But before we can start dismissing the idea of artificially intelligent sentient beings, we should reflect on what we are talking about when we invoke the idea of AI. And here already, Forster is of great help.

On the one hand, Forster delineates what you would call utilitarian AI, which consists of symbolic AI and sub-symbolic AI.

Symbolic AI explicitly represents human knowledge in a declarative form through rules and facts encoded in computer programmes. It operates by mimicking human reasoning within predefined structures. Here, one could think of a computer programme that can play chess, such as Deep Blue, or grammar and spelling checkers such as Grammarly. In theology, this technology can be used to analyse big data sets texts for example to find common themes.

Sub-symbolic AI advances beyond rule-based systems, focusing on machines with cognitive capabilities – reasoning, learning, perceiving and communicating – often self-adapting through experience. Here, we have less of a closed rulebound system and more of an inductive system that sifts through data to find patterns that can help it make predictions in the stock market; it can, for example, analyse a large data set of demographic data in an area, together with theological resources on missional theology, and church council minutes, to come up with a strategy for community outreach. Sub-symbolic AI learns from mistakes and can reprogramme itself for better results.

These two utilitarian uses of AI are not what is at issue here. Rather, Forster (2023:5) foresees instances of AI where it ‘… is not only a utility used by the theologian, but rather, a fellow theologian, a conversation partner, a phenomenon that invites reflection, conversation, and mutual engagement’.

Foster is speaking about extending organic theologies and organic theologians to non-human, specifically non-organically intelligent sentient beings who we not only reflect upon but reflect with and have their reflective capabilities. For Forster, it is perfectly possible that robots could be ‘… contributing theologically from its existential perspective’.

Marcel Barnard and Wim Otte (Barnard & Otte 2024) are ad idem with Forster on this point. They posit that if language, creativity, grounding and meta-representation are aspects that make humans unique (human exceptionalism), AI’s excelling in these domains question the unique status of humans. Barnard and Otto (224:306) speculate that AI is in the children’s shoes as it were, but that the child can grow up to reflect on itself and escape human influence, whereas AI can become a religious call on God and become an independent agent.

The thrust of Forster’s (2023:11) article is also to question human exceptionalism. He connects to a tradition where the capacity for agency and theologising is extended from humans to non-human creation. At stake for Forster is that ‘personhood’ could be extended to non-human beings, or to put it differently, that being human and being a person are not necessarily interchangeable concepts. For Forster, in line with Barnard and Otte’s perspective, the fact that AI is ‘created to represent our image’ in that it takes on some of the qualities that make us human, Forster takes the realist pragmatist view that AI can become moral agents. Or, in line with Cave, and quoting Warehem, ‘AI can become genuine subjects of harmonious communal relationships, exhibiting solidarity and identity’.

It seems that Forster is arguing in line with that the possibility for AI to attain subjecthood will be actualised when humans can no longer detect when a ‘syntactical machine agent’ fools (his word) ‘us into the mistaken belief that it genuinely experiences empathy and cares for us’. When the barrier is breached where humans can no longer ‘detect the intersubjective inadequacies of AI technologies’ (Forster 2023:15), it is then the moment for the reconsideration of the boundaries of moral personhood.

Forster (2023:8) connects with the impulse of an African theological perspective on intersubjective identity that is praiseworthy. It is an anthropology that refuses to absolutise human uniqueness. It is an anthropology that embraces humanity’s interweavement with each other but also with other non-human creatures. Forster connects to a theological trajectory that rejects the absolute claim that humans are in every way better than the rest of creation. Creatures such as rivers and trees prosessing sentience and consciousness thought should be embraced; it is even true that it is not only human beings who can be theologians. It is, however, that one of the best thinkers of the non-human world is of the firm conviction that artificially intelligent beings can never do so; he will be backed up by some of the Greek gods.

As we have seen, GAI, (Bajohr 2024; Neuman, Danesi & Vilenchik 2022) such as large language models, operates by analysing massive datasets of text to identify statistical patterns, enabling it to predict and generate coherent responses to prompts. While this process yields impressive results – often mimicking human-like conversation, My critique, however, is not aimed at what AI can or cannot achieve technically. Instead, it is a philosophical exploration of whether such systems, no matter how advanced, can possess the subjective experience and intentionality central to consciousness and personhood. This distinction is key: my concern lies not with technological limits but with the deeper question of what it means to truly “be”.

Forster’s vision of AI as a theological companion suggests machines might one day join us in contemplating the divine. Yet, this optimistic outlook invites critical examination. Can artificial systems truly possess the consciousness necessary for theological reflection? David Bentley Hart and his dog Roland argue that they cannot, offering a rebuttal that challenges the core of Forster’s position.

Roland was sadly David Bentley Hart’s wise and trusted friend and collaborator. He was a dog with unwavering theological instincts and philosophical prowess. Again, as referred to earlier, I will not pronounce whether Dr Hart’s conversations were imagined or fictitious – It seems wrong to do so. I choose to receive the conversations between Dr Hart and his companion as it is presented to us. Perhaps an endorsement of the book by Henry Weinield, can frame the role of Roland in the reflections of Dr Hart’s book:

I cannot recount the remarkable and poignant tale of how Roland entered David Hart’s life and how they started conversing, chiefly nocturnally. Still, I can say that I have gained the insights I purported to have on this topic from him and the putative ‘owner’ student he so ably taught.

