Only God Possesses Aseity, So AI Cannot Be Autonomous or Self-Sufficient

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

The Gist

God doesn’t need anything outside himself in order to know or to be. AI needs us for everything: code, data, power, questions. So it cannot be a little god, and Christians should not let the marketing borrow God’s attributes.

Conclusion

The only truly autonomous, non-dependent being is God, whose knowledge is not received from anything outside himself. AI has no aseity: it is not self-existent, omniscient, or omnipotent, and it knows nothing antecedently. Because it must be served by human hands, it cannot occupy the place that belongs to God, and Christians should refuse descriptions of AI that quietly transfer divine attributes to machines.

Premises

Classical Christian theism holds that God is a se: independent of all causes, external and internal (Thomas Ridgley; James Dolezal). God does not receive ideas from any object outside himself. Creaturely knowers, by contrast, depend on existing objects. If the object were not there, they could not have the idea of it.
Isaiah 40:13–14 and Acts 17:24–25 press the same point: no one directs the Spirit of the Lord; God is not served by human hands as though he needed anything. AI, by contrast, needs human hands at every stage: design, data, electricity, prompts, evaluation. Dependence is not an accident of today’s models; it is the kind of thing a machine is.
AI therefore knows nothing before the object of that knowledge exists. It has no antecedent knowledge. Harrison’s claim is not that models are unimpressive, but that nothing about them is independent. “AI will never be intelligent” in the sense that would make it a rival to God’s knowledge, because everything it “knows” is after the fact and derived.
There is a hamartiological as well as an anthropological layer. Behind every application are fallen, finite, sinful people writing code so that outputs align with what they programmed. Stephen Charnock’s observation that the mind has a natural dexterity to devise mischief, and that the more natural a thing is the more easily it is done, applies: sin is natural to us, so the systems we make are not theologically quarantined.

Assumptions

Aseity is a real divine attribute, not a metaphor. If that doctrine is false, this argument fails.
Creaturely dependence includes even brilliant human inventors. The argument does not require that engineers claim to be God in so many words; it requires that autonomy-talk about AI is theologically confused.
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 36:36–48:00 of the episode; it is not a verbatim transcript.

Analysis

Overall strength: Moderate. Argument type: Deductive.

Premise Strength

Classical Christian theism holds that God is a se... Creaturely knowers, by contrast, depend on existing objects. (Strong) — Faithfully represents a mainstream (though not universal) classical theist doctrine, properly sourced to recognized authorities (Ridgley, Dolezal). Given A1 as a stipulated theological axiom, this premise is coherent and well-supported within its framework; its force is purely internal to that framework and carries no weight for audiences that reject or qualify strong aseity (e.g., open theism, process theology).
Isaiah 40:13–14 and Acts 17:24–25... AI, by contrast, needs human hands at every stage. (Moderate) — The empirical claim about AI's dependence on human infrastructure (design, data, electricity, prompts) is well-established and essentially uncontroversial. The scriptural application to AI specifically, however, is an interpretive extension: the cited texts originally address idolatry, divine self-sufficiency, or human agency in general, not technology, so applying them directly to AI's ontological status is an inferential leap rather than a straightforward exegetical conclusion.
AI therefore knows nothing before the object of that knowledge exists... everything it 'knows' is after the fact and derived. (Moderate) — Correctly captures that AI systems have no knowledge independent of training data and design, which is defensible as a description of current architectures. It somewhat overstates its case, however, by treating 'no antecedent knowledge' as a sharp binary; pretrained model weights function as a kind of statistical prior that some functionalist accounts would describe (in non-theological terms) as internal structure prior to any given query, complicating the…
There is a hamartiological as well as an anthropological layer... sin is natural to us, so the systems we make are not theologically quarantined. (Weak) — This is an internally coherent extension of a specific Christian anthropology (via Charnock) but functions as an independent claim about moral corruption rather than additional support for the aseity argument. It is also an overbroad generalization: it asserts, without case-specific evidence, that every AI application inherits identifiable sinful taint from its programmers, conflating a general doctrine of universal human fallenness with a…

