Ep55 "Could a brain plugin instantly teach you to fly a helicopter?"

Cosmos with me David Eagleman. I'm a neuroscientist and author at Stanford and in these episodes we sail deeply into our three pound universe to understand why and how our lives look the way they do. Today's episode is about the potability of really coming to understand the tangled forest of eighty six billion neurons in your head and the trillions of connections between them. And if we could do that, could we upload information directly into your brain? Could we

speed up education this way? Now? At the moment, this is all pure fantasy because we simply don't have the technology to allow us to do that. But the question we're going to ask today is whether this is theoretically possible and something we can look forward to around the corner of the next century, and what are the caveats, the things to watch out for, and the unexpected complexities here.

So let's get started some hundreds of years ago and still in many impoverished places in the world, children of the species Homo sapiens reproduce by the time they are young teens. But this situation is it's totally different in modern times and modern societies. Now, young people go to school for their first eighteen years or twenty one years, and increasingly twenty five or twenty six years for an advanced degree, and in fields like medicine, they take another

several years of internship and residency. And in a field like neuroscience research people do a postdoctoral fellowship and then they hope to become an assistant professor, and then an associate professor and finally a full professor. And most people are in their forties by the time they get there. So what accounts for this recent historical change. Why do we do so much schooling for so much of our lives now? Well, it's because we are a runaway species.

We've gone off in a totally different direction than all our animal cousins, and we have made thousands of important discoveries about our world and produced so much art invarious forms, And as a result, there's so much to learn, and so we need to spend decades in institutions of learning, not to mention, reading books and listening to podcasts to understand what millions of humans have devoted their lives to figuring out. But what if there were a way that

we didn't have to do that? What if there were a way to simply upload the information, in other words, to put the information directly into your brain. So let's harken back to this great scene in The Matrix where Neo and Trinity are being hotly pursued by the antagonist, agent Smith, and our two heroes end up on top of a building, and there they spy a helicopter parked on the roof, and Neo asks Trinity do you know

how to fly that? And she replies not yet, And she flips open her phone and she calls Tank, the operator, and she says, I need a pilot program for a B two twelve helicopter, And we see the operator rotate his chair in front of his bank of computers and he quickly types out a bunch of commands, and she closes her eyes and one second later she turns confidently

to Neo and says, let's go. So what happened is that Tank the operator had taken the expertise, the complicated know how of flying a B two twelve helicopter and just uploaded it to her brain. So the question we're going to ask today is is that theoretically possible from a neuroscience perspective, and what will make that straightforward? And

what will make that not straightforward to accomplish someday. Now, in some ways, the whole idea sounds crazy because it seems like we always have to earn things if we want changes to our brains or body. You can't just get something for free. But of course, people for decades have been climbing in into suntan booths instead of spending days outside, and people get botox, which binds to receptors the ends of peripheral nerves and changes the wrinkliness of

your face. And people are increasingly doing things to not have to go to the gym but instead to have your abdominal muscles built for you with electrical stimulation. You just lie on the table and your muscles contract over and over and The idea is that your muscles can grow stronger and look better without you having to do a single sit up. You just lie there. So what

would be the equivalent in the realm of education? Can we imagine a time when you don't have to bury yourself in a book to master some domain, where you don't have to spend hundreds of hours sitting in a flight simulator, but instead you hook something up to your brain and then it is as though you already knew quantum mechanics or electrical engineering, or Persian history, or how to serve for hang glide or repair that model of dishwasher or whatever. Now, how would you push information to

the brain? We currently do this by sitting down dozens of children in front of someone who already has the information in their brain, and that person uses words or pictures, and the students attend to those stimuli and try to translate those words or pictures into changes in their own private jungle of billions of neurons. They try to convert what they're hearing or seeing into storage in their own internal model in a way that makes sense to them.

What learning means is that you very finely change the networks in your head. That's it. That's what we pay lots of tuition for and go off to college for to get someone who already has information in their network to translate it through the low bandwidth channel of language over to your network. So, just to be clear on this, before you know some factor concept, your network is configured in some way, and then I tell you, oh, that dog's name is Nebula, and then you encode that information.

