Future of Science and Technology Q&A (June 2, 2023)

Hello everyone. Welcome to another episode of About Q and A about future of science and Technology. This is part I now have four series going and on Fridays I'm alternating between Future of Science and Technology and my long running Q and A about science and technology for kids and others. So today talking about future of science and technology. So I see there are a bunch of questions saved up here. My gosh, lots and lots of them. Di asks, do you think the latest electric car is worth buying for these days? What is the future of cars? I'm not a close follower of all of the different offerings of cars, so I'm not sure I can comment on that. But the general question of what the future of cars is is sort of interesting. I think there are a bunch of things to say about that. I mean, it's worth so first point is there are cars have been around for a bit more than 100 years. Huge infrastructure built out of paved roads and all these kinds of things for cars. And the and also the effect of cars on kind of the geography of the world and the fact that suburbs could exist and things like this has been quite dramatic. And people have been saying for years, oh, there'll be some situation in which there start to be these excerpts and people living in nice mountain spots and so on that are quite hard to access by car, but there'll be some other way to access them. I think the main thing that'll be interesting to see is when do some kind of personal flying vehicles really come into existence? I mean, flying cars have been predicted. They've been a feature of the future for at least 100 years. You look at issues of Popular Mechanics or whatever from close to 100 years ago and you'll see pictures of the future is flying cars type thing. Flying cars, why haven't they happened? There was a time, I remember, well, 50 years ago or something. Now helicopters were the thing where everybody's going to have a helicopter. And there were occasionally, for example, around where I live in the Boston area, a company, Digital Equipment Corporation, which for a while was the second largest computer company after IBM, had the idea that they would have their facilities dotted around New England and all the execs would go from place to place by helicopter. It's the only company I know that tried to do it that way. Helicopters didn't work out to be nearly as easy to deal with as people expected. They require a lot of maintenance, they're hard to fly. The whole thing is sort of an expensive thing to operate. And so now the question is, will there Be a moment where kind of the flying car ish thing really comes into existence. I think the answer is definitely yes. There's nothing sort of against the laws of physics about having a flying car. The issue has to do with sort of the trade offs between battery density and energy density and batteries and how heavy the car needs to be. You know, I've seen demos, oh gosh, several. I mean I saw one at the Consumer Electronics show maybe six, seven years ago. They had a thing that, well, admittedly it was stationary on the show floor, but it was acclaimed sort of personal drone type thing. So another one again it was stationary on. I've seen several stationary on lawns and things, but I've never seen one actually flying with a person in it, so to speak. I mean the issue is if you want to have a small surveillance drone, cargo delivery drone, it can be, there's, you can just deal with, you know, packages under five pounds, but we humans don't come in packages under five pounds. And so there's sort of a minimum size necessary to be a useful thing to transport humans around. You know, I'm remembering now a company I saw present with business plans for, for kind of personal drone transportation, oh maybe two years ago now. So this is something sort of bubbling up and when will it really happen? Well, so one is battery technology, energy density of batteries. The other is can you actually have a drone type thing that is easy enough to fly, that either somebody can control it themselves or it can be, you know, this sounds a bit scary, you know, centrally controlled and you're just a passenger and you're sort of in the drone. And I suppose the other possibility is piloted kind of self driving cars with actual sort of qualified kind of Uber driver style, but their Uber pilot that style folks instead, my guess is that's not economically viable for all sorts of reasons. But so, you know, a question will be, let's imagine that eventually kind of the sort of drone like individual flying vehicles become practical in a way that helicopters have not been. And presumably it will be the electric version of these things that becomes practical. Although that's not self evidently how it has to be. But my guess is that that's how it will actually play out. And in fact I think it's much easier to have. I mean when one sees drones, they tend to be a lot of the rotor based ones that they are electrical. And I suspect it's easier to have an electric motor respond quickly than it is to have some other kind of combustion motor respond quickly. And that probably makes it Easier to do flight control for such things. But in any case, the question would be, let's say that sort of it was possible to have flying cars, basically. What effect would that have? Well, no doubt it will change real estate values. No doubt it will change commute times change dramatically. People are not stuck in traffic jams. There's a lot more room in three dimensions than there is in the one dimension of a road, so to speak. I think that. But, you know, it's unclear to me the fact that it's going to be an interesting question how long it takes once there really are viable kind of drone flying cars, so to speak, how long it will be before those things become common. And whether that's a five year story, a 10 year story, a 20 year story, I'm not sure. And then how long it will be before that has significant effects on the way people kind of lay out real estate and so on. I mean, there are a tiny number of places in the world where there's sort of a local light aircraft, you know, private plane, airport, and people literally, you know, they land on the Runway and then they kind of taxi their plane to their kind of garage and their house and the houses are all sort of arranged around the Runway. There are a very small number. I've never seen one in person, but I've seen pictures of them of places that are set up that way. But that would presumably become much more the norm. You know, I think the thing that's supposed to be coming soon is, you know, Amazon is trying another effort to make drone delivery work. And kind of the way that that's planned is you kind of pick on a map where you want the drone to land. And it will, it's rather noisy I think right now, but it will, it will come and land there and drop off your five pound package or whatever. I think one might have thought that there would be, it would be better to have sort of, you know, set up the landing pad in your backyard. But I think that it's believed that it's better to just have people sort of select the place on, on a map that so you don't have to mess around worrying about whether, you know, whether the local wildlife had chewed up your landing pad and so that's why you didn't get your latest delivery, so to speak, there. Let's see. Question from Zengith. With technology integration, will we be able to do away with having to sleep in the future? You know, sleep is one of those sort of scientific mysteries of exactly why is sleep necessary? I Personally have long assumed that the real answer is when brains operate, they produce a certain amount of waste product just by virtue of the process of thinking, so to speak. And that one needs to clean that out. Much as, you know, if you exercise a muscle, you produce lactic acid and you have to wait for that to kind of clear itself out before you can go on using that muscle without it being tired. And I think, gosh, what is it? There's a. Oh my. I think it's actually an amino acid that gets produced as a result of repeated neural firings. And sort of one of the common theories of sleep is that you need to clean that out and that it's hard to clean out while the brain is in operation, so to speak, as is so common in biology. Once you have a phenomenon like sleep, biology piles on lots of functions for that phenomenon. And sleep exists down certainly to