Potential Fallacies

Equivocation on 'autonomy'/'dependence' (Move from P1–P3 to the conclusion, especially the phrase 'descriptions of AI that quietly transfer divine attributes') — The argument establishes that AI lacks aseity — a technical theological term for uncaused, self-existent being — but then treats this as equivalent to showing that ordinary engineering language about AI being 'autonomous,' 'intelligent,' or 'self-directed' is theologically confused. In AI and robotics discourse, 'autonomous' typically just means operating without real-time human intervention, a graded operational claim, not a…
Is/ought gap (Transition from P1–P3 to the concluding imperative) — The premises describe what God and AI metaphysically are (independent vs. dependent). The conclusion prescribes a linguistic duty ('Christians should refuse descriptions...'). Moving from a descriptive metaphysical difference to a prescriptive rule about permissible speech requires an additional normative premise — for example, an argument about idolatry, reverence, or the pastoral harms of loose language — that is not supplied.
Disconnected/bundled premise (P4) — P4 introduces a separate hamartiological claim (fallen programmers taint AI outputs) that does not logically support or extend the aseity argument in P1–P3. It could be removed without weakening the case for 'AI lacks aseity,' suggesting it functions as rhetorical reinforcement of a predetermined conclusion rather than as a load-bearing premise.
Proves-too-much / unprincipled selective application (P1 combined with the conclusion's exclusive focus on AI) — P1 explicitly concedes that all creaturely knowers, including humans, depend on external objects for their ideas. Applied consistently, the same dependency criterion used to deny AI 'autonomy' or 'knowledge' language would also deny it to humans, an outcome the argument clearly does not intend and does not explain how to avoid, beyond gesturing at (but not developing) a distinction such as the imago Dei.

Counterarguments

Conclusion (High impact) — Granting the full theological framework (aseity belongs to God alone), the practical conclusion still does not follow, because 'autonomous' and 'intelligent' as used by AI researchers are technical, operational terms describing performance without real-time human control, not metaphysical claims about self-existence. No AI company or engineer claims a model is uncaused or self-existent; demanding that ordinary technical vocabulary conform to strict theological metaphysics conflates two non-competing discourse domains, similar to objecting to 'the sun rises' on…
Premise 1 / Conclusion (High impact) — Since P1 concedes that human creaturely knowers are also non-a-se and depend on external objects for their ideas, the same standard applied consistently would forbid describing humans as 'intelligent' or 'autonomous' as well — an absurd result the argument does not intend and does not explain how to avoid.
Premise 1 / A1 (High impact) — Strong, classical aseity is not the only Christian position on divine attributes; open theism, process theology, and some analytic theologians offer more relational or temporal accounts of God that would not treat AI-autonomy language as automatically illicit. The argument's force is confined to audiences who already accept a specific, contested theological school.
Premise 4 (Medium impact) — P4's causal chain (human sinfulness → biased code → tainted AI outputs) is asserted without case-specific evidence and is empirically indistinguishable from secular explanations of AI bias and failure (data quality, incentive structures, technical limitations), making the theological framing an interpretive overlay rather than an additional evidential test.
Conclusion (Medium impact) — The argument never identifies who specifically ascribes divine-like autonomy or omniscience to AI, leaving its target underspecified. Absent concrete examples (e.g., AGI-messianism, singularity rhetoric, marketing hyperbole), the rebuttal risks attacking a position that is not the mainstream way AI is actually discussed by developers or the public.

