Re:Wolfram's new book and my thoughts on reality
on
Wolframania
·
· Score: 1
I do computational neuroscience, and I run into those people all the time. No one I work with believes them, not because we can prove them wrong, but because there is no convincing theoretical reason or experimental evidence to believe that they are right. Other than the fact that they really really want it to be true.
However, I disagree with you that there is a discrepancy between performance of brains and estimates based on number of neurons, etc. This might have been true prior to us understanding more about computation that occurs in the individual dendrites of each nerve cell. Neurons can do a lot, and we still don't know exactly how much. But there's no evidence of mystical physical phenomena or profound ideological deficiency in current paradigms, it's just a complicated system that is very difficult to approach experimentally and is going to require some time to figure out.
Re:You even had to mention the Lorentz system...
on
Wolframania
·
· Score: 1
The original poster meant to say "Lorenz," not "Lorentz." The Lorenz attractor is the classic demonstration of sensitivity to initial conditions in a coupled system of 3 or more nonlinear ordinary differential equations. This was the first (or one of the very first) demonstrations of chaos. A great mathematical insight that was made by a meteorologist, not a "real mathematician."
You can PCR out fragments longer than 3000bp, but it's hard, and it depends on a lot of factors. There are special "high-fidelity" polymerases that people use for these purposes. There is a whole field in industry dedicated to messing with aspects of PCR, mutating polymerases, and that kind of thing, to try to get longer PCR products, etc.
To address your comments, the reason why cloning animals is difficult has a lot to do with what is called epigenetics, or information stored in the chromosomes that is beyond the mere sequence of the genome. For example, reversible modifications to the chromosomes that activate or inactivate certain regions during the process of cell differentiation. "Cloning" a multicellular organism is the name given to de-differentiating existing adult cells all the way back to the differentiation state of a fertilized egg, then growing them back up into adult organisms. However, even if you have the genome 100% correct, problems with the structure of the chromosomes will lead to horrible birth defects (some textbook examples of this in humans are Angelman's and Praeder-Willi syndromes, which cause mental retardation and characteristic deformities).
The thing is, what they are proposing to do with this extinct marsupial is actually not cloning, but actually synthesizing DNA molecules and building one of these animals FROM SCRATCH. This makes the problem of chromosomal structure MUCH more significant than it is when you "clone" an animal using standard techniques. To address your second point, however, it is extremely unlikely they would try to turn a related animal into the extinct animal like they did in Jurassic Park. First of all, this would not work, and second, even if it did, you would just get a different third animal. The real obstacle here is creating artificial chromosomes out of individual genes, and this is where the work would have to start. Most likely, this would not be done in cells from some exotic mammal, but something very standard like a mammalian tumor cell line or maybe even yeast.
As for your third point, a female mammal can be cloned from a male in theory, and you could breed them. Obviously this is not a perfect arrangement, but it is the case with certain strains of genetically identical lab mice, and they're generally healthy. In any case, the small problems here pale in comparison to the giant problem of assembling artificial chromosomes. This hasn't even been done with bacteria, much less a multicellular eukaryotic organism.
They don't need two individuals, just one male. A male has an X chromosome and a Y chromosome, while a female has two X's. They can artificially create a female from that male by replacing the Y with an X. While this is beyond current technology, it is a lot less beyond current technology than splicing together entire chromosomes out of individual genes, which is what they will have to do in order to clone this animal. The fact that they have only one animal to work from will not by itself make the species go extinct again. Certain strains of lab mice are inbred to the point where all of them are genetically identical (except for the X and Y chromosomes between males and females), and they do fine.
PCR is NOT cloning. You can clone a PCR product by putting it into a plasmid (a small artificial chromosome) and inserting that into bacteria, but just doing the reaction by itself is not cloning.
Actually, much more than 20 or 22 are found in proteins, they're just not genetically encoded. This basically means that the protein is made with a certain amino acid in that spot, and it gets converted later. One of the classic examples is hydroxyproline, which is a modified version of proline (the only amino acid that's technically an imino acid). There are lots of hydroxyprolines in collagen, and conversion of proline to hydroxyproline is vitamin C dependent. So if you don't get enough vitamin C, you have lots of defective collagen which leads to teeth and hair falling out when you have scurvy.
The significance of the whole 22nd amino acid thing is only that it's the 22nd GENETICALLY ENCODABLE amino acid found in nature, and it's in some weird bacteria, not in humans. This means that it has its own tRNA. There are more than 22 amino acids found in proteins, even in humans, and in the lab they've created "alien" bacteria that use two or three completely artificial amino acids to incorporate fluorescent tags into protein molecules and that kind of thing.
Clinically speaking, there are many other causes of hemolytic anemia that are much much more common than anything involving the translation machinery at the level you're describing. It's also the same thing with hyper- and hypo-thyroidism. I'm not saying you're wrong and these things couldn't cause those diseases, only that they would be relevant in a tiny minority of cases.
