 All right. Hello, great. Thanks for joining me today. How are you doing? I'm good. How are you? I am great and we're here to talk about your awesome book that I stumbled across recently. So for those who haven't met you yet, can you kind of explain, I guess the first question I'll ask, you're a computational neuroscientist, right? Can you kind of explain what that is? Yeah, it's a little bit about what the book is supposed to do because yeah, most people have not heard the term, they don't know what it is. So I think people are familiar with the idea that there are some scientists that use math to understand the world and usually people think of those people as physicists, like physicists write down equations and it corresponds to real things in the world, but you can apply that to almost any area of science, really any area of science. And so a computational neuroscientist is just someone who uses math to understand the brain, to either model different parts of the brain or do data analysis on data that comes from the brain. Just generally, it's about using math to understand the brain. Yeah, so and yeah, so I'm not a math guy, but surprisingly, like your book, like it actually it made a decent amount of sense to me. Like I felt like I was able to keep up a little bit, but that's something else I'm curious about. Like whenever I have authors on here, like when you wrote this book talking about the models of the mind and all that, who was the target audience you had in mind? Was it like other academics, people, you know, like college students, like getting into neuroscience and all that? Like, who are you thinking of as you kind of wrote this book? Yeah, so I wanted it to, you know, as you say, kind of be readable by people without a math background and also without a neuroscience background, you know, it doesn't assume that much knowledge. And I wanted to just let people who are interested in the brain, which it seems like there's a lot of people who are, you know, books about the brain are popular, people want to understand the brain works, they want to understand psychology, they want to understand kind of all levels of it. I wanted to let those people in on this way of understanding the brain because it's becoming increasingly more important in neuroscience amongst the scientists who are actually doing the work of understanding the brain. They're using these mathematical methods, but most popular science books about the brain don't really get into them or they just like kind of gloss over very quickly. They say like a computational model was built and it supports this idea, but you don't want people to know how that works or why or anything like that. So it was really targeted yet to people who have an interest in the brain, showing them this other way of understanding that can be really fruitful and interesting, even if you don't know all the details of the math. And then on the flip side, there are people who know kind of math or come from physics and engineering who have an interest in the brain. And this kind of maybe puts it more in their language. So I think it can kind of serve a dual purpose because it is this, you know, cross-disciplinary field. If you come from the math or from an interest in the brain, you can see where they merge here. Yeah. Yeah. No, absolutely. And I think you nailed that. And you do, and you do a great job by kind of explaining some of the history and like the main figures who are like figuring out some of these models and everything, because my background, I got into neuroscience and we'll talk about this in a little bit. Like, I have an addiction background. I got sober in 2012 and I was like, okay, what's going on with me? Why am I different and stuff? And I started learning about, you know, mainly about like neurotransmitters and like dopamine systems. And I got interested in, you know, how various parts of, you know, the brain are involved in like depression and anxiety and, you know, all these other things. So that's kind of where I came in. So, so yeah, your book, it kind of brought in this other aspect. I was like, oh, okay, some stuff started making sense. But one, one thing I wanted to have you help educate me on and maybe some of the others who are more from my side is, for lack of better words, I'm like a skeptic because I come across all this other stuff, but I know there's science and I know like there's ways to research this. So the book, you know, you talk a lot about neurons in there. And as I was talking about, you know, like I'm more familiar with like neurotransmitters. So real quick, can you break down what is the main difference between neurons and neurotransmitters? What are their different responsibilities? Yeah, so neurons are kind of the main cell type in the brain. They don't actually, they're not the majority cell type. There are other cells called glia, but they're the things that as you could guess from the name, neuroscientists focus on we think that neurons are the cells that are doing a lot of the important computational work of the brain and like the information processing and all of that kind of thing. So they're just the cells that are in the brain. And the main thing that they do is communicate with each other. So they send within the neuron, there's an electrical signal that's sent within the membrane of the neuron, it uses electricity to send it from one end of the cell to the other because neurons can actually be quite long. There's like neurons running from the bottom of your spine down to your foot. So they can be quite long. So they use electricity to send messages inside a single neuron. But usually when you want to send a message from one neuron to another, you use a chemical signal, which is a neurotransmitter. So there's like at the end of one neuron, you release this chemical, it crosses this little gap called a synapse, it touches the next neuron and that starts an electrical vent in the next neuron. And so neurons go from electricity to chemical to electricity, but that chemical that they're using is a neurotransmitter. So neurons release neurotransmitters. And sometimes they release them just from one neuron to the next in this little gap, as I said, but also they can release them more broadly. And usually when they release them more broadly, they're a type of neurotransmitter called a neuromodulator. And it's not just about directly trying to send a signal from one cell to another, it's about kind of broadcasting to a broader community of neurons. Oh, this is kind of the state of the brain right now. This is the state that we're in. I'm going to send this broad signal to everybody just so we're all on the same page. So they're completely linked. You can't really have neurons without having neurotransmitters. You definitely can't have neurotransmitters without neurons because they're usually the cells that are producing them. So you need the neurotransmitters to allow the neurons to communicate. Got it. So if I'm understanding you correctly, so when the neurons are sending these neurotransmitters to the neurotransmitters, there's a variety of them like dopamine, endorphins, and all that kind of stuff. Is that what I'm getting like? Yeah, there are different types of neurotransmitters. There are ones that cause the second neuron to have this electrical event and there are other ones that try to shut it down. So it's kind of sending a positive signal or a negative signal. And then yeah, these more broader neuromodulators are sometimes like dopamine can be considered a neuromodulator. Other ones that people hear about a lot might be more the neuromodulator category than in the pure neurotransmitter. Gotcha. And totally random question. Just the other day, I introduced my 12 year old son to the movie The Matrix and you're talking about