Latest / Art of the Question / Can We Upload The Human Brain? - Dr. Randal Koene - #28

Transcript

Can We Upload The Human Brain? - Dr. Randal Koene - #28 · Art of the Question

• Julius Moorman: currently doing a bit of writing ⁓ on the side and I also always like to make my own media, ⁓ part of which is this podcast, which is mostly about science, technology, philosophy, and psychology. And I believe the topic we're going to discuss today touches all four of them, ⁓ one way or another. And yeah, thank you. Thank you very much for the invitation. It's my pleasure to be here. ⁓ it's very kind to have me on your show. Yeah, no, I'm I'm I was very excited to to have this discussion because I think it's incredibly interesting what you do and I ⁓ yeah, I'm very eager to hear ⁓ more about it. ⁓ but before we dive into the specifics, I'm ⁓ very curious. When did you be f when did you first become interested in whole brain emulation and why? Hmm. Well you I think it's ⁓ it's fair to describe that in two different phases because I was interested in the idea that ⁓ what we are isn't really dependent on necessarily the biology that's underneath it, but could be something else for a very long time, probably since I was about thirteen years old or so. But ⁓ but it really transitioned into understanding that, okay, the way to do that or the way to look at and and discover this, figure out the problem, is through this concept of whole brain emulation, which by the way is a a term I introduced back in 2000, and I'm very, very happy that so many people have picked up on it and that brain emulation is a is an actual science now. This is awesome. But that yeah, that that probably happened ⁓ around nineteen ninety-six or so when I was Finishing my my master's at ⁓ Delft University and ⁓ when I then decided I was gonna do ⁓ a PhD in computational neuroscience, because at that point it was very clear to me that the approach is in fact through computational neuroscience. And so it's about reconstructing in terms of models that which is going on inside of the brain. yeah, so about that long. Yeah. Yeah. And and this idea that we like the self if if I'm going to to to put it that way. whether you want to describe it as a mind or or you our consciousness. ⁓ but but the idea ⁓ that we are a self, ⁓ an entity. the idea that that doesn't necessarily have to be a biological thing, but but can be copied it it Goes all the way back. You were thinking about this when you were thirteen years old. Was it inspired by any ⁓ books or or TV shows? Or did you just come up with this concept all by yourself? I would say it was informed from two different directions. one of these was my my dad's profession. ⁓ my father is a particle physicist. ⁓ he was working at CERN and and the way that he described things in the universe is is pretty much everything is ⁓ relational. There are all these different equations governing what's going on and everything about matter. It's not about which atom is doing what. They're all pretty much replaceable. You know, you could replace one atom for another one, and you can see similar functions all across different scales. And and when you choose, for example, which equation you're going to use, if you're just trying to describe how a rocket launches and where it's going to go, You don't need to use general relativity to do that or quantum physics. You can use Newtonian physics, even though Newtonian physics isn't entirely correct. It's an abstraction. So depending on what it is that you're trying to capture, what the goal is, the level at which you describe what's going on can be different, and it is not always the wisest choice to ki to choose the lowest, most complicated level. So that's one side of it that ⁓ That really played into the notion of substrate independence, into the notion that you can s separate the platform from what the platform is providing, just like you can have many different operating systems running the same type of software. And that's one side of it. And the other is indeed from from reading science fiction, but not the wild science fiction that that is more fantasy-like, but the science fiction of really well-informed scientists like Arthur C. Clarke. Who was describing pretty much the same type of idea in a in a story of his, which ⁓ you know, it it's called The City and the Stars. And this is a story in which, in fact, ⁓ you know, there is uploading happening. ⁓ there is this machine that can store people and then bring people back to life and all sorts of things like that. And it's all described in a believable way because it's based on on Arthur C. Clarke's understanding of physics, and I mean Arthur C. Clarke was someone who was involved with getting the first satellites into orbit and things like that. So this is a you know, it's not a fantasy writer. It's a someone with a science background. No, for sure. Yeah, I I love Arthur C. Clarke and it's always yeah, he he he was more than than just a science fiction writer. But even still, like ⁓ b a lot of things that were that started out as pure science fiction. ⁓ tend to become reality a couple couple well dozen years years later, at least some ⁓ sometimes. so you I I heard you make the comparison with with software, so ⁓ that our mind could be in in principle be ⁓ similar to software in the sense that if we given the fact that we would have the right hardware just like we can install ⁓ Photoshop, for example, on on on a computer and multiple copies of it, ⁓ our ⁓ our minds ⁓ would be similar in that way. ⁓ is there any evidence that supports this theory or or how how do you come to that conclusion? Yeah. But before I get into evidence that supports the theory, I just want to point out that I'm not claiming that the brain is like a von Neumann computer that has A set of instructions and memory of data and then combines the two. That is not how the brain works. ⁓ neural networks are very different. The brain has a type of neural network. As computational neuroscientists, we like to use the term neuronal network to distinguish it from the artificial neural networks that are much more abstract and simple. But it it is about a system where memory and what is being done to that, that memory, that data. Is happening in a very localized form, so spread out all over the brain. ⁓ But the ⁓ the evidence that there is a level of abstraction where you can say the function of what is emergent from the brain, the things that are being generated by the brain, which we might call the mind, the evidence comes, for example, from studies with lesions, where you see that the effect that the ⁓ the impact on a specific part of the brain when some part of the brain is being cut or destroyed or disrupted has very specific effects on the mind, on cognition, on the ⁓ on on what happens to that patient. and and then since then there have been many, many more points of evidence, such as, for example, being able to insert memories or s drive memories by just activating cells in a certain part of the brain. ⁓ so much evidence that brain function is directly linked course to the mind, but also that there's a functional divide between ⁓ things that are happening at a very molecular level in the brain and the activation that is at the level of electrical activity flowing through the brain from one neuron to another. It also makes sense when you think about it from an evolutionary point of view. And this is important to get into because there are certainly neuroscientists like Annel Seth and Miguel Nikolais Who critique this, who say you can never get something like uploading because biology is so analog, so complex, that ⁓ you have to basically model everything down to some insane level of complexity if you wanted to be able to replace what's going on. There's so there are these crit critics in that sense. But ⁓ you can imagine that from an evolutionary point of view, you don't really want a system that is that unpredictable. That is that purely analog chaotic. Because just like in a computer, computers are also analog. Computers have analog transistors. So it's really an analog system, of course. The only reason it operates like a binary, ⁓ not analog system is because you set up a whole bunch of constraints. You say that above a certain voltage, this is what we call a one, below that voltage it's a zero. There's an area in between that we don't Trust and don't get into. We have clocks that separate what's going on. We have parity bits that can correct when there's an error in memory. All these extra little protections are in there to keep the system predictable. And the same has to be true in the brain, and you can see the evidence of it too. It has to be true because you have lots of different regions in your brain that are communicating with one another, and they need to understand each other, so the neural code cannot be completely