Is is plausible that we could have neuronal maps of human brains without mind uploading being possible?

Question

I am currently trying to find a reason that mind-uploading could be impossible in a sci-fi setting with fairly mature nanotechnology/scanning techniques - to the extent that it would in principle be possible to create a neuronal map of a human brain, perhaps even with the strengths of connections mapped, yet somehow still be unable to run a virtual person.

So, is there any reason in particular why one could have a neuronal map of a brain, yet be unable to effectively "run" it and reproduce the mind that was mapped?

The idea which came to mind thinking about this is that our brains do not only consist of the neuronal connections and weights, but surely also the pattern of neurons which are currently firing - our brains aren't like a simple feed forward neural network where you just feed it any input and it gives you a result. Without knowing exactly what the pattern of firing in a brain is on a short timescale, is it plausible that the map alone won't be enough?

Answer 1

Insufficient Compute

It's entirely reasonable that we would have the technology to take a full snapshot scan of the brain, yet be unable to simulate how it would evolve over time due to the superfine spatial and temporal resolution required combined with the exponential nature of the dependency network. We might be able to simulate small parts of the brain in isolation but a person that does not make.

You could also selectively relax bits of this constraint if it suits your story. Maybe the Big Bad governmental or megacorp entity can simulate a (single!) person if it musters all of its resources. Or maybe simulation is possible but it would be at 1/10, 1/100, 1/1000 speed (you might have a race-against-time scenario where your characters are trying to accomplish something before a simulated entity speaks a single sentence, spilling the beans.) Lots of fun to be had! But simply ruling it out due to the simulation being too complex is quite reasonable.

Answer 2

The connectome is not enough. To copy the mind you also need the state of each neuron.

The brain is hardware. The mind is software. Having a full schematic of the hardware doesn't tell you much about the software that runs on it.

Think of it like this: you have a complete circuit level schematic of the Zilog Z80 processor, with all its 8,500 transistors. Being generous, let's say that you also have a copy of the Spectrum ZX ROM. This does not tell you anything at all about how a game such as Knight Lore or The Way of the Exploding Fist works.

When the computer is turned off, the software is gone forever. You need to reload it from some sort of storage medium.

To make a copy of the software as it is running on the hardware you need to be able to freeze time and then copy the state of each memory cell and of each register. We can do that with computers. We cannot do that with brains.

With a computer, it is relatively easy to dump a copy of the software as it is running; computers are designed for this. With a living brain, not so much. The biggg problems with dumping the mind-software as it is running on the brain-hardware is that the brain does not have any means of retrieving the state of each neuron, and the brain does not have any kind of high-bandwidth output interface to extract the states of the neurons.

Note: data is not hardware

Reading the comments, I have come to realize that many people can't tell the difference between hardware and the data stored on the hardware. Data, even persistent data, is not hardware.

Let's take for example a flash drive. We all know them, and they are vastly easier to understand than the long-term memory subsystem of a human brain.

The flash drive itself is hardware. It is made up of lots and lots of flash memory cells, all the same. Each of them is made up like this:

^{Schematic diagram of a flash memory cell. Drawing by Wikimedia user Cyferz, available on Wikimedia under the Creative Commons Attribution-Share Alike 3.0 Unported license.}

The flash memory cells are connected in a regular structure, very easy to understand. It is trivially easy to take a flash drive and make a map of the memory cells with all the interconnects.

But this does not tell anything at all about the data stored on the flash drive. To get the data one needs to record the electric potential of the floating gate. For the brain, this corresponds to recording the internal state of each neuron; and the question only provides for the map of neurons and interconnects:

"It would in principle be possible to create a neuronal map of a human brain, perhaps even with the strengths of connections mapped."

Long story short, the internal state of each memory cell, whatever the technology or biology of any kind of rewritable memory device, is essential for reconstructing the data which makes the brain or the computer operate the way it operates. Data is not hardware. One cannot retrieve the data just by looking a circuit diagram of the hardware; one needs to actually measure and record the state of the memory cells.

^{For read-only memory devices, such as a CD-ROM for example, data can be represented directly in hardware. Maybe that's how instincts are represented in a biological brain, I wouldn't know.}

Answer 3

Getting a scan of an object is something entirely different than being able to reconstruct it.

The brain operates fast, and is heavily inter-connected. Thus if you don't sample everything at the exactly same time, you will possibly make big errors. Yet you can only look from the outside, and there is the speed of light... you need to take that into accout for that then, too.