fishes and possibly even lower animals than that. There are sort of separate states of being awake and alert and not being awake and alert. I mean, there are some creatures that have multiple parts of their brain. I think cartilaginous fish, sharks and so on, I think have one of these. Half the brain is asleep while the other half is awake type type deal. And we humans have our left and right sides of our brains do somewhat different things. So that wouldn't work so well to be alternating between sides when sleep and wakefulness, I guess, unless you said I have my nighttime activity and I have my daytime activity and I'm really kind of a little bit of a different person during the night when I'm kind of sleeping on the left side of my brain. And during the day when I'm sleeping on the right side of my brain, I'm different. But that's not how we, not how our species is set up. But in any case, given that biology has sort of piled on functionality onto sleep, perhaps its primary function was just sort of cleaning out waste products of neural firings. But then it got other functions as well. Another function of sleep is probably something to do with memory consolidation. Probably the case that to really fix something in long term memory, one has to kind of replay it somehow during sleep. And possibly even just the maintenance of memories that were around from, I don't know, 30, 50 years ago, whatever in your brain you probably remember remembering them, so to speak, rather than having sort of a primary memory going back that amount of time. And it could very well be that some of that memory kind of gets replayed for one and re solidified in sleep, possibly in dream sleep. And so On, I think if you look at the sort of evolutionary explanations of sleep, one of the ones which I must say I've always found deeply suspicious is the theory that a lot of animals sleep just to keep them out of trouble during the night. And that's been a quite popular theory among evolutionary biologists. My guess is if there is any such effect, it is a piling on of yet another reason why it's a good idea to sleep that biology has done. Now the question is, is there a way in which one could avoid the need to sleep? You know, I know people do all kinds of biohacking type experiments and certainly even in the distant past, people had, who had very regular schedules. I know this was true of some monk scientists who would go and say, well, I'm going to sleep one minute less every day over the course of a decade, and then I'll be down to only needing three hours of sleep per night. I think it's, you know, it's unclear whether we work biologically successfully on that. And certainly there's, I think, decent evidence that all sorts of bad things that can happen to older brains are more likely to happen if you haven't had enough sleep. So the sleep is probably of a value to maintaining sort of the health of the brain. And certainly one knows if you actually forcibly don't sleep for long enough, you eventually die. So that's kind of a sign that sleep is necessary for things to keep working. But any case, the thing that is, can one sort of reduce the amount of sleep one needs? I have to say, I personally have never tried to do that. I find that I'm only able to sort of have a brain that's operating decently if I have a pretty normal amount of sleep every day. One can ask the question between the different phases of sleep, rapid eye movement sleep, deep sleep, light sleep, et cetera, et cetera, et cetera, whether perhaps some of those are more essential than others. And people argue a bit about what that story is. But. And maybe there are parts of the sleep cycle that one could do away with and sort of save those hours of sleep time. But then there's the possibility, oh, maybe there's a drug often thought to be the solution to lots of things and sometimes is, but often isn't. You know, maybe there's a drug you can just, you know, pop a pill and it will do the kind of cleaning out of, you know, in 10 minutes it'll sort of clean your brain out, do some kind of brain cleansing thing, and then you'll be Good to go and can keep kind of operating without sleeping. I'm dubious that that will work partly because I suspect that biology, even if there was one primary function for sleep, which you could get in some other way, that biology has piled on enough other functions for sleep that there'll be things one loses by doing that. So I think in the question of whether there are sort of things that, you know, if one had a direct neural connection to one's brain, could would that affect how sleeping works? You know, I have to say, when I was about 10 years old, I had this. I had read something in some future science book of some kind that there was this great idea of learning things while you were asleep. Well, in those days it was not. There weren't computers that were accessible or other kinds of things like that. So the only way you could like, you know, try to learn something when you're asleep was hook up a tape recorder and a whole sort of alarm clock type thing and to switch the tape recorder on, et cetera, et cetera, et cetera. So I did try this just on the theory of experiments and recorded some random stuff that I wanted to remember, probably some school related kinds of things. The problem was with this experiment. The experiment, I would say, was a total failure. Well, I wouldn't know whether the experiment was a failure or not because I don't remember what might have been kind of, you know, what I might have been exposed to in sleep. But I have a good enough memory that just the very act of recording a bunch of things that were intended to be sort of sleep played was probably enough for me to remember them, thereby completely sort of bowing up this experiment. But you know, people do that. There is the question, you know, if you have a brain and it's sort of more or less operating and you kind of want to get mileage from the brain, can you. Could you like connect something to the brain when it was asleep and still get mileage from it? I mean, it kind of like, I feel like, you know, when you say I can do that with my eyes closed or I can do that when I'm asleep, so to speak. You know, maybe one gets to actually test that. I'm skeptical of whether how well that would work, but it's certainly an interesting thing to imagine. You know, could you start. You've got this brain that's asleep and you're. But it's still got its whole neural net there and you can kind of poke at it and you can kind of use the time when you're asleep to Kind of be constructing, I don't know, a digital twin of your brain and reading out different things and not sort of having to be exposed to that going on. I don't know. Just a. Just an idea. But I don't. I don't immediately have a. Have a thought about that. And I am. I'm very skeptical about the take this pill and you won't have to sleep type thing. I am pretty sure that was in the books that I was reading when I was a kid about visions of the future. And it certainly hasn't happened yet, and I'm skeptical that it will. Gosh, lots of interesting questions here. There's a question from Tinkerholly here. As far as human evolution goes, do you believe the human race is still evolving, or have we peaked as a species? What's next in the stage of human evolution? You know, one of the things about us humans is that a lot of what changes for us is things like technology. And in a sense, that's a layer of evolution beyond the biological, so to speak. In many species, if you want to have something that is kind of passed down across the generations, the only way to do that is genetically, so to speak. If you want this particular type of fish to be really good at being able to find food inside crevices and rocks or something, that skill is presumably something that has to be transmitted genetically from one generation to the next. Whereas as you go to sort of higher organisms and things, there is an increasing sense in which you can learn something in the course of life. You can be kind of a blank slate at the beginning. And then in the first some amount of life, you learn this or that thing. I mean, some aspects of our brains, we know, work that way. For example, vision. It's pretty certain that our sort of neural net that does visual perception really doesn't know how to see anything when we're born. But as we start exposing it to the actual world, with the actual objects that exist