Science & Technology Q&A for Kids (and others) [Part 134]

Sam. Hello, everyone. Welcome to another episode of Science and Technology, Q and A for kids and others. And I see all sorts of questions stored up here. One from Aaron. If human reaction speed was faster, would that be helpful? How much faster could it be? Is the limiting factor the nerve signal relays or brain processing time? Okay, interesting question. So, first of all, what is human reaction time? It's like, you see a light flash, you press a button. Takes about 0.3 seconds, give or take, and 300 milliseconds, 300,000th of a second. I think it varies a bit with, you know, Probably varies a bit with sort of age and alertness and all that kind of thing, and people who play video games all the time. Maybe it's faster. I'm not sure. I'm not sure how trainable it is. I do know that, for example, if you're tapping your finger, that's one of those things where the speed of tapping your finger people typically are at their fastest at age 20 or something. They're not at their fastest at age five. And it sort of steadily declines from there. So I'm told the. I think the question of what's taking that 0.3 seconds, it's not the nerve impulse time, that is, ooh, let's see, we can probably work it out, but it's a small fraction of that time. I believe, I remember working this out at one point that if you just look at nerve conduction speeds. So in order to. Essentially, what has to happen is, you see the light flash, things have got to happen in your brain. Then your brain has got to generate an electrical signal that goes down nerves that come from your brain to your spinal cord, out from your spinal cord into the nerves in your arm, and eventually to the nerves that actuate the muscles in your finger. And then there has to be an electrical signal that causes the fibers in your muscle to contract, and then you move your finger. And I think that the kind of. The effective distance that one goes is around 100 meters. That is that it's like. It's as if we were a huge animal. You know, the brontosaurus or something. Maybe brontosaurus quotes don't exist anymore. In addition to having gone extinct as dinosaurs. Dinosaurs. I think the naming of dinosaurs got changed as a result of discovering that museum specimens weren't assembled the way they were supposed to be assembled and so on. So the brontosaurus may be doubly extinct. But anyway, it was a very long animal that had a very big distance from the brain in its head to the tip of its tail. And you could imagine that the reaction time for flipping its tail would be determined by the time for the signal to go from the eye of the brontosaurus in its head down to the tip of the tail, so to speak. And that could be a long distance. It won't be 100 meters. There's no creature that big, I think. No, even whales are not that big. But. And maybe I should use the example of a whale as something which has a sort of fairly long distance to get from its brain to its tail, so to speak, in any case. So part of the time is spent in nerve conduction from the brain to the finger, but that's a small fraction of the time, at least for us humans. And most of the time, it is, in effect, signals bouncing around inside our brains. And what tends to happen is that you have to kind of form the thought that you're going to move your finger. And you can actually measure, if you look at electrical signals in the brain, there is a motor strip, I think it's at the top one's head around here. That is the part that is sort of actually, in a sense, connected to the nerves that go to fingers and so on. And there's a premotor strip which starts showing electrical activity sooner than the motor strip. I think it happens at like 0.1 seconds or something, whereas it takes 0.25 seconds or more to get sort of proper activity ready to actually drive your finger, so to speak. And back in the day, people were thinking about having things that will pick up electrical signals from the brain as a way to control military planes faster than they could otherwise be controlled, things like that. It's in the category of might not be a good idea because it might be the case that you're starting to form the thought, I should press the trigger, so to speak. But then some other signals inside your brain, maybe from your frontal lobes or some such other thing, come in and say, no, no, no, that's a really bad idea. Don't do it. And that all happens in the full 0.3 seconds. But in 0.1 seconds, if you were to pick up kind of the first beginnings, the formation of that thought, to press the button or something, then you would end up concluding you should press the button a lot of times when actually it was a really bad idea to do that. But in any case. So I think that the vast majority of the time of reaction time is spent sort of bouncing around inside your brain, kind of trying to form the thought, decide what to do. Set things up so that you're going to send that signal to the nerves that go to the finger, so to speak. Now one can think about, as one understands more about how neural nets are put together and so on the brain. Every neuron kind of, it's like a relay. There's one neuron fires, produces an electrical signal. It affects another neuron, another neuron, another neuron. There's this whole sort of train of neurons, and the brain is probably arranged in sort of layers of neurons, a little bit like the neural nets that exist in the chatgpts of this world and so on. And you can ask questions like, how many layers of neurons is it going through? What's the delay going from one neuron to another neuron through the synapse, where there's a connection, where one neuron has a sort of a piece, an axon, that sticks out from the neuron that then has actual ions that go from one, from one, sort of transmit the electricity from one nerve cell to the next. And that happens, I think, more slowly than sort of the. Than the sort of conduction down the wire that has one piece of the neuron. And typically, neurons, if you look at neurons in the brain, for example, they have sort of a central body, and then they have lots and lots of kind of very long tentacles, maybe 10,000 long, long kind of tentacle pieces. Sometimes they just go straight from the body, sometimes they branch and so on. And those are the things down which electrical signals are transmitted to reach other neurons. And the fact that there are sort of so many connections from each neuron is kind of one of the big things about brains is that the neurons are connected to many other neurons, and there's a very complicated network of connections. That's what we see in chatgpts of this world, where there