electricity. Is that like kind of the idea of what they were saying? Like the machines are sucking that kind of electricity from us. Is that what they're saying? It's been a while since I've seen The Matrix. I assume that's the best they could have gotten is that that was the electricity they needed. It is not, I mean, the brain is very energy efficient, but I don't think it was the best way to get energy. Yeah, that's why I'm here. Like I read books about, you know, just like supernatural beliefs and like skeptics and debunking it and like they talk about like our bodies being like magnetic and things like that. And like, from what I understand, like the electrical signals going through are like nowhere near strong enough to really do much. Is that correct? Yeah, I mean, there, yeah, there is certainly the electrical signal along the single neuron is weak compared to, you know, electrical signals we use in the world. Yeah, powering our house and all that. Yeah. So, so yeah, so, so now that I have a better understanding about that. So where my skepticism comes in, like coming from a mental health background, like, you know, I, I, you know, I try to learn a lot about, you know, like how antidepressants work and different medications. And a lot of them talk about, you know, manipulating, you know, like serotonin and all these other things. But there's a whole camp of people who are like, no, we don't know this stuff. Like, you know, we can't really, you know, change how much how, how the neurotransmitters are reacting or one of the big ones is like, we don't have really have a way of measuring it. So I was like, Grace will probably have some answers for me. So like, as far as neurotransmitters, I want to talk about neurons in a second too, because it's a lot of neuron talk, but neurotransmitters, can you, can you help me understand, like, how is it measured? How do I know, like an SSRI is doing what they say it's doing, and it's not just mainly a placebo effect? Yeah, so those are actually two very different questions, kind of how we can measure neurotransmitters and know about them versus SSRIs in humans as antidepressants, because you know, a lot of the basics of what we know about neurotransmitters and how neurons communicate are done in animal work. So in a lab that studies animals, and, you know, we kind of assume based on good reason, and you know, this has been verified, that a lot of the principles we learn in animals can be applied to humans as well, especially on this kind of cellular level of just the basics of how a neuron works. A lot of those things do translate very easily. So a lot of our understanding of just the basics of neurotransmitters and neurons comes from animal work. And so there, you know, you can kind of directly access the brain, there are markers you can use to kind of see neurotransmitters and that kind of thing and stains and all of these things that are like, you know, Petri dish biology kind of stuff. When it comes to humans, you know, we can't get that kind of access in human brain, because it requires, you know, removing the skull and then doing very invasive work that we are not going to let people sign up to be a part of. So there just the experimental access is much less. And so some of the, you know, findings are verified in animals. And then there are indirect ways of measuring things in humans. So specifically about SSRIs, that's not my area. So I don't want to speak to, you know, how people do that clinically, but there are methods for kind of indirectly measuring neurotransmitter levels in humans. And obviously there's, you know, the actual output that you care about, which is a change in depression or whatever disease you're interested in. And so that, you know, you don't, you don't technically have to measure the neurotransmitters. You can just say, did this drug make this person feel better? And maybe the exact route of which, you know, by which that works, we're not fully understanding, but people need help. And so we want to have something that works for them. And we can like, you know, keep working on trying to understand exactly how it's helping them to make them better and all of that, you know, to make the drugs better and more targeted. But, you know, the direct thing that people care about when making these drugs is just does it help the patient? So yeah, so there's, you have to, it's very complicated because yeah, some people say like, we don't know anything about how the brain works just like generally, you know, but obviously there's a lot of stuff. I mean, I wrote a whole book about how the brain works. And I think it has information in it. So we do know some things, kind of just when people say we don't know, or we can't be sure, that means different things to different people. And it's like, yes, in some extreme sense, there's a lot of uncertainty in science. That's what science is. You're like at the edges of knowledge, we can still be confused, we can think we know how something works. And then in 10 years, we'll find out that that's not how it was working all along. And so yes, there is definitely uncertainty and measurements are complicated and all of that. But there are ways by combining different sources of information, doing different types of experiments to kind of triangulate what you think is happening. Yeah, that's actually really interesting. So recently I had the professor, she's a evolutionary biologist, her name's Nicola Rahini. She recently wrote a book, The Social Encyclopedia. Anyways, she helped explain that to me like studying animals and how that kind of relates to humans. So now it's kind of clicking with what you're saying. So basically, if I'm understanding you correctly, this test, the more invasive tests that we could do on animals and what we know about animal brains compared to ours or how the cells kind of react, we can kind of infer like we kind of guess like when we're doing these studies or testing the medications like okay, the work over here where we do have that knowledge, it's most there's a good chance it's working over here. Is that is that what I'm gathering? Yeah, that's definitely what you know how a lot of this research proceeds. And there are, you know, there are ways of getting more invasive information from humans. I mean, again, I'm not going to speak specifically about SSR is a depression. But in generally, generally, there are ways so when people have epilepsy, sometimes they have to have a surgery where their skull is open anyways. And so we can do some kind of basic science at the same time, you like ask the people will you be willing to participate in an experiment that's not actually about your epilepsy, it's just because we don't get this opportunity to understand human brains as well, will you let us, you know, record from your neurons and that kind of thing. Or, you know, when people die, you can section the brain and study it, and then you're getting a more cellular level understanding of the brain, even though it's no longer living, that can still provide some information as well. So yeah, there are all just many different methods for, you know, studying something directly in animals, and then verifying that what we're learning in the animals can apply to humans. And, you know, sometimes it doesn't apply. But that's why you got to do this back and forth. But yeah, people are definitely, you know, trying to make those connections and using the best information they can while still, you know, keeping our, you know, experiments with humans ethical and safe and all of that. Yeah, yeah. Okay, cool. That that that definitely helps me understand. And then and then I'm guessing like, I'm guessing, but you can help me break this down. So with neurons, like there's a ton of neuron talk in this book and like seeing how they