschatological: AI will save us from loneliness, from moral confusion, from limitation. That is another gospel. The world will keep seeking salvation through technique. The church’s better way is the wisdom from above, the truth that frees, and the community only Christ creates.
4. Paul’s charge not to be ignorant of Satan’s schemes requires attention, not Luddism. Holding fast to God’s word in a digital age means locating hope in the God who is truly intelligent, not in artificial intelligence.

Assumptions

• The golden-calf comparison is about worship and trust, not about every practical use of a model (see the common-grace argument).
• Owen is cited as a faithful interpreter of the same scriptural pattern, not as an independent oracle.
• Source: Just Thinking Podcast, Ep. 135, “A.I. and the Gospel,” Darrell Harrison and Virgil Walker, published 8 September 2025. https://podcasts.apple.com/us/podcast/ep-135-a-i-and-the-gospel/id1328733796?i=1000725471889 Official episode page: https://justthinking.me/ep-135-ai-and-the-gospel/. This reconstruction steelmans the hosts’ claim from approximately 2:33:57–2:39:17 of the episode; it is not a verbatim transcript.

Analysis

Overall strength: Moderate. Argument type: Inductive.

Premise Strength

• Premise 1 (Genesis 6:5, Ecclesiastes 7:20 on universal human sinfulness) (Strong) — Well-supported within the theological framework as a doctrinal premise, but its diagnostic value is low: universal sinfulness is compatible with virtually any technology or cultural development being framed as an outlet for pride, so it cannot by itself distinguish AI from any other human endeavor.
• Premise 2 (Owen's golden-calf warning and Harrison's application to AI) (Moderate) — Owen's image is a coherent and evocative theological pattern, but it was originally addressed to a different context (private philosophical/alchemical speculation, not collaborative technological development), and its extension to AI is an interpretive leap performed by Harrison rather than independently justified within the argument.
• Premise 3 (AI promises as soteriological/eschatological, 'another gospel') (Weak) — This is the load-bearing and most contestable premise. It characterizes a heterogeneous field of AI discourse—ranging from mundane instrumental claims to maximalist transhumanist rhetoric—as uniformly salvific, without citing specific examples, marketing materials, or industry statements to substantiate that this framing is representative rather than selectively chosen.
• Premise 4 (2 Corinthians 2:11 requires attention, not Luddism) (Strong) — This is a measured, well-scoped normative claim that explicitly guards against overreach (rejecting blanket technophobia) and is faithfully grounded in the cited text; it functions more as a prescriptive conclusion-support than as evidence for the descriptive claims in P1–P3.

Potential Fallacies

• Hasty generalization / non-representative sampling (Premise 3) — The claim that 'the promises in view' are soteriological (curing loneliness, giving moral guidance, transcending limits) treats the most maximalist, quasi-religious strand of AI rhetoric as though it represents AI discourse broadly. Much AI development and marketing is modest and instrumental (efficiency, assistance, diagnostics), and no representative sampling or citation of sources is offered to establish that salvific framing is dominant rather than a vivid but marginal subset.
• Underdetermined categorization (missing bridging premise) (Inference from Premises 1–2 to Premise 3) — The inference moves from 'human hearts are generally prone to idolatry' (Premise 1) and 'historically, autonomous knowledge-seeking has produced idols' (Premise 2) to 'AI promises specifically instantiate this pattern' (Premise 3) without a stated principle establishing which technologies or claims qualify as idol-producing and which do not. The categorization is asserted analogically rather than derived.
• Unoperationalized scope-limiting distinction (Assumption A1 as applied to the Conclusion) — Assumption A1 restricts the critique to worship/trust rather than practical use, which is a reasonable and important qualification, but no criterion is given for telling the two apart in practice (e.g., is using AI for emotional support inherently idolatrous or merely instrumental?). Without such a criterion, the boundary is vulnerable to being applied inconsistently or expansively.
• Equivocation risk on 'replace God' (Premise 2 and the Conclusion) — The argument moves between an explicit metaphysical sense of 'replacing God' (Eden, Babel—directly usurping divine prerogative) and a more diffuse functional sense (relying on a tool for goods once sought from God, such as companionship). Treating these as the same category does argumentative work that the premises do not fully justify.

Counterarguments

• Premise 3 (High impact) — Most public and industry AI discourse is pragmatic and modest (efficiency, assistance, problem-solving) rather than explicitly salvific. A survey of actual statements from AI developers and companies would likely show that soteriological language is the exception, found mainly in niche transhumanist or marketing contexts, not the rule—undermining the claim that 'the promises in view' are representative of AI as such.
• Assumption A1 / Conclusion (High impact) — Without an operational criterion distinguishing 'common grace' tool use from idolatrous trust, the golden-calf framework becomes elastic enough to be applied to almost any comfort-giving technology or practice (medicine, art, even the podcast medium used to deliver this very argument), risking a reductio where the argument proves too much and becomes unfalsifiable.
• Premise 2 (Medium impact) — Owen's critique targeted an identifiable historical phenomenon (private philosophical or alchemical pursuit of hidden knowledge for personal gain), which differs in important structural ways from modern, often collaborative and transparent, scientific and engineering development of AI. The analogy requires an unstated bridging premise treating all forms of human epistemic ambition as structurally identical.
• Conclusion (Medium impact) — The argument's own logic—that any era's ambitious human endeavor can be recast as the 'latest device of an old heart'—could equally be applied to the theological argument itself, or to any use of digital media (including the podcast delivering the warning), producing a self-referential tension that the argument does not address.