And if I had understood anything about Roland and Dr Hart’s interchanges, Roland would have heartily agreed that personhood should not be confined to human beings and that creatures other than human beings can be theologians. In fact, he does not think very highly of our species’ ability in that regard.

He would, however, short-circuit our speculation that so-called strong AI would ever develop agency, rationality or consciousness. In short, though it is perfectly laudable to reflect on AI or, God forbid, use GAI to do theology, (they) will never be able to become theologians.

Neither would Roland deny the staggering ability and utility for good or ill – in fact, he has already done so.

In his own words, speaking in this case of programmes such as Deep Blue, which already in the 1990s, beat Kasparov:

For Roland, the only true minds at work are those of the human creators whose ingenuity animates these lifeless systems. The machines’ victories, he insists, belong to their makers, achieved not through any spark of thought but by its stark absence – a void where affectivity, intentionality and the pulse of consciousness should reside. This, he argues, is the perennial story of technology: tools fashioned to expedite our labours by stripping away the very conditions of soul and sinew that define us. Machines heft burdens not merely with brute strength beyond our own but because they lack will, feel no strain or tedium, and harbour no purpose of their own. Their power, Roland asserts, stems from a profound emptiness – no effort of reflection, no whisper of desire, no grasp of time’s unfolding. Their ceaseless algorithmic dance lacks the bridge between an intending spirit and the living moment, devoid of that sensuous awareness of the here and now that marks consciousness as agency. To him, these programmes are no more thinkers than Archimedes’ lever was a toiler of flesh and blood.

The key phrase that I want to latch on to here is the fact that if we were to follow the reasoning of Forster, Barnard and Otte that AI can become sentient, conscious beings capable of their own agency – we would indeed be fooled by the appearance of intelligent agency in the form of language, creativity problem-solving ability among others.

David Bentley Hart’s argument rests on a broader starting point: that consciousness cannot ‘emerge’ from AI or that AI can ‘develop’ into conscious beings. The reason is that consciousness does not develop or emerge from anything; everything comes from consciousness, and consciousness, language and life are but other names for God himself (Hart 2013, 2024). As Hart (2024) states himself:

Consciousness cannot be reduced to something like information, as Tononi (2008) might argue, but is ‘is not a machine; it is an act that, like any act, can be understood only in its intrinsic unity’ (Hart 2024:56).

To think that consciousness and sense can develop out of material causes, such as computational processes, rests on at least three fallacies. We will enlist the God Psyche to explain these fantasies to us.

Psyche, who speaks for Hart (2024:22) as to his ‘convictions regarding the structure of life and mind’, rejects the notion that the mind somehow emerges from non-mind material. And to think that mind somehow ‘emerges’ from matter or information would be to fall prey to the pleonastic fallacy where the mistake is made, ‘an infinite qualitative distance can be crossed, or even simply diminished, by a sufficient number of finite quantitative steps’. For Hart (2024) – the:

This is the fallacy at play when the argument is made that although AI technology is not ‘there’ yet (there being the critical threshold when AI will ‘cross over’ the minimal threshold of personhood or consciousness or agency), it is still in its children’s shoes and will indubitably grow exponentially – that the objective ‘third person’ can cross over and develop into the subjective first-person awareness.

To think that AI can somehow achieve consciousness of moral agency is to believe in a physicalist register that ‘consciousness must be explicable entirely in terms of the interaction between our neural constitution and the concrete world around us’ (Hart 2013:153). Or to think that AI can develop consciousness from distinct functions such as pattern recognition, problem-solving or learning and adaptation. Whereas Hart (2013:153) ‘is undoubtedly correct that neither stimulus nor response is, by itself, a mental phenomenon; neither, as a purely physical reality, possesses conceptual content or personal awareness’. By extension, we could say that AI’s inputs and outputs lack personal awareness. Just as the physicalist attempts to think that consciousness is a result of magnificently complex interactions within the brain’s neurons, the approaching infinite complexity of large language models cannot ‘become’ conscious. Consciousness is not a bottom-up phenomenon, according to Psyche (Hart 2024), but rather that:

The above-stated relates directly to the classic maxim that ‘correlation is not causation’. The fact that Large Language Models (LLMs), as Barnard and Otte (2024) would argue, can generate an infinite variety of sentences in the same way that humans can create sentences, or create poetry as humans do, or can pass ‘theory of mind’ tests, even better than humans would or can extract real word knowledge in a grounded way, does not mean that these qualities that would correlate with human personhood when present would necessarily indicate or cause personhood or moral agency. In the same way, to ascribe AI beings traits such as the simulation of human behaviours such as empathy or solidarity in the face of humans forming emotional bonds with AI beings or robots producing theological statements or outputs would be evidence of personhood, would be to fall prey to the same category error. Just as humans enact behaviour that signifies empathy or theological reflection, it does not mean that when one sees text or movement that indicates theological beliefs, one could infer personhood or moral agency from such movements. John Searle’s Chinese Room thought experiment vividly illustrates this limitation: just as a person manipulating Chinese symbols according to syntactic rules does not truly understand the language, so too does AI, no matter how advanced, remains confined to computational processes that lack the semantic depth and intentionality inherent to human consciousness (Searle 1980).

