Lecture 2-2: Does Machine Intelligence Threaten the Image?

Jeremiah 9:23-24 (World English Bible):

Romans 3:27 (WEB):

Neither text was written with computers in mind, obviously. What they were written for is more relevant, not less: they name wisdom — cognitive excellence itself — as exactly the kind of thing a person is tempted to "glory in," to treat as the ground of their standing. Jeremiah's oracle assumes wise people will be tempted to boast in their wisdom the same way strong people boast in strength and rich people boast in wealth. Paul's argument in Romans, aimed at a different target (works of the law), makes the same structural point: whatever grounds a person's standing before God, if it is achievement rather than gift, becomes something to boast in, and Paul says that boasting is excluded categorically, not just discouraged.

Restating the question precisely

It is easy to ask "does AI threaten the image of God?" as if it were one question with one answer. Lecture 2-1 already shows why it is not: whether AI is a threat to the image depends entirely on which doctrine of the image is being assumed. This lecture works through each of the three in turn.

The substantive view, under real pressure

If the image is what Aquinas said it was — a capacity for reason and will, an intellectual nature that sets humans apart from the rest of the created order — then the substantive view has staked the image's distinctiveness on exactly the territory where a large language model now competes. An LLM cannot love, want, or understand in the sense Aquinas meant (chapter 1 already established that an LLM predicts the statistically likely next token given context, with no model of truth, only of plausibility — a claim developed further in chapter 3). But it can now produce, on demand, competent essays, working code, passable legal summaries, and fluent argument in multiple languages — outputs that, a generation ago, most people would have accepted as sufficient evidence of "reasoning" if produced by a human stranger whose insides could not be inspected.

The tension for the substantive view is this: if human distinctiveness is grounded in a cognitive capacity, and a machine can now produce the outputs that capacity was thought to guarantee, defenders of the substantive view face a choice. They can retreat to a stronger claim about what is actually happening inside a human mind — understanding, intentionality, genuinely grasping meaning, as opposed to statistical correlation — which is a real and defensible move, but one that requires the more technical argument chapter 3 makes in full (via Searle's Chinese Room and the philosophy of mind), not simply asserting that humans are different because Scripture says so. Or they can hold their position by relocating the substantive image to something AI genuinely cannot replicate — will, moral responsibility, a soul in the fuller theological sense — which is a coherent move but concedes that "reason" alone, as originally specified, was not quite the right marker after all. Either way, the substantive view has work to do that it did not have to do fifty years ago. This lecture does not resolve which retreat is correct; it only shows that a retreat is now required, which is not nothing.

The functional view, largely untouched

Return to Middleton's reading: the image is not a capacity housed in a person but a vocation — the appointed office of ruling the earth as God's representative, extended by Genesis 1:26-28 to the whole human race. Ask directly: has a large language model been appointed to that office? Has it been commissioned by God, entrusted with dominion, placed in the position ancient Near Eastern kings occupied as the visible representative of an invisible sovereign? Plainly not, and nothing about a model's fluency changes that. Office is not earned by performance; it is conferred. A skilled advisor who drafts a king's speeches with great competence has not thereby become king. On the functional reading, AI simply is not a candidate for the image at all, however capable it becomes at any task, because the image was never a performance to be matched — it is an office to be held, and holding an office requires being addressed and commissioned, which is precisely what does not happen to a language model.

The relational view, largely untouched

Return to Barth: the image consists in existing-in-relation, first to God, then to other humans — an I addressed by a Thou, capable of covenant, capable of love and betrayal and forgiveness. Has a language model entered relationship with God? It has not been addressed by God as a covenant partner, has not sinned against God in the sense Scripture means by sin, has no history with God that could be reconciled. Has it entered relationship with a human, in Barth's sense? A user can feel they are relating to a chatbot — chapter 12 examines this feeling directly and takes it seriously as a real pastoral phenomenon — but feeling addressed by something is not the same as that something existing in the kind of mutual, responsible relation Barth has in view. On the relational reading, too, the image is not a territory a machine enters by getting better at generating text, because relationship is not a text-generation task.

Herzfeld: relationality as the safer ground

Noreen Herzfeld, in In Our Image: Artificial Intelligence and the Human Spirit (2002), makes an argument directly relevant here, and this lecture treats it as the strongest theological statement of the position that the functional and relational views, not the substantive one, should be regarded as the safer theological ground for the image in an age of intelligent machines. Herzfeld's concern is not primarily that AI will "become" human or acquire the image by accident. It is that a Christian anthropology which has quietly let intelligence do the work the image was supposed to do will have no stable answer when a machine matches or exceeds human intelligence on measurable tasks — the anxious popular question "are we still special?" is, on Herzfeld's diagnosis, evidence that the doctrine of the image had already drifted toward the substantive account in an unreflective, culturally inherited way, without most people ever having decided that on purpose. Relationality, Herzfeld argues, was always the more secure ground, because it does not compete with a machine on a scale a machine could ever climb.

This lecture endorses Herzfeld's diagnosis of the anxiety — that the appearance of a threat to the image is itself evidence of an underlying substantive assumption that most people never examined — without thereby declaring the substantive tradition simply wrong. Aquinas's tradition has its own resources (see above) to meet this challenge; the point is that it has a challenge to meet, where the other two traditions largely do not.

The prior, deeper problem: boasting was already excluded

Here the lecture presses past the AI question to something more basic. Even if AI had never been invented, grounding human worth, standing, or dignity in any competitive capacity — intelligence, strength, wealth — was already a theologically unstable foundation, and Scripture said so long before anyone imagined a machine that could write an essay. Jeremiah 9:23-24 names wisdom explicitly, alongside might and riches, as something a person is tempted to "glory in" and explicitly forbids that glorying, redirecting it instead toward knowing God, who acts in "loving kindness, justice, and righteousness." Paul's argument in Romans 3:27, aimed at a different debate (whether standing before God comes through law-keeping or through faith), makes a structurally identical move: boasting grounded in one's own achievement is "excluded," not merely relativized.

The implication is not that intelligence is bad, or that excelling at reasoning is wrong to enjoy. It is that a doctrine of human worth that quietly rests its full weight on being the smartest kind of thing around was never theologically sound to begin with — a truth obscured for most of human history only because no serious rival to human cognitive performance existed to expose it. AI has not introduced a new theological problem. It has stress-tested an old foundation that should not have been carrying that load in the first place, in exactly the way an engineer discovers a bridge's real weight limit only when a heavier truck finally attempts the crossing.

Where this leaves the chapter

2-1 named three traditions without ranking them. This lecture has shown that they are not equally affected by the arrival of capable machines: the substantive view must do new defensive work; the functional and relational views largely need not. It has also shown, via Jeremiah and Romans, that the deeper theological error was never really about AI — it was about mistaking a competitive capacity for the ground of worth, a mistake available long before any machine existed to expose it. 2-3 will not adjudicate the philosophical dispute between the three traditions from 2-1 — that is explicitly left open, as the "Contested points" note for this chapter states. What 2-3 will do is offer a constructive, pastorally load-bearing account of the image that does not depend on winning that argument: vocation and belovedness, conferred rather than achieved.