Support the show to get full episodes and join the Discord community.
Check out my free video series about what’s missing in AI and Neuroscience
Luiz Pessoa runs his Laboratory of Cognition and Emotion at the University of Maryland, College Park, where he studies how emotion and cognition interact. On this episode, we discuss many of the topics from his latest book, The Entangled Brain: How Perception, Cognition, and Emotion Are Woven Together, which is aimed at a general audience. The book argues we need to re-think how to study the brain. Traditionally, cognitive functions of the brain have been studied in a modular fashion: area X does function Y. However, modern research has revealed the brain is highly complex and carries out cognitive functions in a much more interactive and integrative fashion: a given cognitive function results from many areas and circuits temporarily coalescing (for similar ideas, see also BI 152 Michael L. Anderson: After Phrenology: Neural Reuse). Luiz and I discuss the implications of studying the brain from a complex systems perspective, why we need go beyond thinking about anatomy and instead think about functional organization, some of the brain’s principles of organization, and a lot more.
- Laboratory of Cognition and Emotion.
- Twitter: @PessoaBrain.
- Book: The Entangled Brain: How Perception, Cognition, and Emotion Are Woven Together
0:00 – Intro
2:47 – The Entangled Brain
16:24 – How to think about complex systems
23:41 – Modularity thinking
28:16 – How to train one’s mind to think complex
33:26 – Problem or principle?
44:22 – Complex behaviors
47:06 – Organization vs. structure
51:09 – Principles of organization: Massive Combinatorial Anatomical Connectivity
55:15 – Principles of organization: High Distributed Functional Connectivity
1:00:50 – Principles of organization: Networks as Functional Units
1:06:15 – Principles of Organization: Interactions via Cortical-Subcortical Loops
1:08:53 – Open and closed loops
1:16:43 – Principles of organization: Connectivity with the Body
1:21:28 – Consciousness
1:24:53 – Emotions
1:32:49 – Emottions and AI
1:39:47 – Emotion as a concept
1:43:25 – Complexity and functional organization in AI
Luiz 00:00:03 As you know, it’s not being entangled, that you can’t disentangle. It’s, it’s really that it’s inherent nature is entangled in the sense that we don’t have these, these clear cut boundaries, these, this more modular view that has been so prevalent, so dominant in neuroscience for forever. Because the connectivity is so massive, it helps us understand that we need to go moving away from a space of anatomical connectivity to a space of functional relationships. But if we want to solve problems that are similar to what embodied animals solve in intelligent ways, I think we need to solve a couple of problems that bring us to emotion. And, and here’s my way of thinking. My way of thinking is that
Speaker 2 00:00:59 This is brain inspired.
Paul 00:01:01 That was Luis Peso. Luis is a professor at the Department of Psychology at the University of Maryland College Park, where he runs his laboratory of cognition and emotion. His latest book is The Entangled Brain, how Perception, cognition, and Emotion are woven Together, which is aimed at a more, uh, general audience. In broad strokes, the book argues that we need to move past the traditional modular view of the brain, which is still prevalent. You still come across statements like Brain, area X does cognitive function y or you know, for example, the amygdala does emotion, uh, and so on. Uh, but instead, we should account for the highly interactive and integrative functional organization of the brain. So we discuss some of the implications, uh, of approaching the brain from a complex systems perspective. Some of the principles of organization within brains that Luis believes, uh, will help, uh, reshape how we study and understand brains and brain function.
Paul 00:02:01 And Luis has, uh, long been interested in the interaction between emotion and cognition, which you can learn more about in his previous more technical book, the Cognitive Emotional Brain. So we discuss emotion and cognition and why, uh, emotion might be important for artificial intelligence, and we discuss a lot more as well. Um, many of Louise’s ideas are in line with those of Michael Anderson, who’s been on the show recently. So if you liked that recent episode, uh, with Michael Anderson, you’ll probably, uh, like this one as well. Show notes are, uh, brand inspired.co/ 1 55, where you can also learn how to support the show through Patreon. Thank you to my incredible Patreon supporters. Okay, here’s Luis.
Paul 00:02:46 Uh, Luis, the, the, the book is, uh, the Entangled Brain, how perception, cognition, and emotion are woven together. And, uh, there’s, there’s a lot. Uh, the book does a lot and there’s, there’s a lot for us to talk about. One thing maybe I would just start with, um, is that I find the structure of the book interesting because you, you go from a modular, you know, modular descriptions of the brain, um, and you’re warning the reader the entire time that you’ll eventually destroy, uh, that modular view. Yeah. And then you build up from, um, what you describe as, uh, a minimal brain, um, to more complex modular view of the brain tracking evolution along the way. Right? Right. Um, and then you come to the destroying part where you argue for new ways to think about how to connect brains and behavior. Um, so <laugh>, so that’s a lot to unpack right there. But, uh, maybe, we’ll, we’ll just start with the, the, the title. Why Entangled? Why the Word Entangled?