they do so many things and like, you know, a lot of these computational models that you talk about, is this something similar or because something I was wondering as reading your book, like you mentioned, like, you know, kind of petri dish studies, is any of this like done, like, like you mentioned, like, you know, people who donate their brain to science or after a person dies, like, like, how does this work? How do we, you know, find these computational models? Is this also mainly tested in animals? Or can we do it with brain samples and stuff like that? Yeah, so a lot of certainly when it comes to how individual neurons work, a lot of that is done in animals, because even if you have a post mortem brain, the neurons are no longer sending signals. So you can't study that process in them, you can kind of study their molecular makeup and their anatomy, but you can't study their actual signaling, which is what a lot of people who build these mathematical models are what they're kind of ultimately interested in is like, how are the neurons sending signals and how does that lead to intelligent behavior, everything else that we do, you know. So a lot of the work does come from animals, but there also is a lot of computational models that are influenced by fMRI imaging. So that's when you put someone in a scanner, and you can get a broad sense of which parts of their brain are active. Now this is very, this is like, you know, you're looking at basically like a million neurons in one little pixel of this image. So you're not getting fine resolution, but you're getting a broad sense of kind of what different brain areas might be doing when a person is doing some sort of task, you know, they're laying in a big fMRI scanner, and they're seeing a screen, and they're doing some sort of task that the experimenter asked them to do. And so that is a way of using, you know, human live humans to get at an understanding of how the brain is working. And then some things, we can actually get a lot of information just by collecting behavior, just by, you know, observing the choices that people make and that kind of thing, that actually can constrain a model of how the brain works, because it's like, okay, if I give it this input, it produces this output. There's only so many steps that can lead from that input to that output. So that can be helpful. Certainly, as a starting point, you know, you want to get a good characterization of the behavior. And then you can start thinking about what the brain is actually doing to produce that behavior. Got it. Okay, cool. And since we're kind of talking like, you know, technology and everything, one of the most interesting parts of the book, I'm like, okay, it is kind of nuts. You talk about, you know, there was, I believe it was a woman, she was paralyzed from the neck down and using like some computational neuroscience, I guess they made like this chip and it helped her move like move her arm by envisioning and training this chip. Can you explain it far better than I am right now, Grace? But I think, you know, part of the interest is a lot of people are curious about where the technology is going and how this can be used for, you know, medicine and helping people and all that. So, so yeah, I found that one really, really, super interesting. Yeah, so this is a woman, yeah, who was paralyzed from the neck down and she was part of this study called Brain Gate, which is at Brown University, which does this research on, it's usually called brain computer interface or brain machine interface. And what she was able to do was she was able to learn to control a robotic arm. So she couldn't control her own arm because those nerve connections weren't there. But what the scientists were doing were was putting an electrode in the motor cortex part of her brain. So the part of the brain that normally is controlling movement of the body, they were able to read out the neurons in that area. So when she thought about moving, those neurons are active. And even though they can't actually move the body, they're kind of still sending signals, which is fortunate because you can imagine if you're paralyzed for a long time, that part of the brain might just become used for something else or not be sending motor signals. But there's still enough information in those neurons that normally control the body to have her learn to control a robotic arm. So it doesn't happen immediately. It's not like you just record from those neurons and all of a sudden you can control an arm. It is a process of learning both for the human and also the algorithm that the scientists are using to translate the neural activity into robotic arm movement. So it's not like you just kind of plug in and go. And even once you have the movement, it's not perfect or smooth or necessarily exactly as you want it to be. But for someone who's paralyzed, it's a big deal to be able to do that at all, of course. So yeah, she was able to control this robotic arm to pick things up, bring it to her mouth, that kind of thing, grasp, let go just by doing what we do to control our bodies, which is user neural activity. Yeah, so that makes me curious, because something they've been trying to upgrade like prosthetics and stuff. Is this research kind of leaning in that direction or they're hoping like someone who's lost an arm or leg that can kind of help in that sense and be used for that? Is that what a further goal is down the line? Yeah, so a lot of times if people have lost just an arm, say from like the elbow down, their brain is probably still sending signals down to that elbow as though there was the rest of the body. So rather than having to go into the brain itself to get this neural activity, you can put an electrode in the arm and pick up some activity right there and use that to control. So at least for something as simple as kind of open or close the prosthetic hand, which is obviously a big action that you want to be able to control. So yeah, you don't even have to necessarily go to the brain all the time to do this if you just have kind of a local amputation where the local nerves are still available to record from. If you're still sending signals down that far, you kind of want to go almost to the place closest to where the real muscle was that you can still get signal. But if you're paralyzed from the neck down, that's basically the brain. So it depends on the specifics. But yeah, it's the same general idea for sure. You put an electrode in, you try to record neural activity and you use it via an algorithm. You use it to control prosthetic limb. Got it. Okay. That's super cool. And it just crossed my mind too. So Elon Musk has been working on this neural length thing. And you are an expert in this field. Like, what are your thoughts around that? Because as me, as somebody who's not too familiar and I'm trying to learn more, I see debates like people like you're insane and other people like, oh, this is so cool. What are Grace's thoughts around neural link and what Elon Musk is trying to do? Yeah. So I think there's a difference between the hype and what people are saying neural link is and what it will be and the reality of it. And even in the reality, there's some nuance. So the idea of putting an electrode into a brain and trying to read out the neural activity to control a body or to just do some sort of mind reading or whatever it is. I mean, this is what neuroscientists do to understand the brain. Like we record from neurons and then try to make sense of what the neurons are saying. And so that basic concept is not at all revolutionary. This is, you know, this is what we do to try to understand the brain's computations. So that basic thing, you know, doing it in animals is already what we do in the lab and all of that. So the basics of it are not