There is some degree of truth in what you're saying, and this realization is the reason that there are lots of monte carlo-based modeling methods out there for intracellular processes. Basically, if you're modeling using differential equations and concentrations, that's implying that concentration is continuous while it actually is not. Sometimes it's actually necessary to model individual calcium ions, for example.
However, in talking about cracking a "code," you're implying that we somehow have all this information but we can't figure it out. The truth is that we don't have the information yet about large scale protein interaction networks in cells--I think this is the real problem that's closest to the "code" you're talking about, though it's not really a code as much as a jigsaw puzzle in high dimensional non-Euclidean space where we don't have all the pieces. Also, the reason that there has historically been lots of emphasis on chemical reactions on substrates, etc. in terms of dissecting biochemical pathways is because in the past that was all we could do. Since it's fairly easy to use techniques from organic chemistry to quantify rates of chemical reactions and so forth, and you know the next enzyme downstream is the one that acts on the product from the prior one, you can piece these things together with techniques that are primitive by todays standards (hence those giant evil charts they make you memorize in undergrad biochemistry). But we have a lot more advanced techniques now, so the focus has shifted more towards signal transduction (i.e. how the cell "knows" various things are going on, how they communicate with each other, etc.) as opposed to housekeeping functions like making sure the cytosolic ATP concentration doesn't drop too low (like old-school biochemistry). Obviously what I'm saying is a slight generalization/oversimplification, but I think the main theoretical problem in biology right now has more to do with giant networks of interacting proteins and figuring out the qualitative behavior of large networks from incomplete data that only tells you about what one tiny piece of it is doing. The theoretical underpinnings of this problem are likely to be grounded more in nonlinear dynamics than CA, IMHO, but what's holding us back right now is the fact that we can't get enough data, despite what everyone says about the big explosion of data in biology.
I still stick by my point that the main contribution of CA in biology will be modeling extracellular processes in developmental biology, though I'm not developmental biologist.
I'm doing a PhD in biology, so trust me: the map from gene to protein sequence is 100% fully understood and has been for many years. It was the basis of the 1968 Nobel Prize in Physiology or Medicine. In biology it's referred to as the genetic code. So when people throw around the term, they should understand that it has a very specific meaning and that it is well understood. Anyway, the mapping from protein sequence to protein structure is the one that is not well understood. Well, it's not an issue of whether or not it's well-understood, it's just so complex that doing it based on the physical properties of the actual atoms involved requires profoundly unattainable amounts of computing power. So structure prediction methods generally fall into two categories: ab initio with fancy mathematical shortcuts, and "threading," in which evolutionary relations between proteins are exploited so that you can use known structures to help predict similar unknown structures. CA are unlikely to help with this.
CA models ARE likely to be helpful at the level of multicellular organization. Hepatocytes in the liver somehow self-organize into a hexagonal lattice...how? This type of thing happens all the time in development, but it's hard to understand how cells can self-organize without any kind of overarching supervisory signal (though such things sometimes exist in the form of concentration gradients of signaling molecules, etc.) Anyway, I don't see any immediate application of CA-type methods to the study of intracellular processes, which is not to say that there aren't any.
I think it's becoming widely acknowledged that these kinds of things are very important, but I have no idea what the NIH has done specifically. It's usually not mathematicians from the math department that end up collaborating with biologists in my experience, but physicists and applied mathematicians who are often in engineering departments.
I'm not trying to argue that the NSF's positions aren't important, only that there is a certain amount of jurisdiction in terms of what the NSF and NIH will fund. The NSF spends very little money on breast cancer research. Not because it's not science, and not because they think it's not important, but because the NIH funds that. I agree with pretty much everything you're saying, I just don't believe that the NSF increasing funding to mathematics applied to biology indicates in any way that the lack of advanced mathematical techniques outside the grasp of people in CS is the underlying bottleneck in bioinformatics. Math is obviously important, but I think that improved mathematical analyses will be much more important in other branches of biology. I say this as a computational neuroscientist, but I'm referring more to systems biology. Like you said, there are a lot of blurry lines.
You make it sound like what the NSF does reflects what science as a whole thinks is important, but the NSF funds physical sciences. Biological sciences are funded through the National Institutes of Health (NIH). I'm not disagreeing with your main point, though I think you're confusing computational biology with bioinformatics. In any case, the hard part theoretically is dealing with the sheer complexity of biological systems (dealing with a gigantic number of coupled equations which are individually fairly simple), which is different than most problems in theoretical physics, for example. This requires a different way of looking at problems than the way most people usually learn math. However, in my opinion, the main bottleneck in biology right now is still an experimental one, not a theoretical one, despite all the data people have been generating with various high-throughput experimental methods.
For those who don't know what near-field optics (scroll down and look at the diagrams) are, it's a way to overcome the fact that normal far-field optics can't resolve things smaller than the wavelength of the light you're using. In near-field, you use a tiny fiber optic probe that's smaller than the wavelength of the light so you can see things smaller than the wavelength of the light, i.e. pack data more densely. Imagine you had terrible vision, such that you could only see at the resolution of a baseball at arm's length. By looking through a straw and moving the straw methodically, you could see with much higher resolution.