Luiz 00:03:49 Right, right. So, but lemme just say one thing before that, that that the structure of the book is something interesting. I, I really, I really struggled with because I wanted to write a book to a general audience. I did not want to write, I had written a book many years ago, 10 years ago or so, that was for a specialized audience, and it was, was fine. It was, I was published, and I think it was, I think is a, a fine book, but, but what I really wanted was to write something that sort of reminded me, like when I was an adolescent and reading books, I was really into physics and math and other things. And I, I would read these books that I thought was really fascinating about really super complex topics. I don’t know, like, uh, advanced physics and whatnot, that, that, obviously I was not even in in college. So I wanted to write something that perhaps could resonate with someone like me at, you know, whatever number of years ago along many decades. Decades ago. So I really struggled on how to write it, and I, I, I did definitely want to write it from, from, for a general audience. So I decided to not start with, uh, something like about complexity theory and build it from there, and even leave the entire past behind and just like, just forget about a more modular view of the brain. I wanted to,
Paul 00:05:15 Was that a, was that a struggle for you to do that? Because No, it, you sort of, you know, okay.
Luiz 00:05:20 No, no, actually, somehow it, it, it was natural. It was natural, but it, but I, I don’t know if it worked, because I’ll tell you a couple things. One, I’ll tell you in part because it’s a little delicate <laugh>, that an initial re an initial set of reviewers was really confused about, it says, this book is not working, because you were suggesting this is a complex system. This is an entangled brain. That, going back to your original question of Entangled, the idea is that this is kind of intertwined to the extent that we don’t have these clear cut boundaries between these standard domains that we always talk about. Perception, cognition, emotion, motivation, action, so on. And I wanted to have a framework that would sort of dissolve these boundaries. And at one point I was, what’s the title? And Brain Networks, the Network Brain, the Highly Network Brain.
Luiz 00:06:17 I mean, that just didn’t, didn’t really work. And the Network Brain had been something that people had been talking about for 10, 20 years even. And I threw it out on Twitter of all things. And, and, and there were suggestions. And I, and among one of the suggestions, I, to be honest, I don’t even remember if I said, is Entangled good or someone suggested entangled? I think more likely someone literally suggests, how about Entangled? And I said, oh, that’s, that, that’s perfect. So in a way, it is perfect because I view it, I think it’s, it’s quite adequate, but it’s as, as, as you, as you know, it’s not being entangled that you can’t disentangle like some knot that you’re gonna disentangle and, oh, I fixed it now, and it’s all crystal clear and Unentangled disentangled. Hmm. It’s, it’s really that, it’s inherent nature is entangled in the sense that we don’t have these, these clear cut boundaries, these, this more modular view that has been so prevalent, so dominant in neuroscience for forever, for a very long time.
Luiz 00:07:25 That being said, that obviously since the historical beginnings of neuroscience, there’s always movements for networked views from the very beginning, and more anti localization is views. So it’s, it’s this, this continuum swing of a pendulum going more, more modular, more, more distributed. And the extreme would be something like lashley equi, equi potential views that it doesn’t matter what you take out scoop out is just proportional to the volume that you take out, which is obviously not the case. So, yeah, long, long, long, long answer there. Entangled is really an idea to try to move us away from thinking, having almost this reflected, re reflexive way of thinking about the brain in this very modular clockwork machine-like type of analogy that we often make. So that was, that was the goal, is to have a title and someone, uh, there were some suggestions and, and then, and I think that worked well.
Paul 00:08:31 Oh, I was all set to congratulate you on coming up with a new term that, that fits, but maybe, but you don’t even know if I should be congratulating you because you outsourced it and you don’t know. Yeah,
Luiz 00:08:41 No, I think it’s, it’s even better. It’s outsourced. Uh, I think it’s was the collective there. It’s, uh, <laugh>. So,
Paul 00:08:47 So going back to the, the, the structure, um, it surprises me, well, your answer surprised me that it was really about reflecting on your journey and your early, um, fascination with these kinds of works. Um, you know, you just mentioned that the, the modular view is the reflexive kind of view. Um, and I thought it was maybe based on that, because that’s where we all kind of start, right? Yeah. And it’s like mo the most natural. Yeah. And it’s kind of unnatural to think, to start, or, or, or just to think in terms of complexity and those sorts of issues that complexity brings up. Uh, so I thought maybe that’s why you wrote it, wrote it like that.
Luiz 00:09:30 Yeah, no, it is, it, it is, it is that I, no, I think you’re right. No, you’re right. I wrote it like that because of this, this way that we were taught, but that is very natural for people to think, at least in, in the West and the way people think in, right, right. I, I, I, I think that that somehow is very natural. And I, I, I was, I was part of a, a very large group of, of people at the University of Maryland trying to create an institute at some point, starting maybe 10 or so years ago. And there were people from the humanities, and we would often start talking about the brain. And when people would talk about the brain, it was, it was so natural for them to think, okay, this part of the brain does dysfunction, and this other part, there’s this function.
Luiz 00:10:13 So if you have this problem in the Amy Mik or in this other region, you can be anxious or you can, so it, it was, it was so natural for them to think in, in that way, that I just wanted to convey something that we would try to break this naturalness. And if I started from, okay, there are all these physics and mathematicians that created all these complex systems and dynamical systems theory, and let’s just adopt that. And that’s, that’s chapter zero. And then we built from here, I think we, I would lose, I would lose the ability or the possibility of sort of pushing back, like, okay, we, we have these areas, but what really are areas? I mean, what do they compute? Do they compute one function? Is it one-to-one mapping? Is it one to many, many to many? What kind of functional repertoire areas have?