might be, in a chatgpt, there might be a few million neurons, but 100 billion connections. So every individual neuron has a very large number of connections in the chatgpt artificial neural net, as well as in things like brains. So I think another question you might ask is, when you look at human reaction times, how much variation is there? I mean, if you're doing the. Press your finger, press your finger, press your finger, seeing the light, so to speak, and you do that many times. How much. What is the kind of curve of, you know, do you ever hit it in 0.1 seconds, or is it always 0.28, 0.31, you know, this kind of thing? I saw those curves actually, fairly recently. They. They kind of come up fairly steeply and then they have a pretty long tail. So it isn't a distribution where it's sort of a. Oh, there's an average kind of reaction time and it's a kind of bell shaped curve, a Gaussian curve, for example, around that average reaction time. It's a, it's a, it's a thing. I'm not quite sure what the distribution is, but it's a distribution that is sort of lopsided. The chance that you have a very, very short reaction time is lower than the chance that you have an unexpectedly long reaction time. Actually, it's sort of interesting because you can kind of ask the question, is it almost as if there's sort of an artificial temperature to the brain where, because there's curves that describe, for example, the speeds of molecules in a gas, things like this, and the distribution of reaction times looks rather like that curve, the Maxwell distribution. I don't know whether that's totally coincidental or not. I'm suspecting it might be rather coincidental. But now the question is, well, what if our nerve transmission speeds were higher? Well, so in the case of actual nerves, the myelin sheath around the nerve is important to having fast transmission speeds. And so in diseases like multiple sclerosis that destroy that myelin sheath, one of the problems is that that transmission process does not work as well. I think it's more leaky, but I think it's also, I think it's also slower, but it's certainly more leaky. And so I think that's an important. You can imagine how would you get faster transmission of nerve impulses down a nerve? Now, I think that the squid has this giant axon, very big piece of a nerve cell. And I think one of the things that a squid is really good at is you poke one end of the squid and the other end of the squid responds. So you poke one end of the squid and the squid will splurt out a bunch of ink or something like this. And I think one of the things that is true about the giant axon of a squid, much, much thicker, bigger transmission line than we humans have. Very convenient for people who do experiments on nerve cells because they can use the axon of the squid. It's much easier to stick an electrode into that than it is into little tiny nerve cells that exist in things like humans and mammals and mice and whatever else. And I think that one of the things that's true about the squid giant axon setup is that it has faster transmission of nerve impulses. I Have to say, I'm not completely immediately sure. Let's see. Can I work out why that is not immediately, and it may not be obvious why it works that way, but. So one could imagine a situation where you have sort of bigger, thicker nerves that transmit data more quickly in us humans, I think, to imagine, could you speed up the brain? Could you make it so that the nerve signals get transmitted more quickly? When you put various kinds of chemicals, drugs, and so on into the brain, they definitely don't tend to have that effect because they affect the transmission of nerve signals from one nerve cell to another. They affect kind of the environment of the nerve cells. And they can do things like, for example, alcohol effects transmission from one nerve cell to another, and I think makes that transmission less effective than it would usually be. But I don't think anybody knows. Oh, you take this chemical, you put it in your brain, and suddenly all the transmissions were more important. Actually, no, that's not right. There are things that enhance the transmission from one nerve cell to another. And that's not a good thing, actually, because that means it's kind of like, well, you can have more thoughts, but having more thoughts will tend to be more confusing rather than. I think a key thing in the way that we think is we tend to think in this kind of linear chain of experience. And if you start saying, well, let's just have more thoughts happening, then the thing tends to get sort of more confused and it's less coherent kind of process of thinking. But I suppose the question might be if you. So I don't think sort of increasing the rate of transmission from one nerve cell to another is probably something that in normal circumstances is a very good thing to do. Increasing the speed of transmission along a nerve fiber. I'm not sure how you would do that other than by having a different nerve, nerve fiber or nerve, different collection of nerves. I do wonder a little bit for the very big animals how the transmission of things down these nerve fibers works. Now, by the way, it's worth remembering that squids in the deep ocean can get very, very big. Squids have this feature. It's kind of convenient if you're a conically shaped animal that you can just keep growing. It doesn't, you know, there's. Your shape is a kind of a similar shape. It just keeps growing, growing, growing, growing, growing, getting thicker and thicker. Kind of like a tree can potentially keep growing with its trunk up and up. It doesn't. So squids can get pretty big. Now, it's always said that lots of dinosaurs for example, were sort of said to have multiple brains. That was a very common kind of claim, at least when I was a kid. And like the stegosaurus, it's got a little tiny brain in its head. And you can see it's a tiny brain because its skull doesn't have much room to have a bigger brain. And then the question is, well, how did it, how did it wag its tail? I have to say a stegosaurus has this big kind of assembly of spikes at the end of its tail. So wagging might not be quite the thing for the stegosaurus. The happy stegosaurus might not be a tail wagger, I'm not sure. But in any case, the thing that was often claimed is that it sort of had a secondary mini brain at the back that would sort of help deal with things happening at the back of the stegosaurus somewhat independent of what's happening in the main brain, so to speak. Now, the fact is that all animals, the walking motion, for example, is partly controlled by ganglia, clumps of nerves in the spinal cord. And