Mind uploading

Mind uploading is usually explained in three verbs.

Scan the brain. Run the scan. Wake up in software.

Each verb hides a research programme.

We cannot scan a living human brain at synaptic resolution. We do not know the minimum biological detail a faithful model would require. We have not built a whole-brain emulation of a human or another complex animal.

Even if those technical problems were solved, one question would remain: would the digital person be you?

Mind uploading may be relevant to the distant future of biostasis. It is not an established route to revival.

A wiring diagram is not a mind

Neurons communicate through large networks of synaptic connections. A connectome maps those connections.

Building one requires imaging tissue at very high resolution, reconstructing cells across enormous datasets and correcting the mistakes made by automated tracing.

The progress is real.

In 2024, the FlyWire Consortium published a whole-brain wiring diagram of an adult fruit fly. It contained 139,255 reconstructed neurons and 54.5 million chemical synapses between those neurons.

That is a remarkable result. The authors estimated that proofreading the reconstruction required about 33 person-years of manual work.

It did not create a digital fly.

The 2025 MICrONS dataset went further in a different direction. Researchers combined calcium imaging from around 75,000 neurons with an electron-microscopy reconstruction containing more than 200,000 cells and about half a billion synapses.

The volume was roughly one cubic millimetre of visual cortex from a single mouse.

A human brain contains tens of billions of neurons and an immense number of synapses distributed across many specialized regions. Scaling is not just a matter of buying more storage.

Axons cross long distances. Cells are densely packed and easy to trace incorrectly. A human dataset would also need to preserve and reconstruct the particular features that distinguish one person from another, not merely a species-typical circuit.

MICrONS is valuable because it links anatomy to observed activity. It still covers only part of one brain, and many reconstructed processes require proofreading.

These projects show that detailed structure and function can be mapped at increasing scales.

They also show how much work separates a map from a working organism.

We do not know which details are essential

A connectome records which neurons connect. A functioning brain also depends on how those connections behave.

Synaptic strength matters. So do receptor distributions, ion channels, neuromodulators, gene expression, glial cells and the changing chemical environment around neurons.

The body supplies hormones, sensory input and internal signals that influence thought and emotion.

A future emulation might reproduce some of these variables directly and infer others from preserved structure. Some may turn out to be unnecessary.

At present, science does not know the minimum sufficient description for reconstructing one particular human mind.

That uncertainty supports caution in both directions.

It is too confident to say that preserving the connectome must preserve everything important. It is also too confident to say that every molecule and momentary electrical state must be captured.

Brains remain recognizably continuous through sleep, anaesthesia, metabolic turnover and ordinary changes in activity. Personal identity does not appear to require every physical component to stay fixed.

But that observation does not tell us exactly which information can be lost after ischemia and cryopreservation.

The prudent objective is broad preservation: cellular architecture, synaptic organization and as much relevant ultrastructure as current procedures can retain.

That is why CT, physiological case data and electron microscopy answer different questions in biostasis quality checks.

A model would still have to work

Suppose a future laboratory acquires a detailed map of a preserved brain.

The map is data. An emulation needs rules that make the data behave.

The model must reproduce how neurons integrate signals, fire, adapt and change their connections. It must operate across many time and distance scales without small errors gradually creating a different mind.

Then somebody has to test it.

Validation is easier when the biological system can perform the same task for comparison. A preserved patient cannot complete a test before the scan.

Writing, video, medical records and testimony could check broad memories and behaviour. They could not verify every private memory or decide whether an unfamiliar response reflected error, adaptation or an ordinary part of the person.

The body creates another design choice.

A digital mind might use a simulated body, a robotic body or an interface unlike present human life. Each option changes sensation, movement and social interaction.

“Scan and run” skips acquisition, reconstruction, modelling, validation and embodiment.

Naming those missing stages does not prove they are impossible. It prevents a slogan from being mistaken for a plan.

Would the upload be you?

Now imagine the engineering problem is solved.

A digital person wakes with your memories, habits, voice and attachments. They believe they are you.

Is their belief enough?

Psychological-continuity views connect survival to the right continuation of memory, character and intention. Biological views connect persistence to the continued organism or original physical brain.

These are serious competing accounts of personal identity, summarized by the Stanford Encyclopedia of Philosophy.

The copying problem makes the disagreement concrete.

It also separates functional success from personal survival. A model could speak convincingly, remember your childhood and continue your work while leaving open whether your own stream of experience continued.

Engineering evidence may establish the first claim long before philosophy or personal intuition settles the second.

If a non-destructive scan creates an upload while the biological person remains alive, both cannot be numerically identical to the one earlier individual in the ordinary one-to-one sense.

Destroying the original during scanning removes the visible duplicate. It does not demonstrate that identity transferred rather than one person ending and another beginning.

There may never be an experiment that settles the metaphysics.

That does not make a digital continuation worthless. Someone may care deeply about the continuation of their projects, relationships and values even if numerical identity remains uncertain.

Another person may care specifically about being the future subject who opens their eyes.

They are not evaluating the same outcome.

Preserve options before choosing a route

Biological repair and mind uploading should not be treated as interchangeable.

Biological repair aims to restore function in the original physical brain. Uploading aims to reconstruct relevant information in another substrate.

The first faces enormous problems in rewarming, toxicity, repair and medical recovery. The second adds destructive high-resolution mapping, computational emulation and unresolved identity.

A preservation procedure should not assume that only one route will matter.

Keeping more structure intact gives future researchers more choices. Future evidence can show that some preserved variables were unnecessary. It cannot recover information destroyed because present theory dismissed it too early.

The possible recovery routes are compared in how we might achieve revival.

For now, mind uploading remains informed speculation.

Connectomics is advancing. Neural models are improving. Neither development shows that a whole human mind can be extracted, emulated or transferred.

Tomorrow.bio’s informed-consent disclosure states that reanimation may never become possible.

Uploading is not an exception hidden inside that warning.

Further reading

• What physical information may support identity
• What damage could mean for a recovered person
• The barriers that remain before revival