chaotic. It just can't. There has to be some level of abstraction through which neural code from one area of the brain can communicate with another area of the brain. You don't want a situation where you're out in the bush and suddenly you see something and you misrecognize you think, ⁓ this is my grandmother, but really it's a tiger. You don't want that. You need a reliable system. So nature also imposes constraints that make a system more reliable. It uses entire populations of neurons to represent things instead of in single neurons. It uses things like spikes and bursts to give very clear signals to what's going on. There are rhythms in the brain that cause neurons to be more likely to fire at a certain time and not at others, so that you could say that in every region of the brain, in certain modes, that there's like there's a clock and these clocks are all synchronized with one another. So that the output from one region will arrive just when the other region is ready to receive it. All these sorts of things are in the brain too, constraints that make the neural code more understandable. So it's extremely likely that there is a level of abstraction. In complex systems theory, this is called finding the ⁓ level where you do ⁓ can do scale separation or a separation of scales. And and this applies to the complex system of the brain as well, of course. There is a question of where is that level? And ⁓ it's quite likely that the the place where you have to do the abstraction is different in different parts of the brain. Because in some parts of the brain, you're really dealing with these population codes. And in other parts, like the auditory system, where you have to be able to pick up on very specific locations of where a sound is coming from. You're trying to do triangulation of where the sound is. It depends on a very fine fine timing detection. And that requires very precise activity of specific neurons. So that you would expect that the level of the neural code there is different than, for example, in a memory system. Yeah. And ⁓ w ⁓ I just ⁓ while you were explaining it also had to think about ⁓ cell replacements, in the sense that I believe that we ⁓ Kind of similar like snakes, but like re regenerate our cells throughout our body. Like every ⁓ in in two a couple years from now, I believe like pretty much all of my cells have like individually died and be regenerated. Like that that's a like a continuous process that's happening inside the body. I assume that's also true for for there's a there's a little caveat there. So all of the cells in our body do get replaced. brain cells are somewhat of an exception. Brain cells live a lot longer than other cells in the body. Now there is ⁓ for example in memory, when you look at hippocampus and consolidation of memory, there is a process where memories get consolidated and then they're reactivated and re-encoded. So there's a way that you could get the same memories or similar. Memory is not like on a tape or or or disc or something. ⁓ you could get that be reinstated on new cells. So there is a way to do that when you're replacing cells. But even if we don't take into consideration neurogenesis, which is still happens in adults, but mostly just happens in the dentate gyrus, one particular part of the brain, even if you don't consider the replacement of cells themselves, the atoms that make up your cells, the proteins, all of these components of the cells get replaced all the time. So even if those brain cells themselves have a very long life, the parts of those brain cells get replaced just like the planks on the ship of Theseus. Yeah. Yes, that's the that's the ⁓ a comparison I was ⁓ I was aiming for indeed. And before we we we dive into what whole brain emulation actually is, can you maybe give me your definition of the mind? So w what is the mind? My my definition of the mind is ⁓ the mind is all of the things that the brain can do. The mind, and another way to look at it, instead of coming from the brain, because this is already the assumption that the brain creates the mind, let's go from the opposite direction. The mind is all of the things that are required for us to have the experiences that we have. The experiences that we have, awareness by definition, is whatever's going on consciously. But all of those things that go on consciously depend on a huge amount of subconscious processing. There is a lot that we have to do to turn sensory input, ⁓ the signals from our own body, ⁓ whatever thoughts had already been going on and were stored in memory and that need to come back and turn all of that into representations that are consciously available. So all subconscious cognition and all conscious cognition make up part of the mind. Okay. And ⁓ is according to to your definition, is the mind ⁓ separate from consciousness? Is it related to it or is it practically the same thing? To me, consciousness is something that the mind contains that is part of the mind that is generated by the brain in a very similar fashion as any other cognitive process, like for example, the generation of emotions or emotional responses. behavioral responses and so forth. Right. So it's it falls under the umbrella of the mind. Definitely under the umbrella. It's not a separate thing. And I wouldn't even know which sensible theory and neuroscience to go to where you would end up with a separate thing. Yeah. No, okay. No, but I think that's important if we ⁓ before we move into the next part of of the discussion of whole brain emulation. what what is it exactly ⁓ what what are we talking about? What is you coined the term ⁓ well twenty-something years ago. ⁓ yeah, w what do we mean by that? The idea is that if you have an emulator, let's say an emulator that we're used to, like emulating a Macintosh on a PC or emulating a Nintendo system on some other computer, ⁓ we have lots of different emulators. You can emulate A flight simulator is an emulator of sorts. You're trying to create. So an emulator always has an objective that has something to do with a particular type of experience. In an emulator, you're attempting to recreate the experience that you have on one system in another system. ⁓ And it's important to keep in mind that emulators can have different goals. So, for example, if you're doing brain emulation where you're trying to emulate that which creates the mind. You're trying to emulate it in some fashion on a different platform that is not the biology that you're that you're using as the data, the record from which to create the emulation. You could have as a goal that you want to build an emulator because you want to test out new pharmacological drugs. And so then your emulator needs to very specifically have an implementation of how these pharmacological substances affect what's going on in the cells. Otherwise, you can't use it as an emulator for a brain for that purpose. If your emulator, on the other hand, is to create a a ⁓ version of, say, a fruit fly drosophila that is like an original fruit fly drosophila, then you have to package in there all of these little pits of cognition that a fruit fly has to be able to perform all of its behavior, et cetera. If it's a human emulation, Then you also really have to care about the internal experience. So it's not just an external behavior, but you want to you're looking for, you're aiming for, recreating the experience of being a human with a brain, meaning you have the memories, the drives, the characteristics that belong to that system that you're emulating. So that particular brain, that particular person, and of course consciousness and subjective awareness, in other words, identity. Yeah, so would a successful emulation of the entire brain be by definition conscious, or doesn't that necessarily have to be the case? Like I said, emulations are made for different reasons. So they don't necessarily need to be conscious. If you're making an emulator because you want to test out new BCI devices or you want to test out new drugs, that doesn't need to have consciousness unless, of course, you're testing a drug And you want to see what effect it has on the conscious experience. In that case, yes, you would have to include that as well. But yeah, for for the kind of whole brain emulation that is mostly discussed because it's aimed at this idea of mind uploading, ⁓ you know, as a basically to build a a an artificial brain, a neuro prosthetic for a patient in order to survive and continue living. If we're aiming at that, then yeah, that consciousness is it's absolutely essential. Yeah, no, I I agree. And ⁓ yeah, we we briefly touched on ⁓ specific use cases for whole brain emulation. You mentioned ⁓ testing out different drugs. ⁓ but what what what are some of the the use cases ⁓ of this ⁓ technology? Aside from that and ⁓ and the building of neuroprosthetics, where neuroprosthetics could