This connection in your brain gets strengthened and this one gets weakened, and this synapse unplugs, it replugs over there, and this happens over millions of synapses, and then you know something that you did not know before. And for deeper knowledge, like flying a B two twelve helicopter, this requires not just the memory of a fact, but of a procedure. And so those changes happen in different brain

areas and they're more widespread. But what is required in all these forms of learning are simply changes in the patterns of your network, presumably just the synaptic connections, but maybe other details as well, like which neurotransmitter receptors are being expressed on the membranes and whatever. But that's it, that's what it means. To learn something. So is there any way to implement those changes besides the old fashioned way of sitting for a semester in a classroom or

spending hours in the helicopter flight simulator. Well, there's been a lot of excitement about brain machine interfaces, such as the brain electrodes that are implanted robotically by the company Neurallink. So I'll just take a quick moment to clarify the landscape of electrodes in the brain. Even though neuralink hit the news recently. The first thing to note is that these brain machine interfaces have been around for many decades

since people started inserting electrodes. These are just thin metal wires into the brain. The idea is that you just insert this electrode into the neural tissue and you listen to the electrical activity of the cells. And researchers pretty quickly figured out that if you send a little bit of electricity down the wire down this electrode, you can stimulate the cell to make it active where it pops

off its own little electrical spikes that travel around. So you put in some electricity and it goes And this is the technology behind, for example, deep brain stimulation you might have heard of this. Take Parkinson's disease. There's a tiny brain region called the subthalamic nucleus, and it was discovered starting from work in the nineteen seventies that you can insert your electrode into this area and zap it with a bit of electricity and you get these amazing

effects of the movement problems of Parkinson's essentially disappearing. And by the way, the reason you can stick an electrode into the brain is because the brain doesn't have any pain receptors, so you can just dunk the little metal wire right in there after you've opened a little portal

in the skull. So what's happening when you put these little bursts of electricity in is that the cells fire, which has effects on the rest of the network that those cells are connected to, and it also changes the electrical oscillations. And why this works so well in Parkinson's is still a bit of a mystery, but you get what you want out of it, and people have been using this sort of brain stimulation for all kinds of purposes.

For example, my colleague Helen Mayberg puts electrodes directly into a very specific area near the singulate gyrus, and she stimulates and can pull people out of deep clinical depression

this way. So there are many labs and clinics using the technique of stimulating individual cells in the brain, and the direction of the technology over the past couple of decades has been getting more and more electrodes implanted, so that you're not just hitting one or a few cells at the tip of the electrode, but you're instead exciting tens or hundreds or eventually thousands of cells by using

a whole specific collection of electrodes. And companies like Neuralink have become famous in the public eye because of the idea of sewing these electrodes very finely into the brain and getting a thousand of them and soon more than that. And in all these cases, the electrodes can read and write, in other words, they can record the activity in the brain cells, but they can also stimulate the brain cells

to put activity in there. So once you have the electrodes in there, could you just send in the right zaps of electricity in just the right pattern, spread over millions of neurons with precise timing of your patterns in such a way that you shape the network so that you can fly a helicopter. Now, all that sounds pretty exciting as a theoretical possibility, but I think there are two major technical hurdles here to be able to stimulate lots and lots of cells in the brain in the

way that you might want to upload helicopter instructions. The first is simply a physical challenge. The brain is very delicate, and so Mother Nature has surrounded it in the armored plating of the skull. So it's very very hard to get at this fragile, delicate tissue of the brain, and so if you want to insert an electrode, you have to actually drill a small hole in the skull to expose the brain and then you can put your electrode in.

The difficulty is that there are eighty six billion neurons, and at the moment, even with our fanciest technology, we can only get to say a thousand of these at any time, and so that is useless in terms of actually having access to the whole system. It would be equivalent to if you really wanted to say something to all eight billion people on the planet, but you only had one hundred followers on social media. The huge majority of the world will have no idea that you've ever

said anything, or that you even exist. And that's the situation. When you zap a few hundred neurons, the other tens of billions of neurons don't even know that you're knocking on the door. So to actually insert information into the brain, you'd somehow need to access all or at least most of the neurons to make any meaningful change. Now, I'm not yet addressing how you would know what you want to change, where I'll come back to that in a moment.