new. It's my understanding that the actual kind of hardware that they've developed, this type of electrode they have and how they implant it are pretty big advances in electrode technology. So people in neuroscience are kind of excited about the prospect of having better electrodes that are implanted easier and can last longer and have higher signal quality and all of that. But, you know, that's it's a very technical advance. They made a cool, good hardware thing that hopefully will spread through neuroscience and people can use it. The idea that humans are going to have this implanted and they're going to, you know, interact with the world through their brain. That's very far off one because there is still, you can't get around the fact that you're doing brain surgery and that that's a big deal. I don't think a lot of people are going to sign up for brain surgery just for like a casual computer interface experience. So there's just the biology of it. And also, it requires that we understand how the brain works, really, to be able to just kind of read out thoughts. That means that we know when you look at a neuron, you know, what a neuron does, I said, it has this electrical activity that it makes, it makes these little blips of electrical activity. And so basically when you look at a neuron, you just see kind of like a, you know, just a tick, like a series of ticks, like tick, tick, tick, tick, tick is just like, that's what it's doing. So if you can't look at that and say, Oh, that person's thinking about a banana, then it's not going to be useful. You need to actually be able to understand the neural activity to do something useful with it. So there are, and you know, this example of the woman with robotic arm shows that we can do that to some extent, we can try to understand her intentions and her desire to move by recording the neural activity, but it takes a lot of effort and there's a lot of other parts of the brain that if we record that neural activity, we're not going to actually have a good sense of what the person is thinking or what they're trying to do. And so you can't just magically start controlling or interacting with the world through neural activity. We need that translator that says this is what the neural activity means before it can start being a really useful way to interface with the world. Yeah. So yeah, it's kind of, it seems kind of far off and we still have a lot to learn. And so this segues perfectly into my next question because I asked a lot of people and you spoke about this a little in the book, it's like artificial intelligence, right? So I am really interested in like cognitive psychology, decision making, you know, all these other things, right? I'm actually reading right now, Anil Seth's book on consciousness, right? There's a lot of stuff and like, we still don't know a lot. And, you know, earlier, like, you know, when I started the podcast, I had Kate Darling on from MIT to talk about her book, and I asked her about AI. So there's a lot of like random fears like, Oh my God, AI is going to like take over the world. And like, based on my remedial knowledge of how the brain works and how much we don't know and decision making and how just wacky, you know, even humans are and irrational and just random, like, I'm curious your thoughts on like AI becoming this kind of like, you know, conscious thing or even AI like being like an extremely useful tool because we were talking about, you know, kind of machine learning aspect of, you know, reading the neurons and everything. So, so what are your thoughts like, like for the for the public who was like, Oh my God, terminators going to happen and we're going to be taken down. Like, what are your kind of thoughts around the current state of AI and how far off that is? Yeah, I mean, I think, I think sometimes the fears are misplaced, not that there shouldn't be any concerns about how we use AI in society, but the idea that it's going to become, you know, self aware and all of a sudden have a goal to destroy humans or whatever it is. Like, I just don't see where that would come from in the model. Basically, you have to put everything into the model that you want the model to kind of produce, you know, you tell it the tasks you want it to solve and you give it the data to train on and all of that. And you know, there is some extent to which it's doing something on its own because it is machine learning. It's we can put in data and we can train the network and it kind of figures out a way to solve the tasks that we give it. And that's why it's so powerful because you don't actually have to hand design every solution that you want. You can have a system that can learn on its own. But that doesn't mean that it's going to all of a sudden have desires or something like that or have motivations or goals that just came out of nowhere. So that kind of element of oh, it's going to take over or it's going to be sinister or it's going to try to trick us like I don't think any of that is really a concern. But you know, there are issues when you train these models and you are actually deploying them in the real world. There can be consequences and side effects when you don't fully understand how they're working. And you know, we see this kind of in terms of like things that are determining if people get a loan or not using kind of information that we wouldn't want to be included in the decision. And sometimes that's just because there are things in the data set that correlate with other things like if you put in someone's zip code to see if they should get a housing loan. It might be that that tells you something about their race. And because you know, because it is the case in the world that people live in segregated areas and all this kind of thing. So without fully understanding the entire system that you're trying to use an AI as part of, there can be unintended consequences that are happening in the model. And so that's an issue for people who are actually deploying these models. They need to be sure that the model is kind of doing actually what they want. Because even when you train a network on very simple tasks that don't have much to do with like the complexities of the real world, those models can find weird ways to solve the problem that isn't how you or I would solve the problem. And if you just look at the high level, just like, Oh, it's solving the problem. Great. I can go put it in the world. It's going to solve the problems in the world. But if it encounters, you know, something that is too different from what it was trained on, and it's solving the problem in this weird way to come up with a weird answer. And if you're not checking, that could have bad consequences, basically. So people do need to be considerate of if they're training their models appropriately, if they're testing them appropriately before they actually start deploying them in the real world. That is something to consider. But that's just kind of that in a way that's, you know, a mundane issue. It's not this global super intelligent AI is going to take over the world. It's just no, we as people need to be careful on how we use our tools. It's just it's the same with any kind of tool that you might use, especially at a large scale. You just need to be a little careful. You need to reflect, you need to think about how the tool works, the impact it's having and making sure that everything's actually going how you want it to, and not, you know, glossing over the fact that the model isn't actually doing exactly what you want it to. Yeah, that's that's actually a topic I've been really interested lately with like algorithmic biases, right? Like you mentioned, like, you know, loans and, you know, just even issues of like, like sexism just built into a model. Like there was a that famous story