Luiz 00:11:08 How do we think about areas and what are even areas? How do we find boundaries and how do we define them? So I wanted to, the breaking was part of the process of building it, but I do think that my first drafts on the complete worked out drafts left that a little bit not well worked out enough that the reviewers had problems. And I think in part it was justifiable, just more than justifiable. And so I went back and I kept on, like you said, I kept on reminding the re the, the reader that we’re gonna break it, we’re gonna break it, we’re gonna break it. So that it seemed to have some flow, perhaps unusual or unexpected, but that at least didn’t leave the reader confused about, well, there was all this modular stuff, and then you turn to complex systems in chapter emergent behaviors in chapter eight. There, shouldn’t that be chapter one? I mean, there was some chapter two, some re some reviewers said this should be chapter two. Hmm. Yeah.
Paul 00:12:09 How long, when did you break yourself and are you comfortable now? Oh, go ahead. Answer that, and then I’ll, I have a, oh,
Luiz 00:12:16 Yeah, no, I mean, actually to me, it’s, it’s coming back home in a way. So it’s, to me, it’s the opposite. And obviously neuroscience is large enough and diverse enough that there are multiple views. This view could be very productive, or perhaps not, because the modular view is defended by many people still. So I’m not saying that it’s, we should just completely abandon it. So, and I’m not saying that I’m having this vision from early on, that was ahead of my time. But I started from that perspective because I started reading things like then it and the mind’s eye. And, and then I did my PhD in, in the department of, uh, cognitive and neurosystem at Boston University with Berg’s group. And so e everything was already, and I was really into reading about dynamical systems. I was doing, uh, a major in computer science and math, and many of my closest friends were doing honors thesis or already in graduate school doing something on dynamical systems.
Luiz 00:13:17 And so that’s sort of a dis distributed parallel distributed processing was, was something that, in a sense seemed natural to me. And then it got only solidified with my PhD. And so when I started, when I went away from computational neuroscience after my PhD to more experimental neuroscience, it’s, it’s when I really got to learn more how biologists and neuroscientists in particular think, and I learned of this way of, of bringing things a little closer in terms of the mappings one-to-one, or perhaps not exactly one-to-one, but where a double dissociation type of experiment, when you do these lesions and you investigate the respective types of deficits, it, it, it can lead to more conclusive statements about the specificity of the deficit and the, the way the areas and carry out or parts of the brain carry out specific functions. So I, I, I really learned these things after, in more depth, at least at the very least after graduate school.
Luiz 00:14:23 And I started learning this vocabulary in this way of thinking. And so it’s kind of, the book is kind of going back all the way from to my first readings about, I don’t know, I should take a look again. The, the chapters in the Mind’s Eye Do, do you know that book The, by then edited by Deni and Hof Starter that has all these essays on collective behaviors and termite colonies, and how do they implement things or carry out behaviors in this, the system oriented distributed way and things like that. That was basically the first few essays that I was reading in the very, very early days of starting to think about the brain and learning about the brain.
Paul 00:15:07 Well, once you do get to the, um, we’re just gonna jump around. So once you do get to the complexity, uh, parts later in the book, you know, you, you list some of the, um, principles or implications for how to think about, uh, brains based on, um, it being a complex system. And I’m just, I’m gonna list them off here. Uh, and then I then I want to ask you <laugh> a follow up question. So, so I, I think you list five or six, so, um, interactions between parts, uh, levels of analysis, dynamics, you know, we have to think of over time, as, you know, in aural way, uh, decentralization and use the word hierarchy, which is, uh, not, not to be confused with hierarchy, right? You wanna define hierarchy for us real quick?
Luiz 00:15:55 Yeah, sure. I mean, I think hierarchy is just, it’s, it’s really the, in my view, it’s the absence of a, in a clear cut hierarchy. It’s, it’s, so, it’s, it’s just one thinking hierarchically is not a constructive way of thinking.
Paul 00:16:08 Yeah. I think that, um, Mike Anderson, Michael Anderson used that term when I had him on the, uh, podcast a while back also, uh, and then emergence and complexity and, you know, I buy into all of these things. Um, but what I want, and, and maybe we can dive into, you know, more of them individually, but what I struggle with is to think of them maybe cuz I’m from the western world, I don’t know, sort of holistically and, and have them all in mind when I’m thinking about a cognitive function or a brain area or process. Have, so have you, you said it’s kind of natural for you in, in your coming back to that way of thinking, um, after, you know, going learning more. But do you feel comfortable, um, like when you, uh, think about, well, let’s say emotion or something new that you learn about, can you bring all of these, um, implications or principles to bear on the problem in a sort of holistic way? Or do you kind of step through them and think about them in slices? Um, right. How, how do you, because I’m, I’m not comfortable yet having them all in mind. Yeah,
Luiz 00:17:13 Yeah. No, I don’t, I, I, I mean, I wouldn’t claim that it’s, it’s, it’s natural in some instinctive sense. I, I, I think that I, I wasn’t exposed to, to, to some types of solutions that neuroscientists had considered and had thought about at the very, very beginning. I mean, I, I was, I was being exposed at the same time to these, to these multiple ideas. So I always thought that it basically, to a large extent, they’re empirical questions. So it’s, it’s how is this implemented in the mammalian brain or in, in humans versus mice? So there could be some more modular solutions to a problem in some species for some types of functions or, or not. So it, it, it’s something that I don’t, I don’t, I’m not saying that it’s, it’s, it, it’s the type of reasoning or a type of thinking or type of architecture I should say, that is completely non-existent.