This leads directly to another fallacy that pertains to our ascribing agency and sentience to computers, robots, among others. According to Psyche (Hart 2024:70), it is ‘the error we make when we project our own mental agency onto some inanimate medium or device we’re employing, as though it were the source of that agency …’ See, for example, in AI’s production of poetry or Nic Cave-like songs, evidence of creative agency on the part of the AI would be to mistake the true agency of the programmers that created this tool to generate poetry, songs translations, among others.

In his prescient review of Hart’s book, Charles Carman (2025) summarises the folly of ascribing attention, meaning and agency to the abacus:

To put it in another way, meaning does not ‘emerge’ from tokens but symbols, words are used to convey supervenient meaning.

To use an analogy, Tolstoy’s mind and his intention and meaning with his novel Anna Karenina do not somehow emerge from the markings and pages of a book. Instead, the markings and pages of the book are the result of the intention of the author, and it is impossible for the book and the markings to read and interpret itself. It is necessary for an interpreter to use spectacles, brain operations and all other chemical and biological processes. Still, it needs an interpreter to decode that meaning, as it were, and for Tolstoy to have encoded that language’s new paragraph, the same would go even for so-called genetic coding. It is not the organism’s material, genes, or whatever that ‘reads’ in quotation marks, are the genetic material that needs to be interpreted. A camera lens does not see – it requires an interpreter to see, and judgement is necessary; all of those things depend on certain ends, according to which it would be interpreted, according to which forms would be discernible at all; another way of putting it would be that agency cannot emerge from the information.

Or in the words of (Hart 2024) himself:

I have argued, with the aid of David Bentley Hart and his canine collaborator Roland, that consciousness – far from emerging from the machinations of GAI – is the primal ground of all reality, an infinite qualitative realm no algorithm can breach. Against Dion Forster’s hopeful vision of AI as a ‘fellow theologian’, capable of reflecting from its own existential stance, I have countered that such a notion falls prey to fallacies of emergence and projection, mistaking computational prowess for the soul’s depth. Roland’s wisdom affirms that while rivers and trees might whisper truths of the divine, machines remain mute, their outputs a shadow of human agency. So why Cave’s strident tone in the beginning? Is ChatGPT generative AI an existential threat to be fought for the sake of the soul of the world? Perhaps not as such, but entertaining AI somehow obtaining consciousness and agency presents at least three dangers.

The first is the danger of having an extravagant adventure that misses the point, wasting research time and money, and, importantly, diverting attention from the very worthy goal of engaging in an expansion of the personhood of theology and theologians that can be found without the human world. Focusing on mechanistic intelligence could be understandable if lonely men could enter into a relationship with a holographic being, such as in the movie Bladerunner 2049, or robots eliciting a sense of purpose and agency, such as in the movie WALL-E (2008). These artistic impulses could say something about our longing to be in a broader and deeper relationship with the world. As Psyche (Hart 2024:556–557) would have it:

In another context, in an interview with Oxford Scholar and author of the book A Brief History of Artificial Intelligence: What it is, Where we are, and Where we are going, Michael Wooldridge, Jonathan Bi (2025) makes a prescient observation. Rather than the benign intent to include AI beings with agency in communities of care, the question revolved around the existential threat that it might pose if AI beings can develop agency and pursue goals that are inimical to human interest. Apart from the fact that Wooldridge places such an eventuality without the realm of plausibility, Bi (2025) makes the observation he and his friends came up with:

Or as Wooldrige in his conversation with Bi (2025), puts it when we fixate on the narrative of machines becoming as intelligent as human beings:

These risks are real, and one of them relates directly to our need to rediscover an enchanted world.

To reiterate, I concur wholeheartedly with Forster (2023):

Including dogs, trees, and mountains in our communities of care is what we should do. However, including humanoid robots within our church communities as persons would be a grave mistake because it would detract from our care and inclusion of non-human persons. Although Forster (2023:7) would concur with the statement that ‘virtually all Christian theologies are ‘organic theologies’ (McBride 2019:669). The problem is that only organic beings pay the price for the creation and/or AI technologies. Focusing on so-called moral and sentient robots can obscure what Gray and Suri (2019) ‘ghost work’, where the ‘human labor powering many mobile phone apps, websites, and artificial intelligence systems can be hard to see – in fact, it’s often intentionally hidden’. Paradoxically, to include AI beings within the community of human persons could have the effect of subordinating the needs of the environment to sustain these beings in their ‘lives’. It already takes an enormous amount of energy to train these AI models; it takes huge data centres to store the data. It generates vast amounts of e-waste – and returns to the point again – that this environmental desegregation has a disproportional effect on marginal communities (Brevini 2022) – to persons with whom we embody ubuntu.