that means some aspect of the way one walks sort of isn't directly controlled by the brain. It's something where your brain is just saying, okay, go ahead and walk. And the actual motion is being orchestrated by some sort of lower level set of things in the spinal cord. And presumably that's kind of the sense in which the stegosaurus had a secondary brain. It wasn't that it could have a discussion, you know, okay, head, what do you think? Tail, what do you think? And it could have sort of a back and forth. I don't think that was what was happening. Just like I don't think we get to have a back and forth with the ganglia and our spinal cord. And in fact our impression of the world, our impression of consciousness is there's only one place where things are happening, so to speak. I mean, we don't have an internal sense of where that place is, but the impression from sort of the science side of it is it's all happening in the brain. It's not like, oh, you know, I'm trying to solve this problem, let me solve it in a ganglion in my spinal cord. I don't think so. I don't think that's how it works. I think that those ganglia are used merely for kind of the. Once you've sort of decided what you're going to do, make it happen. Now, some of those things happen independently of the brain. Like for example, the standard knee reflex where you where you kind of hit right below your kneecap and your leg kind of pokes out. That is a reflex that happens purely in the spinal cord. Interesting question is how much faster that happens than something that has to knock around your brain. My assumption is it happens significantly faster. And in a sense, that's an example of, well, you don't think about it. And so it can sort of just use the pure nerve cell loop rather than having to go through all of that complicated stuff in the brain where you say, should I press the button or not? So anyway, some thoughts on kind of reaction times, nerve cells, nerve transmission, stegosauruses and other kinds of things. Let's see. Anon asks, do I find it odd that on the earth, animals with bigger brains are more intelligent, but in technology, smarter computer chips always seem to be smaller? I'm not sure that either of those claims is quite right. First point is if, okay, what are our brains used for? Part of it is controlling all of our muscles. You need a certain amount of brain just to be able to control all the muscles you have. So if you're a whale, you need a big brain just because you're going to have kind of the base connections, the base neurons that are going to deal with the muscles that are in some piece of your tail separate from muscles that are somewhere else. So I think the scaling of brain size with body size is a little bit inevitable just because you have to kind of have the base place where each muscle is going to be controlled. And an interesting question, when the animal gets bigger, why doesn't it just get bigger muscles rather than more muscles? I think it does get bigger muscles, but I think it also ends up with more muscles. And exactly how that scales I don't know, and I don't remember ever seeing something about that. So the question of once you have a brain of a certain size, is it the case that the sort of the bigger the brain, the smarter the critter? I think there's some evidence for that, but it's a little bit unclear. Like for humans, there was a big fashion a hundred and. Well, a bit more than 100 years ago, there was a whole field of science called phrenology that had to do with the shapes of people's heads and conclusions about the personality of the person based on the shape of their head. It's one of those areas of science that fell into absolute disrepute. I don't know whether there's any sensible, meaningful things that could be deduced from it. What tends to happen in Science is people get very enthusiastic about something and then sometimes the enthusiasm is too hyped up and the thing just crashes completely. And then it's lost for 100 years or 200 years. And people say, oh, no, no you don't. You know, that's a bad idea type thing. I suspect phrenology was used for some rather silly purposes and you know, things like, is this a good person or a bad person? Let me see whether there's a bump on the side of their head in this particular place. That's a very unconvincing direction. But the. So, you know, there was a whole lot of work that was done on kind of, on a whole lot of kind of these crazy maps that showed, oh, you know, if your head shuts, if your skull juts out in this place, that means you're a good forward thinker or something like that or whatever. But I think the question of the sort of correlation between kind of brain size and mental capacity is a rather unconvincing one. I mean, for example, there are people who have the misfortune of having high pressure of fluid in their brains, which means that the sort of central region inside the brain that has fluid in it is big and the actual brain is sort of squashed to the outside of the skull. I know that there are people who are perfectly, happily leading their lives and doing all sorts of clever things and they had no idea that, that the volume of their brains was actually quite small until somebody did an MRI and discovered that. And, oh, that's interesting, but doesn't really make much difference to them. So there's, you know, that's sort of a counterexample to the kind of idea that the brain volume is what determines sort of mental capacity. It could be the case that in some of those situations the total number of nerve cells is still large and that what matters is the number of nerve cells. Now, you know, one of the things that was for a long time a real mystery was how many nerve cells do we actually have in our brains? And you might say, well, why don't you just look and see? Well, the problem is that if you look under a microscope at a little slice of brain, you'll see all kinds of neurons of nerve cells there. But I mentioned before that each nerve cell can have like 10,000 sort of tentacles sticking out of it. And so being able to figure out what the kind of connection between this nerve cell and that nerve cell and the other one is from microscope size is really difficult. There's a big sort of citizen science project, trying to trace some of those things. But I think what happened recently, and I was always mystified why this wasn't a generic thing to do, is the brain has all these nerve cells. They're all kind of held together by all sorts of sort of connective material. If you could just sort of dissolve everything, wouldn't you just wind up with some number of nerve cells and you can just count them, and it's rather