One problem with this is that contrary to what you're suggesting, it is unlikely that they would ever be able to make multiple layers on a disc because you can't focus the light onto lower layers. It's also questionable how durable the disc would be because you have to put the probe so close to the surface of the substrate storing the actual information that you couldn't have a (relatively) thick layer of plastic protecting it like on a CD. There are also things that make this difficult in practice, like the fact that it's easier to focus light precisely than it is to move a microscopic probe precisely.
Another thing is that it's not entirely obvious that these discs will be faster than CDs or DVDs because the process of pulling data off them is intrinsically more difficult. You can't assume that you can spin the thing as fast as a CD and it will work flawlessly. In addition, there are ways to increase the speed of CDs or DVDs by rapid scanning of the light beam, so they will always have the potential to be faster. I agree with you, though, that it's stupid to compare this to CD burners.
I think you're being a little too harsh...first, the things you're saying aren't unique to CA. In any case, if Wolfram is right and the universe is organized that way, the limitations you describe are inconveniently imposed by the nature of reality, not by the mathematical language used to describe it. If you think math is powerful enough to change that, you've been watching Pi a few too many times.
More importantly, it's important to realize that just because there are nonlinear differential equations involved, it doesn't mean the attractors are necessarily chaotic. And even if chaos is involved, that doesn't automatically render us powerless to predict or understand. I'm skeptical about the universal power Wolfram claims his ideas will have, but there are certainly areas where CA will be useful. Personally, I think it's the future of theory in developmental biology. For example, there are complicated but highly stereotyped arrangements of cells in the cortex of the brain, and it's unclear how individual cells can "know" how to arrange themselves like this in the absence of some overarching presence telling them where to go. I doubt Wolfram's book will unravel the mysteries of the universe or anything, but I think it's premature to write off cellular automata as "useless."
Well, I guess it depends who you ask. Although protons technically bond covalently to the atoms of the side chains, very few biochemists would think of that as becoming part of the protein. Plus, the protein is translocating them. Well, in any case, this new thing is a big achievement, and the molecules involved are certainly easier to work with than something like bacteriorhodopsin.
There's actually been some interesting work done on that based on this nonlinear phenomenon called stochastic resonance. There's a link to a paper here.
BTW, the phenomenon is called hyperacuity, and it specifically refers to having higher resolution than should be allowed by the size of the photoreceptor cells in the retina. I state this explicitly because it is not something due to a more optics-related phenomenon like the fundamental wavelength limit of the light or anything like that.
We're all skating on some thin semantic ice here, but bacteriorhodopsin DOES operate mechanically. It actually does mechanically pick up individual protons and move them from one place to another. The protons are not somehow magically absorbed with new protons being produced on the other side. If you don't believe that, there is a related protein called halorhodopsin that does the same thing with chloride anions.
It's true that these molecules exist in nature, but that doesn't change the fact that people are experimenting with using them for all kinds of completely non-natural things like high-density optical memory, bizarre types of sensors, etc. If you type bacteriorhodopsin into the patent database, there are 197 patents associated with it, and probably half to two-thirds of those are not biological applications. Would you feel differently about the molecule if someone synthesized it using a peptide synthesizer? I don't know if that's been done or not... And I'm not saying that the springboard thing isn't different and potentially more useful, just that bacteriorhodopsin is a molecule people use that directly transduces light energy into mechanical energy.
But you're pulling energy out of the system by moving the board, and that energy has to come from somewhere. I don't know enough about the physics of what's really going on, but this is simple conservation of energy. However, I doubt that the force comes from electromagnetic dipoles. I think it's more likely that the light bumps up an electron, causing that atom to have a different type of orbital, bringing it into conjugation with other orbitals, which changes which conformation of the molecule ("long" or "short") is most stable, or something along those lines. But I am not a physicist or physical chemist by any means. However, bumping the electrons up to higher energy orbitals DOES directly convert the energy of that photon into a different form, and those electron orbital changes are the basis for the movement, so this system IS directly transducing the light into mechanical energy.
Unless you count bacteriorhodopsin, a photosynthetic protein in certain bacteria. Light energy is used to change the physical shape of the protein and move hydrogen atoms from one side of a membrane to the other. This is direct conversion of light to mechanical energy (although it is not usually used that way), and since the 70's there has been a lot of work with this protein making different sensor arrays and so forth.
If it weren't energy conversion, you would only be able to bend the board once. Otherwise you would be getting energy from nowhere, or "using up" the molecules. But if you were doing that, the thing would only work for a very short period of time, not a whole day. There are other mechanisms to explain the photodamage to the azobenzene. I don't know anything about this molecule in particular, but in general irreversible light damage occurs through either free radical reactions catalyzed by the light or bumping the electrons up to an energy state where they can fall to two (or more) separate states. One of these is the one you want, and the other one isn't, and each time a certain percentage of the molecules will get stuck there and not work any more.