be for the entire brain or could be for parts of the brain, because if a patient has a deficit somewhere, you might want to be able to replace that piece of the brain to give them back full function. Aside from that, ⁓ it's also very interesting for computational neuroscientists and neuroscientists at large, because if you have an entire map, an entire system that not just is a map but also works, then you can investigate precisely what's happening when we think, when we experience, when we feel pain, ⁓ when we process sensation. How we store memories and retrieve memories, all those things that neuroscientists have been interested in for so long. And for artificial intelligence, it could be a nice comparative system where you can say, ⁓ so this is how things are done right now in AI, and this is how we process information and how we think in a in a human brain. Or we could discover certain circuits and functions in the brain that would be useful for AI. So that whole field is called neuroAI, actually. ⁓ and yeah, I can't even imagine. There must be a numerous other things you could use it for that we can't even imagine right now. And you'll note that I haven't touched upon the ethics of the whole thing. That might be a whole separate whole separate topic. I I w I was going to touch on on the ethics for sure, but ⁓ first I I was actually more interested in exploring where we're currently at with this ⁓ technology because I think that's Very important ethics are very important to to discuss. And like I said, it's definitely something I wanna explore later. But ⁓ of course ethics get the ⁓ the ⁓ they get more important, the more advanced this technology is, obviously. ⁓ that doesn't mean we shouldn't have to think about the future, right? Because eventually the goal is to create a whole brain emulation that automatically comes with some ⁓ interesting questions. but maybe it's it's not as relevant right now because where are we currently in in the the technology? Because you mentioned the the fruit fly c ⁓ example there that obviously made ⁓ a lot of headlines but I also read on your website that ⁓ it was ⁓ Yeah, kind of misleading in in some ways. Maybe yeah, maybe not the authors of Yeah, yeah. Let's talk to that. Yeah. So first let me just say where we are. And it we're actually in a very interesting place. This is much, much different than where we were, say, 10 years or 20 years ago, because now the ability to collect data from the brain is so much better than it used to be. ⁓ we're still Not quite at the place where you could record active data in a living brain from every neuron altogether. That's that's still very hard, ⁓ involves. There's some ways to kind of approximate that for cortical structures to get close enough by making a a window in the skull and putting in a two photon microscope and and doing calcium imaging for a bunch of cells. But it's that's not quite there yet, ⁓ but much better than it used to be. ⁓ but the really great breakthrough is in ⁓ structural data collection. So the ability to get high resolution data of brains using electron microscopes and in the future also with something called expansion microscopy and licon. I won't get into that right now, but in the electron microscopy approach, you're taking ⁓ a brain, you've ⁓ prepared it as a sample, and then you work your way through slice by slice at a very high resolution. Taking many, many images with an electron microscope so that you can then put them back together again and trace out neurons, axons, dendrites, find where synapses are, even zoom in and observe the little vesicles containing neurotransmitters that you could count to get an idea of how strong that synapse is. So there's a lot of information coming now from this structure of the brain. ⁓ if we knew enough About how every neuron in the brain works and how every synapse works and how you can correlate structure to function. If we could really do that well, if we could always estimate function really well from structure, then the structural scan might be enough to be able to reconstruct a model that actually works and that you can validate that it works. We're not quite there yet, but this is way, way, way better than it used to be. And so now the fruit fly comes up a lot because. The fruit fly is one of the few organisms for which an entire brain has been scanned in this manner. And in fact, genalia research now has the scans of multiple fruit flies, so you can actually compare between different individual fruit flies if you want to. But the data itself hasn't been studied in nearly enough detail at this point. so you mentioned this ⁓ you know, how the fruit fly really got into the news recently, and ⁓ this was based on a two 2024 paper ⁓ that Philip Shu had written, where he took the fruit fly brain, the conectome that had been produced by genalia, and he replaced every one of these biological cells that you could see in the connectome with a very simple model of a neuron, so-called integrated and fire neuron, very, very simplistic, and he used the exact same model for every one of the cells. And then estimated weights, estimated strength of connection between each one of the cells by counting the number of synapses. So he didn't even look at the size of the synapse or how strong each individual synapse might be. He kind of estimated it from a count. And he left out ⁓ any kinds of rhythms that are in there, he left out inhibition, ⁓ so there was no, you know, force pushing down activity. there were many things that were abstracted away. It's a very, very simple ⁓ model. And the whole purpose of the model in the paper was to show that if you followed the input of activity from the sensors, like the eyes or the proboscis, that's this thing that the fruit flies eat with, or little antennae, ⁓ that you could follow these inputs and find out which cells downstream were receiving input from that and would be connected, would be in some way involved. So that he in the paper he made predictions for some cells and said, ⁓ it looks like this cell should be active when the fruit fly is behaving in this way or seeing this or smelling that. And then you can go to an experiment and you can record from those cells and then verify and validate and say, ⁓ yeah, that's right. It in fact that neuron does activate. That's what the paper did. And that was totally solid science. Very good. ⁓ well, and then there was a a demo where The company closed the loop, took this brain model that he had produced in 2024, and then dropped that into an environment simulation where you have a virtual fly walking around and ⁓ smelling or sensing food and then trying to eat the food, and and a body model that makes this walking possible. The body model is really an essential component of this demo and of what happened, because the body model itself contained all of the functions for making the legs move. Making the proboscis extend, making grooming possible. These were in principle just switches where you turn on and off. Now move left legs, now move right legs, now eat. Super, super simple in that way, but giving a an impression that looked very convincing. Super real, even if it was really just five buttons. So all that the model was doing, in a sense, was taking the input where the environment is giving you a stimulus. And then piping it through that set of connections, what he what he had been predicting in the 2024 paper, from this input to a neuron over there that ought to be activating. So if that neuron ought to be activating, you can use that in out as an output to drive one of those switches. So then if you put that model, that simple model of the Drosophila fruit fly that doesn't really show anything about how the internal mechanisms and the cognition of a fruit fly works. You just connect that to a body model, you get something that looks like a fly really doing behavior, walking around and grooming and eating. So that's what I what I was commenting on, that just a body model is not a great way to to evaluate an emulation. It's much better to really dig into the brain model and see what's going on there. ⁓ and I commented also that in principle you could take the brain of some other animal, and if you could do the through the sort of through lines from input to output, you could connect that to a fruit fly body. And get the same results. And in fact, there was a a ⁓ researcher from a German lab that went ahead and did this. They took the brain of ⁓ the nematode C elegance and they connected that to the environment and the fruit fly model. And whoa, look at that. Fruit fly behaves beautifully. So what does this mean? All brains are the same? Nematode brains are fruit fly brains, are human brains. It's all the same. Yeah, it's it's just not a good way to evaluate the actual outcome. That was the issue. Yeah, and and you briefly mentioned it, but what