Mind Uploading

← Encyclopedia

A worm with three hundred and two neurons has been completely mapped for nearly forty years, every cell and every connection written down. No one has yet made a copy of it that behaves like a worm. The nematode Caenorhabditis elegans is the simplest nervous system science possesses in full — 302 neurons, 959 cells in the whole animal — and OpenWorm, the open-source project launched in 2011 to grow that wiring diagram into a living digital simulation, was by its coordinator’s own 2015 estimate “only 20 to 30 percent of the way towards where we need to get.” The map exists; the mind, even this faint flicker of one, does not follow from it for free. That gap is the whole story of mind uploading in a single small fact.

Mind uploading is the proposal that a brain could be scanned in enough detail to reconstruct, as software, the person it carried — the self lifted out of its tissue and run on a computer, intact. The field’s foundational document, the 2008 “Whole Brain Emulation: A Roadmap” by Anders Sandberg and Nick Bostrom of Oxford’s Future of Humanity Institute, gives the canonical statement: take a particular brain, scan its structure in detail, and build a software model so faithful that, run on the right hardware, it behaves in essentially the same way as the original. The roadmap insists on a borrowed distinction. A simulation mimics the outward results; an emulation mimics the internal causal dynamics, the machinery itself. What uploading wants is emulation — not a program that acts like the person, but the person’s own processes reinstantiated in a new medium.

Everything stands on a single load-bearing premise: substrate independence. The thesis is that mind is a kind of information, an organization of process rather than a particular stuff, so that what runs on neurons could in principle run on silicon. Philosophy calls the family this belongs to functionalism — a mental state defined not by what it is made of but by what it does — with its companion thesis, multiple realizability, that the same state could be carried by neurons, by circuits, by anything occupying the right role. That is the license for moving a mind to a new home. It may be true. It is not established. The roadmap is candid that its deepest technical bet is what it calls scale separation: the assumption that somewhere between the atomic and the macroscopic lies a level of detail below which finer structure no longer matters to function, so the brain can be modeled without simulating every molecule. If no such cutoff exists high enough, the authors note plainly, emulation “would be severely limited or infeasible.” Whether the brain has it, and where, is an open empirical question.

What science has actually accomplished is the mapping, not the running, and the figures form a ladder whose rungs are very far apart. Past the worm, the largest complete connectome of a whole brain belongs to the adult fruit fly: the FlyWire project, published in Nature in October 2024, traced 139,255 neurons and more than fifty million synaptic connections in Drosophila, an animal that walks and sees. In April 2025 the MICrONS consortium reconstructed one cubic millimeter of mouse visual cortex — about 84,000 neurons, over five hundred million synapses — which it noted is only about one one-thousandth of a whole mouse connectome. The frontier of human work is the H01 fragment, one cubic millimeter of human temporal cortex in Science in 2024: some 57,000 cells, 150 million synapses, on the order of a millionth of a brain by volume. Against all of this stands the human brain’s roughly 86 billion neurons — and the distance from a complete fly, or a speck of human cortex, to a whole human brain is not more of the same effort but many orders of magnitude in imaging, storage, and compute.

And a connectome is not an emulation. A wiring diagram is a frozen photograph of structure; running a mind requires the dynamics that structure carries — synaptic strengths and signs, ion-channel states, neuromodulation, the constant remodeling by which the brain rewrites itself over weeks. The Columbia neuroscientist Kenneth Miller has argued that reconstructing structure is “far from sufficient” for this reason, guessing genuine duplication might lie several hundred years off — a pessimism brain-preservation advocates dispute, which is the point: the timeline is contested, not settled, and the loudest dates come from people invested in the outcome. Hans Moravec, whose 1988 Mind Children made the idea famous, expected hardware matching the brain in cheap machines in the 2020s; Ray Kurzweil’s The Singularity Is Near fixed 2045 for the human–machine merger and a nonbiological immortality through uploading — but only, he conceded, if consciousness can be uploaded at all, which no objective test can confirm. The roadmap, more cautious than its advocates, refuses to name a year. No mind has ever been emulated, and as a working technology mind uploading remains almost entirely theoretical.