in the world, so we start to wire up our visual cortex to have the feature recognizers and so on, that we need to recognize the things that we see today. Actually, a very interesting question is whether, given that people grow up in an environment that is different in terms of its visual appearance from the environment that existed earlier in human history, does that affect the very way we see? My guess is the answer is yes. There's a famous experiment done in the 1960s or 70s with cats who were reared in an environment with only vertical stripes. And the cat, just as it's a stripy environment, and they have some kind of collar to prevent them sort of seeing things at an angle, et cetera, et cetera, et cetera. It's kind of a weird experiment. But the result of that experiment is those cats didn't learn to see horizontal stripes when they, you know, were exposed only to vertical stripes. You show them something moving horizontally, they're like, I don't notice that. So it's an interesting point, which I'd never thought of before, that modern humans raised in cities and things like this, the visual scenes that they will get will be somewhat different from the ones that people have ever had out in the wild in the past. And how does that affect what they learn to be able to see? It's reminiscent of when one has image generation systems, AI, generative AI, image generation systems, whatever it is, the latent diffusion models or these other kinds of things, where one is going from a textual prompt to a picture of a cat with a hat or something. Those work by basically having seen lots and lots of images of the world and finding the statistics of how those images work and then saying, okay, I'm going to make an image that follows the statistics of images in the world, but I'm always going to make it with the condition that it's ultimately going to be described by an image captioner or something as being a cat with a hat. And. But I'm going to sort of. There are many, many. There's a sort of almost infinite number of possible images that can be made that respect the statistics of natural images. And in that case, you'll be sort of guided to make an image that follows the caption that you want. But that means that those image generation systems, they have learned from the actual images that people have taken, you know, have put on the web and so on. Those actual images certainly contain lots of technological artifacts. And it's sort of an interesting thing that the statistics of those images depends on the fact that we live in a technological environment, so to speak. And so it's an interesting thing that one can imagine that even at the level of the actual sort of wiring in our brains, at a pretty low level, below the conscious level, we can be affected by the environment that we happen to find. But at a much higher level, the big innovation of our species is language. And that's allowed sort of increasingly abstract concepts to be passed down from generation to generation. The primary form of evolution that we have is knowledge evolution, intellectual evolution. Evolution that is about ideas rather than about the details of, you know, how are our thumbs jointed or whatever else. Now, you know, talking about biological style evolution, well, it's always a story of natural selection. It's, well, almost always a story of natural selection. I mean, one of the things that's been an achievement of recent times is that lots of people born with various kinds of issues who in the past would have died and would never have been able to pass down their genetic material are now, as a result of modern medicine, able to, you know, live long and prosper, so to speak, and have children. So that, that means that certain traits which might have been kind of ground out of the, of the population have not been. Now, of course there are other funky things that happen because, you know, there's, there's the question of, in the natural selection, the sort of winner of the natural selection stakes are the folks who have the most children. And in modern times it's a funny set of different sort of cultural groups that end up being the large children groups. And there's certainly ones where that might be, you know, you might say they're a wonderful group, but unfortunately they're not having any children. And so in the biological evolution stakes, they're just going to lose. And obviously in the current state of the world, there are lots of countries where there's rapid growth of population and where the per capita GDP and things like that is quite low. And then there are parts of the world where the GDP per capita is really high and the population growth is zero or negative. So in that sense there is human evolution. And it's kind of funky because it depends on these kind of cultural factors that determine how many children people have. Now in terms of, I think there's a, I don't know what that selects for. That's a funky thing. And that is the primary form of natural selection style selection that we currently have. I think in terms of, well, what else could happen? I mean, there are much wilder things that could happen. For example, with gene editing, it's perfectly possible to put new and different genes into the kind of germline, into the, into the actual, you know, into what is passed down to future generations. I think people generally think that's a really bad idea, but bad ideas don't necessarily not get done. And I think that's the other thing one can imagine is something where we have a more than artificial selection. I mean, so normally in animal breeding and things like that, one has a form of artificial selection where one's deciding that's a really good looking cat, we'll breed that cat and get lots of kittens. From that cat and not from this cat that looks kind of ugly over here. That for the concept of a sort of external force doing that for the human population is viewed as a really bad idea. I mean, it was popular 100 years ago with sort of theories of eugenics and so on, but people, I think, have universally thought that's a bad idea. So then the question is, you know, that that force of human evolution is not happening. It could happen, but it's been thought to be a bad idea. Just as one could imagine something even more direct than artificial selection, which is literally doing gene editing is something that could in principle happen, but probably people think it's a bad idea. So I think it's hard to predict what any of those things would lead to very rapid biological level evolution. But I think otherwise most of the evolution is at the level of ideas and at the level of sort of the technology world, which in a sense is co evolving with us humans. I mean, as technology. You know, the whole sort of lineage of technology. I mean, it's like one doesn't get a microprocessor from nothing. A microprocessor is the result of a big tower of evolution, of technological evolution, just as we humans are the result of a big tower of biological evolution going back a few billion years. So a few thoughts about sort of the future of us as a species. I mean, I would say that the kind of the question of to what extent we evolve our ideas and our technology versus to what extent we evolve our biology. My guess is that in terms of the natural evolution of our biology, and it's at best confused right now and not something where we can realistically expect. I mean, you know, partly because through technology we are shielding ourselves from lots of kinds of issues. I mean, we might imagine that, oh gosh, you're living in a hot climate, you know, you'd get, you know, you'd somehow end up evolving, you know, built in air conditioning or something. But then it turns out that in many parts of the world, at least some parts of the world, you live in a hot climate and you have air conditioning. So external technologically provided air conditioning, so to speak. And it's also the case that as the mortality curves for people get to be much more. I mean, it used to be the case that life expectancy was quite low and was comparatively low compared to childbearing ages. And that meant that there was sort of much more sensitivity to things that might kill you and the effect of things that might kill you on whether you have children. Whereas if life Expectancy mostly is extended far past childbearing age. It's like that means that sort of anybody gets the chance to have children, so to speak. And there doesn't