Future technological advances may allow us to instantiate high-resolution models of our mindbrains on machine substrate, or even create de novo persons. Critics point out, quite rightly, that machines are digital while mindbrains are analog. From this insight, they conclude that machines won’t be able to recreate the detailed processing of neural wetware.
However, the critics miss the fundamental fact that we can approximate analog processing with high resolution digital processing. Nature already does it.
Genes are discrete (digital) while phenotypes are continuous (analog). Continuous traits can be approximated with a large number of genes that each contribute a small amount to the outcome, or by one gene with a large number of alleles that each tweak the outcome by a small amount. A continuous phenotype, such as the spectrum of adult human heights, is determined by a set of discrete genes. Even if we controlled for all other influences on human height and looked at a single hypothetical gene that controls growth hormone output, we can see that, by implementing a large number of alleles, each one resulting in slightly more or less growth hormone output, a continuous spectrum can be approximated. If adult human height ranges over, say, one meter, and our gene has 1000 alleles, than we can specify height with millimeter precision. If our gene has a million alleles, than we can specify height with micrometer precision, and so on. (Of course, in reality, genes merely produce organisms whose traits are differentially influenced by environment, and environmental influence is analog.)
The brain is actually analog and digital. Synaptic firing is digital, and synaptic organization allows for signal processing through logical operations much the same way that transistors do. But the events that aggregate to induce synapse firing are continuously additive or subtractive. They are analog. This is what the critics are talking about.
If approximating analog events is possible with digital events, then we only need to achieve a sufficently high resolution digital model of pre-synaptic events to produce accurate models of neural processing at any arbitrary level of organization. Nothing makes this physically impossible. It’s all a matter of having the technology and money to do it.