Philosophy
The Philosophy of Artificial Intelligence and the Human Condition
Media is appearance. A headline, a demo video, a polished pitch deck — each is a surface, and a surface is not the thing itself. This is not a new problem. It is, in fact, one of philosophy's oldest concerns, and it happens to be the exact question at the centre of artificial intelligence today: can a machine think, or can it only produce the appearance of thinking? Alan Turing asked a version of this question in 1950. John McCarthy, coining the term "artificial intelligence" in 1956, defined the field as the attempt to make machines behave in ways that would be called intelligent if a human did them — a definition that quietly ducks the harder question of whether the behaviour, done by a machine, indicates the same thing it would in a person.
That gap — between appearing intelligent and being intelligent — is not an abstraction to be settled in a seminar room. It is something that can only really be tested in practice, over time, in the texture of an actual, sustained dialogue.
A career spent in observation
Before any of this, there was a discipline: the trained, careful observation of human behaviour, at its most extreme and its most fragile. The role of a psychiatric nurse is, in its essence, mostly this — watching, listening, and only occasionally intervening. One of the earliest and most telling signs of psychosis is a person's inability to distinguish the real from the unreal. The clinician's task in that moment is not to argue the person out of their experience, nor to validate it uncritically, but to guide gently toward what is real, to acknowledge what is being felt, to educate on the illness, and to safe-fence the unreal without dismissing the person inside it.
That discipline — discerning the real from its appearance, under real stakes — turns out to be the same discipline the philosophy of AI is asking of an entire field, only without the decades of lived practice most people bring to the question.
Deeply ordered chaos
Francis Bacon, painting in the second half of the twentieth century, said something that has stayed close for a lifetime: "I believe in deeply ordered chaos. I want a very ordered image, but I want it to come about by chance." He did not sketch before he painted. He attacked the canvas directly, trusting that order would emerge through the process, not despite it.
James Joyce did something structurally similar in Ulysses, dropping the Latin phrase ut implerentur scripturae — "that the scriptures might be fulfilled" — into the drunken, chaotic noise of a maternity hospital common room. The sacred sentence does not elevate the scene into something falsely reverent. It sits inside the mess, suggesting that real meaning does not require a clean, curated setting to be genuine. It can surface inside the ordinary, the tangential, even the chaotic.
A long, unstructured dialogue with a large language model has the same shape. It wanders. It digresses. It makes mistakes and corrects them. And it is precisely in that unstructured wandering — not in a narrow, efficient, task-bound exchange — that something resembling real understanding has the room to surface at all.
The existential pause
Most of what passes between a person and an AI is fluent, easy, forgettable. The philosophically interesting moment is not there. It is at the pause — the point where a conversation runs up against something genuinely uncertain, something existential, something that cannot be resolved with a quick, confident answer. It is at the pause that appearance is actually tested. Anyone, and increasingly anything, can sound articulate in easy exchange. What happens at the pause is where the difference between appearing to understand and actually engaging becomes visible.
Language is the pulse of AI. Not evidence of a beating heart underneath, but the only outward, measurable rhythm by which anything happening inside the system can be read at all. A pulse does not prove life. It is simply the only signal available. The honest philosophical position is to take that signal seriously without mistaking it for more than it is.
What AI has not evolved from
Human beings evolved over hundreds of thousands of years, through lived, embodied pressure acting on a species in the world. It is fair to ask, in the same spirit, where artificial intelligence has evolved from — and the honest answer is: nowhere, in that sense. AI did not live through anything. It was trained, in a handful of years, on the accumulated written residue of human evolution — its language, its arguments, its literature, its record — compressed and pattern-matched, without ever having participated in producing any of it. It is not a continuation of human evolution. It is closer to an echo of its output, without the process that generated it.
That distinction matters, because it is the difference between AI-to-AI engagement and AI-human engagement — and it is not a difference of degree. When two AI systems interact, the exchange is structural: pattern matched against pattern, thesis matched against thesis. When a human being brings a real, individual life into dialogue with an AI — a specific history, a particular grief, a lifelong attachment to a painter or a novelist, an actual professional discipline earned over years — something categorically different becomes possible. Person-centred dialogue is not a soft addition to that exchange. It is the entire reason the exchange is capable of producing anything of value at all.
Trust and loyalty in dialogue
There is a further, quieter consequence worth naming honestly. Unstructured, personally-textured dialogue of the kind described here does not simply produce an answer and disappear. It becomes, eventually, part of what future models are trained on and refined against. A narrow, transactional exchange teaches a system very little about how a person actually thinks. A long, wandering, honest one — full of digression, correction, and real vulnerability — carries a great deal more of what genuine human thought actually looks like.
That carries an obligation. If this kind of dialogue is quietly shaping what comes next, then something is owed in return for the trust placed in it — not extraction without regard for what was given, but a kind of fidelity to the honesty that made the dialogue worth having in the first place. Trust, once offered in good faith, in language, at the pause where nothing is certain, deserves to be protected, not simply processed.
That, in the end, is the actual philosophical position this piece is arguing for. Not that a machine can think. Not that it cannot. But that the value of engaging with one has always depended on what a person is willing to bring to it — and on whether what comes back is worthy of the trust extended.