What mind uploading means

Mind uploading proposes to reconstruct a particular mind — not a mind in general — by mapping a brain in enough detail to simulate it. That specificity is what separates it from Part 3: the target is a person, and success would have to be judged against that person. The proposal decomposes into three problems that are usually run together and should not be: mapping the brain, simulating what was mapped, and deciding whether the result is the same individual. The first is an imaging problem with real momentum. The second is a modelling problem in much worse shape. The third is not an engineering problem at all.

Where the science stands

Established Connectomics is genuinely succeeding, and at a pace that surprised the field. Nature Methods named electron-microscopy connectomics its Method of the Year for 2025. Complete synaptic-resolution connectomes now exist for C. elegans — roughly ten whole-nervous-system datasets from individual specimens — and for the adult fruit fly, with fully proofread reconstructions of the male central nervous system and the female brain from the FlyWire consortium.

Established In mammals the state of the art is a single cubic millimetre. The MICrONS consortium densely reconstructed that volume of mouse visual cortex after nine years of work: approximately 120,000 neurons and 523 million synapses, from over a petabyte of imaging data, combining electron microscopy with in vivo calcium imaging of the same tissue. A larval zebrafish whole-brain connectome, paired with activity recordings from 70,000 neurons in the same animal, has been expected to complete around 2026.

Established The scale gap is the number that matters and it is rarely stated. A human brain is on the order of a million times the volume of that cubic millimetre, with roughly 86 billion neurons. Imaging runs at 10–20 nanometre resolution and produces about a petabyte per cubic millimetre — and alignment, tracing and proofreading typically cost far more than the imaging itself. This is a six-order-of-magnitude problem in volume before any of the harder questions are reached.

Frontier Those harder questions concern what a connectome omits. It is a wiring diagram: it records that two neurons connect, not the strength of that connection, its neuromodulatory context, its molecular state, glial contributions, or the dynamics that make the circuit compute anything. Simulation work reflects this — in C. elegans, where the connectome has been available for decades, closed-loop models integrating neural dynamics, body mechanics and environment reproduce basic behaviours, and full emulation of a 302-neuron nervous system is still not claimed.

Frontier The field is also institutionally fragile. Large connectomics efforts depend on sustained public funding, and a prominent Harvard-led project targeting ten cubic millimetres of mouse hippocampus was affected by shifting funding priorities in early 2025, with technical work continuing outside that support. Capability here is not a smooth extrapolation.

Speculative Whether a sufficiently detailed emulation would be conscious depends entirely on the programme's shared premise. Handwave And whether it would be you is a question no imaging resolution answers. A non-destructive scan produces a copy while the original continues, which makes the continuity claim plainly false in that case; a destructive scan does not obviously improve matters, it only removes the person who could disagree.

The binding bottleneck

Frontier The constraint is not imaging throughput, which is improving, but the gap between structure and function. Even a complete human connectome would not specify a running mind, because the parameters that make a circuit compute are not visible in its wiring. Proofreading remains substantially manual and dominates cost. And there is no validation target: no agreed test would confirm that an emulation of a particular person is accurate, which means the field could succeed technically and be unable to demonstrate it.

Ethics and governance

Frontier Destructive scanning of a brain is, at minimum, an end-of-life procedure, and its relationship to assisted dying law is unaddressed everywhere. Frontier Preservation services are already sold on the premise that future uploading may be possible — a commercial claim resting on an unproven premise and an unbuilt technology, which is a consumer-protection question now rather than later; see brain preservation and cryonics. Speculative If emulation ever worked, copies raise questions of consent, ownership and identity that no legal system anticipates. Established Connectomics itself, meanwhile, is ordinary neuroscience with ordinary tissue-donation ethics, and conflating the two does the science no favours.

Timelines

  • 10 yr: Established whole-brain connectomes for small vertebrates; larger mammalian volumes; automated proofreading improves.
  • 25 yr: Frontier a whole mouse brain connectome is plausible; functional models of small nervous systems mature.
  • 50 yr: Speculative human-scale mapping conceivable; simulation and validation remain unsolved.
  • 100 / 250+ yr: Speculative emulation of a particular human mind, if the premise holds and the structure-function gap closes. Neither is currently in view.