It seems likely you'd have to process as you go, and not store it all. Or have one hell of a datacenter/buffer like the LHC does, holding it until it can be crunched down to a "simple" network instead of voxel data.
Eh, you can guess for a lot of it, and you at least have a map of where all the inputs actually GO. If it synapses in the olfactory bulb, probably don't hook up your simulated eyeball and all that.
Brain plasticity can cover some wiring sins - cochlear implants are pretty awful for connecting where they should go, but it's good enough to get some degree of hearing going. Get enough and the simulated patient can help you out telling you if you've got it right or not.
Like, I get that these guys might be - maybe even probably are! - snake oil salesmen, but what about brain digitization is actually impossible? There are certainly a fixed number of neurotransmitter types, a fixed number of channels, a fixed structure to a particular operating brain, and so on. Anything that exists can be reverse engineered and, in principle, rebuilt.
You can electroshock a brain, completely scramble the "active patterns" and the person will... well, reboot, for lack of a better term. If you could get the weights without activity, I see no reason you couldn't just initialize in a random state and let it figure itself out. Figuring out the weights is a technical problem, but technical problems are made to be solved.
Finally, dualism is juvenile bullshit. A neuroscientist who thought the mind WASN'T the brain would be absolutely useless.