easy to count cells. There are all sorts of devices where you let cells drop down in cell cytometry, Where you kind of drop individual cells down and you have kind of laser devices that are just counting the individual cells as they drop down. And why couldn't you just do that, sort of dissolve a brain and just see how many nerve cells there are there? Well, finally, people have done that, and the claim is for humans that There are around 86 billion nerve cells at least. I've seen that number a bunch of times. I don't know whether that's one particular brain or whether that's sort of closely. Whether all brains are sort of closely connected with that number of nerve cells. But it's on the order of 100 billion nerve cells. And I don't know how much variation there is. And that would be something one could ask about. But I think then the question is, if you have more nerve cells, is it a good thing? Well, we have a little bit of evidence from the chat GPTs and the large language models and the artificial neural nets that it does seem to be the case that having more weights, more connections between artificial neurons does seem to have an effect on the performance of these large language models. And maybe there's even a threshold below which it's kind of hard to have enough linguistic information, enough common sense knowledge, to be able to sort of do human like discourse, so to speak. And maybe that also is a limit. Maybe it's the case that sort of the reason cats can't talk like humans in part is they just aren't above that threshold in the number of nerve cells. Given the complexity of the language that we humans have developed, which after all we got to develop because we have that number of nerve cells, so to speak. And maybe when the sort of superhuman aliens arrive, not that I think that's quite how it will play out, that have, you know, brains that are 10 times the size of ours, maybe they'll have a language that has more words in it. I'm not sure. I don't tend to think that's how it's going to work, forget the absence of the aliens. But even if, for example, you know, let's imagine, and there's no evidence that this is the case, that human evolution is going in the direction of sort of bigger brains, there's no evidence, to my knowledge, that that's the case. There's no evidence that over the course of recorded human history that our species has changed the size of the skull that encloses the brain. Which, again, is kind of. If you believe that being smart is in some sense is kind of of value to an individual in the species, you might think that in the course of biological evolution that you would see sort of bigger brain cases to enclose bigger brains for the smarter humans, who are the more successful humans, so to speak. There doesn't seem to be much evidence for that. So I think, although our species does have a larger brain enclosure size than previous hominid species, But I think. So this question of sort of more neurons, bigger brains, is it better? Do you get to know more if you have a bigger brain? It's not even obvious that. Well, I have to say that my own personal anecdotal feeling would be the more, you know, the easier it is to solve problems. But I'm not sure that that is people's universal impression. Sometimes it's the more, you know, the more distracted you get. But personally, I don't think that's the case. But that's just an anecdotal claim based on my personal experience rather than some sort of scientific conclusion about sort of what it takes to solve problems versus what it takes to know things up front, so to speak. Now, the other part of this question had to do with computer chips and so on, and whether as they get smaller, they get smarter and not really. I mean, the sort of. The chips that now have like billions of transistors on them, these are pretty big things. I mean, they're several inches across. Their size is usually limited by things like the size of piece of silicon crystal, which you can have without having defects in the crystal and things like this. Although there are technological ways around that where you make something that has defects in it, you test it, you realize, oh, that part doesn't work properly. This is commonly done for memory chips. That part doesn't work properly. So let's just blow out some sort of fuses on the edge of the chip that will mean that whenever we're sending data into the thing, it avoids the plots that don't work. So it isn't necessarily limited by that. But the thing that does happen is that one tries to pack sort of denser and denser circuitry into microprocessors and sort of the big threshold now is 5nm, which is that's essentially the size of the wires. So that's. Five billionths of a meter is the, is the size of the size of the features, size of essentially the wires. Now it's good to have smaller wires because the smaller the chip is, the. The less time it takes for electrical signals to go from one part of the chip to the other. And the smaller the wire, the less there's kind of capacity, the less sort of electrons that have to build up to send that signal. By the way, makes me wonder, in the case of brains, if you had a really big brain, maybe you would think really slowly because you would have to wait longer for the nerve impulses to go from one side of the brain to the other. I'm not sure that's a significant effect because I think that the typical time for, well, let's see, we can probably work it out, it's a few milliseconds for one between firings of different nerve cells. So, yeah, that's interesting. So I'm not sure if it's a significant effect to say, oh, you have. Well, if it had to go through more junctions, so to speak, to get from one side of the brain to the other, if sort of your brain, instead of being a length 50 sort of relay, chain of nerve cells, was a length 300 relay chain of nerve cells, so to speak, maybe you would think more slowly because you'd be waiting for what does the other side of my brain think about that? And sort of coming to consensus in what your brain wants to say next, so to speak, might take longer. In the case of an artificial neural net, for example, the difference between ChatGPT, GPT3 GPT4, all these kinds of things, it's bigger networks and the thing that can happen. I don't think we know precisely how some of these systems are built, but in the case of ChatGPT, the original version at least, there were sort of 400 layers that were involved. So it took kind of 400, a sequence of 400 computations. And each of those 400 computations was pulling in data from maybe a million different artificial neurons. But there's 400 sequential computations that have to get done to get