Suppose, though, that every engineering wall fell — a scan captured enough, the model ran, and the thing on the computer reported a full inner life and passed every test. A harder question would still be waiting, and it is not engineering at all. Is the upload the person, or a copy? Consider a molecule-for-molecule duplicate: qualitatively identical — same memories, same character, same conscious states — and yet, as David Chalmers puts it in his 2014 analysis, not numerically identical. It is not the original, and if the original is destroyed while the duplicate is preserved, the original has died. Personal identity, he concludes, is not an organizational invariant; reproducing the pattern does not by itself reproduce the person. A non-destructive scan makes this plainest: leave the original intact and run the copy, and there are now two beings with equal claim to the name. On this reading uploading is not survival but duplication.

Philosophy has worked this ground for half a century, and the views sort cleanly. The psychological-continuity tradition, descending from Locke through Derek Parfit, holds that a person persists as whatever future being inherits their memories and mental life — and so leans toward calling the upload the same person, if uploads are conscious at all. Animalism, the biological view, makes a person an organism whose persistence is the life of the body, and calls uploading a death with a copy for a survivor. Parfit’s own move is the most radical, and for the uploader the most consoling. In the branching cases — split a brain, transplant each half, and two people walk away each continuous with the one who lay down — strict identity breaks, since one thing cannot be two. Identity, Parfit argued, is therefore not what matters; what matters in survival is what he called Relation R, psychological continuity and connectedness, which can hold to a duplicate with no fact of identity at all. One should care, on this view, that someone carrying one’s mental life exists, whether or not she is strictly oneself. It is the strongest case for the upload — and a revision of what survival has always meant, not a proof that uploading delivers it.

The dilemma is ancient under the new vocabulary: it is Parfit’s teletransportation paradox — scan a traveler, destroy the body, build a replica at the destination, and no one can say whether the person traveled or died and was copied — moved from physical to functional continuity. Which is why the method matters. Destructive copying — scan the brain, in the act destroying it, then switch on the model — is where the duplication worry bites hardest. Gradual replacement feels, to many, like genuine survival: the idea, traced to Moravec and made rigorous by Chalmers under the name nanotransfer, is that tiny devices replace neurons one at a time, each doing exactly what the neuron it supplants did, until no tissue remains and the mind runs on the new substrate, the system never noticing the break. Chalmers’s argument from fading qualia presses the case: as functionally identical parts swap in, consciousness cannot plausibly switch off at one neuron, nor fade into a being that behaves normally while half its experience has drained away — most likely it rides through unbroken. But this argues only that the upload is conscious. Whether it is the same person is a further step, and the gradualist’s “this is still me” is what the biological theorist denies. Chalmers himself, sympathetic to the psychological view, is genuinely unsure about destructive uploading and says he would hesitate to undergo it.

The objections to substrate independence are serious and unsettled. John Searle’s biological naturalism holds that consciousness is a concrete biological phenomenon, tied to the causal powers of brain matter the way digestion is tied to a gut, so that a flawless functional emulation could be a permanent zombie — perfect in behavior, dark within. The hard problem of consciousness frames the worry generally: nothing in a functional description says why performing the function should be felt, which leaves room for a system that does everything a mind does and experiences nothing, with no objective test to tell the difference — the doubt Ned Block’s China Brain dramatizes, wiring a whole population into a brain’s functional organization and finding the intuition refuse it any inner life. None of these refute uploading; Chalmers’s own fading-qualia argument pushes the other way, leaving a standoff rather than a verdict.

Set against the Hermetic and Platonic inheritance this archive keeps, the proposal takes a familiar shape. Substrate independence is, in effect, a secular doctrine of the separable soul: the self as the information state of the brain, an essence detachable from its body. The traditions documented here held, each in its own vocabulary, that something of the person outlasts the flesh; the transhumanist holds the same shape with the theology removed — pattern in place of pneuma, a server in place of the heavens. And the question trailing uploading from its first statement is the one that trailed the doctrine of the soul for as long as anyone held it: whether the thing that survives is the person who died, or only something that remembers being them. Yet the two are not the same doctrine: the old one rested on a soul the new one declines, and the new one rests on an engineering nobody has performed. The worm is still mostly unrun, and the deepest doubt about uploading is one no scan could ever settle.

→ Related: Transhumanism · Cryonics · Immortality · Hard Problem Of Consciousness · Soul