Let's just imagine for now that you know exactly what you want to tweak in the brain. Now, I do think that in the future there may be a very different solution besides electrodes to this issue of manipulating the network, because I don't think the idea of dunking electrodes in there is ever going to be a long term solution. When I squint into the future, I think the solution is something like nano robots. So what are nano robots. The idea is that you use atomically precise three D

printing to make little molecular machines out of atoms. Essentially, you make little robots that carry out some functions, so they're like little robots, but they're microscopically small, built out of individual atoms, by the way, which is what proteins are. Anyway, you could make these super durable, for example by printing them out of carbon, making them diamond robots. The idea,

and this is probably not for several decades. The idea is that you swallow a pill with tens of billions of these little nano robots in there, and they float through your bloodstream and you give them the right FedEx labels to pass the blood brain barrier, and once they're in there in the brain, they wiggle their way into your neurons where they can read the activity and they can cause the cell to spike to fire signal whenever

they need to. So, with proper signaling between the nanobots, using for example, mesh networking, you could in theory generate whatever patterns you needed to across the entire brain, and if your science is really advanced, then you hit the correct brain wide patterns that will cement in the knowledge of how to fly a B two twelve. Now, although this is not happening anytime soon, it certainly seems plausible that this could be in our future. But wait, there's

actually a difficult twist to this story. I said before there are two technical hurdles, and here comes the second. And that hurdle is that there won't be a single program for flying a B two twelve helicopter. Why not, because the brain inside each of us is totally unique. We each have a massive forest of eighty six billion euro on each with ten thousand connection points reaching out

and interacting with other trees. And it's a living forest such as each connection, every twig on every branch finds its place in life based on the exact details of what you have seen and heard and experienced in your life. You born in your hometown, with your family, your neighborhood, your culture, your moment in history. All those things determine

the exact wiring of your brain. And your brain has a network that is different from his brain over there, and her brain over there, and everyone else's brain on the planet. And the exact wiring is what makes you you. So in the proposed future of the Matrix, the operator Tank would have to specify that he wants a program to pilot a B two twelve helicopter that is specified exact exactly for Trinity's brain, that is bespoke for her

neural network only. And if Tank tried to upload the same program to Neo's brain or Morpheus's brain, who knows what that would result in. Because if the program alters the way that neuron nineteen million, three hundred fifty six three hundred and two is talking to its neighbors, and it does this over a million other neurons with high specificity, that might teach Trinity how to fly a helicopter, but it certainly would not work for someone else whose brain

is different. So how do we get around that problem, the problem of everyone having a unique neural network. Well, the answer will have to rely on what is called system identification. This is an engineering approach where you have some complicated dynamic system and you measure lots of input output pairs, as in, when I put this in, what happens? Okay, now it happens if I put that in. So imagine you find a really complicated machine and you don't know

exactly what it does. So you tap one of the keys and you see how it moves, and then you tap three of the keys at the same time, and you look at what it does as its output, and then you hit a series of the keys and you see what results. And you do this over and over and over to try to figure out what is the structure under the hood. This system identification approach is used in lots of fields. For example, in economics, let's say you want to figure out the guts of the stock market.

So you take lots of inputs like gross domestic product and inflation and unemployment and interest rates and blah blah blah, and you look at all these as inputs and you look at the reaction of the market this way, and you develop better and better mathematical models of what the machinery of the stock market is doing, even though you can't see it. Okay, So the question is, could you do system identification on a human brain. No one's ever really done this because there's no purpose for it now,

but someday it might make sense. So the idea is you go into a super futuristic brain scanner and you get lots of inputs, and this sophisticated brain imaging device measures the outputs, in other words, which cells in your brain are responding. So you see a rapid series of images and you hear words, and you feel touches on your body, and you smell smells, and you run through thousands or maybe millions of little micro experiences while your

brain is getting measured. And in theory, this is how a scientist could say, Aha, Trinity's brain is organized like this, while Neo's brain is laid out like that, and Morpheus's brain has a slightly different pattern, And you might find that for teaching the operation of a B two twelve helicopter, in his brain thinks about it in analogy to riding a horse and controlling it, which let's say she grew up riding horses, while Neo's brain would learn the helicopter

in analogy to the way a motorcycle feels, which is, let's say how he grew up. And for Morpheus, the actions of piloting emerge from his deep knowledge of surfing, which is how he grew up and what is stored in his brain. Now, it's not clear how many inputs you'd have to ping in there to get high enough resolution to make all the little changes you need, but