I used to work on visual system neurophysiology, and I just want to say that I agree completely about the long-term future of retinal implants. People with a lot of exposure to electronics and computer technology and very little exposure to real biological science (e.g. the average/. reader) have a strong tendency to think that we will be able to replace or improve upon biology. However, most biological systems are already running at a much higher efficiency than pretty much anything humans have ever been able to build. For example, the retina can go out on a bright sunlit day and look at stuff, then go into a dark room and 30 minutes later be fully adapted to SINGLE PHOTON SENSITIVITY. This is an extremely complex process that occurs at the level of the retina--it's not just your pupil changing size. Another thing is that there is retinal circuitry for anticipation of movement. If you work out the math, a person's waving arm can travel 2-3 feet by the time the signal gets from your retina to occipital cortex in your brain, yet you perceive its position accurately because of compensatory mechanisms all the way down to the level of the retina. No artificial retina design has even begun to think about stuff like that.
I'm no expert on CCDs, PMTs, etc., but I seriously doubt that the technology to build an artificial device capable of that yet implantable in the eye will be available in my lifetime. Putting new cells in the retina, on the other hand, will most likely be possible in less than 25 years. Which are patients going to prefer?
One of my med school professors is the ophthalmologist for the Yankees (IIRC...a major league baseball team, anyway). He told us that he went and did eye tests on the whole team, and their average eyesight was 20/7.5, meaning that at 20 feet they can see what a person with 20/20 vision can see at 7.5 feet. I think that's insane.
The behavior of receptors (the actual membrane proteins that do something when neurotransmitters bind to them) themselves is actually quite straightforward and easy to quantify in a large percentage of cases. The big problems come afterwards. One of them is understanding how the input from the receptors is added up into an output in terms of firing. If you have access to a university library, you could get a copy of a review article here. However, firing is NOT binary or remotely computer-like in any way. The whole idea that there are useful brain-computer analogies to be drawn is entirely misguided and has poisoned the way too many people think about brain function. Firing is not necessarily a discrete event because there are multiple types of firing (burst spikes vs. normal spikes, etc.). Also, the temporal PATTERN of spikes matters and influences how later information is processed both pre and post-synaptically. And how a neuron processes the spatiotemporal pattern of inputs it gets from the thousands of other neurons it's connected to is not understood. It's extremely complex, and unfortunately not the kind of thing you can just read a web page and pick up. A good low-level neuroscience textbook is Kandel & Schwartz's _Essentials of Neural Science_, but if you have a strong quantitative background and are serious about understanding these things, you might want to check out Koch's _Biophysics of Computation_ (which sticks mostly to computation in single cells as opposed to networks of them) or Dayan and Abbott's _Theoretical Neuroscience_.
The quantum holography I was referring to is something completely different, where you perform holography with quantum entangled photons. I was afraid you might be referring to something like what you're talking about. As far as I know, the whole quantum uncertainty thing underlying brain function goes back to Sir John Eccles, a Nobel prize winning neurophysiologist who did some very good work, but was a very religious Catholic, IIRC. He desperately wanted to find God in the brain, and I think he was the first one to suggest it might be in some sort of quantum indeterminacy thing. He wrote a book with the philosopher Karl Popper (of "falsifiability" fame) called _The Self and Its Brain_ back in the late 70's where they discuss this. Then Roger Penrose came along and made a rather flawed argument as to why the human brain can do things that no computer can do in principle, and he's really the one these ideas are most commonly associated with now. Stuart Hameroff is a biologist who works on microtubules, and he's another person more directly associated with the whole microtubule hypothesis. A number of people outside of neuroscience believe these ideas, especially physicists who know very little about brain function. However, Penrose's argument suffers from a number of logical flaws (try typing "penrose wrong" into google), and Hameroff's arguments about the mechanism of action of gas anesthetics are laughable in light of huge amounts of ion channel data. Ultimately, no one knows if Penrose and Hameroff are right, as that's an empirical matter. But there's absolutely no experimental evidence or convincing theoretical reason to believe that they ARE right, other than the fact that many people desperately want them to be. And this makes sense--the whole notion that the brain is a mere machine subject to more boring laws of physics is something that threatens many people's self worth. Alternatively, there is decades of evidence supporting more conventional models of brain function. There are, of course, a few interesting anomalies. You will probably like this.
This comes from a Hank Wesselman [ph.d.] book - yet it's about the thing you try to repudate.
The above statement comes however from the man who founded this organization [noetic.org] - an Apollo 14 astronaut and theoretical physicist.
Theory? Maybe. But I want the proof either way, that is what science is about right?
I find it interesting that you think an astronaut/physicist would have insight into brain function than people who actually study the brain. Maybe we should start founding organizations to tell everyone the truth about how atoms behave and what's in outer space? At any rate, science is not about the "proof either way." When someone comes up with an interesting but highly implausible idea and they want other people to take it seriously, they have to have something more supporting it than "you can't prove it's wrong." Whatever that might be, Penrose, Mitchell (the founder of the organization you mentioned), and all those other people do not have it yet. Maybe they will someday, but I doubt it. The situation is much simpler for all the spoon-benders and mind-readers, who are consistently unable to demonstrate their abilities under controlled conditions.