would have been a better way to to do this? You have to ⁓ you have to be able to see that the activation inside of systems and circuits of the fruit fly brain is what you also observe when you're recording from a real fruit fly, so that you can see, ⁓ here we get these oscillations, here this circuit is feeding information back and forward again. Here's the how the data gets processed and turned from ⁓ input that is light into identifying features, into identifying something you have to avoid and then becomes avoidance behavior, for example. So you really have to dig into the brain and examine the circuits and the circuit activity and compare that to the activity in the real fruit fly. But overall, this question of how do we evaluate an emulation, that's an unsolved problem. We don't really know exactly how to evaluate it because we don't have Good ground truth. There is no single emulation of a brain that actually works at this moment, that has been validated to work for the entire behavior of that brain. So there isn't a single brain where we can say, here's the ground truth against which you can test your methods for doing this brain emulation. And that's actually the challenge that ⁓ that we chose now at at the Carbon Copies Foundation to focus on. So we're building this ⁓ brain emulation challenge, a a benchmark suite where we give bunch of examples for people to try to do their brain emulation on and then evaluate how well it goes. But I think that probably goes into a another question you want to ask later on, maybe. Well yeah, no, I was definitely going to to ask about ⁓ about the competition ⁓ you were doing but also right now wondering what is the yeah you you practically ⁓ already answered it but I was just curious what are the the the current the biggest bottleneck ⁓ in in rolling this ⁓ technology out and ⁓ what all needs to be solved in order to eventually get to all whole brain emulation? Yeah. There are definitely at least two big bottlenecks. One of them is scaling up. So fruit fly brains are really tiny. They have about a hundred and twenty to hundred and forty thousand neurons. C elegance even tinier, 302 neurons. ⁓ getting to anything that is, say, the size of a mouse and beyond that, we don't really have the technology lined up to collect the data. ⁓ for that something that scale. It it right now for a human brain this would require a warehouse full of microscopes and all of them coordinated. And then it would still take five to ten years to collect all the data. And then you have so much data, and how are you going to process all that and dig through it? So the scaling problem is a really big deal. But that's kind of an engineering problem. ⁓ it's not a scientific problem. It's ⁓ it's a scale problem. The other, much bigger problem from a scientific perspective is that we just don't understand everything yet. There are still cells that haven't been properly properly tested. ⁓ for which we don't know all the dynamics. There are neurotransmitters that we don't exactly understand what's going on. ⁓ there are interactions between cells in circuits that we don't fully understand. And and on top of that, ⁓ we just don't really understand how to make that transition from structure back to function, or how to properly align our libraries of dynamical functions for cells with the structure, do estimates for parameters, and test it all out. So Altogether, I would just call that the problem of going from collected data, which is just a pile of data, to models with specific parameters and architecture. So going from the data to a system that works. It's a computational problem, it's also a scientific problem because of some of the unknown unknowns in there. So I think that's the biggest. But solvable. There's nothing in there that's theoretically unsolvable because Most of the things that we just don't know right now yet is because someone hasn't done an experiment somewhere and recorded ⁓ and observed, for instance, how a certain neuron responds to stimuli. ⁓ these are all techniques that already exist in neuroscience. It's just that they haven't been applied at that scale. Now, when it comes to the reverse engineering of an entire brain and knowing how to best to carve it up into subsystems so that you can do one system at a time and validate whether that system is working. There are some strategic questions there of how you should do that. The decomposition depends on things like that scale separation problem we were talking about before. So in a particular system, what is the type of neural code that's being used there? So there are scientific questions there, things that are unknowns. They're not of the kind where we go, ⁓ my gosh, this is as mysterious as dark energy or or something like that. It's not it's not one of those things where people only have really high hypothetical little ideas somewhere and and no clue about how to go about it. They're not of that kind. But they are large and complicated and require a lot of funding and a lot of work by a lot of people. Yeah are there are there currently ⁓ how how many people are working on on this? ⁓ because it does seem like you mentioned that compared to ten years ago, there's also of course Different fields that aren't specifically looking into whole brain emulation that are moving the field forward by doing other research, for example, by mapping out the brain to ⁓ for example, create BCI ⁓ implants. For example, I can understand that ⁓ this science is also yeah, profiting from from other breakthroughs that are being made. But I'm just curious to know how how big is the field. at the moment. Yes, you're up you're absolutely right. So of course this field benefits a lot from adjacent fields that also work with neural signals or work with mapping parts of the brain or connecting with neurons. ⁓ so all those fields are relevant and then fields outside of neuroscience like the advances in machine learning and just the available compute power, all of that helps work in brain emulation and it it doesn't have to come from the field itself. But field itself has grown a lot. ⁓ and the biggest ch difference is that the notion of being able to do brain emulation at all, beyond maybe in the worm, was something that was really not something you could propose as a as a something you could get a grant for a few years ago. You would sit around ⁓ with a beer with your professor to discuss how wonderful it would be to do something like that. Because they understood that theoretically it's possible. There's no reason why you can't emulate what's going on in a brain. It's a piece of matter that r behaves in a certain way. ⁓ But now this is a very real field of research. And there are ⁓ there are labs at MIT and Stanford and also in Europe that are directly involved with either understanding how to emulate what's going on in memory and decode memory. ⁓ for different kinds of animals, or that are involved with ⁓ the emulation drive directly that are trying to do that for very small animals and then moving up to humans. Most recently, maybe just a few days ago, the Allen in the Allen Brain Institute itself ⁓ posted a new yeah, an article in which they describe ⁓ the route towards human brain emulation and they allow you to play around with a little with levers to try to determine what is the timeline to get there if you do different things, how you would approach the problem. So this is a, you know, among among the big institutes out there, Brain, the Allen Brain Institute, Genalia Research, those are kind of the largest in the United States that are not directly part of a university. And and they've just, they're starting to engage with it completely. There's no doubt that they're they're looking straight at emulating mouse brains. Then maybe primate brains, human brains. It's absolutely a goal now. So I'd say That's awesome. We're talking about thousands of researchers now, where in the past you'd have what twenty. So it's a it's a big deal. Yeah, that is a big deal. And that's awesome. That's great to hear. And it it seems you you mentioned the the stepping stone, which would seem very logical to me, ⁓ by starting out with the simplest of creatures and then slowly move our way up to the the human brain would be which would obviously be ⁓ the holy grill of ⁓ of emulation, if you will. Is that also the approach that's currently being taken? The the the stepping stone ⁓ approach I think that's everyone's approach. ⁓ that's not entirely true. Sorry, I should back up there. All of the people I normally talk to are using that approach. There are groups outside of that who are looking at Brain emulation as something that perhaps could be done at with an enormous amount of abstraction, where you almost treat the entire brain as a black box, ⁓ where you're just observing behavior, observing input output, collecting a lot of information