presumably that would get figured out with enough experimentation. Okay, so let's say we as a society grow to a point where we can do system identification on an individual's brain and then use nanobots to upload knowledge of helicopter piloting. I need to emphasize that this is not right around the corner, but it certainly seems the theoretically plausible. Another century of advancement, and suddenly the network that makes you

can get directed and shaped in a bespoke manner. And if we come to a point where we can do it, that's possibly the biggest societal change. I can imagine you say to your three year old kid, Okay, we're gonna upload first grade now. Great, Now, go play outside for an hour, and then we're gonna upload second grade after lunch. Imagine that by the end of the week, your three year old knows as much as a full professor does. Now, so what becomes of society and the way we run

it now? You may think the analogy here is to look at super smart, genius kids in our current world, But these kids often go off to attend college at twelve years old, and they very often end up lonely and socially misplaced, because really what they want is to play with their colleagues other kids their age, But they get stuck with a bunch of older kids who have gone through puberty and are running deeply carved evolutionary programs that cause their brains to be taken over by sexuality,

and that software hasn't yet turned on in the heads of these young genuses, and as a result, they can't mesh with what is happening around them, and they can feel very lonely in these contexts. But the future scenario of uploading knowledge is totally different because now every single kid can stay among colleagues. But the question is if education is uploaded, what do the kids do all day? Do they launch startups at the age of six, do they write epic novels by the time they're eight years old?

Do they return to reproducing as teenagers like their distant ancestors did? And is it dangerous that they have all the knowledge of decades of schooling but without the maturity. The most slowly developing part of the brain is the prefrontal cortex, and this underlies our ability to simulate possible futures and think about consequences. So imagine a kid with an undeveloped prefederal cortex who has all the knowledge that

Albert Einstein commanded at midlife. But this child lacks the ability to simulate consequences, so they think something like, wouldn't it be hilarious to build a small nuclear bomb and blow up my neighbor's porch, Or wouldn't it be a crackup to disrupt the presidential broadcast by hijacking the frequency

and imposing a video of me twerking or whatever? Because children don't yet have a fully developed profederal cortex that can't simulate consequences the way an adult can, and this is why it could be dangerous to inject the knowledge of an adult into a child's brain. Now, perhaps I'm being shortsighted here, and we could somehow upload maturity as well.

We could figure out the learning that translates to morally complex situations and simulate those over and over do the synaptic equivalent of working through the possibilities and feeling the consequences. Maybe you could massively speed up emotional learning that way. After all, as my father would always tell me, the wise person learns from experience, but the wiser person learns

from the experience of others. So maybe there could be enough uploaded knowledge where a kid understands various possible scenarios and outcomes, and the good decision making simply results from a deep knowledge of previous examples, things that have happened to other people, all of which have been uploaded. So maybe the maturity problem could be taken care of, but still we're looking at massive societal shifts that would render

our current civilization totally unrecognizable. Now, we all like to be very thoughtful about the future, but it doesn't matter what we speculate about it, because we are guaranteed to be wrong. We can only envision what we're capable of, in this case, a cartoonish version of a bunch of super intelligent kids running around while their parents go off to their jobs. But the world is likely to be

very different by then. Presuming that everything is massively sped up by artificial intelligence, it seems very possible that society is going to evolve exponentially faster at a pace that we really can't conceive of here in the first third of the twenty first century. I mean, just imagine that AI knocks down scientific problems rapidly, such that we move from our current state of pretty wide spread ignorance to perfect,

wonderful models of everything in the cosmos. Just think about the incredibly slow pace between the Stone Age and the Bronze Age, and then the Bronze Age to the Silver Age. Now imagine this pace goes up by a thousandfold or a millionfold. So we find ourselves a few decades from now in the Diamond Age, where we can manipulate carbon atoms however we like. And then a few years later we're past that and into a new era where we can entangle photons and find ourselves in the quantum age

and so on. Like everyone, I love to speculate about the future, but the truth is that it is impossible to picture what things will become and how quickly. And I want to share an example. Last month here in Silicon Valley, I saw a black and white photograph from nineteen forty. It was a man on horseback ambling up a dirt road on a hills and there was nothing particularly special about this sandy hill with its scrubbrush. So I was intrigued to read the caption and find out

that this little dirt road was sand Hill Road. Now you may know that sand Hill Road is nowadays a road almost as famous as Wall Street in New York. Sand Hill Road is where many of the world's most elite venture capitalists do their business. They invest hundreds of billions.