Luiz 00:18:18 But I think that I, I always thought in terms of a spectrum of, of, of ways of being organized from very hierarchical, let’s say visual cortex goes themus and thalamus goes to v1 and v1 V goes to v2, and V2 goes to v4, and V4 goes to it, and, and so on all down the line. And then association areas and association areas have some fibers all the way to prefrontal cortex, to an apex region, all the way there where everything comes together and is put together in a way that reflects your past experience, your current experience, your perceptual and abstraction puts everything together, right? Puts everything together. Yeah. So that could be, that could be that, that could be how the brain works. And, and, and that could be even somehow how some brains work. But I was always attracted to the multiple types of solutions because, for instance, I started seeing that I had a lot of colleagues in Brazil, I’m originally from Brazil, and, and, and I had a lot of colleagues and mentors in Brazil originally that were doing a lot of work on anatomy and the physiology of the visual system.
Luiz 00:19:29 And, and in fact, they, they happened to be collaborating very strongly with Leslie Underliner from the National Institutes of Mental Health. And long story short, that’s why I ended up doing a postdoc there. But they were, they were finding all these, these bypass connections and from v from from v2, it goes all the way to it, or, and, and from, so these parallel type of processing was also very evident. And, and because I was sort of in a computer science type of background, and we were talking, we were talking a lot about in the end of the eighties about these, these parallel architectures. It, it, it seemed natural. And, and so that’s the sense, I, I think that, that I, I think about the being natural, but going back to your question about how do I put all of them together? I don’t think I put them all of them together at the same time, but it, I I view them as, as, as more guiding principles every time that I’m trying to think of something and I start to think of something too much of a, as a thing in itself, I remind myself, okay, but it’s al it’s always temporal.
Luiz 00:20:38 So we should try to let go of that and think of it more as a process. So where’s time in this? How does it evolve? What sort of trajectory does it in some state space type of situation, scenario, does it evolve through or along? And, and, and so this is, so this is in, in, in, in one part of the, in the space of those concepts that you mentioned, the, the, the, the, the time and process part, the hierarchy is, it’s partly hierarchical, but it’s to a good extent also violating hierarchy. The latency across the visual system all the way to frontal eye fields and whatnot, completely violates the sequential propagation of signals. So, oh, okay. That means that in addition to this se to, to an existing and undeniable possibility of sequential processing, there are other forms of processing that are happening in parallel together with feedback, because we know those exist.
Luiz 00:21:37 They, they’re ubiquitous, that they’re everywhere. They’re, they’re, they’re an underlying characteristic of the architecture itself. So, so kind of reminding yourself of these things that at every time that you, you, you think that there’s, it’s something in, in more, one more, in more of one of the ends of the spectrum, from modules to distributed to more linear, to non-linear, to simple to complex sort of a spanning this, this range of, of, of possible architectures that different brains might carry out. But I do think that these questions have to be empirical, right? So a lot of my colleagues really think that that’s, there’s a just a lot of hand waving, saying, saying that the kinds of things that we say, I, I completely understand that I do think that there is at least an equivalent amount of hand waving on the other direction in the other direction, because okay, so like, I I literally just shared this question on Twitter the other day.
Luiz 00:22:40 So exactly what is an area? So we, we don’t even know how to define an area. So it’s, if we def an, an area is defined by a, a multivariate by a multiple set of criteria, the borders don’t coincide exactly if it’s cell density or if it’s receptor density, if it’s gene expression, if it’s connection, it’s, there are these gradients and whatnot. So some people say, okay, there are gradients. And so it’s not exactly a sharp boundary, but I do think that both ends, there has to be pushback because both ends seem to be having to address important questions. And I think that this standard story claims that more distributed processing, emergent type of processing is hand waving. But I kind of see the equivalent on the other, other side of things. But I mean, that’s my view at least.
Paul 00:23:41 So, okay. So, uh, I don’t, I’m not sure where I want to go with this, but okay. You mentioned frontal eye fields, um, uh, which was one of my area Yeah. Areas of expertise, right? Um, and you know, how it gets, uh, um, spikes occur in it earlier, uh, than you would think if the brain was strictly hierarchical. And, you know, that’s been known for some time, you know, but then, so let you know frontal eye field projects to superior colliculus another brain area. And then probably, because my training, I think, oh, superior colliculus, it does blank, right? It, you know, it’s involved in attention, eye movements orienting, um, fight or flight, you know, all the, the stuff. And you talk about it a lot when you, um, described the minimal brain, right? Um, early on in the book. But that’s my natural inclination is to think, okay, well, what does it mean? Or that frontal lay field gets early, uh, skip connections from, you know what, whatever v1, let’s say, oh, okay, well, V1 does this early. So that’s like my natural inclination. And, um, do you still think that way at all? Like, you know, well, what does it mean that superior colliculus and area and frontal life field and area are connected? Right? No. So I, you know, do you think like front of life field, does this superior colliculus does that, or have you thrown that away from your thinking?
Luiz 00:25:00 No, I, I don’t think I think that way. I, I would say at all, but maybe I’ll back. Yeah, I think it’s possibly at all. But let me think through as I answer <laugh>, because I always think in terms of circuits. So for instance, the superior colus is embedded in some kind of circuit together with the, the striatum or something, or, or, or the mid-brain or some other kind of, um, a circuit that it forms momentarily depending on the behavioral needs and, and, and requirements. And so I, I, I mean, it, it, I think it has some kind of flavors. I think like Mike Anderson, Michael and Michael Anderson said a few weeks ago, uh, in, in, in, in, in your, in one of your sessions, that it really momentarily forms the, this circuit momentarily forms. It doesn’t mean that the superior Colus or any other re other region has infinite capacity to, to implement anything depending on that very same moment created from nothing, creates some, some new kind of function in the context of that circuit.