about like Amazon, it wasn't hiring like any women based on this like algorithm they created. So with, you know, all the math involved in like computational neuroscience, is there any crossover with, you know, the work that you do or anybody in your field with kind of these algorithms that are being created? Or is like, for example, like, is, is this something like Google's looking into, you know, and stuff, because even like searches can be kind of skewed and bias and, you know, so many things. And that that's something like, like you mentioned, I don't think it's like a world ending thing, but it can, you know, for as far as like societal issues and, you know, wealth inequality and racial issues and sexism and all sorts of stuff. So, so does your field crossover with that at all? And, and what, what can we be optimistic about that aspect? Yeah. So right now, there is like a ton of interaction between computational neuroscience and artificial intelligence. It's like really a booming period for that. And this, you know, historically, it's gone in and out, they've kind of been aligned, and then they've gone their own separate paths, and then they've come back together. And so right now, there's a ton of interaction. And the same types of models that kind of form the basis of a lot of these machine learning systems that people are using in the real world and industry, those same models can be used as models of the brain, at least very similar models, you can use kind of the same approach to build a model of the brain as you do to build an artificial intelligence. And the reason for that is just because right now, kind of the dominant form of artificial intelligence is artificial neural networks. And so they're kind of based very simply on how real neurons work. And this comes from back in the 40s and the 50s work that where at that time things were kind of close again at that time. And so the people building AI were interested in how the brain works. And so they came up with these ideas of artificial neural networks. And now we're in a time where that's really coming to fruition. And then these artificial neural networks can do a lot of interesting things. So there's definitely overlap my personal research, I train artificial neural networks to do visual tasks. And it's the same kind of style of artificial network, as Instagram would use to analyze pictures that it has on its service, that kind of thing, it's, you know, it's what's used in computer vision now. So there's definitely overlap. And there is some shared interest in understanding how these networks work. So understanding how they train and how once they're trained, kind of what functions are they using to solve the task. And just, yeah, kind of wanting to know how they work so that we can use them as models, but also just kind of a general interest in how these systems could work, which could then be applied to, you know, understanding them in the real world and making them more robust and better in the real world. So there definitely is shared interests, there's shared methods. People in computational neuroscience are paying attention to what's happening in artificial intelligence and vice versa. So maybe that's optimistic, because, you know, we're supposed to be good at trying to understand these kinds of things. But there's their hard to understand once they're trained, you still can't look at a trained artificial neural network and just be like, Oh, I know what it's doing. I know what task it's all being. It's it's hard because it's this giant system of bunch of numbers. And yeah, we need methods for for better understanding them. Yeah. And so so with, you know, in your in your book, like I started seeing how, you know, just these different like mathematical models can help us with like predictions and understanding. And I, I, you know, I've been reading books on like ethics around like, you know, machines and all this other stuff. But I'm curious, you know, like, as we were talking about like these unintended consequences with kind of the models that you're using. So for example, like you're talking about like visual training with some of these algorithms. And for example, that's off the top of my head if we're talking like about Instagram, there's, you know, there's been, you know, issues where I've seen like, you know, like, women speak up about how like it's like censoring or labeling, like their images, which it wouldn't label like a guy and everything like that. So, so are there mathematical models that you guys are using that kind of work on these kind of issues and try to predict certain unintended consequences? Or how do you, as somebody who's like focused on like the, the, the mathematical models, look at that to make sure that you're like taking things into consideration, like for like this visual style of learning. Yeah. So I mean, I would say that, you know, while there is a lot of interaction, there still are different motivations and goals. So when I build these models, I'm trying to build models that will match neural data that's recorded from animals and kind of has the same features of neural processing that those do so that I could use this as a stand in kind of for studying the animal brain and, you know, figure out how the model works and see what happens if you perturb the model and all that. So the goal is still to connect it to the biology and understanding the frameworks versus a company that wants an algorithm that can find images that shouldn't be on their website. Their goal is to train a system that does that. And this kind of relates to this general idea that you need to be careful about your tools and you need to do proper testing because if it's flagging the wrong kind of stuff, that means that it hasn't been trained correctly. And so, but that can be for a lot of reasons, it can be about the type of data that they use to train it. It can be about the type of model or the specifics of the training algorithm. And that really is kind of what a lot of the open questions in machine learning are right now is just how do we best train our models? What do we have to think about for the data? How do we test them? And it's just a lot of just kind of practical open questions in the world of applied machine learning. And they relate to some of the shared questions that computational neuroscientists and machine learning people have in just in terms of better understanding the models and understanding how they learn and all of that. There is some overlap at the more kind of theoretical level. The assumption is, you know, if we have a better theoretical understanding of how these were kind of in the abstract mathematical form, then that would translate to, you know, tips for on the ground training a model that you're actually going to deploy. We should be able to come up with better ways of kind of controlling how that process works if we have a better background understanding of how this training works in general. Got it. Okay. Cool. And that helps with my optimism a little as well. Yeah, I like seeing how those interact. So, you know, a little bit, you know, earlier in the conversation, we're talking about, you know, who the books like for and all that. And, you know, one thing that you mentioned to I'm curious from what you've seen where like I'm a curious person, I read so many books on different topics, it's nuts. And I'm always like trying to I'm like, why am I even interested in this? But like you said, they're like, there's like this growing, you know, curiosity about the brain, right? There's like brain and like psychology and all this stuff. So like, from what you've seen in your experience, you know, and you know, you have a decent size following on Twitter, like I'm curious like what kind of people like you think like, are curious