sort of the next token of result. And so one could imagine in the case of a human brain that there's sort of this question of how long is that relay? And the Longer that might be, the slower the brain might think. So that's an interesting question. I don't know the answer to is what the reaction times of critters like whales and so on are. My kind of impression would be they're quite slow, but I don't know for sure. Although, as I say, the squid is fast, I think. So in terms of microprocessors, computer chips and so on, the individual features are getting smaller and that allows the chip to run faster, as in. So the way a chip works, it's doing sort of succession of instructions that come from the program you've specified. And the question is, how fast can it do those instructions? And modern chips will do maybe 3 billion instructions per second, so to speak. So there's a question of how fast can you have successive instructions be done? And the smaller the features, the lower the transmission times, the faster you can have successive instructions be done, because you can't really have the next instruction happen until there's been time for an electrical signal to go across the chip to deal with finishing the previous instruction. Now, in actuality, chips, modern chips have all sorts of cleverness and all sorts of things where you're storing data very close to where it's going to be used. So called levels of caches, L1 cache, L2 level 1, level 2, level 3 cache that are sort of progressively higher speed. You know, the level 1 cache is the highest speed, sort of closest in to where the data is going to be used cache. And what can take a lot of time in a, in a CPU is fetching data from memory. And so if you can prefetch things, if you know what data is going to be used and you can get it ready, then you don't have to wait that time. So a typical thing would be if you have a whole array of data you're working on, well, let's say you're working on some image or something like that, and you kind of know you're going to need this next piece of the image. You can kind of load that into this fast cache memory in advance of when you actually need to compute with it so that the actual computation can happen more quickly without having to say, oh, now I need this piece of what's for memory. I have to spend a lot of time going out to main memory and going to fetch that. And similarly, there are all kinds of tricks in microprocessors where things like, well, what instruction are you going to do next? Well, let's guess what instruction is going to do next and let's be going to be done next and let's just go ahead and do it. And even if there's a sort of branch where it says, well if this happens, then go do that instruction, if that happens, then go do that instruction. It's pretty common in modern microprocessors to do so called speculative execution where you just follow all the branches. It's kind of like the multi way systems that we talk about in our physics project where you're kind of following all the paths of what can happen. Or like a game graph where you're looking at playing a game and you're saying what are all the possible moves you can make? Speculative execution unfortunately also ends up being a source of all sorts of computer security issues as you kind of watch what happens on the speculative branch and all kinds of different things happen and you can kind of kind of deduce things you weren't supposed to be deducing based on what happens on that different branch. Let's see other comments here. Let's see, DES is asking about neuron connections allowing sort of precognition of. I think sort of the best one could hope for is figuring stuff out more quickly, not figuring out things from the future of time, so to speak. Let's see, Harry is asking, do you think the brain can be trained or not? Like a muscle can be trained? You know, I think when you train your muscles and you get stronger and all that kind of thing, I think you're getting additional fibers in the actual muscles and that that's a big part of what's going on in the brain. It always used to be said that you were born with a certain number of nerve cells and all that would ever happen in the course of your life is that nerve cells would die, you'd never get new ones. That turns out not to be the case. There are new nerve cells that are produced. The rate of new nerve cell production and the amount that sort of thinking every day does to make more nerve cells get produced, I don't think anybody knows. I think that the question I have to say, I know in my own very anecdotal case, I do feel that I think better for having thought more, so to speak. The fact that I have sort of do a lot of thinking helps me to do a better job of thinking. But I suspect part of the reason for that part of it may be physiological. As in more nerve cells are produced. Maybe, possibly, I'm not sure. Part of it I think is that one has certain patterns of thinking and one gradually learns to sort of say, well, if I'm going in this direction, I'm going to end up here. Well, let me remember that. If I go in this direction, then I end up here. I'll just remember that directly. So then the next time around, I don't have to go through and work out this and this and this. It's like I go, you know, well, I could go one plus one plus one plus one plus one. And I know I can add up all those ones and I can conclude at 7 or something. Or I can just remember the rhythm of those ones and say, oh, that's seven. And I can kind of then go directly. It's kind of a. How do you beat sort of the computational irreducibility of what you have to work out? Well, you just remember the answer. And I think increasingly, if one has certain types of things one thinks about. I mean, in general. Yeah, I mean, in general, one of the things to say about brains is that training. One can get trained to do all sorts of things. If one's doing some complicated kind of activity, whether it's, well, riding a bicycle, flying a plane, doing some kind of programming, whatever else it is, gradually, as one does more of it, one can get to the point where one can both do it more quickly and do it more reliably, but particularly more quickly. And my guess is that happens because we're basically just establishing memories that are, if this, this, this happened, boom. You know the answer. You don't have to go think it through. You just know the answer. It's kind of like a caching strategy where you're. You're just saying, I just store the result. I don't have to compute anything. So that's my suspicion here. JH asks, how will brains change through neural links connecting directly to AI? I've talked about