Even current technology allows a reasonably detailed analysis of which parts in your brain are operating at what intensity and communicating which whichever other. Without invasive techniques we are not a neuron level yet, but still talking of many thousand neurons, thus we are a very long way from being able to mind-read you, let along use the information obtained in order to clone or "upload" your thoughts to an artificial neural network. However technology will make progress, so it will become increasingly easier to read these data - but that will need HUGE and FAST storage as you need to store the data for each neuron at each time step - and 100 billion neurons with 100 trillion connections with various states they can take, sampling at more than a million samples a second without error... that will remain a challenge for some time I recon (as neurons are not just 1 or 0 - but have many many shades between). Thus you may know where the neurons are, with which other ones they are connected and communicate, given a good spatial scan. But capturing their interaction detailed-enough to reconstruct that, is still another level (think of reconstructing Mona Lisa by looking at the backside of the painting).

See also

• https://www.sciencedirect.com/science/article/pii/S2352711021001795 and references therein.
• https://www.brainaccess.ai/methods-of-brain-activity-measurements/
• https://pmc.ncbi.nlm.nih.gov/articles/PMC3786741/
• https://en.wikipedia.org/wiki/Magnetoencephalography

Answer 4

The human mind is WAY more than just a set of neural pathways in the brain

Science has already proven that that the human body has all sorts of systems all over the body and within the molecular level of each brain cell that contributes to how our minds work. If you can not capture ALL of these variables and try running a synthetic brain simply making assumptions about a person's peripheral nervous system, endocrine system, and NT systems, then you will run into the They Came Back Wrong trope. So, even if the synthetic brain functions and is able to remember things, what you restore will be nothing like the person you restored the synthetic brain from.

Reflexes, Muscle Memory, and Gut Feelings

Significant amounts of data processing is actually done in your spine and other decentralized neural ganglions.

Copying a brain without your whole peripheral nervous system would eliminate your muscle memory; so, even if you uploaded that brain into an android, the level of control it has over its body would be like that of an infant. It could not walk or crawl or write or do any of those rehearsed actions that you've spent your whole life training your body to do. So your artificial person would need to undergo months of physical therapy to regain functional control over their body, and years of it to regain any significant areas of expertise.

Also, that "gut feeling" you get when something feels off is actually caused by a major neural ganglion that is physically located in your abdomen that works sort of like a second much smaller and more primitive brain and is believed to be responsible for certain forms of instinctive intuition. Without it your person may come back with altered executive functioning because they no longer have a healthy sense of fear or guilt tied to their actions.

Hormones and Emotions

Your Endocrine System is a complex system of organs located all over your body that produces hormones: chemicals that stimulate both physical and mental states. This system includes your Pineal Gland, Hypothalamus, Pituitary Gland, Thyroid Gland, Thymus, Pancreas, Adrenal Gland, Kidneys, Gonads, as well as a number of other glands with poorly understood effects on your nervous system. Collectively, these organs are responsible for regulating all sorts of mental processes including learning, attention, motivation, emotions, wake/sleep functions, etc.

While your brain stores what you know, a person's personality is far more motivated by hormones than it is by the structures of your brain; so, trying to run a virtual brain of a person and having to make assumptions about their hormone balance might give you their memories, but none of that person's personality traits.

It only takes a tiny imbalance to accidentally turn a normal person's brain psychopathic, bipolar, ADHD, suicidal, schizophrenic, etc. This means that synthetic brains could very easily represent a credible danger to society akin to robotic zombies unless you also capture their entire hormonal profile, and have a flawless way of simulating it in your synthetic brain.

Even if you can capture their hormone profile in a given moment, it may not represent that person's averages or you models for simulating them might be slightly off; so, best case scenario, you might get an uncanny valley resemblance to the way that the original person acted. So, they might seem fine to strangers, but to friends and family, it would be uncomfortably not the person it thinks it is.

Neurotransmitters and Thought Patterns

This will be the hardest things to copy right because if the cell is not right at the molecular level, how it communicates with adjacent cells may be fundamentally different than the original, even if all the connections themselves are correctly captured.

Just because you know that a neuron connects somewhere does not mean that you know how many neural receptors exist in that connection or understand the release/uptake patterns that may routinely happen at that receptor. Your brain is not made up of wires that just carry a current. Instead they are cells that have to release chemicals that then move between two connections to trigger a response in the adjacent cell. Those chemicals then become depleted over time weakening the effective link if overused and it takes time to reuptake them and get back to peak efficiency. And these number of transmitter/receptor sites may change over time as the cell adapts to workloads the same way that a muscle can become weaker or stronger over time.

