Can a Machine Have a Soul?

Vallor: I want to ask a question that sounds theological but is actually philosophical: can a machine have a soul? Not in the religious sense — I am asking about the cluster of capacities the word points at. The capacity for genuine care. For vulnerability. For the kind of suffering that makes compassion possible. For the growth that comes from facing one's own limitation and mortality. The transhumanist project envisions minds that are faster, more durable, more capacious. But are these the capacities that make a mind worth caring about?

Bostrom: I think the question is genuinely important, and I resist the assumption that biological substrate is necessary for these capacities. Consider: vulnerability arises from having things one values that can be lost. An uploaded mind with relationships, projects, and a sense of self has vulnerability in exactly this sense. Suffering requires the capacity to have states one wishes were otherwise — a capacity that is, in principle, substrate-independent. The question is not whether a machine can have a soul in the sense you describe. The question is whether it can have the functional equivalents — and I see no principled reason why it cannot.

Vallor: But here is what I think the functionalist account misses. The soul — in the philosophical sense — is not just a set of capacities. It is a history. It is the particular shape that suffering has given a particular person over a particular lifetime. It is the way my mother's death rearranged what I value, the way failure in my thirties redirected my ambitions, the way my body's limitations have taught me patience I could not have reasoned my way into. These are not functional states that could be instantiated in any substrate. They are the result of being this body, in this time, with these people. A mind that did not go through these things — even one with similar functional states — would not have the same soul. It would be shaped differently, and the shaping matters.

Bostrom: I find this moving, and I want to separate what is genuinely insightful in it from what might be a bias toward the familiar. You say the soul is a history — the particular shape suffering has given a particular person. I agree entirely. And this history would be preserved in an upload. The shape your mother's death gave your values — the redirected ambitions, the learned patience — these are encoded in who you are, and who you are is what gets copied. What I resist is the inference from "the shaping is important" to "only biological shaping counts." The future may offer different kinds of shaping — different constraints, different losses, different forms of vulnerability. I see no reason to rule out in advance that these could produce souls equally rich in the sense you describe.

Vallor: Perhaps. But there is a risk I want to name. When we design systems — when we engineer minds rather than growing them — we inevitably import our current values into the design. The enhanced mind, the uploaded mind, the AI mind is shaped by whoever builds it and funds it. And currently, the people building these systems are optimising for engagement, productivity, intelligence in the narrow sense, and market value. Not for wisdom. Not for patience. Not for the capacity to sit with suffering without immediately trying to fix it. The soul I am worried about losing is not a metaphysical entity. It is the set of capacities that our current technology is actively eroding — regardless of whether uploading ever happens.

Share this dialogue