this a few times. I think that this whole question, first of all, there's the whole question of what does the neural user interface look like? And what's the way that you get this collection of electrodes sort of connecting to a piece of your brain? How do those electrodes sort of present their signals to the brain? And it's kind of like our eyes, in some sense are doing this. They're a bunch of photoreceptors in our eyes where particles of light photons hit the retina of our eyes. They produce little pulses of electricity which go down our optic nerves. And in the bad design of alien brains, they go all the way through the brain to the visual cortex. That's at the back of the brain. And then they start producing electrical activity in the brain. But so in a sense, we have a neural link. It's our eyes that go from light to nerve impulses, electrical impulses in our nerve cells in our brains. And so there's sort of. And our eyes are arranged in a certain way, and they present a certain neural interface to our brains. And as we start thinking about kind of additional kinds of interfaces to our brains, we have to say, well, for example, eyes present their neurons in a physical array that is the arrangement of pixels on our retinas. So there's a place on your brain, if there are pixels that go from left to right, that is actually mapped into particular nerve cells that kind of go maybe it's even from right to left. I think it inverts that in one's brain. And for example, from one's two eyes, the nerve signals from one's two eyes are mixed together. And if you kind of look at kind of which nerves are paying attention to each eye, it's kind of a pattern that looks a bit like a zebra stripe pattern, possibly because it's made from a reaction diffusion process, not unlike the process that makes zebra stripes, but playing out in terms of collections in the brain. But that means that you're kind of getting that sort of part of the neural interface from our eyes to our brain is that the two eyes are sort of mixed together in that way. So first step is, what's the technology of the neural interface? And how do you present the information that's coming in on this neural interface? And then also, how does the neural interface get something from our brains? How do we learn to send it the information it needs to be sent? And I suspect that that's will be ultimately very similar to how do we learn to type? How do we learn to do any of these other things, which are kind of a way of taking a thought and actuating it in something which involves some orchestrated arrangement of nerve impulses. My guess is that we'll kind of have to learn by the same sort of feedback mechanisms and so on, that we learn to do lots of other kinds of activities. There'll probably be a kind of a training of how to do sort of connect to your neural interface type thing. What will it feel like? I think it will kind of feel like, well, any kind of sort of feedback that you have. Like you move your arm and you feel this thing, and that's sort of a thing happening from the outside world. It's kind of like you, or even you do something on your computer and your computer responds. That's like you've done something, you've actuated something. You press this button and something comes back. My guess is it will feel very much like that, that it'll be that you feel, I'm doing something. It isn't something where you can see what you're doing. It's not like, oh, I can see that finger moving. It's something that you're sort of imagining going on. And it will then come back from what you think, so to speak. Something will come back from that, just like kind of. I wonder what will happen if I press this button and then you do it and something comes back and it happens to go through your eyes. But I don't think it will be that different going directly into brains. And I think there are many kind of more junior versions of this that one starts to wonder about. Like, for example, as soon as we have sort of good augmented reality capabilities, there's the question of, can we see things around us that are beyond kind of the ordinary images that we see with ordinary visible light? Very obvious question. Can we see into the ultraviolet? Can we see into the infrared? Can we see more precision of the frequencies of light that we're seeing? Not just is it red, green, blue, but which is what we currently have sensors for in our eyes. But, you know, is it producing light right at that particular frequency that tells us that it's a particular material, for instance? And the thing that. One wonders is, okay, you're having this kind of enhanced vision. How do we present that enhanced vision to our brains? Well, we could have a direct neural interface that goes beyond kind of the interface that we get from our eyes. Our eyes are delivering signals to the visual cortex at the back of our brains. Okay, well, what piece of real estate in the brain are you going to connect? Let's say you had a direct neural interface. What piece of real estate in the brain are you going to connect that information to? That's going to be an issue of, well, whoops, I've already used that part of my brain for this. You can't connect to that. Then you'd use up all my knowledge of this or whatever it is. So that's sort of a question. I don't know how that's going to play out. But that's one. One thing would be to take the infrared channel, so to speak, and play it into a piece of real estate in the visual cortex. Another possibility is to kind of multiplex the images that you already see and kind of Modify them in some way. Like, for example, let's say you're seeing this image and then something is kind of flashing in that image. Let's say there's a thing that's really hot. You could see that instead of just seeing the thing as a sort of piece of metal or something could be flashing a different color. You could still tell, oh, it's a piece of metal, because you can kind of see it through the flashes, so to speak. But. But the flashing, you know, the using the time part of things could give you a better way to kind of be able to perceive sort of what sort of additional information sort of multiplexed onto the visual scene that you already have. So I don't know what you know. Do you get to use more real estate in the brain that is sort of unused and send signals indirectly, or do you get to use the current sort of nerve channels and sort of put more information on them by, for example, changing what you see in the visual field and so on? The. Harry is commenting, I think some parts of our brains adapted to modern laggy typing, so we don't really perceive it anymore. Yeah, I mean, one thing that