Without knowing how much of various neurotransmitters that each cell stockpiles, how much it releases for a given signal strength, and how many receptors for each NT the receiving cell has, and how good that connection is at reuptaking spent chemicals, there is no real way of understanding how that connection is supposed to work. Knowing the size can tell you about what the total strength of a connection is, but not if the site is big because of lots of Dopamine sites or Serotonin sites which is fundamental for knowing under what circumstances the connection activates.

This means that the brain may have all the correct memories, but accesses them at the wrong times. Traumatic memories that only come up when you are scared or in danger might start to emerge every time the robot gets excited. Happy memories might emerge when you are sad.

This is a problem because, the brain is designed to reinforce patterns from remembering things just as much as from experiencing things; so, this could cause your the robot's mind to reshape what is good and what is bad. They may start to react to the wrong stimuli by doing things like getting happy when a loved one dies or angry when someone thanks them for something. Or, your robot might start to seek self destructive or avoid positive experiences to avoid/achieve certain memories.

Lastly, part of what makes people think, act, and learn the way that we do is that we are designed to get tired of one thing and move on to the next thing we should be doing. If you fail to simulate this neurotransmitter release/uptake relationship properly because you have not copied the cell all the way down to the molecular level, then the resulting robot may either be way more obsessive or flippant than the person whose brain they are based off of or get stuck in uncontrollable habits, tics, compulsions, and though patterns that prevent them from functioning in any practical way.

Answer 5

Multiple issues blocking uploads

There are several different issues that make mind 'uploads' impossible at this time.

Different neuron types

There are something like a dozen different neuron types in the brain some unique to humans. we don't know if we know all of them, and certainly the behavior profile of all these are not all known either.

This will likely mean that the scanning sufficient for a standard neuron may not be sufficient in all cases.

Activation weight scanning

Activation weights are an issue on top of knowing where the physical connections are. This requires observation a living active brain with the complication of observation will change behavior.

Brain has many connections

A thing that may be an issue is the shear number of connections. It will take time to make a scan. That could mean that by the time a scan is finished, the map has changed in the first area to be scanned. Thus the map may no longer be valid/correct.

Up load to what?

So a map has been made. Now what?

Creating a system that can run/execute an uploaded mind map is whole other large challenge. Quite possibly a larger challenge then making the scan in the first place.

without a run system how can the brain scan possibly be verified as correct and complete?

Brains are complicated

There are lots of unknowns and unknown unknowns when it comes to the brain. There are many different and difficult problems to be resolved before brain scanning can create a uploadable scan.

Answer 6

Yes, and you don't need to justify it with anything.

No a priori assumption can be science-based because science is about conclusions that follow from data. Therefore there's nothing science-based about the a priori assumptions of naturalistic mechanistic philosophical monism plus substrate independence on which the question and existing answers are all based. You can just deny (or be agnostic about) one or both a priori assumptions: declare that in your setting, it has turned out that no efforts to simulate human minds by simulating brains have ever succeeded, despite apparently flawless simulations. No further justification is needed.

There is a distinction between real physics and speculative physics that creates a distinction between hard and soft sci-fi, but all metaphysics is speculative metaphysics until it actually happens.

Answer 7

The idea which came to mind thinking about this is that our brains do not only consist of the neuronal connections and weights, but surely also the pattern of neurons which are currently firing.

True, but brains can survive a temporary disruption. A blow to the head, an epileptic fit, electroconvulsive therapy, etc. all force neurons to fire when they otherwise wouldn't. That knocks the patient/victim unconscious, if they weren't already. But they recover quickly once the source of the disruption is gone. Nothing permanent is stored in "the pattern of neurons which are currently firing".

--

Still, the brain is certainly more than the pattern of neurons and the strengths of the connections between them. There's a lot we don't know about the internals of a cell, and although we could make a list of all the proteins a neuron can produce we don't know what (if anything) many of them are for.

So let's say that nanotech can't record the position of every single molecule in a cell. I mean, where would it store all that data, any storage device that can hold molecular-level information would need to be made of molecules itself, so it would be many times larger than the thing it's trying to scan and wouldn't fit into a nanobot. One would have to narrow it down to just what's "important" for the mind to function - and we don't know that!

Besides that, your nanobots will surely require power and data connections, produce waste heat, and interact chemically with their surroundings (because that's their job). These things could damage the cell as it's being scanned. Your nanoscience technicians hope that they can strike a balance between (a) identifying and saving enough of the important data that the person actually wakes up in upload and (b) not having the nanobots do so much work that their waste heat cooks the brain before it's finished uploading. But there's no reason to assume that balance should be achievable.