tend to be this sensitivity. The things that might sort of bump you off before you get to childbearing age. Let's see. Oh boy. So many questions. MADOB asks, will we ever have technology that will allow us to learn why we sleep? I don't think we know. Good question. I don't think we know. I don't think we know whether what the role of sleep in terms of memory consolidation is. I don't think we know whether our sensory inputs are active in sleep. That one would have thought somebody must know by looking at FMRI imaging of people while asleep. But I don't know that. And certainly sort of anecdotally, one doesn't have the impression that sort of sensory stuff is open while asleep. One doesn't. I don't think people often report that they, you know, if they were in an environment where there's a bunch of noise going on, I don't think that impinges on their dreams, for example. I don't think that's a common thing. I think one, you know, dreams are affected by the. By the way in which one wakes up and that's mostly sort of compressing the time in the dream into. Into the time while one was in those moments of waking up. I'm always very bad at things about dreams because I essentially one of these people who remembers very few dreams they have when they're asleep. So I'm not an expert about that. Let's see. Ah, spare parts asked. We're coming back to an earlier question. Noise seems to be almost inevitable when it comes to flying. Is there a way to solve that? You know, I don't think noise is inevitable to flying. I think that noise is primarily produced by turbulence, which is produced by kind of rapid motion. There certainly are, you know, planes that people built with propellers that are these very kind of nice wooden propellers, big propellers and so on that have produced much less sound when they fly. They're much less. I think the primary source of the sound is turbulent flow. And turbulent flow is not a necessary feature of flight. Certainly not. Well, little insects don't have it, but I think that that's something one can avoid. So I think. And one of the things that certainly happened over the course of the last one or 30 years is Jet engines have gotten quieter through all kinds of technological innovations, but particularly in the way that the fluid flow works. Through them. So my guess is that flight doesn't have to be loud and exactly what it takes to make it not be loud. I'm not sure. Obviously a balloon is. Well, you would think a balloon was silent, except a hot air balloon has this great big burner in it that's making very loud growling sounds, so to speak. But in principle, a balloon is not making sound. And obviously gliders, well, again, don't make sound like that. But I don't think it's the case. I'm sure that there will come a time when, when there's. When there are quiet drones and so on. How big they have to be and what all of those physical constraints are, I'm not sure. Nunes comments. There's a Japanese hover bike for half a million dollars. Interesting. Let's see. Okay, there's a comment here from Jamie about how do you optimize the sky for regular air travel to accommodate flying cars? You know, one of the things about sort of the evolution of cars and roads and all that kind of thing is how smart are the cars going to get? Because, you know, if cars were smart enough, they don't even have to all be definitively driving on the right or left. They could be using the whole roadway and just they know, oh, there's a car coming in the opposite direction, I have to move this amount, et cetera, et cetera, et cetera. That's perfectly imaginable with sort of smart enough cars. I think the issue with cars, and this is sort of a thing about someone asked before about sort of the evolution of cars and such like, one of the questions I suppose is to what extent do you have individual sort of self driving cars? To what extent you have whole fleets of cars that are communicating? And then what does it even look like when cars are kind of negotiating with each other for, oh, this car, you know, if you go this way, I'll not cut you off at this intersection, et cetera, et cetera, et cetera. And you get this very elaborate negotiation between cars. It's all happening outside of human awareness, probably. And that where the, where the human user has said, I want to get from here to there as quickly as possible, I need to get there by this particular time. And then the cars can kind of negotiate. Well, you know, we can open up this thing here and not put so many cars through this road, which will cause it to slow down, etc. Etc. Etc. I think the. Another big feature of cars and sort of communication between them is traffic jams. Traffic jams are A very mysterious phenomenon. I mean, you might have thought that, you know, you've got a big road and you've got a place where everybody's slowing down, you know, they're looking to the side of the road because there's something bizarre on the side of the road or something and every car slows down by five miles an hour. And you might have thought, well, why doesn't that just mean people, you know, there's still the same flow rate of cars? Well, it turns out there's this instability that develops in road traffic that propagates backwards. And even though everybody at one point was just slowing down by five miles an hour, you go far enough back, you get stop start traffic, you eventually get to the stop start regime where you've got traffic jams. And this jamming phenomenon is why does it happen? Ultimately it happens because of driver delays. If everybody could instantaneously respond to the speed of the car in front of them, everybody would just slow down a bit and there wouldn't ever be this kind of stop start jamming type phenomenon. So that phenomenon one would expect if every car was self driven, that one could sort of prevent that phenomenon and just have the cars always full of traffic would be much higher than it is now. And I think that's quite achievable. And there's sort of the question of how that will be, you know, how, how that will come into existence. Will it be the case that there are lanes, special lanes on freeways and so on for self driving cars that are connected into the sort of global traffic brain and that will not have this kind of jamming phenomenon? Undoubtedly, yes. But then there are questions about, you know, what does the transition look like as one's got some self driving, some non self driving cars. What does it take to make the self driving cars really work? I think on freeways it's something that is realistic. Even now in complicated city streets, it's something very challenging that I don't really think has been, seems still fairly far away from being firmly achieved. I think that it's something where, so you know, this whole question of how do you manage sort of, you know, right now road traffic is uncontrolled in that sense. In the sense, you know, there are traffic lights and things like this that try to control road traffic. There another interesting story of, you know, how do you time traffic lights? What's the scheme for doing that? Do you, how much sensitivity do you have to the actual cars that are around and how much is just this light changes every 30 seconds? Or whatever it is. The, the theory of sort of optimal traffic light timings is still quite imperfect. And there's sort of a question. There are these little tricks people have, like, you know, one car per green entering the freeway. So sort of a trick for spacing out cars, or back when I was a kid at least, they invented the, this idea of these, you know, boxes and intersections where to prevent gridlock. So it's, you know, you're not supposed to drive into the box unless you can get out of the box type thing. And there are these little sort of tricks to try and improve road traffic flow. But I think there's still the question of what could you do if you. Another, another trick is these brake lights up high so you can see through a bunch of cars to the sort of future brake lights of cars, Many, many cars in front of you and so on. But there's a question, for example, one question you can ask is if you put in special seed cars, could you prevent traffic jams? I think the answer is no. I think that there's no way. Once, once all the cars have this driver delay phenomenon built into them, there's nothing you can really do to sort of prevent that and to make sort of a firebreak type thing that prevents the traffic