does tend to be the case is that that people get used to things often more quickly than they ever expect. So it's a famous thing in user interface design, system design, that you tell humans, oh, your computer is adapting to you. And they're like, okay, great. That's why it's getting so much easier to use the system. But actually, the computer didn't change at all. It was the humans that learnt more about how to use the system. They just had the perception that the system was adapting because they were adapting. And that was, in a sense, by telling them the system adapts, it was just sort of an inducement for them to persist longer in using the system. I think the thing that's often surprised me, like, for example, at times in the past, hasn't happened recently when computer keyboards were laid out slightly differently and I would switch from using one keyboard to another. I was amazed at how quickly my sort of touch typing got rewired. I mean, literally it took like seconds for one to, oh, that's a different keyboard. Okay, that's there and there and there. And it's like, okay, now I'm typing, I'm not looking at my fingers. I don't think I am, at least. And yet I was able to use sort of the rewired keyboard. And it happened really fast. And I was surprised at how much one adapts to that. And I think that's the thing we have going for us as we learn. And often some of these kind of sort of skills oriented things are things that we learn quite quickly. I mean they're different parts of our brains that are associated with these kinds of manual skills. The riding a bicycle type thing as opposed to the how do I figure out the answer to this philosophy question type thing. So let's see. Ah, wolf comments. Something I didn't know. The claim is that the brains of whales have fewer neurons than humans, but the neurons are big. It's interesting. I didn't know that. Let's see, Cryo is making an interesting statement that societies and groups are like larger scale brains developing. Interesting claim. I mean it's kind of like the ant which has 50,000 neurons per ant brain. You might say the whole ant colony is kind of achieving actually. How does that compare? How many ants are there in an ant colony? I would guess, let me think, 100,000 maybe. So let's see, if that's right, then that's. I see at 100,000. If I'm right, that, that's right. That's roughly right. Then that would mean that there are 10 billion neurons in the whole ant colony. If there are 50,000 neurons per ant, roughly so still the ant colony doesn't have as many neurons as a single human brain. But you know, an interesting question is if, if you look at the 8 billion humans, each of whom, let's say has 100 billion brain cells, that's, you know, that's a big factor, you know, one's looking at. And then the question is how efficient is it to have sort of the coordinated effect of sort of a whole group of humans as compared to the sort of thinking of an individual brain? I have to say that the sort of, there's this question of the coherence of thinking of an individual brain versus the coherence of action of a whole group of people. And there's sort of a question of how that works. And often one thinks, well, gosh, things, you know, when all the humans are kind of doing things in lockstep, that's not very interesting. We want them all to kind of be, be learning separately. It's kind of the dream of, kind of the inventiveness of the humans. It's like with all those humans trying to invent things, they'll be more successful in inventing things than if it's just one human doing it. Of course, what we actually know in practice is that human organization, in order to sort of keep everything organized there's more structure when you get more humans trying to do things. And often it is harder for that one human who has the great idea to kind of implement the great idea, because there's this big institutional structure, so to speak. Now, I wonder whether things like that happen in brains too. I mean, in other words, I wonder whether when you have an idea, does that idea initiate with a small number of neurons and then they have to be fighting their way through your brain to say, yes, that's really a good idea. Is that the right picture of what's going on? It's an interesting question, something where I think one could start to investigate that with artificial neural nets, sort of, where does the idea originate? And is it the case that you can kind of trace where the. I mean, there's a little bit that's been done, and I've even done. And we've even done a little bit of looking at the sort of LLMs and asking questions like, can you identify where in the LLM some thought starts to develop? It's not so easy to do that. But there's a few places where you can kind of see some sort of definite brain circuit in action. But one might be able to do that. One might be able to say, where did the thought originate? What did it take for that thought to sort of take over and be the thought that ended up being the, you know, great. That thought won. It was the next word. Interesting question, actually. Something to look at. But then the issue would be, what is the comparison between the way that we kind of sort of initiate and promote ideas in society versus the way pieces of our brains initiate and promote ideas? Maybe there's something great we'll learn. Maybe that's the next fad of corporate organization is organize your company like the neurons in your brain are organized. But I don't think we know what that is yet, so. Well, we know some things about how brains are organized, and they do tend to form clumps and then sort of super clumps and so on. So maybe that's kind of like the hierarchical organization of typical companies and such like. But that's a. It's an interesting question. What is the comparison between kind of idea generation in an idea development in brains and in societies? And is it the case that sort of. That's another question, you know, is bigger better? You know, history does not support the idea that bigger is better in terms of the creation of new ideas and all sorts of kinds of progress. It is not the case that, you know, now sometimes the bigger country Wins because it simply has more resources and more ability for people to have the opportunity to do things like invent or get their inventions paid attention to. But it's certainly not a constant of history that the bigger country is the more successful country sometimes. It's just there are an awful lot of people and it's really hard to deal with that number of people. And in order to deal with that number of people, you have to organize things in that way. Let's see, Cryo is asking, would