that can include video, can include speech, can include all this sort of thing, and then use that to train an AI to become more and more like the observed animal or human. Mostly this is human, they'll just jump straight to human to to persons. And sometimes this is given labels like digital twin and things like that. ⁓ the the more neuroscience oriented people who are attempting that sort of an approach don't necessarily call it that because that's often considered just building avatars for online use or something. ⁓ but there there are people who believe that you can capture what's going on inside of the mind through enough of that observation. And in theory, they're correct. If you could watch every possible behavior, if you could watch for an infinite amount of time or an entire lifetime, you could capture everything that a person has done in that lifetime and try to evaluate how they figured that how that behavior occurred and come up with some kind of model. There's still two problems, which is if you then were to extend That lifetime, if you were trying to look at the next thing and you put that person into a circumstance that never before happened, you might still be predicting wrong because you don't have the internal mechanisms that are going on. Plus, it sort of assumes that behavior is everything. And ⁓ and we're not necessarily happy with that because we also care about internal states. We care about the internal experience, consciousness, feelings, stuff like that. So you don't necessarily want a big table that's just going through all the possible inputs. And producing all the possible output. ⁓ we we care about the I.O. of the subsystems as well, the the input-output of the subsystems in the brain. Yeah, so I don't really agree with that approach, but ⁓ yeah, it's it's sometimes seen as ⁓ this is more feasible, something we can do right now with the AI that we have. Yeah, it still sounds pretty pretty difficult. Well y you mentioned AI there. Wha what role does AI currently play ⁓ in this field of research? Because I can imagine that ⁓ yeah, we've seen major developments over the past few years. ⁓ has it sped up the research in any significant way? It does. Yeah, it helps. I'd say every neuroscientist I know uses AI day in, day out, in one form or another. ⁓ I mean, often this is just about helping analyze data or go through piles of papers and find what you're looking for, or it's ⁓ directly helping with programming, or it's using an AI method when you're analyzing your data directly, so that it's it's it's really an issue of ⁓ with these large ⁓ transformer networks, for example, or reinforcement networks, that it's possible to describe functions in a different way than we used to. We don't have to derive a symbolic equation quite the same way. So there are a lot of things that apply machine learning and AI directly to computational neuroscience. ⁓ but that's always as a tool, as something that we're using to explore the brain and then implement what's in the brain. It's not as if we're necessarily saying, okay, let's just abstract out this part of the brain and replace it with some AI similar thing or something like that. Yeah. No, that makes ⁓ that makes sense. And let's say we are able to make ⁓ a perfect copy. We we ⁓ scan the the slices of the brain or we find some other way to to create ⁓ yeah an exact digital replica of of the brain and we therefore succeed in at least making the copy What kind of computing power would we need to actually run the emulation if we're talking about a human brain? Has I'm I'm sure that someone has run the numbers on that. Yeah. So if in the end, I expect that you could run the brain on a computer that's about the size of a head at about twenty watts. Why am I saying this? Because that's what we can do now. So we know there is a an implementation, a hardware that can do that. ⁓ so that's what you eventually as the end result can expect. ⁓ but yeah, in the meantime, ⁓ no, these are these are all ⁓ this is all at supercomputer size because because everything that we're doing right now is expressed in software, not in hardware, not in hardware that's tuned to what you need the most and how to run it at the lowest power or something like that. It's It's all software because the most important thing at the moment, when you're still understanding and researching, is for something to be as flexible and editable as possible. You want to be able to change what you've got. And if you build it in hardware, you cannot change it. So we do everything in software because we need to be able to modify because all of our hypotheses are always wrong. That's normal, right? If you have hypothesis, you build a model, it's wrong. You're going to have to change it. ⁓ so that's what it's like right now. ⁓ But yeah, there's a there's a future where that would be different. But there's one thing I wanted to add. ⁓ when people talk about the computational requirements, typically they ask, what is the requirement to run this thing? And and that feels like that's the big question. But the bigger computational question is, what does you take? How much computation does it take to build the thing? Because the amount of compute that you need. To run multiple different attempted models where you say maybe it's this architecture, that architecture, that architecture, that architecture, and then estimate the parameters and keep on doing estimates to try to find the right parameters, and then do validation. The work to get from the pile of data that you collected in a brain to a model that works and has been validated to work correctly is way, way, way larger than running the in final model for years. That's where the the real compute comes in. is this whole optimization process. All right. All right. So ⁓ yeah, no, I I I can imagine that that must be even even more taxing. It's funny that you mentioned the 20 watts ⁓ there because I I recently did an episode with ⁓ Evelina Kurtis and she's ⁓ scientists ⁓ working on a ⁓ wetware computer ⁓ a brain ⁓ computer if you will so there ⁓ in Switzerland it's a company called Final Spark. I don't know if you've heard about it, but they're essentially trying I know of several doing similar work. Yeah. So you've you've seen the the videos where where ⁓ a bunch of cells ⁓ place doom or or or for example. Yeah. Yeah, yeah. ⁓ so she was working on that and and they ⁓ basically their goal is to make a biocomputer ⁓ because they're much more efficient, right? They they the our brains only use twenty watts of power, so if we were to Be able to create a computer that only uses 20 watts of power, as powerful as the brain, that would be pretty huge. ⁓ so I I I knew I knew that number, so that was funny. ⁓ but if we if we talk about ⁓ let's assume for a second we now live in a world where ⁓ whole brain emulation is is possible. And you've also mentioned before that for some use cases. the goal would be to simulate consciousness or create consciousness consciousness even if you wanna try some yeah test a few things. What is what does that mean? What does that entail ⁓ like ⁓ ethically? H what is ⁓ what is yeah, what is your stance on this? Wha how what how should that make us feel? So you're Are you asking specifically about the ethics of it right now? Or are you asking about how we know that we can do it or or or what is consciousness? Or so which part of this question are you really trying to get to at the moment? That's a good it's it's ⁓ it's a difficult question. So maybe just start at the beginning. how would you describe w what is consciousness? What is your definition of consciousness? So I ⁓ of course in the beginning I knew nothing, right? And then I started looking into it more and more because we really do care about being able to identify consciousness in other animals, in babies, in AI, in wherever. And we we want to be able to understand what brings it about, ⁓ what would make it personal so that you have a subjective experience, ⁓ how this relates to identity and all those things. ⁓ so we spent a long time now, a number of years, kind of looking at the various literature that exists in the field of consciousness and working our way through that, but also working to try to connect the the reductionist approach where you take the brain and you're trying to come up with the various cognitive processes and then determine what is generating aspects of what is necessary for consciousness. with the bottom-up phenomenological structure of consciousness. So why does it feel a certain way? Why do we experience it a certain way? And some of the best work that we found and then built on was the work by Thomas Metzinger. he has a couple of books in which he describes what the requirements for consciousness are and he goes through it from so many different scientific and philosophical perspectives that he's really covering a