about this stuff? Or why do you think that interest is for like, the average person like, like me, like I have nothing to do with brains, like in my regular daily life, right? So I'm curious, like, what, where do you think this, this broader kind of interest comes with how our brains work and all that? Yeah, I think it can come from a lot of different directions. I think some people maybe they or someone they know had some sort of neurological disorder or psychiatric disease and they're drawn to it from that perspective, like they want to understand what's happened to them or the person they know and to understand how that person's brain works and why. Some people, I think, do come at it from the kind of AI side of it, like, I want to know how to make intelligent machines and the most intelligent thing that I can study is the human brain. And so they want kind of reverse engineer it, because they want, you know, computers to be smarter and all of that. I think there's also just, you know, an inherent interest in understanding yourself and other people. And so that's, you know, usually, you know, at a kind of psychology level, you don't necessarily have to talk about the brain and how the brain works to talk about human behavior. But sometimes, I mean, this was kind of what happened to me, I was interested in psychology in high school, but then I kept thinking like, okay, that's kind of a description of what's happening, but what actually causes it, you know, like what's the mechanism, and then you have to go into the brain and think about how the brain works and how it's producing human behavior. And so I think it can just come from a very, you know, there's a lot of reasons why you want to understand people. I mean, they make up a big chunk of the world around you. So if you want to understand your world at all, you're going to have to understand humans and how they work. And so that naturally leads to a desire to understand psychology, which could lead to a desire to understand the brain. Yeah, yeah, it's funny because it's almost like you're just like, like you've been spying on my own background, because that's basically what happened to me. Like I mentioned, like, you know, I came from addiction background, that I want to learn about mental illness. And then like, when I was learning about the behaviors that I got deeper, and started learning a little bit more about like the neuroscience behind it and the different like, you know, when I got into mindfulness, for example, I started learning about like the prefrontal cortex and, you know, and like the amygdala and all these other things. And so, so yeah, I could definitely see that. And, and so one of my goals with this podcast too is like, just because I'm, you know, interested in all these topics and trying to see how they relate to the average person because I want more people to be interested in these kind of things rather than just, you know, as you know, like in the middle of COVID and this Delta variant spike, like there's an issue with like science communication and people's lack of understanding of science. One of the best things you said, which is very true is how science we're constantly learning and something we know now maybe different in 10 years and all that. So, so like with your book and these topics around like computational neuroscience, what would you say like the main benefits? So, so aside from the people we were just talking about, like I live in Las Vegas, I go down to the strip. Why should, if I go up to somebody, how do I sell them on the idea like, Hey, you should care about computational neuroscience. Here's why. Like what kind of real life stuff, you know, Yeah, I think maybe there are kind of two routes and one, it can just simply be put as the brain is very complicated. And I think people don't always appreciate just how complicated the brain is and therefore kind of how complicated they and their biology and their behavior are. And, you know, some of that is because there are these kind of popular science narratives that make it very simplified. And, you know, they try to make it seem like, Oh, it's just like left brain versus right brain or your lizard brain versus your primate brain or something like that. And the truth is it's just it's far more complicated and our knowledge is evolving. And to not, you know, hold on to any belief about how your brain works with so much certainty. You know, it's, it's, it's something that is complicated and very multifaceted. And if you're trying to understand yourself or others, you know, you can, like, acknowledge that complexity and it might help. So even if you don't have an interest in biology itself, just kind of the general acknowledgement that the brain is complicated and our understanding is evolving might make people see things a little bit differently. They don't have to force everything into some existing narrative that they've been told about how the brain works. And then the kind of computational side of it, in particular, I think that it would be great if more people understood kind of the very practical benefits of mathematical modeling. And it relates to the fact that things are complicated. And it's really just like, you can't hold all these ideas in your head at once. And so, you know, it's like, if you have a to-do list, it's like, I, I, okay, I have to go to the grocery store. I have to buy milk. One thing, okay, you can remember it, you go, you get it. Oh, wait, I also want bananas. I also want eggs. Eventually, like, ah, shit, I gotta write this down. It's the same thing with mathematical models, except you're just writing it down as an equation. It's like, okay, well, that neuron connects to that neuron. And that neuron has this kind of activity, but that other neuron, it kind of, it stops the other neurons from firing. It's like, oh, wait, I can't hold this all in my brain. I need to write down in code or in math what I think is happening. And you can apply that to anything in life, just the idea that when you have something complex with a lot of interacting parts, you gotta kind of be explicit and just write down how you think it's working. And like, you know, you can, it's sometimes it's just like a little diagram, but if you really want to be precise about it, it's going to have to be math eventually. You have a little diagram, but you have to be specific about, you know, how the different parts interact, you're going to eventually get to an equation. And that's all that, you know, in applied math, like people think of math as being like, so, you know, oh, I couldn't possibly understand it, but it's really, it's just like, no, it's like too much going on for me to just think it through. I just want to write it down exactly what I think is happening, and you write it down in an equation. And that's kind of the basics of it. So I think it's broadly useful. It will explain, you know, especially like with COVID and these models that were predicting cases and all of that, that's just like, that's a mathematical model that somebody's saying, I expect people to interact with this many people in their lives and the disease spreads at this rate. And if I put that together and I look at the population of my country, then this is what's going to happen over time. And it's just people having a thought about how the world works, and they're just putting it down in math, and then you can make predictions. So it's pretty powerful, but it's also not that complicated, that idea of just writing down in math, if you want to be explicit about them. Yeah. And Grace, that was, we should have started with that, because you just, you resold me on the book. Like it is, like there's so much stuff and understanding about, especially like just the different models. Like for example, I have never