Luiz 00:26:08 It does have some clear flavors of type of processing, given its connectivity, given its types of receptors, given the types of, obviously the inputs and outputs are completely determined, or I should say, are he incredibly important in determining the kinds of things that it can, the signals it determines what it, the signals that it receives and what it can influence, obviously, by definition. So it’s not that it has an infant capacity to carry out any function. It obviously has some, some flavor, some kind of repertoire that has some theme to it. And I think that sometimes the, these, the theme can be quite complex and can be quite multivariate in the sense they can expand quite different things. But nonetheless, it, it’s, it’s, it, it doesn’t have this, this, this complete flexibility that’s, that’s, that’s a factor that’s clear. But I do think, when I do think in terms of a region, I think of it, what circuit is it embedded into, and what’s the role of the amygdala or the sub nucleus of the amygdala, let’s say now that it’s actually really important in this, in this circuit for unlearning some relationship that it had learned before that some stimulus was fear invo evoking fear inducing or threat related.
Luiz 00:27:32 And now it’s, it’s participating, these other circuits are participating in some other circuit that, that actually is about the unlearning this relationship. So of course, the complexities there, different subpopulations of neurons, and, and that’s, a lot of things need to be worked out. But to the extent that we think of a little bit more macro regions, even sub regions being considered in this context, macro enough, I’m thinking that they exert their functions as a function of the coalitions, the coalition of regions. It’s grouped with that it creates this ensemble with, at momentarily in the function of solving some behavioral problem.
Paul 00:28:16 One answer to the question I’m about to ask you. One great answer is read my book. Uh, because, uh, in large part, your book does begin to train one’s mind on how to think from, you know, that more complex systems theory perspective. Uh, but do you have actionable advice on students, you know, who come up in what is still, I, I suppose, the dominant paradigm in neuroscience and thinking in this modular fashion of how to train themselves to get out of that way of thinking and come over to the Luis Paso side?
Luiz 00:28:52 <laugh>? Yeah. Right. Uh, I, I, I think that, I think that it is, it’s, it’s really important to be exposed, and I think more and more people are being exposed to ideas of complex science and network science, right? So that’s something that has clearly changed in the past 20 years. That, that,
Paul 00:29:14 Let me interrupt you. Yeah. Is that true? Cause I was talking with, um, a, a friend and, and, uh, someone prominent in neuroscience, and they said that they think it’s getting worse, um, that, you know, not necessarily just the modular view, but that sort of a, connects to b. Um, and so a does this, that that is actually getting worse in neuroscience, and that’s not my view, but I don’t know, I’m kind of in the bubble that you’re
Luiz 00:29:37 In. No, I think, I think those two statements could be correct at the same time in this, I, I don’t in the, in the following sense, in the sense that perhaps neuroscience training is still very much like that. Mm-hmm. But for instance, sometimes, uh, when I’m teaching a, a, a class, a course on functional m i or some other course that I teach, I, I ask sometimes when I introduce these concepts a little bit about if, if it, I happened to discuss this in the course, I always am curious to ask them, because I’m teaching students from psychology, and some are from, uh, a neuroscience major, and some are even from other majors, but they’re many from psychology. And, and, and so they haven’t been, they don’t come from a, let’s say a physics background. So I, I always ask them, have you encountered these kinds of ideas?
Luiz 00:30:26 And, and many of them mention, uh, environmental sciences, studying ecology, uh, ecology and, and, and some other, other things that they might have studied, but also have heard of this via reading books or reading other things from other areas. So not necessarily about the brain, but the interconnected economy or interconnected water and food supply systems of the entire planet, or what have you, and how these interdependencies are highly complex and, and how, how this breaks a little bit, this, this, the linear way of thinking that is again, is, I think is, is is supernatural for everyone. So I do think that because of this wave of network science, that at least in terms of popularizing some, popularizing some of these ideas with, with, really with books that were quite, that were quite popular and is sold extremely well. And, and were, were, were, I think, influential, influential in the, in general audience education about some of these ideas.
Luiz 00:31:35 I think that has happened to, to, to a greater extent than, than perhaps, you know, when I was at the, that same stage of going to grad to, to going to college or in, in the early years of co in of, of being in college, had I not been in computer science and thinking about distributed architectures, because I happened to be immersed in some environment that one of my mentors had studied AI and was, was, was trying to, was I was doing an internship and, uh, a buddy of mine, friend of mine was I was doing computer graphics related things, and he was doing ai, and, and we were talking about these architectures, these AI distributed architectures for ai. And so super computers were becoming really, uh, a notion that was starting to pick up in the eighties. And so I do think, I think, so my su going back to your question, I think my suggestion is, is to, is, is really try to, to find ways of reading some of these, this, this, this, and I wouldn’t say literature because like, it, it finding books that, that discuss these issues about highly interactive complex systems and exposing oneself to this, these ways of thinking that are, that are perhaps a li a little different from, from, it’s something that we learn in high school.