Slashdot Mirror
I do computational neuroscience, and I run into those people all the time. No one I work with believes them, not because we can prove them wrong, but because there is no convincing theoretical reason or experimental evidence to believe that they are right. Other than the fact that they really really want it to be true.
However, I disagree with you that there is a discrepancy between performance of brains and estimates based on number of neurons, etc. This might have been true prior to us understanding more about computation that occurs in the individual dendrites of each nerve cell. Neurons can do a lot, and we still don't know exactly how much. But there's no evidence of mystical physical phenomena or profound ideological deficiency in current paradigms, it's just a complicated system that is very difficult to approach experimentally and is going to require some time to figure out.
The original poster meant to say "Lorenz," not "Lorentz." The Lorenz attractor is the classic demonstration of sensitivity to initial conditions in a coupled system of 3 or more nonlinear ordinary differential equations. This was the first (or one of the very first) demonstrations of chaos. A great mathematical insight that was made by a meteorologist, not a "real mathematician."
You can PCR out fragments longer than 3000bp, but it's hard, and it depends on a lot of factors. There are special "high-fidelity" polymerases that people use for these purposes. There is a whole field in industry dedicated to messing with aspects of PCR, mutating polymerases, and that kind of thing, to try to get longer PCR products, etc.
To address your comments, the reason why cloning animals is difficult has a lot to do with what is called epigenetics, or information stored in the chromosomes that is beyond the mere sequence of the genome. For example, reversible modifications to the chromosomes that activate or inactivate certain regions during the process of cell differentiation. "Cloning" a multicellular organism is the name given to de-differentiating existing adult cells all the way back to the differentiation state of a fertilized egg, then growing them back up into adult organisms. However, even if you have the genome 100% correct, problems with the structure of the chromosomes will lead to horrible birth defects (some textbook examples of this in humans are Angelman's and Praeder-Willi syndromes, which cause mental retardation and characteristic deformities).
The thing is, what they are proposing to do with this extinct marsupial is actually not cloning, but actually synthesizing DNA molecules and building one of these animals FROM SCRATCH. This makes the problem of chromosomal structure MUCH more significant than it is when you "clone" an animal using standard techniques. To address your second point, however, it is extremely unlikely they would try to turn a related animal into the extinct animal like they did in Jurassic Park. First of all, this would not work, and second, even if it did, you would just get a different third animal. The real obstacle here is creating artificial chromosomes out of individual genes, and this is where the work would have to start. Most likely, this would not be done in cells from some exotic mammal, but something very standard like a mammalian tumor cell line or maybe even yeast.
As for your third point, a female mammal can be cloned from a male in theory, and you could breed them. Obviously this is not a perfect arrangement, but it is the case with certain strains of genetically identical lab mice, and they're generally healthy. In any case, the small problems here pale in comparison to the giant problem of assembling artificial chromosomes. This hasn't even been done with bacteria, much less a multicellular eukaryotic organism.
They don't need two individuals, just one male. A male has an X chromosome and a Y chromosome, while a female has two X's. They can artificially create a female from that male by replacing the Y with an X. While this is beyond current technology, it is a lot less beyond current technology than splicing together entire chromosomes out of individual genes, which is what they will have to do in order to clone this animal. The fact that they have only one animal to work from will not by itself make the species go extinct again. Certain strains of lab mice are inbred to the point where all of them are genetically identical (except for the X and Y chromosomes between males and females), and they do fine.
PCR is NOT cloning. You can clone a PCR product by putting it into a plasmid (a small artificial chromosome) and inserting that into bacteria, but just doing the reaction by itself is not cloning.
Actually, much more than 20 or 22 are found in proteins, they're just not genetically encoded. This basically means that the protein is made with a certain amino acid in that spot, and it gets converted later. One of the classic examples is hydroxyproline, which is a modified version of proline (the only amino acid that's technically an imino acid). There are lots of hydroxyprolines in collagen, and conversion of proline to hydroxyproline is vitamin C dependent. So if you don't get enough vitamin C, you have lots of defective collagen which leads to teeth and hair falling out when you have scurvy.
The significance of the whole 22nd amino acid thing is only that it's the 22nd GENETICALLY ENCODABLE amino acid found in nature, and it's in some weird bacteria, not in humans. This means that it has its own tRNA. There are more than 22 amino acids found in proteins, even in humans, and in the lab they've created "alien" bacteria that use two or three completely artificial amino acids to incorporate fluorescent tags into protein molecules and that kind of thing.
Clinically speaking, there are many other causes of hemolytic anemia that are much much more common than anything involving the translation machinery at the level you're describing. It's also the same thing with hyper- and hypo-thyroidism. I'm not saying you're wrong and these things couldn't cause those diseases, only that they would be relevant in a tiny minority of cases.
There is some degree of truth in what you're saying, and this realization is the reason that there are lots of monte carlo-based modeling methods out there for intracellular processes. Basically, if you're modeling using differential equations and concentrations, that's implying that concentration is continuous while it actually is not. Sometimes it's actually necessary to model individual calcium ions, for example.
