The Undesigned Brain is Hard to Copy

By Kyle Munkittrick | January 17, 2011 10:47 am

UPDATE: Hanson has responded and Lee has rebutted. My reaction after the jump.

The Singularity seems to be getting less and less near. One of the big goals of Singularity hopefuls is to be able to put a human mind onto (into? not sure on the proper preposition here) a non-biological substrate. Most of the debates have revolved around computer analogies. The brain is hardware, the mind is software. Therefore, to run the mind on different hardware, it just has to be “ported” or “emulated” the way a computer program might be. Timothy B. Lee (not the internet inventing one) counters Robin Hanson’s claim that we will be able to upload a human mind onto a computer within the next couple decades by dissecting the computer=mind analogy:

You can’t emulate a natural system because natural systems don’t have designers, and therefore weren’t built to conform to any particular mathematical model. Modeling natural systems is much more difficult—indeed, so difficult that we use a different word, “simulation” to describe the process. Creating a simulation of a natural system inherently means means making judgment calls about which aspects of a physical system are the most important. And because there’s no underlying blueprint, these guesses are never perfect: it will always be necessary to leave out some details that affect the behavior of the overall system, which means that simulations are never more than approximately right. Weather simulations, for example, are never going to be able to predict precisely where each raindrop will fall, they only predict general large-scale trends, and only for a limited period of time. This is different than an emulator, which (if implemented well) can be expected to behave exactly like the system it is emulating, for as long as you care to run it.

In short: we know how software is written, we can see the code and rules that govern the system–not true for the mind, so we guess at the unknowns and test the guesses with simulations. Lee’s post is very much worth the full read, so give it a perusal.

Lee got me thinking with his point that “natural systems don’t have designers.” Evolutionary processes have resulted in the brain we have today, but there was no intention or design behind those process. Our minds are undesigned.

I find that fascinating. In the first place, because it means that simulation will be exceedingly difficult. How do you reverse-engineer something with no engineer? Second, even if a simulation is successful, it by no means a guarantees that we can change the substrate of an existing mind. If the mind is an emergent property of the physical brain, then one can no more move a mind than one could move a hurricane from one system to another. The mind, it may turn out, is fundamentally and essentially related to the substrate in which it is embodied.UPDATE: Hanson’s main claim is that we can make “economically-sufficient substitutes for human workers” at some point because “we already have pretty good signal-processing models of some cell types; we just need to do the same for all the other cell types.”  I acknowledge non-biological minds may be possible, but not within this century.

The problem is two-fold. First, we don’t understand the neurons in sensory organs well enough to replicate them, we just skip over broken ones with an entirely different system and let the brain figure the data out. The mind can “see” with almost any sufficiently nuanced set of information. As Lee notes, we can do that because we already know how to make an artificial ear, the problem was connecting it to a brain. Second, understanding the neuron does not mean we understand the network, nor how the network uses each neuron to process and retain information. Lee’s protein example is designed to show this micro/macro incongruence.

Finally, even if I grant Hanson’s point that the mind was “designed,” evolution is a messy and lazy inventor. As such, replication of biological structures is notoriously difficult. Perhaps unplanned is a better word?

I’m eager to see how this debate progresses, but it seems to me that Hanson takes the computer analogy a bit to strongly.

Image of DTI Sagittal Fibers via Wikipedia


Comments (15)

  1. Adam

    You can always (with enough CPU time) do a complete simulation of the underlying hardware (if you have sufficient knowledge of the laws of physics governing the underlying hardware). That is, even an undesigned system which runs on physical hardware is emulatable because you can write an emulation of the underlying hardware. I’m not saying this is practical at this point in time (for instance, doing a complete quantum-chemical simulation of an entire brain far exceeds the total power of all computers on the Earth at the moment), but in principle it can be done.

  2. Congnition is an emergent process. An emergent observable can be modeled, but its emergence cannot. Emergence is an extrinsic property. Emergence arises from a discontinuity of aggregation. How does one model something that should not be there, e.g., Wolfram’s few cellular automata with serial contingent yet random outputs?

    Create a deterministic brain that has only predictable bad ideas and you have a Republican. Create a random brain that has no ideas at all and you have a Democrat. Omit the brain entirely while retaining operating expenses and you have a Congress. Nullius in fidelis.

  3. @Adam
    That was also my initial thought. Then I thought perhaps the architecture of the hardware could be replicated. Neuron-like computer components, which operate in the same way a neuron would. Then wired together in the same(similar) architecture as a brain. Of coarse this would be disproportionally inefficient, in both hardware and power. Which results in my thought that nature brought the brain to its current state as the most efficient way it could. At least in terms of hardware mass/complexity vs. Energy Consumption vs. Computational output.

    If we’re going to ‘simulate’ the underlying physics to accomplish such a complexity intensive system. I’d recommend using real particles. Not much use in precisely simulating the physics of the universe to simulate a complex system within the universe. Clearly it’s more practical to just build a brain within the existing universe. That would be redundant computational power. Not to mention inaccurate. Which isn’t to say rebuilding a brain is an easy task.

    Now. A little more on topic of what this article is trying to convey. The more we know about how the brain works, the more it seems we don’t know. I totally agree. Do I think the reverse engineering of the brain is too complex a problem to solve? No. Maybe not in the next couple of decades; But i’m absolutely certain this feat is within our reach. The human mind is probably(in my opinion) evolution’s most successful feat. I don’t expect to figure it out over night.