Luiz 00:32:59 We, we don’t learn quite a, quite so much about that perhaps, unless some, someone’s taking some classes in environmental science and they start environmental science starting to get a lot to those issues because they start talking about water cycles and population dynamics and other kinds of
Paul 00:33:20 Yeah.
Luiz 00:33:21 Problems that, that mapping into these kinds of formalisms. So
Paul 00:33:28 Sorry to keep this kind of high level bef, because I want to get into more concrete, uh, topics in the book. But, you know, one of the, so I, I, I’ve had Eve martyr on, and I know that, you know, Eve Martyr’s work mm-hmm. <affirmative>, and one of the things that she has discovered is that there’s this, um, that structure does not, uh, dictate function necess, you know, it’s necessary for function, but it’s not, uh, sufficient, right? So you can have one structure and multiple different functions. And this is also connected to, you know, your ideas of about interacting brain parts and, and the neural reuse ideas Yeah. And coalitions that, that evolve over time. And I think when I first, you know, came across these kinds of ideas, my first reaction was, oh, this is a, a problem because it’s, it makes it more complicated.
Paul 00:34:16 And, and it’s a problem relative to the way that, you know, I’ve been taught neuroscience, um, you know, region A does this, region B does that. But then, um, I thought, no, this is actually, I, I started to think maybe this is more like a solution, but it’s not really a solution so much as a principle. I, I think what I’m trying to say is that these kinds of principles as, as I sit with them over time, um, I get more and more comfortable with them as being an answer instead of a problem or, you know, at least like toward a solution or a direction. But, but in some sense, oh, this is a, a productive solution, not a roadblock to understanding. Does that, uh, does that ring true to you?
Luiz 00:34:58 Yeah, I think it does, but I think it does. But I think that there’s one challenge that is a, that feels like a roadblock, which is
Paul 00:35:07 Just one
Luiz 00:35:08 <laugh>. Yeah. Well, yeah, exactly. Like one, yeah. Well, there’s one clear, a clear one, which is, it’s something that I’ve, I’ve been thinking about a lot and I, I haven’t been able to still put on paper and, and, and other, but a lot of other people have been writing about, which is many of these systems are incredibly context dependent. So I’m not just talking about the brain, I’m talking about these complex systems in general. And, and so that becomes a little bit of a, that becomes a, an important roadblock, I would say, because in science we are trained to do the exact opposite, right? We abstract the way context, we fix everything, and we study that system so that we can have, make conclusions. Because if everything is varying, I mean, we can make, we cannot conclude anything because obviously costly, it just doesn’t make sense that kind of reasoning right at the first, first pass.
Luiz 00:36:05 But at the same time, if you’re dealing with a system that is incredibly nuanced and sensitive to this specifics of the context that this, this queue was encountered in the past and together with other queues that signaled that it’s actually safe only in that context, but not in other contexts. So it, it starts behaving in very different ways that you, it’s experimentally and, and, and theoretically become incredibly challenging. So I am a big believer in this enormous context dependence of brain behavior, and it, it, it does feel daunting sometimes because it feels that, how do you make sense of these, of this richness, richness of expressions that that would not be there were the system much less context dependent, right? So you drive a car and your car is, is made to be relatively insensitive to lots of contextual differences. You don’t want it to drive hugely different over gravel versus asphalt or what have you.
Luiz 00:37:24 Obviously it’s different, but, and we engineer systems to be relatively, to buffer us from these contextual differences. It’s raining a little bit. It’s, it shouldn’t be a major problem. It it, it’s, it isn’t. So it’s not a good example, but a but the, the kind of engineered systems that we engineer, that we, we create are created for that, that to have to be context in as a context, independent as possible, or only to the, and only depend on context when we really want, because okay, you want to go really fast and now, so press this button because you bought a super expensive car, <laugh>, I mean, maybe those exist. I don’t, don’t have an extensive car. So, but yeah, there is contact, contact contextual dependence, con contact dependency, but it’s of a, of a, a completely different qualitative nature than in biological systems. And I think that is something that is very difficult to wrap our ha heads around because even our training has experimentalists tries to remove that. And so I don’t think we are good. I don’t think there’s just period. We don’t even have good ways of dealing with that. I think that we have to change the way we do quite a bit of our science and training different ways because we are abstracting too much of context when in some sciences, I think context can be everything. And, and this is not just the brain, it’s, it’s, it’s about logical systems and many other complex systems that have these, these properties.
Paul 00:38:57 See, so now you’re, you, you were kind of a downer just now. To me, you were reigning on my parade because especially with that example of context dependency and context sensitivity, because I’ve become to really appreciate context as, you know, such a huge part. And, and that principle then, um, you know, allows me to take the next step and really appreciate that our brains and our behavior and, you know, the context within we’re operating, how it affects those things that Oh, wow. You know, that is an amazing principle that shows, you know, the adaptability and plasticity of our abilities, right? And that itself is a, uh, glor glorifying principle.
Luiz 00:39:36 Yeah. I, I, I I,
Paul 00:39:37 But it’s, but you say it’s a roadblock.