However, in talking about cracking a "code," you're implying that we somehow have all this information but we can't figure it out. The truth is that we don't have the information yet about large scale protein interaction networks in cells--I think this is the real problem that's closest to the "code" you're talking about, though it's not really a code as much as a jigsaw puzzle in high dimensional non-Euclidean space where we don't have all the pieces. Also, the reason that there has historically been lots of emphasis on chemical reactions on substrates, etc. in terms of dissecting biochemical pathways is because in the past that was all we could do. Since it's fairly easy to use techniques from organic chemistry to quantify rates of chemical reactions and so forth, and you know the next enzyme downstream is the one that acts on the product from the prior one, you can piece these things together with techniques that are primitive by todays standards (hence those giant evil charts they make you memorize in undergrad biochemistry). But we have a lot more advanced techniques now, so the focus has shifted more towards signal transduction (i.e. how the cell "knows" various things are going on, how they communicate with each other, etc.) as opposed to housekeeping functions like making sure the cytosolic ATP concentration doesn't drop too low (like old-school biochemistry). Obviously what I'm saying is a slight generalization/oversimplification, but I think the main theoretical problem in biology right now has more to do with giant networks of interacting proteins and figuring out the qualitative behavior of large networks from incomplete data that only tells you about what one tiny piece of it is doing. The theoretical underpinnings of this problem are likely to be grounded more in nonlinear dynamics than CA, IMHO, but what's holding us back right now is the fact that we can't get enough data, despite what everyone says about the big explosion of data in biology.
I still stick by my point that the main contribution of CA in biology will be modeling extracellular processes in developmental biology, though I'm not developmental biologist.
I'm doing a PhD in biology, so trust me: the map from gene to protein sequence is 100% fully understood and has been for many years. It was the basis of the 1968 Nobel Prize in Physiology or Medicine. In biology it's referred to as the genetic code. So when people throw around the term, they should understand that it has a very specific meaning and that it is well understood. Anyway, the mapping from protein sequence to protein structure is the one that is not well understood. Well, it's not an issue of whether or not it's well-understood, it's just so complex that doing it based on the physical properties of the actual atoms involved requires profoundly unattainable amounts of computing power. So structure prediction methods generally fall into two categories: ab initio with fancy mathematical shortcuts, and "threading," in which evolutionary relations between proteins are exploited so that you can use known structures to help predict similar unknown structures. CA are unlikely to help with this.
CA models ARE likely to be helpful at the level of multicellular organization. Hepatocytes in the liver somehow self-organize into a hexagonal lattice...how? This type of thing happens all the time in development, but it's hard to understand how cells can self-organize without any kind of overarching supervisory signal (though such things sometimes exist in the form of concentration gradients of signaling molecules, etc.) Anyway, I don't see any immediate application of CA-type methods to the study of intracellular processes, which is not to say that there aren't any.
I think it's becoming widely acknowledged that these kinds of things are very important, but I have no idea what the NIH has done specifically. It's usually not mathematicians from the math department that end up collaborating with biologists in my experience, but physicists and applied mathematicians who are often in engineering departments.
I'm not trying to argue that the NSF's positions aren't important, only that there is a certain amount of jurisdiction in terms of what the NSF and NIH will fund. The NSF spends very little money on breast cancer research. Not because it's not science, and not because they think it's not important, but because the NIH funds that. I agree with pretty much everything you're saying, I just don't believe that the NSF increasing funding to mathematics applied to biology indicates in any way that the lack of advanced mathematical techniques outside the grasp of people in CS is the underlying bottleneck in bioinformatics. Math is obviously important, but I think that improved mathematical analyses will be much more important in other branches of biology. I say this as a computational neuroscientist, but I'm referring more to systems biology. Like you said, there are a lot of blurry lines.
You make it sound like what the NSF does reflects what science as a whole thinks is important, but the NSF funds physical sciences. Biological sciences are funded through the National Institutes of Health (NIH). I'm not disagreeing with your main point, though I think you're confusing computational biology with bioinformatics. In any case, the hard part theoretically is dealing with the sheer complexity of biological systems (dealing with a gigantic number of coupled equations which are individually fairly simple), which is different than most problems in theoretical physics, for example. This requires a different way of looking at problems than the way most people usually learn math. However, in my opinion, the main bottleneck in biology right now is still an experimental one, not a theoretical one, despite all the data people have been generating with various high-throughput experimental methods.
For those who don't know what near-field optics (scroll down and look at the diagrams) are, it's a way to overcome the fact that normal far-field optics can't resolve things smaller than the wavelength of the light you're using. In near-field, you use a tiny fiber optic probe that's smaller than the wavelength of the light so you can see things smaller than the wavelength of the light, i.e. pack data more densely. Imagine you had terrible vision, such that you could only see at the resolution of a baseball at arm's length. By looking through a straw and moving the straw methodically, you could see with much higher resolution.