lot of ground. And there are some very important bits and pieces. So for example one thing that's extremely important is that you have to have conceptual representations because what we experience in consciousness is not all the details of our memory or of any moment or sensation. It's it's a rough sort of conceptual thing where we know, ⁓ this is happening or I'm feeling this now, or I want to do that now, and then here's an event happening. It's a conceptual representation of the world and of yourself. Now, not all systems, not all animals necessarily have those. You might not need to have conceptual representations like that just to be able to live and survive. So it's possible that there are living systems that have only minimal requirements. They have only met certain requirements for degrees of consciousness, and not all of the requirements that we have. But what you need to to be a human with a human type of consciousness is you need to have a model of your world, a model of yourself in it. So, in in essence, you're creating a virtual reality. You're using a very ⁓ limited amount of data. You're you're compressing the data, you're taking all the data and you're abstracting it away to its very basics. You have attention, so you're only focusing on one part of that at a time. You have to shift attention all the time. We can't attend to everything at once. And so you have relationships between these concepts that are in your In your world model, and you have relationships between concepts that define who you are. One of the hypotheses about how you learn who you are is that it's the part of your world representation that's always with you. So even as you move around, as time changes, as things change, some of the things you can always, those concepts are always there. And those then are, for example, things that are related to your body, your thoughts, your memories. So you start to associate those with yourself. And these relationships, therefore, are all Based on ⁓ in the end, they become based on relationships from these things identified as self to other. And that explains that's a kind of coordinate transformation. It changes what is otherwise a universal coordinate trans ⁓ coordinate system of some kind into one that is based where the center is on yourself. It's on on that self model, on who you are. So obviously that's going to be subjective. So this is this explains why you would have a subjective awareness as part of this modeling process. But ultimately, it's like a virtual reality. You're generating a virtual reality of the world. You don't have awareness of everything. You're getting a very limited amount of data in that virtual reality. This isn't certainly not, you know, consciousness, if you dive in and just focus on your consciousness, it's not going to be a deeper understanding of the universe because how would you have access to that data? Where did that come from? It's it's a filtered amount of data. It's always less than the totality. It's a very abstract virtual representation. And then you have this virtual representation of yourself in it. It's like you're you have a flight simulator and you're also simulating the pilot. But you're not always simulating the pilot. This pilot flashes on and off. ⁓ the ⁓ researcher ⁓ sorry Susan Blackmore. She did some really great experiments where she even takes audiences through the process of exploring how they don't really have a stream of consciousness, that the reason why it seems like you have a stream of consciousness, even though you keep turning on and off, you're you're often not there. You're often on automatic. Is that every time you ask the question, wait, am I aware? Then the answer is yes, because the asking of the question turns on that virtual pilot, turns on yourself. So it's like when you open the refrigerator and the light is always on. So you just assume the light's always on in the refrigerator, but there's no real evidence that the light's always on. It actually turns off. So consciousness is not a stream. The self is a virtual representation. And all of this is being generated by the brain. It's also a very rudimentary thing. It's like if you have an ocean of processing going on inside the brain, most of that ocean is invisible to consciousness. It's as Metzinger says, transparent. So that's all transparent, and there's this little puddle on top that is consciousness. The information that gets in there is this very shallow little bit. And and is often more similar between people, you know, what you're like when you're happy, what I'm like when I'm happy, just recognizing happiness versus say sadness, much more similar than all of that big pile of data that contains your very specific memories and characteristics, drives, goals, all of this stuff hidden in your subconscious pile of information that's in the brain. The consciousness is kind of like this little abstract virtual reality that's being generated on the moment. So a very important insight from the Metzinger approach, but not just Metzinger, many others as well, is that you shouldn't think of consciousness or self-awareness as a thing. It's not a thing you can isolate and and take out and transfer and and whatever. It's a process. This is a process of generating that virtual reality. And not only is it a process, It's always a little late. As Libbet already showed in other experiments after it, what ends up in our conscious report of what's going on is almost half a second later than things actually happen, including your own decisions. So your decisions are not being made in your conscious mind. You're not consciously making your decisions. They've already been made, and then you're reporting them to yourself. This conscious thing is like a little story and narrative that's super useful. Because not only can you use it to create virtual worlds about the now, you can also make virtual worlds about the future and use them to plan. You can make virtual worlds remembering the past, remembering what the world was like at a certain time. You can dream all these sorts of things. So you can keep reusing that same circuitry that's makes these virtual worlds and not necessarily use them for a report of what has just happened, but also for other things. Does this give you enough background? I really love the fact that you know and have s thought so much about consciousness. One of the really interesting thought experiments that I would love to get your take on ⁓ about consciousness is the following one. and it it touches on the the process exp aspect of consciousness. and that it also has to do something with our our memories because consciousness is always happening now. It's ⁓ moments of or half a second after now. Or half a Generating what we call a phenomenological now. That's part of the process. That's actually one of the requirements that Thomas Metzinger also wrote out. To have human type consciousness, you need to be able to generate a phenomenological now. Yeah. Yeah. So okay, give or take half a second, but at least very ⁓ it's it's happening in the now. However, let's assume you were jumping out of a plane and your memory would be non existent. You would have zero memory whatsoever of anything that happened even zero point zero zero one second before that moment. If you f are falling out of a plane, would you even understand? that you're falling. Probably not, right? Because in order for you to understand that that you're falling outside of discipline, you would have to know that you were ten meters higher up and that you are now lower than you were five seconds ago. Yeah. No, literally literally if you if you lost all your memory ⁓ while that's going on, ⁓ that would include memories of understanding how the world works or anything else, right? It would ⁓ it would eliminate All understanding. So it's like you're a baby that was born with a lot of wind coming into your face and just feeling that. And you just think this is normal. That's that's like that's your experience. You're learning, ⁓ there's a l there's a lot of pressure on my face and on my body the whole time. And vision is like this. Okay, interesting. Yeah. ⁓ fluffy, cloudy things. But would that constitute ⁓ a consciousness? Or or would that be something different? Well, if the process Is still generating a virtual world for you, a representation of what's around you and yourself in it. Now, I don't know if in if you've lost all your memory and you're jumping out of a plane, if there's enough time there for you to learn enough about the ⁓ what is being generated by that process as your virtual world representation to start to recognize who yourself is in that world. Is there enough time to create those relationships by seeing yourself in different situations? I mean, there's no different situation. You're falling the whole time. So there may not be enough there to