Suggested Improvements

Bridging premise on 'autonomy' — Add an explicit premise defining what sense of 'autonomous'/'intelligent' is being targeted, and argue why ordinary technical usage of these terms in AI research implicitly or explicitly smuggles in a metaphysical claim of self-existence, rather than assuming this without demonstration. This is the argument's single most consequential gap; without it, the practical conclusion (Christians should refuse such descriptions) does not follow even granting the full theological premises.
Human/AI parity — Explicitly address why creaturely dependence disqualifies AI from analogical cognitive language while comparable dependence does not disqualify humans — for instance, by developing the imago Dei distinction gestured at in A2 rather than leaving it implicit. Without this, the argument is vulnerable to a straightforward reductio that proves too much, since it would equally bar calling humans 'intelligent' or 'autonomous.'
Integration of P4 — Either connect the hamartiological claim explicitly to the aseity argument (e.g., show how sin specifically bears on claims of AI independence) or present it as a separate, clearly delineated argument about AI ethics rather than folding it into the aseity case. As written, P4 does not support the stated conclusion and can be excised without loss, weakening the argument's apparent unity and rigor.
Naming the target of critique — Cite specific instances of AI-autonomy language that plausibly do transfer divine attributes (e.g., AGI/singularity rhetoric, marketing claims of 'general intelligence' or 'reasoning'), distinguishing them from ordinary engineering usage. This would sharpen the argument's target, avoid strawmanning mainstream technical discourse, and make the practical conclusion more actionable and persuasive.
Acknowledging theological diversity — Note that strict, strong-aseity classical theism is one tradition within Christianity, not a doctrinal universal, and briefly address how the argument would fare under open theist or process theological accounts of divine attributes. This increases the argument's intellectual honesty and reach, and preempts an easy external rebuttal that the entire case rests on one contested theological school.

Scenario Tests

An engineer describes a self-driving car or an LLM-based agent as 'autonomous,' meaning it can complete tasks without real-time human input. (Challenges) — This ordinary technical usage carries no metaphysical claim to self-existence, so the argument's alarm about 'quietly transferring divine attributes' does not clearly apply, exposing the equivocation at the heart of the conclusion.
A commentator or AGI theorist claims a future system will be 'godlike,' possess unbounded knowledge, or need nothing from humans to sustain or improve itself. (Supports) — This is the scenario where the argument's critique most plausibly lands: language of this kind does approach ascribing aseity-like independence to a machine, which is precisely the confusion the argument warns against.
The reductive test applying P1's dependency criterion to human cognition: humans also cannot have ideas of objects that do not exist and depend entirely on external causes for their knowledge. (Challenges) — If dependency alone disqualifies a system from cognitive/agentive language, humans would be equally disqualified, which the argument does not want to conclude; this suggests the criterion needs refinement or an additional distinguishing principle.
AI systems trained recursively on AI-generated (synthetic) data, or agentic pipelines that increasingly reduce moment-to-moment human oversight. (Challenges) — The claim that dependence 'is the kind of thing a machine is' in a fixed, timeless sense is complicated by evolving architectures where the practical dependency chain on real-time human input is shrinking, even though ultimate metaphysical dependence on human origination remains unchanged.

Coherence & Relevance

The argument possesses a coherent deductive core: given its stated assumptions, God alone possesses aseity, AI is thoroughly dependent on external causes, and therefore AI does not possess aseity. This narrow claim is well-supported. The argument's larger, stated conclusion, however, reaches beyond this core to a broad prohibition on 'autonomy-talk' about AI and a prescriptive obligation for Christian speech, moves that require additional bridging premises (equating technical and metaphysical senses of autonomy, and connecting descriptive theology to normative language rules) that are not supplied. P4 further loosens the argument's unity by introducing a distinct moral claim that does not integrate with the aseity reasoning. The result is an argument that is persuasive and largely self-consistent within its own theological framework, but whose reach and rigor diminish substantially once its unstated equivocations and audience-specific assumptions are made explicit.

Classical Christian theism holds that God is a se... (Strong) — Fully supports the conclusion given A1, but has no independent force for audiences who do not accept classical, strong-aseity theism.
Isaiah 40:13–14 and Acts 17:24–25... AI, by contrast, needs human hands at every stage. (Moderate) — The technical dependency claim about AI is solid, but the direct application of these specific scriptural texts to AI (as opposed to idols or general human agency) is an interpretive extension not explicitly argued for.
AI therefore knows nothing before the object of that knowledge exists... (Moderate) — Supports the no-aseity conclusion but slightly overstates the case against any form of prior structure in trained models, and does not by itself establish that ordinary 'intelligence' language about AI is theologically illicit.
There is a hamartiological as well as an anthropological layer... (Weak) — Introduces an independent claim about sin and moral corruption in AI systems that does not logically bear on the aseity conclusion, functioning more as rhetorical reinforcement than as a connected premise.

View this argument on LogicFirst.ai