you imagine bigger brains will run into heating or cooling issues? You know, I think brains, you know, an awful lot of the heat made by us humans is made by our brains. A lot is made by the liver as well. But, you know, you have to kind of worry about losing heat from your head when you're in a very cold place. And brains, I think they use about. They use about 50 watts of. Power, and that's significant fraction of the heat generation for us humans. Now, if you have a bigger brain, does it use more power? I actually don't know. I assume the answer is yes. I assume that the, you know, when we. We eat to produce glucose to be able to sort of power our brains, I'm kind of assuming that critters with bigger brains have to have to eat more to power their brains, and so there's more heat that gets generated. I don't think the issue that happens with computers a lot, which is how do you dissipate the heat, is so much of an issue for brains. I think the reason is that we have an awfully good cooling system because we have blood vessels that are exchanging heat in the brain. We have capillaries going, reaching right down to 2 millionths of a meter away from everything there's a capillary. Otherwise we don't get to have that piece of tissue live because it has to have oxygen delivered to it. And that. So it's delivering oxygen at that rate to every part of the brain. And it's also being able to take heat, exchanging heat between any heat that's produced by the brain and then that's being sort of taken away by the bloodstream. And I think in computers one has to have these rather elaborate heat sinks and so on. On a cpu, one would have something where there's a. One's trying to get air to flow past the chip to be able to take away the heat from the chip. But one doesn't have something as convenient as a bloodstream of capillaries and so on. I mean, there have been at times computers that were cooled with liquids. And there have been water cooled computers, There have been computers that are cooled by, that are cooled at low temperatures. I remember back in the 1980s seeing computers, there were several. One of the more exotic ones, supercomputers of the time, you could actually see the bubbling liquid was some kind of inert liquid that was sort of flowing through the actual circuitry of the computer to take heat away. And there certainly are. Even today, they're water cooled things, but they don't get in with capillaries and so on. So I think that current technology has it a lot harder than brains have it, because brains have all those capillaries that form. See. Well, let's see. DES is commenting that is nature more efficient at doing things than electronics, so much so that we should use biological cells for computation. That's certainly been a persistent thought. What is true is that biology has succeeded in doing sort of orchestration of things at a molecular scale. We Talk about a 5 nanometer feature size for microprocessors that's getting down there in terms of things the size of big biological molecules. But still one of the features of microprocessors is they still tend to be largely two dimensional, largely just in a plane, not built up in 3D. Whereas all of our biological material, most of things in biology are built up in three dimensions. And I think that the whole orchestration of this molecule goes there, this electron, this ion goes there. That happens in biological tissue with biological molecules is still something where we don't have an analog of that, that we can sort of print out as microprocessor chips or anything like that. So there are pieces to that where there are sort of molecular level orchestration of things that we don't yet know how to do technologically. I kind of feel like there's a direction where just as microprocessors use electrons for everything in biology, it's not just electrons. We're using other atoms and molecules and so on to sort of transport both information and material around. And I kind of think that there's probably a way to build technology around that idea and to make sort of the analog and microprocessors, but transporting, you know, I don't know what iron atoms and sodium atoms and things like this, rather than just electrons around and that we'll end up with something that is sort of more biologically related in that way. I don't think that the idea of using biology as such as the foundation for doing sort of computational kinds of things is a great idea. I think that when we have to interface to biology like we're making medical devices and we have to say this is going to attach to some particular kind of cell, then there's more reason to do things kind of biologically. But if we're just trying to compute the next abstract thing, I really don't think that's going to be biological material. I don't think that's going to be little mini brains of organoids that have been grown out of brain tissue. I don't think that's going to be things that are biologically constructed that let one do that. Things that are molecular scale, yes, but things that are biologically constructed, no. I think no more so than. Saying, oh, let's use horses instead of cars, let's use the thing that didn't work out, flocks of geese or something, rather than airplanes. These were sort of biological solutions that weren't a good fit to the technological needs. In the case where one's dealing with medicine, biomedicine, there may be more push to use sort of truly biological things to fit into that use case. But in the case of pure computation, I don't think that's the right way to go. All right, I see. I'm being told that I'm late for a whole bunch of other things here. I do want to mention that on Monday, I believe it is math storytelling day. Every day has to be some kind of special day. And it's math storytelling day. And I'm going to do a math livestream about math. And I was hoping to at 4:00 Eastern Time, U.S. eastern Time. And I was hoping to kind of be able to explain some math in an interesting way. And if people have math that they would like explained, including things that you were sort of might have been supposed to learn in school or something like that, I'll do my best to try to explain those things and see whether I can do the outrageous thing of sort of compressing sort of math ideas that are supposed to take semesters to learn into a short time. We'll see what I'm setting myself up for there. But anyway, I'm looking forward to having a chance to try and explain some math. 4:00pm Eastern Time on a live stream. And for now we should wrap up here. Thanks for joining me. Thanks for lots of interesting questions. I see a bunch that we'll save up for next time. Bye for now.