jam from occurring. But as soon as there's kind of more global car to car control and so on, you can prevent it. I think. Now the question of, in the case of planes, you know, for many planes there's air traffic control. It's not a necessary feature. A country like pretty much anywhere in the world, you know, you can fly a plane around if you, unless you're in one of these, you know, terminal control areas near an airport, you can fly a plane around without necessarily talking to an air traffic controller. And it's so, and it's kind of like, well, just don't run into anything. And you probably won't because 3D is big. And also, you know, if you're flying a plane, you're supposed to be constantly scanning the horizon to make sure there are no planes you're about to run into. So that's not that different from the story of cars and so on. And I think where it's kind of every driver, every pilot for themselves kind of thing. I suspect one can go an awfully long way in having lots of flying cars before there is serious risk of kind of large scale complicated congestion, kinds of things going on. And my guess is by the time you can have something which will keep the drone flying, you can also have collision avoidance systems that will prevent the drone from running into another one. So I'm guessing that's not going to be a huge issue. I mean it is the case right now in air traffic control that even though sort of the sky is three dimensional, there tend to be these point to point airways that planes will fly from one radio navigation beacon to another. And they'll typically be, if you look at the sort of the plot of where the planes are in the world, you'll typically see these streams of planes going in these essentially one dimensional paths. And there'll be planes are even, even an odd number flight levels of thousands of feet or whatever will be going east versus west and so on. And it is quite like roads in the sky, so to speak. People have often, and I think there are plans to do this, you waste a certain amount of fuel going point to point because it will be better if you could follow a nice smooth path between two airports. And there's a certain question of how do you do air traffic control when the planes are just flying anywhere as opposed to flying on these spec point to point routes. But I think that's something that has been long planned. I don't think it's implemented yet to be able to do air traffic control in that case. Which again gets more into this use sort of smarter control mechanisms to make better use of the resources. Ultimately in the case of you don't have to drive on the right hand side of the road necessarily because all the cars are sort of negotiating with each other where each one will go. The Aaron is asking what about the future of tunnels? You know, I think, you know, Elon Musk has his boring company that's proposing to make tunnels in lots of places. I think my impression is that there are certain innovations in tunneling technology that make it a bit easier to make tunnels. I think it's challenging to make tunnels over, under most cities, which is the place of most interest partly because nobody has any idea what's buried there. And you know, when people started burying cables in 100 and something years ago, it's like did they keep good records? Well, not really. So you know, you start digging and you've no idea whether you're going to run into some critical electrical cable that was buried in, in 1892 or something like this. And so I think that's something of a challenge in most cities. And I don't know if you, you know, can you tunnel down? So you're sort of just like if you, if you fly high enough, you're Kind of out of the main, you know, if you fly at 60,000ft, you're out of the main, you know, airplane traffic levels. Similarly, if you tunnel deep enough, you're out of the place where anybody will have buried a cable in 1892. On the other hand, if you, if you tunnel deep enough, you're probably hitting bedrock. And it's an awful lot harder to tunnel through bedrock than it is to tunnel through something softer. I have to say, I would be skeptical of a. I think it's going to be a lot easier to have things fly than to have things most of the time be kind of snaking through, through the ground and in some complicated tunnel arrangement. But, you know, I think, yeah, I don't think that would. That. That doesn't seem promising. Let's see. Lots of questions here. Crucial comments about flying cars. Flying is dangerous and requires more training and skill and safety than ground cars. Well, it doesn't take, you know, the amount of training necessary to fly a drone is not very large. It's actually less training than required to drive a car. So I think it's really a question of if, if it's an automated drone that is just flying around and, you know, by default it'll just sit there. Is, is. That's a, that's not going to require more training. Is it less safe to have something, you know, flying because, you know, things can go wrong and it can fall. Actually, in many ways, flying can be safer than driving because in driving, something can happen at any moment. You know, it takes only a second for you to have crashed and very bad things have happened. But when you're flying, most of the time you're just, there's, you know, you're not going to run into anything. Now these moments when you're landing and taking off, there's the unlikely possibility that you, you, you know, encounter some other traffic somewhere. But most of the time, it's just sort of just flying. Now, you know, can the thing drop out of the sky? Well, yes, in principle. I mean, one can. You know, there are two small planes that have, you know, if it's going to drop out of the sky, you know, deploys a parachute. And one can imagine that that kind of thing could be done for drones, which would make it, you know, pretty safe, actually. Safer than cars. Cars don't, you know, the best cars manage to do is if they actually crash, they'll deploy airbags. But it's not the case that if the car kind of goes off the road, it'll suddenly puff out some giant, you know, airbag like thing around itself and just sort of roll like a, like a, a ball or something, which, so I, I, I tend to think that that's not so much of an issue. Dugan comments. Flying cars would take an incredible amount of energy. You think it's feasible they would replace ground transportation? Well, it kind of depends what the future of energy sources is. I mean somebody gets fusion working, there's not going to be an issue with that. The, you know, even widely deployed geothermal, et cetera, et cetera, et cetera, that there are, you know, I don't think it's clear what the future of kind of energy sources might be. But I agree that, that it, you know, it takes energy to maintain things that are sort of being pulled down by gravity, so to speak. So yes, that, that, that would be something that would be needed, but I doubt that that's the first thing that kind of limits the growth of that technology. Let's see. Okay, there's a question from Jeremy here. What kind of architecture will be needed in order to build an AI that is as good at math as LLMs are at language? Do you think this will be a fundamentally different architecture than a neural net? If so, how do humans do math in any self consistent way at all? So first point is I think humans are not very good at doing math, at least not doing kind of multi step towers of math. Humans, when they turn math into abstract concepts that can be knitted together one concept to the next and where the thing you figured out is five concepts deep, go from here to there, that feels a lot more like putting together language. And it's something humans are quite good at in doing abstract math. When it's a question of a human sort of running the math for themselves. Humans are really pretty bad at that. I mean the limiting case I suppose is a human who can run a program in their heads and essentially none of us can do that. It's something for which neural architectures are not particularly good. It's something for which the invention of computers as something back from Turing machines on, as something which was supposed to emulate some of the overall operation of humans computing things, but then ended up doing it sort of better, faster, stronger and being able to run a billion operations or something, that's something that it's not realistic for neural nets to do. And the surprising