Tuesday, May 26, 2015

Mapping the brain


RK: Right now… our most complex machines are still on the order of a million times simpler than the human brain. Although we’re making very strong progress in reverse engineering of the human brain — in fact we’ve made some very strong progress even since my book came out two years ago–we still don’t have maps or reverse engineering of 99% of the human brain. We’ll get there because brain scanning, our ability to reverse engineer the human brain, as well as computation, communication and miniaturization are all growing exponentially. 
http://www.kurzweilai.net/nerd-of-the-week-ray-kurzweil

Even if we suppose mind is reducible to brain, I don't think his analysis works. We can only map the brain at scales large enough to be detectable. But matter has an underlying subatomic structure. That pattern can't be mapped. We hit a wall in terms of our inability to detect physical entities below a certain scale. 

The basic principle is that bigger things are composed of smaller things. The question is how far down that goes. Where does it bottom out? Does it bottom out? Is matter infinitely divisible? Is there a smallest physical unit? 

We can only map what we can scan. To my knowledge, brain scans only map macro brain structures. 

You can put slices of a brain under a microscope, but a dead brain doesn't give you all the information you need to map the brain. 

High-energy physics would kill the brain. If you put a human inside a collider, what comes out won't be alive. 

And even then, we still hit a limit. The part of the brain we perceive is the effect of elementary particles we can't perceive. It's a larger pattern, caused by smaller patterns–like nested Russian dolls. Dolls inside dolls inside dolls. 

This is why we can't prove string theory. Even if strings were the ultimate physical constituents, they are too small to detect, because we must use something physical to detect something physical, and what we use is bigger than what we hope to detect. 

The structure of the brain we can scan is generated by structures that elude detection. So the resultant map is a very coarse-grained map that misses all the finer details. Like low resolution photography. 

That's not to knock brain scanning technology. It contributes to medical treatment. But we can only take it so far. 

7 comments:

  1. I'd like to add to what Steve has written:

    1. Presumably mapping the brain down to its subatomic levels is relevant if we assume the mind is reducible to its physical brain, and consciousness arises from the physical brain. Like what if the brain is using quantum entanglement to correlate brain states or processes across distant parts of the brain?

    In addition, given the same assumption as above, since the eyes and ears are in a sense external extensions of the nervous system, mapping the brain at subatomic levels could prove relevant to understanding how our eyes and ears fundamentally work, which in turn would be relevant to creating A.I. For instance, see here.

    2. Neuroimaging using MRIs, SPECT, and PET scans can map the brain, though not how Kurzweil envisions, for neuroimaging only gets us to the molecular or cellular levels at best. It may not seem like it to the general public, but the molecular and cellular levels are still leaps and bounds away from the subatomic levels (which at any rate would be very unlikely to be relevant in terms of medical treatment). Normally neuroimaging just stays at the tissue level or higher.

    3. Also, neuroimaging doesn't map the brain directly, at least not to my knowledge, but in general they do so using analogues (e.g. seeing which cells take up which contrast agents).

    4. Plus, using radioisotopes can carry significant radiation exposure to the person. There could be lots of ionizing radiation involved. For example, the most widely used radioisotope in medicine is technetium-99m, which decays by emitting low-energy gamma rays.

    5. Perhaps we could map tiny portions of the human brain at many different times or across many different persons to dilute or limit the radiation exposure. However, even if we were able to put all of this together into a single human brain map all the way down to the subatomic levels, it's one thing to have a brain map, and it's quite another thing to understand how it works. Just like with the human genome. We have a map of the entire genome, but how does it all work? We still have little idea. (And mapping genes is far easier of a feat than mapping the brain at subatomic levels!)

    ReplyDelete
  2. That also makes the point that mapping the brain is misleading, because that's static, but we're dealing with a dynamic system. Quantum states aren't static.

    There's nothing wrong with neuroimaging. That's great as far as it goes. But it doesn't capture all the dynamic aspects, not to mention, say, structures below Planck length. Patterns are both temporal and spatial.

    ReplyDelete
  3. I'd add:

    The way we would detect or "see" subatomic particles would be to bombard them with photons. Quite apart from whether we are really going to bombard a living human brain with photons (a dead preserved brain won't do, an animal brain wouldn't be the same), how are we realistically going to accomplish this (e.g. the more precision we want then the higher frequency and thus higher energy photon we'd have to use which could potentially injure if not ultimately kill the brain)?

    Also, even if we could detect subatomic particles, wouldn't we quite possibly run into some of the more perplexing parts of quantum mechanics, which if so would seem to severely undermine if not outright defeat our attempts to reverse engineer the brain and so forth?

    I know with regard to brain scanning Kurzweil's statement is highly if not naively optimistic! The technology to scan brains is very cool, and continuously developing, but it's certainly not growing "exponentially" if by this Kurzweil is referring to something like Moore's law.

    I mean, I share a lot of his enthusiasm for technology, but he is just far too unrealistic, I think.

    ReplyDelete
  4. Would the successful creation of Strong AI be problematic for Christianity? I don't see why it would. If it happened, it would do so by God's permissive providence.

    I can imagine an AI computer reading all the literature on philosophy and apologetics and calculating the probabilities of the existence of God. Of concluding the likely existence of God. Of wishing for salvation and immortality. I can imagine it fearing it doesn't have a "soul," or isn't a "soul". Of it petitioning God to grant it immortality; even transforming it into a human like Pinocchio and Star Trek's Data (if Q can do it, why can the true God?). Maybe Steve can write a short story based on this premise.

    If it's theoretically possible that God may resurrect deceased pets, why not God granting immortality to an AI?

    If Strong AI disturbs some Christians, shouldn't the level of sentience some animals rise to do so as well? The claims regarding "Koko" the gorilla's use of language (via sign language) have been controversial for many years, but I wouldn't be surprised if there's some truth to it. There's the claim that no one ever taught Koko the sign for the word "ring" and that she apparently created a word for it by referring to it as "finger-bracelet." If that's true, then that's an amazing creative use of language for an animal.

    The above are just some thoughts to ponder.

    ReplyDelete
    Replies
    1. There's also the scenario of an AI sinning and going to hell. Would the AI need atonement to be saved? Christian philosopher Robin Collins has speculated that Christ may have been incarnated on other planets or universes in the multiverse. Would Christ need to be "digitized," the computer equivalent to incarnation, to atone for a fallen AI souls?

      Delete
    2. This comment has been removed by the author.

      Delete
    3. Some might be offended by the suggestion of Christ being "digitized," but His human incarnation was pretty degrading as well. We're so used to the incarnation that we forget how great an act of condescension it was for the Blessed God to become a finite creature of the species that we are. Strong AI would potentially be more knowledgeable, logical, rational and creative than ordinary humans. In which case, Christ's incarnation as a human is more humbling than would His being digitized.

      Delete