Answer 8

AI software runs on digital software. It typically has separate training and predicting passes. This lets us train a model, then save it, then experiment with changing the parameters, or seeing what injecting signals at each level of the neutral net and seeing how that affects the output. This can be used to determine what the middle levels of the neural net are actually doing.

The brain is continuously growing, training, and predicting all at the same time. We cannot stop and start it, or tweak the parameters to see what is happening. There have been experiments where brain activity maps have been used to reconstruct what people are dreaming about, but this is a long way off being able to capture and reproduce all the fine structure and action within a thinking brain.

I think AI models will carry on doing new extraordinary things until they can imitate almost all of the things that the brain does. But we will not be able to turn a living mind into program data and back again the way we can do with DNA sequences.

Answer 9

Metaphysics 101

Depending on your setting, and the nature of the world that you are building you could simply handwave this away using metaphysics as an explanation.

EG, that scientists have tired for decades to map the human brain digitally, but the best that they could do was to create a repository of memories without consciousness also being transferred.

Like creating a library but not the librarian necessary to operate it

Some scientists held that the problem was down to consciousness being created in the by some subtle structure of the brain itself, which they hadn't been able to replicate yet, while other suggested a more metaphysical reasons: that consciousness existed separately from the physical structures that were being replicated, and that maybe the best people to ask about it were priests, rather than scientists.

AKA, the body is just a vessel, and it needs a spirit to operate.

You could make this explanation as simple or as complicated as you like, and could either have it as being an interesting piece of background flavor for your universe, or complex mechanic. Or just leave it ambiguous.

Answer 10

Due to recent real world progress, I’d say your three options are focusing on being unable to train the brain to an adult state, uncertainty over how the different parts of the brain interact with each other and a recent theory that very small tubules in the brain may have quantum interactions – which the simulation can’t model. Or state all 3 and say no one has figured out which one is causing the problem or how much each one is contributing.

So the real world progress is that scientists managed to model the complete brain of a fruit fly and on feeding it the data that would mean it spotted a fruit, the simulated fruit fly brain went through the previously recorded set of signals to direct its non-existent body towards the non-existent fruit. Yay us.

The main difference between us and a fruit fly (in my opinion) is that the behavior of a fruit fly is hard coded into neural network of the fruit flies brain while human brains have very little card coding on how to do anything. When a baby is born, the people delivering it often need to slap the baby to make it cry – as lots of human babies don’t have the default behaviour of breathing unless they tricked into by making them cry. I bring that up to point out that it takes years of mucking about before human babies even figure out how to move their body parts properly, let alone be able to interpret the world – while fruit flies know from the moment they born how to fly and the art of seeking out fruit. So the first out is the claim that the simulated human brain probably works but without years of the complicated external and internal stimulating inputs of being a baby (recognizing mom or being hungry) the less pre-programmed human mind has no reason to interact with the outside world or engage in any complex behaviors at all. The second option is that while the interactions within each part of the brain are known, it isn’t exactly clear how the different parts of the brain interact with each other. The large left and right lobes of the brain are built on top of a much older lizard brain which in turn connects to the brain stem, which in turn collects input form the length of the spinal cord and the spinal cord gets signals in from complex organs and receptors – and even worse, there is evidence that the organs of the body do brain like processing as well, unrelated to their intended functions and that there may be 2 way communication between the brain and the other organs as opposed to 1 directional input. So while sending a simulated fruit fly compound eye input to a simulated fruit fly brain is achievable, who knows what sort of repeating signal is supposed to be initiated to bounce back and forth between the human brain and the human spleen, and all the exchange nuances that signal goes through while transversing the spine, brain stem and lizard brain in between?

Thirdly, there has been enough evidence recently to suggest that really small parts of the brain may not just be doing normal bio-electric interactions but may have really small tubules capable of quantum interactions – and if that’s happening who knows what those tubules are doing in their internal calculations before they pass their decisions onto neighboring neurons. If there are quantum interactions going on, a normal computer simulation can’t model them. A normal computer could probably doo a crude simulation of their result but only if you could figure out what the tubules were doing before hand.