One problem with this is that contrary to what you're suggesting, it is unlikely that they would ever be able to make multiple layers on a disc because you can't focus the light onto lower layers. It's also questionable how durable the disc would be because you have to put the probe so close to the surface of the substrate storing the actual information that you couldn't have a (relatively) thick layer of plastic protecting it like on a CD. There are also things that make this difficult in practice, like the fact that it's easier to focus light precisely than it is to move a microscopic probe precisely.
Another thing is that it's not entirely obvious that these discs will be faster than CDs or DVDs because the process of pulling data off them is intrinsically more difficult. You can't assume that you can spin the thing as fast as a CD and it will work flawlessly. In addition, there are ways to increase the speed of CDs or DVDs by rapid scanning of the light beam, so they will always have the potential to be faster. I agree with you, though, that it's stupid to compare this to CD burners.
I think you're being a little too harsh...first, the things you're saying aren't unique to CA. In any case, if Wolfram is right and the universe is organized that way, the limitations you describe are inconveniently imposed by the nature of reality, not by the mathematical language used to describe it. If you think math is powerful enough to change that, you've been watching Pi a few too many times.
More importantly, it's important to realize that just because there are nonlinear differential equations involved, it doesn't mean the attractors are necessarily chaotic. And even if chaos is involved, that doesn't automatically render us powerless to predict or understand. I'm skeptical about the universal power Wolfram claims his ideas will have, but there are certainly areas where CA will be useful. Personally, I think it's the future of theory in developmental biology. For example, there are complicated but highly stereotyped arrangements of cells in the cortex of the brain, and it's unclear how individual cells can "know" how to arrange themselves like this in the absence of some overarching presence telling them where to go. I doubt Wolfram's book will unravel the mysteries of the universe or anything, but I think it's premature to write off cellular automata as "useless."
Well, I guess it depends who you ask. Although protons technically bond covalently to the atoms of the side chains, very few biochemists would think of that as becoming part of the protein. Plus, the protein is translocating them. Well, in any case, this new thing is a big achievement, and the molecules involved are certainly easier to work with than something like bacteriorhodopsin.
There's actually been some interesting work done on that based on this nonlinear phenomenon called stochastic resonance. There's a link to a paper here.
BTW, the phenomenon is called hyperacuity, and it specifically refers to having higher resolution than should be allowed by the size of the photoreceptor cells in the retina. I state this explicitly because it is not something due to a more optics-related phenomenon like the fundamental wavelength limit of the light or anything like that.
We're all skating on some thin semantic ice here, but bacteriorhodopsin DOES operate mechanically. It actually does mechanically pick up individual protons and move them from one place to another. The protons are not somehow magically absorbed with new protons being produced on the other side. If you don't believe that, there is a related protein called halorhodopsin that does the same thing with chloride anions.
It's true that these molecules exist in nature, but that doesn't change the fact that people are experimenting with using them for all kinds of completely non-natural things like high-density optical memory, bizarre types of sensors, etc. If you type bacteriorhodopsin into the patent database, there are 197 patents associated with it, and probably half to two-thirds of those are not biological applications. Would you feel differently about the molecule if someone synthesized it using a peptide synthesizer? I don't know if that's been done or not... And I'm not saying that the springboard thing isn't different and potentially more useful, just that bacteriorhodopsin is a molecule people use that directly transduces light energy into mechanical energy.
But you're pulling energy out of the system by moving the board, and that energy has to come from somewhere. I don't know enough about the physics of what's really going on, but this is simple conservation of energy. However, I doubt that the force comes from electromagnetic dipoles. I think it's more likely that the light bumps up an electron, causing that atom to have a different type of orbital, bringing it into conjugation with other orbitals, which changes which conformation of the molecule ("long" or "short") is most stable, or something along those lines. But I am not a physicist or physical chemist by any means. However, bumping the electrons up to higher energy orbitals DOES directly convert the energy of that photon into a different form, and those electron orbital changes are the basis for the movement, so this system IS directly transducing the light into mechanical energy.
Unless you count bacteriorhodopsin, a photosynthetic protein in certain bacteria. Light energy is used to change the physical shape of the protein and move hydrogen atoms from one side of a membrane to the other. This is direct conversion of light to mechanical energy (although it is not usually used that way), and since the 70's there has been a lot of work with this protein making different sensor arrays and so forth.
If it weren't energy conversion, you would only be able to bend the board once. Otherwise you would be getting energy from nowhere, or "using up" the molecules. But if you were doing that, the thing would only work for a very short period of time, not a whole day. There are other mechanisms to explain the photodamage to the azobenzene. I don't know anything about this molecule in particular, but in general irreversible light damage occurs through either free radical reactions catalyzed by the light or bumping the electrons up to an energy state where they can fall to two (or more) separate states. One of these is the one you want, and the other one isn't, and each time a certain percentage of the molecules will get stuck there and not work any more.