Luiz 00:39:39 No, I think, no, I don’t think it’s, I think that’s why we exist, right? I mean, after these one plus billion years, or if we just start 500 million years ago or what, whatever you wanna start, where you wanna start, it’s hundreds of millions of years that that, that, that led to, to this kind of architecture, uh, these kinds of architectures of all, you know, species, existing species on the planet. And, but for us to study that in, in a way that is based on a standard way of thinking, that is largely driven by, well, to a good extent at, at the very least, driven by how physicists have been thinking for the past several centuries, and how engineering has been so successful in the past 100 plus years. And so I think that way of thinking is, is, is very different from a, a way of thinking that would be productive in other kinds of sciences. So I do think that we need to kind of a reset button, perhaps not reset button, but, um, let’s shift this the other direction button and, and get started in ways of thinking that we, we just embrace this adaptiveness, this plasticity, this complexity, this the temporal dimensions, the lack of clear cut hierarchies and, and this naturally interacting nature that sort of determines how many systems work.
Paul 00:41:18 Hmm. And, and startup world, that’s called a, a pivot. If you want a new, a new term, we can pivot. That
Luiz 00:41:23 Is a, that’s a perf that’s a great word. Yeah. We really need a good pivot there. Yeah, yeah, yeah.
Paul 00:41:30 So I actually had a, um, a quote from your book, uh, in my notes here. And, and since we were talking about context, this will segue into our, uh, next little topic here. Okay. Uh, we can think, so this is from the book, we can think of the brain with all its different parts as evolutions solution to the problem of uncoupling inputs from outputs. This observation reflects a fundamental principle of brain function, context sensitivity. See, you didn’t call it a roadblock there, called it a fundamental <laugh> principle. <laugh>, anyway.
Luiz 00:42:01 Well, no, but that’s what I’m saying, right? It’s actually a fundamental principle that poses a huge challenge for us as extreme observers of these systems to work out exactly how, what’s happening. So I didn’t, that actually, I think it’s, it’s really fundamental to the existence of these to, to these bi of these biological systems. But for the experimenter, the scientist in all of us doing this, it’s, it’s a roadblock in the sense that we’re not even well prepared to think of it that way. First thing that we do is, is fix context. Like if you just tremendous advances in, in, in, in learning and memory and fear learning and other kinds of things, decades of research, the, the main goal is to fix the context and make, I mean, it’s literally, like I say in the book, it’s a little box that you put the animal in and everything is restricted.
Luiz 00:42:58 And there were other kinds of approaches, more ethological approaches that we’re trying to see what happens in more natural environments. But they didn’t, they didn’t take off because it’s incredibly challenging. We can’t have a forest here and study animals doing their natural thing. It’s just not gonna, it’s just, and there are too many things that we can’t control. So what, what conclusions do we make? So we started from the premise that we need to, to limit the amounts of things that can happen if you’re gonna conclude anything. But if that’s not a good model for studying this kind of system, and that’s, that’s up for debate, right? Obviously people are gonna say, no, this is still a really good way. And I, this is like, this is completely defensible and reasonable way of thinking. If you, if, if, but if not, then, then we are kind of in trouble. Because doing this research, for instance, in rodents is incredibly important for many decades in a, in an enclosed box, might have really incredibly limited the kinds of conclusions that we can make that actually apply to more real world, world settings. Hopefully not because we, we hopefully not. And again, that’s largely is an empirical thing that we’re gonna discover as we create more neuro techniques that can be more open-ended, more naturalistic environments that have the animals just navigating and, and, and, and living the way that they live outside the lab.
Paul 00:44:23 Yeah. You, you, uh, make more arguments like this in a recent paper. I don’t have it in front of me, but, um, where you argue that neuroscience, um, needs to become a study of more complex behaviors as well because of this kind of context sensitivity and you use threat assessment and other examples, um, in, you know, I, I’ve seen you use it multiple in multiple different places, but as this, um, reason why we need to think about both brain function and complex behaviors in, in these context sensitive kinds of terms, cuz they can change depending on all the factors going on.
Luiz 00:44:57 Yeah, I mean, my own education was one in which I was never exposed to these ideas cuz I didn’t study psychology and I didn’t study, uh, the, the whole breadth of, of, of methods and approaches. So I wasn’t familiar with these approaches until more recently, maybe 10 years ago or maybe a little bit more, more ecological types of approaches that really take the animals’ natural behaviors in their natural habitats as the basis of studying behavior and the organization of behavior. So I was not very highly educated in, in all of this, this type of thinking. And it, it just resonated completely with me because I, I said, yeah, that’s, that’s exactly how we need to do it. I mean, that’s not to say that we can do it because we, we just don’t have the techniques, right? It’s easy to say, whoa, I like to study a a troop of ba of 50 baboons with this complex hierarchy and having this, this, these kinds of fights and this kind of sexual behavior and this kind of, uh, food gathering thing and all in real time. Well, yeah, everyone wants that too, but obviously we can’t do that.
Paul 00:46:05 Take your notebook and pin out there and really look, look around fast. Take a lot of this.
Luiz 00:46:08 Yeah, I mean, yeah, I mean people have done these amazing things, but what I meant is that we’re not having the, the neural recordings at the same time we’re recording all the behaviors, right? Yeah. And so, and yeah, I dunno, maybe one day the water that they drink leads to some net of, of sensors neural
Paul 00:46:29 Dust. Yep. Yeah.
