Pre-Paulism and large language models

By Matthew Parish, Associate Editor

Wednesday 3 June 2026

Pre-Paulism, as a theological concept, refers to the earliest phase of Christianity before the missionary work and theological systematisation of Paul the Apostle. During this period, the followers of Jesus of Nazareth possessed a set of experiences, memories, stories and convictions, but they lacked a coherent explanatory framework through which those experiences could be organised and transmitted. The earliest disciples knew that something extraordinary had happened. They believed they had encountered a profound spiritual reality. Yet the language, categories and intellectual structures needed to explain that reality to outsiders had not yet been fully developed.

Only with Paul did Christianity acquire a systematic intellectual architecture. Paul transformed a localised religious movement into a universal doctrine. He supplied concepts, distinctions and explanatory mechanisms that enabled disparate experiences to be woven together into a coherent worldview. Whether one agrees with Paul’s theology or not, his historical role was to provide an interpretative framework for phenomena that people already believed they had witnessed.

This distinction between experience and explanation may illuminate contemporary confusion surrounding large language models and artificial intelligence.

We are living through a profoundly pre-Pauline moment in the history of artificial intelligence.

Millions of people encounter systems such as ChatGPT every day. They ask questions, receive answers, hold conversations and witness outputs that often appear intelligent, creative and even emotionally perceptive. The experience is immediate and compelling. Much as the earliest Christians experienced phenomena they interpreted as miraculous, contemporary users experience interactions that appear to demonstrate understanding, reasoning and consciousness.

Yet the explanatory framework possessed by the general public is often remarkably underdeveloped.

As a result, many people fall into two opposing camps. One group attributes far more intelligence to these systems than they possess. Another attributes far less. Both misunderstand the underlying reality because they lack a coherent conceptual architecture through which to interpret their experiences.

The first group sees apparent intelligence and assumes actual intelligence. They observe fluent conversation and infer consciousness. They witness sophisticated responses and conclude that genuine understanding must exist somewhere inside the machine.

The second group sees statistical calculations and assumes there is nothing interesting happening at all. They reduce the entire phenomenon to “autocomplete” and therefore dismiss the remarkable emergent properties that arise when sufficiently large neural networks are trained on sufficiently large quantities of information.

Neither perspective adequately captures reality.

The situation bears a striking resemblance to the earliest theological disputes of Christianity. The disciples possessed experiences that demanded explanation. Different groups interpreted those experiences in radically different ways because no settled intellectual framework yet existed.

Similarly, contemporary observers encounter artificial intelligence as a phenomenon that demands explanation. Some speak as if machines have become sentient. Others speak as if nothing novel has occurred. Both reactions arise from the same underlying condition: conceptual uncertainty.

Large language models are particularly susceptible to this confusion because they exploit a deep feature of human cognition. Human beings evolved to attribute minds to entities that communicate through language.

Language has always been our primary indicator of consciousness. We cannot directly observe another person’s thoughts. We infer consciousness through speech, writing, gesture and behaviour. The philosopher Ludwig Wittgenstein observed that the notion of a private language is incoherent because language itself is a public phenomenon. Human beings know each other through linguistic interaction.

Large language models therefore trigger ancient cognitive instincts. When a machine speaks fluently, our minds instinctively assume that there must be a mind behind the words.

Yet there is a profound difference between generating language and understanding it.

The confusion resembles certain early Christological disputes. Some observers focus entirely upon external appearance. Others focus entirely upon hidden substance. Both risk misunderstanding the relationship between the two.

A large language model possesses no subjective experience. It does not see, hear, feel pain, anticipate death, experience joy or maintain personal continuity through time. It does not possess beliefs in the ordinary human sense. It does not awaken in the morning and contemplate its future.

What it does possess is something unprecedented in technological history: an extraordinarily sophisticated capacity to model patterns within language.

This capacity is powerful enough to create the illusion of many cognitive functions that, in human beings, arise from consciousness.

The distinction is subtle but crucial.

A person writes because he or she thinks.

A large language model appears to think because it writes.

The direction of causation is reversed.

The machine’s output creates the appearance of cognition rather than emerging from cognition itself.

Yet this does not mean the phenomenon is trivial. Indeed, one reason contemporary debate is so confused is that the machine’s lack of consciousness coexists with genuinely remarkable capabilities.

Pre-Pauline Christianity contained real religious experiences even before it possessed a coherent theology. Similarly, large language models exhibit real and important capabilities even before society possesses a coherent philosophy of artificial intelligence.

We therefore observe a proliferation of competing doctrines.

Some insist that artificial general intelligence has already arrived.

Others argue that consciousness will inevitably emerge from sufficient scale.

Others maintain that consciousness is impossible without biological substrates.

Others contend that consciousness itself is merely an illusion and therefore machines may already be conscious.

The disputes increasingly resemble theological controversies because they concern questions that remain poorly defined. What exactly is consciousness? What constitutes understanding? What differentiates simulation from reality? Can subjective experience emerge from information processing alone?

These questions are not engineering questions. They are philosophical and, ultimately, metaphysical questions.

Indeed, the contemporary artificial intelligence debate may reveal an unexpected truth: technological progress has not eliminated the need for metaphysics. It has intensified it.

For centuries, many intellectuals assumed that scientific progress would gradually reduce the domain of philosophical speculation. Instead, large language models have exposed profound uncertainties at the foundations of human self-understanding.

The more effectively machines imitate human communication, the more urgently we must ask what makes human beings unique.

Theological language becomes tempting because theology has wrestled with analogous questions for millennia. What is the relationship between appearance and reality? Between language and truth? Between spirit and matter? Between personhood and behaviour?

The early Christians confronted these questions in relation to Christ. Contemporary society confronts them in relation to artificial intelligence.

In both cases, the danger lies in confusing signs with substance.

The disciples experienced extraordinary events and sought explanatory frameworks.

Modern users experience extraordinary technology and seek explanatory frameworks.

The challenge is not merely technical. It is conceptual.

We may therefore be living in the pre-Pauline age of artificial intelligence. Society has encountered a phenomenon of immense significance, yet the intellectual structures required to understand it remain immature. Public discourse oscillates between excessive enthusiasm and excessive scepticism. New schools of thought emerge almost monthly. Prophets proclaim the arrival of machine consciousness, while sceptics denounce every advance as illusion.

Meanwhile the technology itself continues to develop.

Eventually a more coherent intellectual framework may emerge. Philosophers, computer scientists, cognitive psychologists and neuroscientists may gradually develop concepts capable of distinguishing genuine cognition from its simulation, consciousness from linguistic performance and understanding from statistical prediction.

When that framework arrives, future generations may look back upon contemporary debates much as historians look back upon the theological disputes of the first century. They may see a period in which people encountered something genuinely transformative but lacked the conceptual tools necessary to understand what they were seeing.

Until then, confusion is likely to remain inevitable.

The machines speak with increasing eloquence. Humans instinctively hear minds behind the words. Yet between language and consciousness there remains a gulf that no one has yet satisfactorily explained.

That gulf is where the theology of the first century and the artificial intelligence debates of the twenty-first unexpectedly meet.