I used to work on visual system neurophysiology, and I just want to say that I agree completely about the long-term future of retinal implants. People with a lot of exposure to electronics and computer technology and very little exposure to real biological science (e.g. the average /. reader) have a strong tendency to think that we will be able to replace or improve upon biology. However, most biological systems are already running at a much higher efficiency than pretty much anything humans have ever been able to build. For example, the retina can go out on a bright sunlit day and look at stuff, then go into a dark room and 30 minutes later be fully adapted to SINGLE PHOTON SENSITIVITY. This is an extremely complex process that occurs at the level of the retina--it's not just your pupil changing size. Another thing is that there is retinal circuitry for anticipation of movement. If you work out the math, a person's waving arm can travel 2-3 feet by the time the signal gets from your retina to occipital cortex in your brain, yet you perceive its position accurately because of compensatory mechanisms all the way down to the level of the retina. No artificial retina design has even begun to think about stuff like that.
I'm no expert on CCDs, PMTs, etc., but I seriously doubt that the technology to build an artificial device capable of that yet implantable in the eye will be available in my lifetime. Putting new cells in the retina, on the other hand, will most likely be possible in less than 25 years. Which are patients going to prefer?
One of my med school professors is the ophthalmologist for the Yankees (IIRC...a major league baseball team, anyway). He told us that he went and did eye tests on the whole team, and their average eyesight was 20/7.5, meaning that at 20 feet they can see what a person with 20/20 vision can see at 7.5 feet. I think that's insane.
The behavior of receptors (the actual membrane proteins that do something when neurotransmitters bind to them) themselves is actually quite straightforward and easy to quantify in a large percentage of cases. The big problems come afterwards. One of them is understanding how the input from the receptors is added up into an output in terms of firing. If you have access to a university library, you could get a copy of a review article here. However, firing is NOT binary or remotely computer-like in any way. The whole idea that there are useful brain-computer analogies to be drawn is entirely misguided and has poisoned the way too many people think about brain function. Firing is not necessarily a discrete event because there are multiple types of firing (burst spikes vs. normal spikes, etc.). Also, the temporal PATTERN of spikes matters and influences how later information is processed both pre and post-synaptically. And how a neuron processes the spatiotemporal pattern of inputs it gets from the thousands of other neurons it's connected to is not understood. It's extremely complex, and unfortunately not the kind of thing you can just read a web page and pick up. A good low-level neuroscience textbook is Kandel & Schwartz's _Essentials of Neural Science_, but if you have a strong quantitative background and are serious about understanding these things, you might want to check out Koch's _Biophysics of Computation_ (which sticks mostly to computation in single cells as opposed to networks of them) or Dayan and Abbott's _Theoretical Neuroscience_.
The quantum holography I was referring to is something completely different, where you perform holography with quantum entangled photons. I was afraid you might be referring to something like what you're talking about. As far as I know, the whole quantum uncertainty thing underlying brain function goes back to Sir John Eccles, a Nobel prize winning neurophysiologist who did some very good work, but was a very religious Catholic, IIRC. He desperately wanted to find God in the brain, and I think he was the first one to suggest it might be in some sort of quantum indeterminacy thing. He wrote a book with the philosopher Karl Popper (of "falsifiability" fame) called _The Self and Its Brain_ back in the late 70's where they discuss this. Then Roger Penrose came along and made a rather flawed argument as to why the human brain can do things that no computer can do in principle, and he's really the one these ideas are most commonly associated with now. Stuart Hameroff is a biologist who works on microtubules, and he's another person more directly associated with the whole microtubule hypothesis. A number of people outside of neuroscience believe these ideas, especially physicists who know very little about brain function. However, Penrose's argument suffers from a number of logical flaws (try typing "penrose wrong" into google), and Hameroff's arguments about the mechanism of action of gas anesthetics are laughable in light of huge amounts of ion channel data. Ultimately, no one knows if Penrose and Hameroff are right, as that's an empirical matter. But there's absolutely no experimental evidence or convincing theoretical reason to believe that they ARE right, other than the fact that many people desperately want them to be. And this makes sense--the whole notion that the brain is a mere machine subject to more boring laws of physics is something that threatens many people's self worth. Alternatively, there is decades of evidence supporting more conventional models of brain function. There are, of course, a few interesting anomalies. You will probably like this.
This comes from a Hank Wesselman [ph.d.] book - yet it's about the thing you try to repudate. The above statement comes however from the man who founded this organization [noetic.org] - an Apollo 14 astronaut and theoretical physicist. Theory? Maybe. But I want the proof either way, that is what science is about right?
I find it interesting that you think an astronaut/physicist would have insight into brain function than people who actually study the brain. Maybe we should start founding organizations to tell everyone the truth about how atoms behave and what's in outer space? At any rate, science is not about the "proof either way." When someone comes up with an interesting but highly implausible idea and they want other people to take it seriously, they have to have something more supporting it than "you can't prove it's wrong." Whatever that might be, Penrose, Mitchell (the founder of the organization you mentioned), and all those other people do not have it yet. Maybe they will someday, but I doubt it. The situation is much simpler for all the spoon-benders and mind-readers, who are consistently unable to demonstrate their abilities under controlled conditions.