thing about large language models is that the neural nets that are being used there, it's really every token that has to be produced. Essentially a word that has to part of A word that has to be produced. What's happening is you're starting with the data which has been turned into numbers for what the previous words that have so far been produced by that LLM are. And then you're asking, what's the next word going to be? You're taking all that previous data, you're just sort of rippling it through this neural net, this fixed neural net with maybe 400 layers or something. And then you're seeing at the end what results does it get. And that's what determines the probabilities for the, what the next word will be. So it's this very, you know, just, you start at the beginning, you flow through, you get the next word. The only form of kind of sort of more sophisticated looping is the outer loop that just says, okay, you're going to put another word down. Okay, go take everything that has been written so far, feed it into the large language model, use that to generate the next token and keep going. And it does that. A typical large language model does that until it gets to a stop token. And when it generates a stop token, the outer loop, the controlling system says, okay, we should stop now. That's the end of the essay. Or actually slightly more sophisticatedly, we should stop running the LLM. Now we're going to get a place where the LLM is going to call an external tool. And a bunch of the data that the LLM wrote out was actually data that it wrote for feeding to an external tool, like Wolfram Language, for example, feeding it to that external tool. Then there's actual computation that happens. You get the results, you feed it back to the LLM as language, and the LLM keeps going. But so my feeling is that the answer to what's a good architecture for doing sort of the inner computational side of math is pretty much computers as we have them today. Now, there are things that you could do that involve parallelism, where you're trying to tree out many possibilities and you're trying to find a theorem by following many possible paths. I actually have never seen substantial work done on. Am I right about this? Yeah, I think so. On, on sort of these kinds of. It's even happens in logic programming. And so on this kind of, we've got a goal. There are many possible paths that will achieve that goal. Which path is going to work? Actually, that's not true. I do remember when I used to do some work for Thinking Machines Corporation when they were building the connection machine computer, there was some graph pathfinding that was attempted on that parallel machine. But it's not something one commonly hears about going on. It is something where one can certainly in principle use, for example, parallel computation to look at those different paths in different processes and things like this. So a question would be, let's say that you are doing sort of path finding in a graph. You're can you photograph and what is the optimal architecture for sort of finding a path through this graph? And actually now that I think about it, that that is something where if you have sort of a massively parallel processor of the kind that the connection machine computer was, you can imagine essentially laying that graph out on, on different processors. Rather than just laying it out in the memory of one machine, you lay it out in different processors and you're basically looking for communication between processors to try and find those paths. But I don't think that's a, that is something one could imagine having is sort of a graph processor like that. Actually, now that I think about it, I think I know one example of a company that's trying to do that. But. So yes, I could imagine that type of thing being interesting. I think the sort of traditional neural nets, the big discovery is well, they got pretty good at language, they're pretty good at generating images, they're pretty good at a lot of tasks that we humans are pretty good at. And so in this kind of bizarre arc of history of neural nets having been invented in the 1940s to emulate brains. Well, it turns out eventually the technology gets to the point where yes, they can really do the kinds of things that brains can do. But long after we got to be able to do lots of things that brains can't do with more traditional computers. Let's see, Does AI being an interface to books means there will be more subject matter experts or fewer of them. Interesting. I think there will be more pop up subject matter experts, by which I mean one didn't spend a decade learning that field, but one was at some moment one really started caring about it. And one said with the help of one's AI, so to speak, one learnt that field. And the AI can help in two ways. One, in most fields there's a certain amount of mechanical kind of structure that you need to learn and a lot of that or you need to learn if you're going to do it all by hand. And a lot of that is probably automatable and you could probably pretty much tell your AI what you want. Hopefully it will then make some kind of computational language, Wolfram language. That is the thing that allows you to Kind of see, did I really precisely define what I wanted to define? And then you build from there, so to speak. I think the. And it will be less a thing where you say, well, gosh, in order to do homotopy theory or something, I have to learn all this stuff about how to compute, you know, homotopy groups or some such other thing. Well, actually, you need to learn the principles of this and you need to learn to figure out what question to ask. But a lot of the mechanics of how you compute these things might be. Is likely to be automated. So it kind of reduces the kind of the necessary depth of sort of mechanical learning that you have to do to get into these fields, which means that it's much cheaper to get into the field, which means that there'll be more people who are capable of operating within that field. Now, I think, what does it mean to be a subject matter expert? Sometimes there's a certain amount of sort of mechanical stuff one learns to do, a certain number of facts one learns to be able to recall, but there's also a certain amount of how do you think about things in that domain? And that's the thing which I think ultimately is the most valuable aspect of subject matter expertise is sort of, how do you think in that domain? If you're thinking about things as an economist or something, there's a certain way of thinking about how things work, you're thinking about things as an evolutionary biologist, there's a certain form of explanation way of thinking about things that works. You're thinking about things as a computational language designer, there's a certain way of thinking about things and so on. That way of thinking about things is really a valuable part of the subject matter expertise and a part that isn't the part that is sort of most automatable, so to speak. But I don't know how quickly and effectively people will be able to absorb those sort of ways of thinking. One would imagine that kind of. With sort of AI assistance and AI as the teaching aid, where the AI knows what you know and can figure out what it should tell you to get you to sort of take the next step. One imagines that those things will get a bit faster to be able to do. So, I think in terms of. So my guess is that many more people will be able to be subject matter experts in many more things. If by subject matter expert we mean able to think about things and do things in that subject area, when given the appropriate tools, so to speak, will be my guess. And I think that the value of kind of knowing broadly lots of different things is going to increase, and the value of knowing the sort of tall tower of mechanical specifics will decrease. Let's see. Well, Parker's asking several questions here. Pedro is asking will technology carry us away from the human condition or allow it to flourish? Interesting question. I mean, we can look at history for some lessons about that. I think that there were times when it was just harder to keep living. You know, you had to grow your own food. You had to be able to deal with lots of kinds of things which technology or sort of the structure of modern society Made pretty easy for a large swath of people. And I think that. Well, I would say that the history tends to suggest that things getting easier for