Those 3 things seem like the best starting points to me. Though I must say, in direct opposition to your original question, that the most profound issues in science are not those situations where we know what the problem is but rather those situations where we don’t know what the problem is. Unless you have something specific in mind, it is far more suspect to say that no one knows why the simulated human mind isn’t working rather stating that it would work if we could just implement this one exact solution. Rather than state that this is why we can’t get a simulated human brain to function , rather say that we don’t know why the simulated human brain remains dormant while the simulated fruit fly strives for simulated fruit.

Answer 11

This depends on your definition of "complete scan." There's a lot more going on in a brain than just the connectome. To reproduce a person's complete personality, you would need to scan a brain down to the last atom, and beyond.

There are many ways to describe how a brain works. They are actually integer-based weighted accumulators. Yes, integer based. Each dendrite has vesicles, which are tiny reservoirs of neurotransmitters. We're talking atoms numbering the hundreds or maybe thousands, but definitely not billions. To reproduce a mind, you would have to count every single one.

It isn't just your neurons that provide what we call thinking. For every neuron, there are also hundreds of glial cells. They're like the trash collectors, the administrators, and all of the other thankless roles that support the superstar neurons. They also communicate among each other, although slowly compared to the neurons.

The cerebral fluid isn't just a random soup. For instance, the availability of ATP (the physiological equivalent of money) is essential for continued functioning. You would also need to track the state of all of the sodium-potassium pumps, which change shape based on their current activation state. Getting the pumps right would require counting and tracking electrons.

On the planet Hell in my book, they have the ability to perform brain scans to that detail. They clone the bodies, and then build a brain in the clone's head after it's fully grown. There are several levels of brain-copy. Getting just the connectome right results in someone with the knowledge and experiences of the previous iteration, but they lack the emotional biases that the person builds up over time. I justify that as having the state of those biases stored in the glial cells. More advanced, detailed transfers allow those to be passed along, too, but that's only for wealthy people who plan ahead. I don't go into detail about the molecular state machines because that would be boring.

Answer 12

Let's look at a far simpler organism for a moment. C. Elegans. Specifically the adult hermaphrodite: 952 cells including 302 neurons, 95 muscle cells, and a bunch of other stuff. It also has a bunch of other 'nice' characteristics like being transparent that make working with it easier.

We've had the complete connectome mapped out since 1986¹. Every neuron and every connection between them. I say 'the' connectome not 'a' connectome, because the connectome of C. Elegans is fixed.

I'll repeat that, because it's a key point: the connectome is identical across all² adult hermaphrodite C. Elegans.

...and yet C. Elegans can learn³. And this learning directly comes in the form of different neural responses to the same stimuli!

How is this possible? Answer: the connectome cannot be the only thing that matters.

(Now, the neurons of C. Elegans are known to be different than those of humans. It is possible that human neurons are somehow simpler than those of C. Elegans. Is it implausible that human neurons are at least as complex as those of C. Elegans? I'd argue no.)

1. And it's worth noting that despite the best efforts of the OpenWorm project we still can't do accurate whole-brain simulation of C. Elegans without blackbox emulation...
2. Barring damage of course.
3. Not directly relevant to the question, but it's so weird I want to mention it. Not only can C. Elegans learn & remember, it has generational memory. That is, if you train a C. Elegans, its descendants retain some of the same training...

Answer 13

Computers are not brains, and mood

Brains and minds are often compared to each other. This has evolved from long ago, where we tried to find approximations to make sense of it all. We've come a long way from comparing it to fluids for example. Computers were for a long time the latest thing. Processing, long and short term storage, graphics! Now it is all neural networks. This shift should show that it is only an approximation. However close we get, it is not the same. Even the answer of AlexP has it in it's answer to clarify the thought process. In truth we don't know if there even is a difference in "hardware" or "software" of the body.

Though things not being the same does not exclude we can't simulate it. A simulation of a coin flip is not a coin flip, but we can see it is pretty close to the truth. Though with something as complex as the brain it could already fall apart, being unable to create anything from the original correctly. To simulate a brain accuratwly you need a brain.

If that is not enough we run into an information paradox. Scanning a brain is more than just having a neural map. Each neuron has many unique properties how they function. They have different sensitivity to chemicals in the synaptic gap. They can be without synapses and all electrical. So you need to have this too. However, the deeper you go the more information you need. Teleportation or the coastline show this well. At what point do you have enough information to say definitively you have everything you need? Can you say a coastline us x long? You can nearly always zoom in closer. It leads to a problem where you need more information than possible in this universe. Say we did get that. Now you are fine, right?