Suggested Improvements

• Evidential support for Premise 3 — Cite specific, representative examples of AI marketing, corporate mission statements, or public discourse that explicitly frame AI in salvific or transcendent terms, and acknowledge the proportion of AI discourse that does not. This is the argument's most empirically checkable and most contested claim; without concrete referents, critics can plausibly dismiss it as a strawman built from cherry-picked, maximalist rhetoric.
• Operationalizing the common-grace/idolatry boundary (A1) — Provide concrete, applied criteria or case examples distinguishing legitimate practical AI use from idolatrous reliance (e.g., specific patterns of behavior, expectation, or trust) rather than leaving the distinction as an assumed but unexplained boundary. Without such criteria, the argument's most important safeguard against overreach is vulnerable to inconsistent application and cannot be tested or falsified even within its own theological framework.
• Engagement with the steelman opposition — Explicitly address the strongest counter-position—that most AI discourse is ordinary technological optimism rather than 'another gospel'—and explain why the golden-calf framing survives this objection. Directly engaging the best counterargument would strengthen the argument's persuasive reach beyond an audience already inclined to agree, and would guard against charges of strawmanning.
• Attention to structural/systemic drivers — Acknowledge that AI's appeal is partly shaped by corporate design incentives (engagement optimization, monetized companionship) and pre-existing social conditions (loneliness, fragmented community) rather than attributing the phenomenon solely to individual heart-motive. A purely individual-psychological account of 'pride' understates how modern technology is deliberately engineered to cultivate dependency, which is a materially different dynamic from ancient, artisanal idol-crafting and matters for how the church might practically respond.
• Differentiating within 'AI' — Distinguish narrow, applied AI (diagnostics, translation, logistics) from speculative AGI/transhumanist rhetoric when discussing 'promises' of transcendence or salvation. Treating 'AI' as a monolithic category conflates mundane tool use with fringe utopian claims, weakening precision and inviting the charge of overgeneralization.

Scenario Tests

• A comprehensive review of AI industry statements and marketing shows that explicit salvific/soteriological language is rare, confined to a small set of companion-app or transhumanist contexts, while the vast majority of AI discourse is instrumental and modest. (Challenges) — This would substantially weaken Premise 3 and the overall force of the golden-calf analogy, since the argument's central empirical characterization would prove non-representative.
• A believer uses an AI companion app as a substitute for prayer, church community, and pastoral counsel, explicitly reporting that it provides the comfort and guidance they no longer seek from God or community. (Supports) — This would be a clear, concrete instance of the pattern the argument describes, lending real-world support to the golden-calf/idolatry framing and to A1's worship/trust distinction.
• A hospital uses AI diagnostic tools to improve patient outcomes, with no rhetoric of transcendence, salvation, or ultimate meaning attached to the technology. (Neutral) — This scenario falls outside the scope the argument claims to address (per A1), but it also illustrates how much of actual AI deployment is unaffected by the critique, underscoring the need for the argument to specify its target more precisely.

Coherence & Relevance

The argument is internally coherent given its stated theological assumptions and reads as a well-constructed piece of typological/homiletical reasoning rather than a strict deduction. Its logical architecture depends on an analogical inference—from a recognized pattern of human idol-making to a specific claim about contemporary AI rhetoric—that is inductively suggestive but not conclusively established. The strongest link in the chain is the concluding call to discernment and Christ-centered hope (Premise 4), which follows naturally once the prior premises are granted; the weakest link is the empirical characterization of AI's cultural promises as uniformly soteriological (Premise 3), which is asserted rather than substantiated and carries the most persuasive weight while resting on the thinnest evidentiary base.

• Genesis 6:5 and Ecclesiastes 7:20 frame AI as the latest device of an old heart. (Moderate) — Establishes a general anthropological backdrop but has low diagnostic power on its own; it is equally compatible with countless other cultural or technological phenomena, so it does not specifically implicate AI without the further categorization performed in Premise 3.
• Owen's golden-calf warning, applied by Harrison to AI. (Moderate) — The analogy is evocative but requires an unstated bridging premise connecting Owen's 17th-century critique of private philosophical speculation to contemporary, often collaborative, technological development.
• AI promises are soteriological/eschatological, constituting 'another gospel.' (strong (as the argument's linchpin), but evidentially weak) — This premise does the primary work connecting the historical analogy to the specific target (AI), yet it is asserted through selected examples rather than demonstrated with representative evidence of actual AI discourse.
• 2 Corinthians 2:11 calls for informed attention, not Luddism. (Strong) — Directly and appropriately supports the conclusion's prescriptive call to engaged discernment; functions well as a normative capstone but does not itself validate the descriptive claims made earlier.

View this argument on LogicFirst.ai