actually establish a self if you're just doing that. No. Yeah. I'm I I would agree. I think if if you keep erasing the memory and and everything you know is just a slice of now, ⁓ I think it's arguable that that wouldn't constitute as ⁓ yeah, it wouldn't be be considered. Consciousness. But people people have different opinions ⁓ on this. but now that we've established more or less what we what we consider consciousness, what what would be ⁓ let's say we were able to do a whole brain emulation ⁓ that resulted in a a conscious mind. would that be ⁓ problematic? Would that be okay? Is it okay if I take a minute? ⁓ and attend to something for just a minute and come back? Yeah, of course, of course. Okay, I'm back. Yes. So what was your question? My question ⁓ was basically let's let's assume that we ⁓ have the technology to do a whole brain emulation and ⁓ this whole bl brain emulation is now all of a sudden conscious, a conscious entity has appeared, if you will. It will it will be an exact copy of of ⁓ the person ⁓ we we are doing the emulation of. what are the implications of that? Does it does the copy get human rights as well? Or or w what is ⁓ I I'm sure the field must have had some ⁓ thoughts and opinions about this. I mean I think we have some intuitions about this. ⁓ if I ask you, what if someone took your brain, let's say you died and you donated your brain to science and then they went and Scanned it in and built an emulation and then turned the emulation on and the emulation says, ⁓ hey, okay. I'm Julius and I feel fine. This is great. I get to continue living. How would you feel about then them saying, well, we're still doing some experiments and we're going to expose you to this raft of pain stimuli. Okay with that? So I think we have some intuitions about how we should deal with that, right? It's We don't know for sure, of course, ⁓ what happens when you either make an artificial intelligence that you suspect may have some form of consciousness or some kind of ⁓ awareness. And the same is true for brain emulation, where we still need to come up with the right ways to validate whether or not that has succeeded and if there is full consciousness there and all all these things. but but we should always err on the side of caution. And this is even true when we're working with lab animals, because we have a whole raft of procedures for how you have to deal with, say, a mouse versus say with a fruit fly. It's a bit different for the two. ⁓ And I think that you have to err on the side of caution and proceed ⁓ as if, say, an emulated mouse can experience what a mouse can experience. Now, if you're running an emulation of a mouse for a hundred milliseconds. That's like a mouse being hardly having the time to get become conscious and be aware of something. So you can imagine that maybe the emulation of that and then testing out, ⁓ did the neurons respond the way I wanted to? That is probably fine. But if you run that emulation for an hour, then it's kind of like could be, you know, let's just assume, because we want to be cautious about this, that it's like working with a real mouse for an hour. And then you should apply the same types of Ethical concerns. But the reality is, of course, that we're not in a place where there are laws that would force someone to apply these ethical concerns. There are currently no laws. About ⁓ you know, if you've donated your brain to science, they can a person can literally do anything with it. ⁓ there's no law there. And and the same for working with virtual animals, there are no laws there. So the laws haven't caught up. with the requirements for what I would call cautious ethics. Right. Have you ⁓ are you are you a Black Mirror fan? I've watched a few episodes. I wouldn't call myself a fan because I I had to, you know, stop watching the series for time constraints. Okay, okay. Well I'm I'm I would consider myself a fan and there's two episodes that come to mind. ⁓ one is a very happy episode ⁓ and one is a darker one like they usually are. The happy episode is ⁓ called San Junipero and I know that one, yeah. Yeah. And it's it's yeah, it's a it's a happy view of the future where people are able to have the option, it's not force, but people can copy their their consciousness, copy their their mind into the simulation where they can live in paradise, ⁓ potentially ⁓ forever. which Well could be could be ⁓ a future we're we're going to if we if we figure this out. A darker one is ⁓ White Christmas. And that episode actually uses ⁓ whole brain emulation, makes copies of ⁓ suspects, for example, and then they reenact ⁓ a crime scene in order to ⁓ try to make them spill some of the facts about the the murder they may have committed. ⁓ and an even more dystopian one is that people can as a service ⁓ upload their own minds to a computer. And because this is an exact copy of themselves, ⁓ they know all the preferences of their of the original, right? So they know how they like their steak cooked and they know ⁓ what time to open the curtains in the morning and when to turn on the coffee machine. So they have this copied version live in this software and And the software is kind of like a Google Home application where they basically manage the entire household for ⁓ the original version, basically. So ⁓ yeah. That's ⁓ it's it's I don't know those two darker episodes that you were talking about. I s it so happens I know the the very good one. Yeah. Yeah. No, it's ⁓ it's it's ⁓ but do you think like ⁓ it it's it's obviously it's it's premature, but is that something ⁓ we should take into consideration as this field matures, ⁓ that we ⁓ have discussion about possible legislation about how to treat ⁓ digital. Absolutely. Yeah, there's no doubt. ⁓ this is a field where the technology, the possibilities coming out of the science are going to have medical applications, whether that's for creating small neuroprostheses to repair parts of brains. And then of course, thinking about, ⁓ we could augment these to back up data or whatever, ⁓ or doing entire brain emulations for mind uploading purposes. This is a field that if we assume it matures, which it ought to unless something terrible happens to the world, ⁓ then then we have to treat the we have to think ahead and we need to treat this very carefully because why wouldn't we worry about and treat ⁓ you know someone who happens to have an artificial brain? Just like someone who has an artificial heart or leg or something like that, why would we not want to treat them the same way as other persons? And the same probably applies to other kinds of persons that may come along. I think that in the future we're going to be dealing with an a a an environment in which we have lots of different kinds of intelligences, some of which may be conscious. And and that could be whole brain emulation of people, regular biological people. People with some augmentations that are maybe half biological, half not, artificial intelligence that has aspects of consciousness. ⁓ there could be so many varieties, and we need to start treating the whole thing as ⁓ how do we deal with persons, rather than just, ⁓ how do we deal with biological human beings who have never had any implant or something like that? It's just a it seems to me a completely natural thing that we have to broaden our horizons of what we might consider the in group that we need to care about. Yeah, hopefully we'll we'll be ⁓ on time because usually ⁓ legislation lags behind ⁓ reality. We see this ⁓ pretty much rights. Yeah. Yeah, yeah. Some disaster has to happen in order for us to ⁓ to change ⁓ the law. ⁓ I believe Chile is one of the only countries that actually protects our neuro of or has amended their constitution so that ⁓ neuro rights are protected. So ⁓ they basically but that's more against brain computer interfaces so that we ⁓ if if maybe a hundred years from now when everyone has a BCI interface, ⁓ a brain computer interface, ⁓ that ⁓ Jeff Bezos can't run ⁓ Amazon ads right into your brain. But ⁓ which is probably a good idea. so yeah, some some countries are are a bit more ⁓ I don't know, they they're they're looking ahead rather than ⁓ trying to ⁓ incredibly forward looking of Chile. I I I wouldn't have expected that. It's amazing. No, it's it's it's incredible. ⁓ is there ⁓ anything ⁓ be before we wrap it up today that you think is still important to to touch on? ⁓ you briefly mentioned the the competition, for example. ⁓ yeah, I'd like to touch upon that, sure. ⁓ yeah, so we are We are running this competition, a platform, where we instead of having to base ourselves on the ground truth of a biological brain, because nobody really has that yet, not fully. There's still things in neurobiology we don't understand. We're generating synthetic brains, little synthetic brains. You might say, ⁓ here's a new butterfly we just discovered in the Amazon. It's