people and people having to do less just to survive, Things getting easier for people has tended to lead to these times of great advances in our civilization. If by advances we mean kind of the development of intellectual work, the development of creative kinds of things, those are the times. Whether it's the Renaissance in Italy, Whether it's ancient Athens, Whether it's even ancient Babylon, things like this, Whether it's the heyday of the British empire, whatever. I think that the lesson of history is when the going gets easier for people, People do more interesting things. Now, I'm not sure how long that lasts. I think many of those examples from history, it was a burst of energy that lasted for a few decades or a little bit more. And as people kind of got used to, oh, we can just sort of sit back and relax, Things perhaps didn't continue to have that same level of vibrancy. So maybe the moment that's of greatest interest Is the moment where sort of a lot of things that people previously had to do Were no longer. Were no longer things people had to do. I mean, imagine a situation in which most of the kind of economically important work that people do. Let's imagine that most of those things that people have to do, Whether it's in manufacturing or whether it's in sort of administration or whatever else, or just sort of moving papers around the world. Or virtual papers around the world and things. Imagine that there came a moment when, gosh, we've got this AI and it can just do all of those. And in a very short time, A lot of people who were previously having to kind of go into an office every day and do a bunch of work they didn't particularly like doing in order to make a living and so on, Imagine if suddenly that wasn't necessary. And imagine there was. Whether the dynamics end up being Universal basic income or something quite different. Imagine that suddenly a lot of people were in this position where a lot of the kind of the grind of life was removed. What would happen? I think that's an interesting question for modern society. I think for some people that would lead to a kind of flourishing of things that I think we've generally valued as a society. And for some people it would lead to I'm completely bored and I'm going to do crazy stuff or just I'm not very happy because I'm completely bored. I mean, it always. It's always one of these things where when people retire, it's always a very dangerous thing to do. Because when people are doing stuff, there's kind of a, you know, a. A flow of. Of successive days. And when people sort of retire and say, well, I'm not really going to do anything anymore, that's usually not a very healthy situation to be in unless you have something you really, really want to be doing. So I mean, I think this question of if people don't have to do anything, will that be good or bad? The lesson of history would tend to be for at least some stratum of people. And it's a good question, which I don't know the answer to. In these moments of sort of really advancing and flourishing as a result of prosperity, was it the case that sort of at all parts of society that things were good? Or is it just that a small set of people who would otherwise have been grinding, doing something where they couldn't do creative work, for example, were suddenly freed to do creative work? And that's what gave us the main thing that we know from that period in history. I don't know the answer. That's a good question. And I think that. So, you know, the question of whether if we removed sort of the grind, would it be the case that one possibility is there'd be 1% of the population that would flourish in a way that was not possible before and potentially would lead to innovations, creations, et cetera. People say that was a really cool thing to been able to achieve, but that perhaps even 99% of the population with the grind removed, would actually not have such a happy time of it. And then you have to ask the question. It's sort of a very political philosophy utilitarianism type of question. If that was the case, if sort of the removal of the grind led to a sort of dramatic flourishing of a small part of the population, would that kind of, you know, what does one want to have happen there? And I'M not sure. You know, I think that when most kinds of technology people find, you know, generally when, when it's mostly bad and there's very little good in it, it doesn't have that much of a chance to survive as a technology. But when there's plenty of good in it, but nevertheless some bad, it will tend to survive. And, and the bad, you know, it'll be a complicated story whether the bad eventually overtakes the good. But I think, yeah, it's a. I mean, I suppose it depends a bit what we mean by human flourishing. I mean, even, you know, you could argue something like, how does it affect human flourishing? That it's possible to travel across the world in a realistic, for, for many people, in a realistic amount of time, in a sort of economically viable way. Does that mean that people can kind of, you know, maintain human relationships better than before? Does it. Does the fact that there's, there's phones or that there's video conferencing, does that increase flourishing? I think one would assume if one had looked at it from the past, one would have said, absolutely, that will increase human flourishing. I think now it's like, well, I suppose people would. People have adapted to it. People adapt very quickly and they say, you know, again from the past, people say, if only we were able to have blah and blah and blah, the world will be a better place, so to speak. And then such and such a thing arrives and then people kind of get used to it and they just say, oh yeah, it's just the way the world is, and, and then they find the things that are wrong with it. And then they say, well, maybe we should get rid of it. Not remembering that there were all these things that people really thought were good about it. So I don't have a great answer to this question, though. I think it's a, it's an interesting and important question. The DIANA is asking, what does the future of libraries look like? You know, I'm embarrassed at how rarely I go into a physical library these days. When I was younger, I was a huge enthusiast of libraries. I could find all these amazing things that were in books and journals and so on, and they were, you know, that was how I understood things about the world. As soon as there was the web, that became less important. You know, I think the browsability of the web is not. I mean, we have browsers, but we browsability in the same sense that I used to do of. You go look at the books that are shelved in, you know, QC175 you know, 0.1 or whatever it is, you know, that some particular section of a library that's shelved according to the Library of Congress classification or duodecimal classification, whatever. You know, you can live in the 530s in the jury decimal classification and so on, and you can kind of go in a library, you can kind of look down the shelf of all the books that are in classification 530, and you'll find all kinds of things where you say, oh, I didn't know about that, and that's interesting, and so on. That was fun to do. And, you know, there is obviously a way that one can imagine replicating that virtually on the Web. And the fact that I don't know of anybody having done that is kind of a strange thing. I mean, it's sort of an obvious thing to do, is recreate different densities of library. Like, for example, if you go to the library that is the fairly small library with 100,000 books, for example, you look in the section on foundations of mathematics or whatever else, maybe there are three books there. If you go to the library that has 2 million books, then there'll be many more books in that area. And one could imagine something would be a nice kind of app on the Web where you kind of go to different densities of libraries, so to speak, and the. You can kind of browse the shelves and browse these virtual shelves. I see Harry commenting that lots of libraries have ebook checkouts. Yeah, I'm not sure that's quite the same type of thing. I mean, maybe it is. I, I've not seen such an interface where it's kind of the equivalent of what used to be kind of going into that mysterious part of the library that wasn't clear humans had been in for a while, and going and looking at that obscure shelf, there was a certain sense of discovery that