This road is the mecca for startups who are seeking investment. Now, the thing that was so striking to me is that for the horseman sauntering up this sandy hillside in nineteen forty in the hot sun, there's no way he could have imagined that the lonely hoof prints he was leaving would in just sixty years mark this spot of one of the world's economic engines. And there's no way he could have envisioned what advances would get funded on that spot.

The worldwide light speed network that allows anyone on the planet to effortlessly communicate to anyone else, or rectangles that everyone would carry in their pocket like a handkerchief or a tobacco tin. But these rectangles would contain the accumulated knowledge of all humankind. Or satellites or quantum computers or blockchain cryptocurrencies, or large language models that could read every

book ever written. None of these would be even vaguely imaginable to the cowboy in nineteen forty, moving slowly up the hill. We are blind to the future. I often wish I could talk to whoever is listening to this historical podcast in the year twenty eighty four, because the world will be so different by then, and I am incapable of imagining it. And it's not just that we are not being creative about extrapolating technology curves into the future.

It's that there will be new technologies and novel sciences and new convergences that will make it intrinsically unpredictable. There will be serendipitous discoveries and socioeconomic changes and geopolitical events. While we always make guesses based on our current trends and research, the future is shaped by hundreds of things we just can't see. Not only that, but you've heard me speak before about our limited perspective, our inability to see past the fence line of what we already know.

Our current knowledge understanding are based on the technologies and paradigms that exist right now, so it's really hard for us to anticipate breakthroughs or paradigm shifts that are going to radically alter our society in the future. But this idea of putting information directly into the brain, that's it certainly seems like that could be a big shift. So when we think about the future, it's more than just adults like us riding around on a spaceship with a

robot or two. Things are guaranteed to be weirder than we expect. While brain uploads our science fiction right now, assuming we don't blow ourselves up, this inevitably seems like it will become science fact. So let's wrap this up. This episode is not about what's going to happen anytime soon, but I think it is inevitably what will happen in the future. After all, the brain is made of billions of cells, each one of which is very complicated, and

each is connected in very complicated patterns. But fundamentally, learning and memory take place in the changes of connectivity, and as far as we can tell, that's all learning is. So what we talked about is the way that the jungle of neurons in your head is wired up differently than in your friend's head because you have different genetic predispositions,

and more importantly, you have different experiences in life. So in order to upload any changes into your network, we'd have to know your brain in exquisitely fine detail, and we'd have to know those patterns right now, because it's

just a little bit different than it was yesterday. But in theory, if we had this information and understood the language of the connections, we could dial knobs here and there in a million other spots, strengthening or weakening synapta connections, tickling the genome to express a little more neurotransmit or receptor over here, a little less over there, and after that you might be able to suddenly possess some knowledge

you didn't have before. Now, obviously, society will have to be very careful about this technology when that century comes, because in theory, you could use it to implant false memories, or to erase knowledge, or to do any number of nefarious things. So we will enter a very strange time, and like every technology, a whole raft of protections and legislation will grow up around it. Again, this is likely impossible to achieve in our generation because of the size

of the problem. It would take about a zetabyte of information to store the detailed structure of one human brain, and that, by the way, would only tell you the structure of the forest of neurons, but wouldn't even tell you anything about their individual details, like which genes are getting expressed and which proteins are getting put where. So for us, the citizens of the twenty first century, this is likely to be an unsolvably huge problem to capture

a detailed description of an individual brain. But as a species, we're in an interesting situation because we can see that this is all coming, and we can speculate on the size of the changes this will have on society writ large. Now, what I find amazing is our guaranteed inability to correctly picture this future world, even though it will be populated

by our own great grandchildren. Given all this, I think the only specific prediction we can make is that we have more in common with our ancestors two million years ago than we do with our descendants two hundred years from now. In the meantime, go to eagleman dot com slash podcast for more information and to find further reading. Send me an email at podcast at eagleman dot com with questions for discussion, and check out and subscribe to Inner Cosmos on YouTube for videos of each episode and

to leave comments. Until next time. I'm David Eagleman, and this is Inner Cosmos. You and not you. You