not like any other particular animal that's out there. We generated it. ⁓ so we grow these neurons. in a computer and set up architecture and then synapses and then we hide behavioral functions inside of these models. And we do that so that we have a system where we understand all of the structure and all all of the function that's there, even if we're leaving out some details in biology that we don't know about. This means that we can test the approaches that researchers have for how you would go from the data you collect in a brain, how you then Create an emulation. What we do is we apply virtual instruments to these synthetic brains, like electron microscopes, calcium imaging, electrophysiology, putting probes inside. We collect a pile of data and we give that data to a researcher, just like they would have collected data from a biological sample. So for them it's a very similar challenge. ⁓ here's this data. How do we create the emulation of what's going on? But this time, because it's all synthetic, we already know what's there. So we can compare their result directly with what we have and give them a report of what's different and what's similar and how close they got, basically creating scores and leaderboards and a report with all kinds of details. And the idea is for that to be a virtual lab where you can make super simple examples, and then you can make them more complicated by adding more neurons and more layers, or more detail in the neurons, more abstract versus more detailed representations, and more artifacts in the data that's been recorded or not. So you can add a single variable at a time, testing out everything from simple to hard to figure out what certain approaches to emulation are good at, what they fail at, how they compare with others, where you should apply one or the other. And this challenge is then meant to become a benchmark where ⁓ everyone can compare what they've achieved on the same playing field and not have a situation where you try something, write a paper, Claim this is an emulation, but really have nothing to back that up. ⁓ And this is a bit modeled on how things were done in AI, where you also have known data sets of images that have been labeled or things like that, and use that to train out an algorithm before you release it in the in the wild, right? And you compare different approaches and algorithms on the same data. Yeah. So that's ⁓ that's what we're trying to contribute right now because we think this is central to the main challenge that exists in whole breehen emulation at the moment, this problem of going from the data to the system. Super interesting. And is there a a deadline for ⁓ is the is there a deadline for for for companies and researchers to to apply and and is there some sort of price attached to it right now? We're we're just working with a few initial ⁓ researchers who've agreed to test out our system. So those who already have tried some their hand at emulation and said, okay, the same method that I published in my paper, I'd be happy to try it out on the synthetic systems you've built. So we're still testing. We're we're thinking that we'll actually launch the challenge in about six months. And at that point it's going to become an annual thing where we go through a challen through a a set of challenges and maybe if we can get enough funding, prizes, ⁓ at different levels. different degrees of difficulty every year, and then do a summit where we kind of present the winners. And then ⁓ over time I assume that the challenges will become more and more difficult because, you know, people will have passed a certain bar. And so over the years, we get to at some point really complex systems, systems that are more complex maybe than what we find in a fruit fly, sort of mouse like, I would hope. Super interesting. Well I'm I'm very happy to ⁓ to stay ⁓ in the loop of ⁓ of anything ⁓ that comes out of this. I think finally I wanted to know, is there anything that people get wrong when they when we're talking about whole brain emulation? And is there one thing that you wish everyone would understand about this s field of science? ⁓ I think there are two things I would like to make sure people understand. ⁓ Before I say these, I should mention that we're also about to release a whole brain emulation 101 course. So it's a class that you can take online for free. And it'll teach you about the basics of whole brain emulation and understanding the terminology and what's going on. So one of the things I wanted to mention that people often get wrong is they use the terms whole brain emulation and mind uploading interchangeably, like they're the same thing. And and they're not really, because they describe different objectives. Mind uploading describe describes a desired outcome in terms of patient survival. A patient being able to survive with an artificial brain. And that then involves things like, are they still the same person? Is it is it their identity? And and things like that. That's part of the mind uploading objective. Mind uploading doesn't say how you get there. You could have all kinds of different approaches or methods. Whole brain emulation is one technique. one s or set of techniques, because of course lots of scientists try different ways to do emulation, of being able to create a working brain from data about a brain. And that could be used as a way to achieve mind uploading. And as we mentioned, it could also be used for other purposes. So there's a distinction there in the sense that one is a a a very clearly described set of techniques with science and engineering going into it. ⁓ some things already being feasible. And the other is uploading is a is a direction and a desired outcome for people, for humanity, for something like that. So that that's distinction there. And another thing that I find that is important for people to to really dig into, and there's an aspect here that's philosophical, you know, I mean, ⁓ whether you're a functionalist or not, but it's important to think of ⁓ first of all, of the self of yourself, not just being your conscious self. It's all of that subconscious stuff too. Your conscious would be almost nothing without all of the subconscious information. So that all is one part. They're not really distinguishable. You can't separate them. You cannot transfer consciousness. You can't just transfer consciousness. That means nothing. Because consciousness, the your conscious representation, that virtual world you have there, has no form or function without all of the subconscious stuff going into it. So they're not separable in that way. And and the whole idea of therefore transferring it is kind of weird because it feels like it's a virus that has to infect the new equipment or something like that. And that's not how it is. It's a process that emergently generates. So yeah, those are some basic, some often misunderstood things. Yeah. And ⁓ finally, to just respond to the last ⁓ thing you said, I I recently learned about How much of an influence the gut has, for example, on how we feel. ⁓ so of course when we eat something, but also when our stomach is empty and all sorts of things that are happening with our moods, ⁓ very subconsciously ⁓ steered by the gut. So I think that's just ⁓ a a small aspect of it. But ⁓ yeah, that I think a great real world example of ⁓ of of yeah, that we're not just this Yeah, c consciousness is is is is also v heavily steered by things we we can't control or aren't even aware of. but yeah, Randall, this was ⁓ incredibly interesting. Thank you so much for this ⁓ discussion today. ⁓ where can people go if they wanna keep up with ⁓ all the work you're doing? Yeah, definitely go to Carboncopies.org. That's our main website and that's where you'll find newsletters and all the information about journal clubs we're doing or events that are going on or anything really. Or if you want to join us, become part of the team, ⁓ become an intern, for instance. We have often got internships available, ⁓ or just a regular team member as well, that's also available. ⁓ if you want to donate to us, you'll find the site for donating to Carbon Copies there. It's very important for people to donate because we are a five ⁓ one C three nonprofit organization and completely dependent on grants and donations. ⁓ if you want to reach me personally, the best way is to go through the people who watch my email better than I do by sending an email to contact at carboncopies.org. Those are all the best ways, I think. Perfect. I will make sure to ⁓ put ⁓ everything ⁓ in the show notes so ⁓ people can ⁓ can easily ⁓ find it. But ⁓ yeah, thanks again ⁓ for ⁓ for today, Randall. I ⁓ I had a blast. I learned a lot and ⁓ yeah, maybe ⁓ if there are some ⁓ developments in the field we can ⁓ we can run it back someday. Wonderful. Yeah, thank you so much. This was an excellent conversation.