really been a math guy. But I, like you mentioned, like, you know, just even with, you know, COVID or like just remembering stuff when we go to the grocery store, that kind of stuff. Like, I've read a lot of books on just assessing data, right? Because especially during COVID, we're constantly hit with data, right? How many people are being infected? How many people are dying? What are these rates and da da da da. And I don't have time to figure out all the math behind it. But when I'm able to look at data and ask the right questions, understand what they mean and stuff like that, it helps me, you know, like I'm a father, right? It helps me with like risk management. Like right now, there's a conversation around, do you get vaccinated and how much research is there and all this other stuff. But when you can understand some of these models and statistics, you know, and something you talk about in the book is like probability and things like that. When we know basic, like basic things like that, life goes so much better. But one of the things you mentioned too was like, just we have this idea of how the brain works and all this, but we realize how complex it is. Because I think one thing that all of us can use is like this intellectual humility, right? And that's something I get every time I pick up a book, like especially yours, I'm just like, oh man, I don't know nearly as much as I thought I did. So it's something we can all benefit from. But one of the things like you mentioned too, like about memory, I love that part of the book. Can you give kind of a brief explanation of like pruning? Like what happens in our brain when pruning? Why that, like what it is and why it matters? Can you, because I think that is a very practical thing that we should all know about. So lay it on me Grace. Sure, yeah. So pruning is, it's kind of what it sounds like. It's the idea that you kind of, your brain is kind of overgrown and you need to prune it back. And it's particularly a strong effect during development. So like in utero and when a baby is born, they have like way more neurons and connections between neurons than you do as an adult. And so the process of pruning is just you get rid of those connections or some of those neurons die off and you're just left with a kind of smaller network that has fewer interactions. And the reason to do it is basically it's a way of, you know, finding the important connections. So you had this overgrown thing and then you go through this process of like use it or lose it. So if in the course of experiencing the world, certain pathways and certain neurons just aren't sending a lot of signals and it's like, okay, that wasn't important. We don't need that. And so it's a way of kind of creating a system that's adapted to the environment that you're in because you're only keeping the pathways that are helpful for you. And yeah, so if you were in a different environment, there might be different pathways that you need. So that's, there's a bit of that element to this kind of use it or lose it and this where, yeah, you can have more connections than you need and get rid of the ones that you don't need. But the ones that you don't need are going to be kind of dependent on what gets used in this time of pruning. So that's the basic idea of it. And how it's done in the brain. Yeah. And when I learned about that, like that was that was one of the things I learned when I first started learning about neuroscience. And it makes sense, right? Like we learned so much like there's, you know, there's like a rare amount of people who like remember everything, like literally everything and they go insane. But like with us, I'm curious, you in the neuroscience field, like, do you do anything specifically to kind of like hack the pruning and memory system? Like for example, something I've done is like I like just using social media, right? I have, I've pruned my social media. So I'm only getting important stuff. So I'm not wasting cognitive, you know, cognitive energy on nonsense, right? Like, there's certain things that I don't need to remember. I like, I have a finite space in my brain. It's like a hard drive. So it's like, get all the nonsense out. So I'm, I want to learn from the best now. Like, is there any kind of pruning hacks that you use to remember important stuff, disregard other stuff, or any tips that the listeners might be able to benefit from? Yeah, so I think that kind of as we're adults, we think that repetition isn't important anymore. Like repetition is for when you're in grade school, and you're trying to like memorize something. But repetition is this way of, you know, keep, you know, keep activating the same pathways, and then they won't go away. So I think repetition is important. I actually feel like I, I came around to being open to repetition, like exposing myself to the same thing multiple times. Or yeah, like kind of having a social media feed that's some redundant voices or information just to like make sure that I really caught it and that I remember if it's something that's important to me. But I feel like it wasn't, I mean, it's a little bit of neuroscience that makes me appreciate like, yeah, you just sometimes have to repeat things to remember them. But it was also a book by Alon de Baton, which was a religion for atheists. I love that book so much. Yeah, so he's talking about different aspects of religion that actually make a lot of sense in terms of, you know, keeping culture and that kind of thing. And so obviously in religion, there's a lot of repetition, you say the same prayers and all of that. And it's just an explicit sense of like, yeah, you're going to do the same thing multiple times. But it's because it's important and you want to really imprint it on you. So yeah, so repetition is a very basic way just allowing yourself to see or experience the same thing multiple times so that you really get it in there. Yeah. Yeah. And, and you know, that segues perfectly into, you know, one of the last topics I wanted to chat with you about. So again, like I really got into neuroscience because of my addiction background. And one of the last chapters in the book talks about like reward based learning, right? And you talk about, you know, Pavlov, and then you talk about B.F. Skinner, who was like the dude when it came to like behavioral behaviorism. And, and yeah. So, you know, with that, right, like there's cues and rewards and you dive into that and kind of, you know, even a little bit about like our expectations of rewards and all that. And so kind of like what you were just talking about with this repetition, you know, another like benefit is like, I found like, you know, cognitive behavioral therapy, right? A lot of that is about repetition and changing, you know, the way our reward system works and all that. So, so I guess like, you know, from, from what, what we know about the brain, like, even with just like habit building and building good habits, are there like, you know, benefits that we can learn from like behaviorism and reward based learning, if we want to build, you know, good, healthy habits, whether it's, you know, not just overcoming an addiction, but just, you know, trying to stop doing this thing. And what kind of hacks can we use to do the better thing instead, based on what we know about the brain? Yeah, I don't know if they're hacks, if it's just kind of better and you'll be better. Because as you said, a lot of it is habit building and repetition and all of that. And yeah, so the chapter on reinforcement learning, it talks, you know, about how kind of there are things that are rewarding, like primary words like food or something like that. And then there is kind of how we learn to get to those rewards, which in the terms of reinforcement learning is we have this thing