was quite visceral in kind of looking at that shelf where you pull the book down and there's dust flies off from the top of the book. And you kind of know nobody's looked at this in years. But, you know, I think, what have libraries turned into? I, I, you know, I myself don't really know. I mean, I have the impression that there's a certain sort of community character to at least public libraries that seems like I'm in the town that I live in, spent a lot of money building a very beautiful new public library. And I'm embarrassed to say I don't believe I've been inside it since the, since its grand construction. Project was done, which is sort of a strange sign. But my impression is that more as kind of a community intellectual and other center is a potential sort of application for libraries. And I suppose as one potentially sees more education happening outside of the traditional universities and things like that, that's a place where that could be a role. And it was a role for. But many early libraries, I mean if you go back sufficiently far in the past, the Library of Alexandria was basically a university where people went to learn things and you went to the library because you could then, you know, go look at the scrolls on the shelves or whatever and learn things from them. And there were also people there that you could learn things from. And even in much more recent times of 150 years ago, there was a certain amount of kind of library situated education going on. And maybe that's something that gets to come back. Now in terms of physical books, I like physical books. I think I mentioned this before in one of these live streams, I mean I find the concept that you can have sort of this lump of knowledge and you can sort of pick it up and you see how much there is there and you can kind of understand at some visceral level, you know, how much of it you flipped through and how much you haven't. I, I find that a useful user interface. Maybe, maybe we'll get some augmented reality version of virtual books that's just as good as that. But so far at least the physical version has been useful. And even the, even the process of producing a book where you really have to just sort of line up all the ducks, get this thing ready, you're going to make a print run. It is, you know, it's all ready, it's all a thing that's there and there's no sort of tentacles that you can kind of add later I think, I think that helps to define something which is useful. Ah, okay. Prototype says there's a virtual shelf browsing experience called Open Library Explorer. Shows you cover art and titles. Okay, I have to check it out. The. Crab is commenting. The future of library is the anti library with more books being collected than read. I don't know, I think, I mean, you know, I have about 10,000 books I think and I have a one day I'm going to do some live stream tour of my, my book collection and library and so on. I just keep on not getting around to that because it's a bit more technically difficult than what I usually get to do here. But for me, you know, it is really nice to be able to, you know, browsing it is convenient, I would say, that the. The pure collecting of books for the sake of collecting them becomes sort of an investment story and not really one that I'm so interested in. But I think, you know, it's always a funny thing, libraries. There's some sort of secret advantages to libraries. Like, for example, libraries tend to have a nice acoustic environment. And part of the reason for that is that when you have a book on a shelf, there's this air gap above the book and sound gets into this kind of Helmholtz resonator space that is sort of behind the book. And so books are really good absorbers of sound. They're better than, you know, you can have those fluffy things that you put on the wall and so on, and acoustic tile and things like this. Books are a really good sound absorber. So that's independent of their intellectual value or whatever. They create a certain ambience just as a result of things like that. But I think, yeah, I'm not. It's. Well, as I say, I like physical books, but I have to say I'm. I'm not sure what the kind of. Almost what the strategic future of large collections of them is. I have a friend who runs the Internet Archive who has been scanning a large number of books and has giant containers full of millions of books. And he's been wondering what to do with those millions of books and whether. Whether he should create kind of a book theme park or some such other thing that is sort of a destination where you can go and browse 3 million books or perhaps have. It's just unclear what you do with that. Let's see. I should wrap up soon here, but. Boost is commenting that they value their personal library with technical reference books and so on and lots of things that can't be found online. Yeah, it is a funny thing, what, you know, I don't know what the curve is going to look like of the point at which sort of everything that is at all published can be found online. I mean, these efforts to scan books and other documents have been reasonably successful. I would say that the real peak of success I consider to be documents from the 1800s, documents that are out of copyright. I know in historical work that I've done, it's just spectacular what you can get from, particularly the 1800s of all these newspapers and everything. And when will it be the case that anything that was broadly printed can be found online? I don't know what the rate of sort of closing of that process is. There's Certainly a. And it depends a bit on the actual mechanics of scanning things has not really advanced that much in quite a few years. I mean it tends to be that you have, you know, you take a book, you put it on this, you know, usually a, an inverted V shaped glass thing, you have camera underneath it, you, you know, you might have a foot pedal where you press it and it flashes and takes a picture. You go on, you turn the page, you put it down again, you flash it again, etc. And you know, there are different schemes for doing that, but it pretty much amounts to that. And I don't know whether there's been much advance in automatic page turners. I think that's a challenging robotics problem. And when it comes to scanning just random documents, you know, I have taken some trouble to scan paper documents that I have and I think I have now. I haven't counted actually recently, but I'm guessing it's around so probably approaching half a million pages of scanned documents. That are most things that I've ever kind of dealt with. Well, most non scratch paper kinds of things that I've, I've ever sort of dealt with mostly from the early 1980s and well early to mid-1980s and before when sort of everything I did was done on paper. And you know, it's been, I would say that the process of scanning all that stuff is surprisingly difficult and it takes surprisingly much human effort. And you know, you can, if you're lucky there'll be some stack of paper that is all nicely the same size and the same way up and all this kind of, it doesn't matter if it's the same way up but the, the same size and so on and you can just put in an automatic feeder and it'll just go, you know, ripple through and scan each page. But I would say that that's very much the minority experience most of the time you really have to go painstakingly essentially item by item, putting on a scanner, flashing it, going on, getting the next one and so on. It's a surprisingly non automated process to this point and my guess is it is not one that is easy but will easily become robotic anytime soon. So that, but I, you know, so I wonder whether in another 20 years whether it will be the case that sort of anything that's been printed in reasonable numbers you can expect to find online or whether it will still be the case that there's lots of stuff where that isn't true. I think there will be certainly some issues of kind of copyright issues and so on where sort of the alignment, even when people want their thing scanned and they own the copyright, getting sort of the connection between that and the people who are doing the scanning and the distribution of the scans and so on, is still a very imperfect setup at the present time. All right. Well, a few some interesting questions and I see a lot more that I look forward to addressing another time. I think we should wrap up for now and thanks very much and see you another time. Bye for now.