    Do I think we’ll ever reach a ‘singularity’ where the human mind is ported to machine. No. Not because I think it’s impossible. Only that it’s totally impractical. A far less audacious goal would be to maintain the brain’s lifespan a prolonged or indefinite amount of time. As a computer scientist, I see Bio-Mechanical Engineering the technology of the future. Especially when it comes to computing. I see engineered brains being how we make computation a couple decades in the future. For more reasons then the obvious. The human brain consumes about 20volts, and can output a stupid amount of computational power. Not to mention the brains ability to perform fuzzy logic, and many other lines of reasoning. Also sustainability issues. Computers are not bio-degratable. Their components need to be mined, shipped, manufactured, shipped again, used for a sort time span, and the disposal process is utterly atrocious.

    Anyway… Back to work.

  4. Since “the underlying hardware” is a quantum mechanical system (like all other physical objects) a complete simulation would certainly be difficult, if not impossible even in theory.

    I doubt very much it is possible to simulate the brain without taking into account that it is in fact a quantum mechanical system, but this is not known today. And it might in fact never be known, exactly for the reason mentioned in the article: That the brain is not designed.

  5. ‘And it might in fact never be known’- Mogens
    Oh it’ll totally be know. I recommend reading ‘The Blink Watch Maker’-Richard Dawkins.
    The brain may not have a designer, but its design is consistently apparent amongst many humans(animals). If the brain were as random as throwing mud at the wall, I’d agree with you. But its not. So I don’t. Surely the brain can’t be so chaotically different between differing brains that reverse engineering the brain is impossible. The machine may be very complex, but it is not totally random.

    ‘Since “the underlying hardware” is a quantum mechanical system (like all other physical objects) a complete simulation would certainly be difficult, if not impossible even in theory.’-Mogens
    The level of atomic accuracy is totally unnecessary at this level. Signals between neurons are measurable voltages. And the rate of missfires between neurons can be as high as 30-90%. So obviously accuracy is not an issue.

  6. kirk

    Just imagine my world of non-genome computing in 15 years. I have retina implants for heads up display, cyborg joints so I can still run a marathon at 72 etc. At no time do I require the cognitive machinery to piss and shit. My “unextended” phenotype does that now and could continue to operate as a distributed, non-heterogeneous system with Singularity-ness. I want to simulate my own orgasms after all. Meat Puppet Uber Alles.

    Then, in 2050 when this version of my impermanent self wears out – my 8 year old clone puts on the environment I build over time. I run triathlons at 97 while an 8 year old Kirk prepares under careful tutelage to do the same. Is it really “Kirk 1.0” that replaces “Kirk 2.0”. No, actually it is “Kirk 1.97.365” that is replaced by “Kirk 2.8.320”. Epic impermanent self win. I am a different person every morning when my feet hit the floor. Dream on.

  7. Brian Too

    I’d suggest that it may be possible to model a generic brain. It may be impossible to model any specific brain however. It seems to me that there is a significant linkage between the specific neuronal configurations and axon connections, and a specific consciousness.

    Just because we can say that, say, human brains are constructed by X set of biologically determined rule sets, that does not mean that each brain is identical. We can see gross scale similarities but when you look closely, the details can be staggeringly different at the cellular level.

    I’d wager that’s important.

  8. Colin

    Unless quantum states are involved in brain function, what makes brain function completely impossible to model accurately with a sufficiently advanced map of a brain and a sufficiently advanced computer to run the model? You would, of course, have to model each individual separately. Obviously, neither medical nor computer technology are there yet with regards to the specific difficult of scanning this, but what actually makes it impossible? If we can model, then it isn’t a case of moving a mind, but rather simulating a mind at which point, the distinctions between a simulation and a mind become so much semantics.

    None of which is to say that this will happen in any of our lifetimes.

  9. Ken

    My favorite novel on this topic is “Permutation City” by Greg Egan. The approach there is to emulate the hardware; detailed brain scans yield the cell structure and interconnections, then the behavior of the individual cells is simulated by the computer. These “Copies” are self-aware (although some humans refuse to accept this), and Egan explores the consequences of this. For example, a Copy’s own consciousness as it perceives it continues normally even if the software is run on thousands of widely-separated and non-synchronized machines.

  10. I responded to that argument here.

  11. Lee says … “natural systems don’t have designers.” How does he know?

    Kyle says … “Finally, even if I grant Hanson’s point that the mind was “designed,” evolution is a messy and lazy inventor” What evidence is there for this messiness and laziness? Who is in a rush? Where is the mess?

  12. Kyle Munkittrick

    @Ian: The burden of proof is on you, who believe in God, not we agnostics. As the ontological argument and the anthropic principle are the best you’ve got, I suggest you move along. As for evidence of messiness, I suggest you look into how often the creation of life, surely one of your favorite topics, involves mistakes, mutation, damage, and outright failure.

  13. @Kyle, Quite apart from the link to my blog, I’ve made no reference in my contribution to this blog to God. I’m coming at this purely from the perspective of reason. The scientific evidence I have seen so far, and let’s take ‘junk DNA’ as an example suggests that there is no such thing as junk DNA, and that indeed it is responsible for making us who we are (

    Besides as for ‘burden of proof’ – you’re the one stating the original argument without references, I’m merely pointing out that science is based upon the sum of our experience. To state categorically that “evolution is a messy and lazy inventor” assumes you have experienced everything there is to experience. I doubt that very much.

    And you also assume that ‘creation of life is one of my favourite topics’ – you’re wrong.

  14. This discussion leaves out the concept of objective vs subjective. Even if my mind could be replicated/simulated in detail, everyone else might think it was me but “I” would know it was not really me. The same is true for the replicants when THEY are duplicated. They would know when they wink out, it is over for them. This is why teleportation will not catch on when people find out that is just the information to reconstruct a simile that is sent, they won’t buy in. The actual YOU is destroyed in the process.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!


See More

Collapse bottom bar