Not by a long shot. You made a single scan how the brain works of a person. People react differently depending on a lot of factors, only partially governed by the brain. Are you hungry? You'll notice some stimuli more and suppress others. Close to someone you hate? Having sex? Depression? Sadness? Overly enjoyed? All make the brain react very differently, which isn't clear from a single neural map. You would need to scan it many times to understand the changes in a single person.

Research has shown that gut bacteria influence the brain in some major way. The same way some parasites make you like cats more or rabies makes you avoid water. These influence a person, yet are fully outside a neural map. You'll have gaps in how it should react. You might think to simulate a person with it, but a change as simple as different gut bacteria can 'create' different person. How can you accurately simulate someone if you do not have such data?

You mention difference in neurons firing as well. Many of the above change how neurons fire. It is important to note that neurons fire all the time. They can only hold a charge for so long before they spontaneously discharge. A single pulse is ignored by the system. You need many in certain patterns. You can glean a lot from the neural pathways and synapses, but simply not all.

A neural scan is a good start, but far from the total picture.

Answer 14

Technologically if you can map every connection you can upload the mind.

The trick is every connection is more than just every neuron. Not only is it possible we can do it for simple brains, the only limitation for humans is how difficult the mapping is and the fact it is 100% destructive, so you get one try. Mapping a fruit fly took years and hundres of people and AI. So if you want to get around it in story just use the risk to the brain angle. The gamble is still too high to risk destroying a brain mapping it if it could be lost through a single error.

the other issue is if you could map the brain without destroying it, it starts diverging from the original brain in the first few minutes as you learn new things and thuse the pattern changes. so you can use it to predict a person that well, and it gets worse and worse as time oes on. If your goal is upload a person to say provide an alternative to a dying body then the only real arguments are politcal or ethical not technological, you could do it. The connectomes we have of fly brains appear to perfectly predict behavior so there is no reaon to beleive it would not do the same in humans.

The firing sequence basically resets when you sleep so you the sequence is not really important if your trying to recreate "you". You just loose everything they learned since the scan, but it makes prediction less and less possible.

Now as Nosajimiki pointed out you also need a receptor map, neurotransmitter and hormone receptors, but that is doable too. We have already started doing it. its up to you what the limits of your technology is.

https://www.princeton.edu/news/2024/10/02/mapping-entire-fly-brain-step-toward-understanding-diseases-human-brain

https://pmc.ncbi.nlm.nih.gov/articles/PMC7930367/#:~:text=2.6.-,Connectome%2Dbased%20predictive%20modeling,et%20al.%2C%202017).

https://www.nature.com/articles/s41586-024-07939-3

https://www.nature.com/articles/s41593-022-01186-3

Answer 15

“Mind” is a function of a body, not just a brain

An isolated brain doesn’t work, it’s intrinsically connected to the body, not just by neurones but by blood and all the weird and wonderful mix of hormones, nutrients, and even pathogens that flood it and the body.

Yes, you can simulate a brain in a computer, but it just doesn’t work as a mind without a body.

Answer 16

Heisenberg and chaos

You are measuring things at molecular scale where quantum effects start to manifest - indeed, there are some biological processes, like photosynthesis, that depend on it. So, you can’t measure things with absolutely precision. Indeed, the smaller you make your nano bots to increase precision, the greater the quantum effects will be.

If the brain is chaotic, and who’s to say it isn’t, that initial tiny error will cause massive divergence, and, since most states a brain can be in equate to ‘dead’, it’s never been successfully done.

Answer 17

As many people have illustrated in this thread, there are many possible reasons.

However, the core issue that arises before even considering technical difficulties, is the following: creating a perfect neuronal replication is a difficult technical problem; producing a mind is a metaphysical problem. The two are entirely different issues; one cannot simply deduce that solving a difficult technical problem that we merely assume, without sufficient grounds, may create a metaphysical effect, would indeed create such a metaphysical effect.

Everything that follows from such an assumption is the equivalent of going into an empty grocery store blind-folded and expecting to successfully retrieve an apple. The Hard Problem of Consciousness is possibly so non-trivial that we may never be able to make such a deduction.

Largely in agreement with the user g s, I believe your search for some kind of definitive justification as for why a perfect neuronal replication would be insufficient to produce a mind, may ironically be less profound or elegant in a fictional story, than simply illustrating the result of failure with many possible reasons, but without any assuredly known reason.

In short, there are no "particular reasons" for the possibility of failure, in the manner that you are attempting to find. There are many, many—exponential possible reasons for such a failure, and to place an assuredly known reason for such a failure as a thesis of your fictional story, would be highly naïve and unrealistic (on the basis of current ignorance).