called a value function. And it kind of back tracks the reward so that, you know, if you're the example I use, one of the examples I use in the book is that if you're like at an office and people are singing happy birthday, that probably means you're about to get cake. So you, yeah, you hear at the song happy birthday, you might be close to getting cake. And so if you do that a lot in life, which you know, if you attend a lot of birthday parties, you start to form that association. And so the song happy birthday has kind of some value to it, to the extent that it's tied to this reward. And that can kind of keep backtracking and backtracking and you know, it fades a little bit. So like things very far away from birthday cake have little value when it comes to if you want birthday cake, but the things that are closer to it have more value. And that's kind of how you can like follow this value function to get to a rewarding state. But what that means is then, you know, if you have these strong associations, you kind of creating this pathway that's going to lead you to that reward. And if that reward is actually something you don't want to have anymore, like you're trying to lose weight and you don't want birthday cake, it's hard to you don't have you have to kind of recognize that you're not just going to have to not eat that, but you're going to have to understand the kind of full pathway that leads you into that behavioral state. So yeah, it's a little bit. It's interesting this notion of reinforcement learning, which is also a way that you can train these artificial neural networks to have them learn how to navigate to rewarding places. It's a way of, you know, it puts some terms and language to kind of concepts that maybe we are a little subconsciously aware of or whatever. But it kind of gives you a way to talk about it and to think about it that makes clear how you're kind of learning this association and then might give you ideas of what you have to do to break the association. Yeah. So real quick out of curiosity, are you familiar with the work of Dr. Judson Brewer? His stuff. Um, he's okay. So he's he's he does some neuroscience, but yeah, he wrote a book about a lot about this and, you know, how the brain works and he's a huge BF skinner fan. But uh, but yeah, that's kind of where I learned and just kind of replacing behaviors by learning what that, you know, that pathway is and the way he kind of breaks it down is like trigger behavior reward, right? And so part of like CBT, for example, is replacing that behavior. So for example, with your cake example, right, like, uh, like what, uh, what, uh, the best way to do it, like replace it with something else that might give us a reward, like instead, like something similar, that's going to, you know, trigger dopamine or endorphins or, or something like that. Yeah. You could imagine that that's helpful because again, another part of this reinforcement learning and the way that you learn this value function is if, if I hear happy birthday and I think cake is coming and then I don't get any food at all, I have this big kind of negative reward prediction error and that's used to update the value function because it's like, Oh, maybe someone's saying happy birthday, but they weren't actually serving cake. And so now I need to stop associating the song with cake because I'm actually like in a new place where it doesn't mean cake is coming. And so you can kind of just break that association by repeatedly hearing the song and not getting, uh, what you thought you were going to get. And then you won't have this expectation when you hear the song that you're going to get that. Um, so yeah, you can kind of, you know, but yeah, the, the full drop from like cakes and nothing might be hard. So maybe it's like cake to fruit or something like that. But yeah. So there, there is this sense that you, you know, you learn your value function by these social associations happening and you can unlearn it if these associations stop happening. Gotcha. Okay. Cool. So one of the, the last question I want to ask you, you, you talk about, you know, the brain being like this prediction machine and all that. And when I kind of learned about this, it was when I was just really into like reading books about why people have supernatural beliefs and skepticism and stuff. Like, so when you look at, you know, like the brain, like predicting things and you talk about kind of like, uh, you know, uh, how our visual functions work. Like, so does these, like, are there prediction errors that kind of explain, you know, these like, oh, I saw like a ghost or there's a UFO. Like, do you look at that kind of stuff when it comes to computational neuroscience? Does it kind of explain these like supernatural phenomena that happen with people because their brain is just predicting certain things? Because isn't that like pretty much how magic works? They know how to manipulate that. Can you break that down a little? So, yeah, so people having beliefs and kind of conspiracy theories or just things that don't have a lot of evidence and all of that, that feels, you know, to some extent, it's like personal psychology and probably has partly to do with genetics and all of that, but it also feels very social, I think. So this isn't something that I study or know a lot about. And a lot of computational neuroscience, as we talked about, kind of is informed by animal work where you can't really study those kinds of things. Like, I don't know if the, you know, mice that we're looking at, yeah, believes UFOs or anything like that. So it's not, it's not in like mainstream computational neuroscience that people are really thinking about those things. There is, there's a field called computational psychiatry, which is focused on psychiatric disorders and the mathematical modeling of those. And that might, they kind of study the basis of hallucinations and delusions and schizophrenics and that kind of thing. But yeah, I think the kind of social element of more low grade beliefs like that is, it's not super focused on in computational neuroscience. But yeah, certainly studied in, you know, psychology more broadly and all of that. Got it. Okay. Okay. So now I know where to, where to look into some of this stuff. So, so yeah, Grace, you have been awesome. I have learned so much. These are the conversations I love the most because it, it fills some gaps in my knowledge and I love it. So the book is out. So for everybody listening who wants to know more about what you're doing, upcoming projects and all that, two questions. Where's the best place to find you? And do we have any upcoming books to look forward to at all or what you're working on? Yes. So I am on Twitter. My handle is NeuroGrace. So pretty straightforward. I also have a website with information about the book and where to get it. And that's just Grace W. Lindsay, L-A-N-D-S-A-Y.com. I'm not writing another book yet. Taking a little break in between. But I do occasionally write for some online outlets and stuff like that. So that would be advertised on Twitter and on my website and all of that. And I used to have a podcast about computational neuroscience and artificial intelligence, but the episodes are, they were never very timely. So if people are interested in the topic, they can still listen to those episodes. And that's unsupervised thinking was the name of that podcast. It's a little, it's more technical. It's more in the weeds than the book is or anything like that. It was more, you know, other academics would listen to it. But if people want to get into the weeds, that's a good place to start. Beautiful. Awesome. And I will link all that stuff down below. But yeah, again, Grace, thank you so much for your time. And, and yeah, I'm sure we'll be in touch soon. Yeah, great.