Luiz 00:46:31 Some magic dust you throw their way and, and that sends signals back of all the neurons in their brain. And then we, we can have that and, and I dunno,
Paul 00:46:40 I don’t know. Do, do, I don’t know. Would we, would we even know what to do with that? Is, is the perennial question, right? So, right. But yeah, and that’s, that’s what, that’s what these things are getting at. That’s what, you know, the, the, like the topics in your book are getting at and these, these principles will, well, you know, maybe we could frame how to attack these problems using these, um, you know, like dynamical systems theory, which is already being, being done a lot. Um, yeah, of course. And these other topics. Anyway, I I read that, uh, quote, uh, before because it had the word evolution in it. And one of the things that you do in the book is that you use evolution and comparative anatomy, uh, and changes in brain organization, not brain structure. You make the distinction between organization and structure in the book. Um, what, what is the difference between organization and structure, and why do we need to think about organization instead of structure?
Luiz 00:47:33 Yeah, because I do think, going back to the, the themes that we’ve been talking about, I do think that the fun in the functional, the functional landscape or the functional domain of things, we need to be thinking about these, these, these circuits. And so these circuits will come about with these, in my view, to a large extent with these large scale distributed systems that the underlying architecture either provides or doesn’t provide. Let’s say there are several, uh, basal ganglia loops of certain kind in some species. Uh, let’s say, uh, amphibians fishes, uh, reptiles have loops through the basal ganglia that go through the ventral striatum. And only birds and, and mammals have, in addition to those loops, have another set of loops that go through the dorsal striatum, the dorsal parts of the striatum. So this is something that we don’t know yet, because we don’t even, we haven’t been investigating this computationally and with the comp, with the, with the com compar comparative anatomy and comparative physiology, which is, there’s so little work on that obviously cuz of the, the challenges.
Luiz 00:48:48 So we don’t know exactly what these structural components are enabling functionally, but it gives us clues that it’s allowing novel functional repertoires or, or, or, or ways of processing that are not available in other species, in other taxonomic, taxonomic groups. So amphibians don’t have that, so they will not be benefiting from this type of organization that, that mammals have, that mammals, ma mammals and birds both exhibit these, these kinds of loops. Or for instance, we know and have been learning a lot about how the thalamus works in such close proximity and in with cortex, it’s Thor cortical processing is essential. It’s something that has been known for quite some time, but it’s been more and more well characterized and understood. And so these Thor cortical circuits, for instance, are something that is a property, again, not of all vertebrates. So understanding these kinds of structural components in their, in their, their relationships and the pathways allows us to understand what kinds of functional organization they would they end up enabling or supporting. I mean, that’s, that’s, I i, if, if that’s the sense that you were refrain to organization and structure
Paul 00:50:17 Yeah, yeah. Well, structure’s just basic anatomy. Right. But like, they’re, so we’re, I’m gonna step through in a moment. Um, you go through five principles of organization, and maybe we can just say a word about each of them, but you, you said thalamus, I thought you were gonna go, uh, basal ganglia loops and looking at you amygdala loops, um, which we’re learning more about and the comparative, um, anatomy and the evolutionary and the development, um, of, uh, amygdala. You know, looking back, you talked about their specifically about amygdala, how, you know, their, because it develops from different parts. Um, we can think of part of the amygdala is closer to cortex, part of it is closer to a subcortical structure. So calling it the amygdala is kind of confusing in that respect. And these are based on the organizational principles of the loops that are coming through in the connections with other areas,
Luiz 00:51:07 Right? Right,
Paul 00:51:08 Mm-hmm. <affirmative>, but, okay. So maybe, maybe we could step through, uh, some of the principles, if you don’t mind, we could step through some of the principles Yeah. Of organization that you write about in the book. Um, and I, I, you know, mainly I wanted to mention the, uh, the loops, uh, since you mentioned thalamus, because you go into some detail in the book. Um, if people wanna learn more, um, about, you know, the, the implications for the different, you know, loops and structures, um, in the brain in that organization. Principle one. Go ahead. Did you read? Okay, so I thought, I thought you can, yeah, go ahead.
Luiz 00:51:42 Okay. So principle
Paul 00:51:43 One. No, no, I was, I’ll read it. I’ll read you don’t, you don’t have to memorize. I, i I gotta right in front of me. I I can do it. I wasn’t, I’m not testing you here,
Luiz 00:51:50 <laugh>. No, go ahead. No, no. Ok, go ahead. <laugh>,
Paul 00:51:55 That’s staying in probably pr uh, principle one, massive combinatorial anatomical connectivity. So, and, and, and I’ll just ask you to say a word about each of these as we go through.
Luiz 00:52:05 Yeah, sure. So I think one of the things that struck me incredibly, so around 15 years or so, I really started, uh, studying anatomy quite a lot, quite a bit in anatomy of, of vertebrates, uh, across taxonomic groups, not just let’s say humans or primates or, or just, or even mammals. So one of the things that really struck me and continues to grow with this computational analysis of connect homes and other kinds of things, I, is really this, this, um, incredible massive backbone of connectivity that links in a very structured way. So it’s not everything connects to all by any means. So it’s extremely structured, but in nonetheless this enormous backbone of connectivity that creates this, what I, what I was just terming this kind of combinatorial ability of linking A to B via multiple ways, directly or indirectly in all sorts of ways that I think is so important because it really helps us understand how we, because the connectivity is so massive, it helps us understand that we need to go from moving away from a space of anatomical connectivity to a space of functional relationships.
Luiz 00:53:34 Hmm. Because doing my research and, and interacting with, with a lot of peoples in the, in the area of emotion and cognition and things like that, there was always a lot of discussion that that to me sounded a little confusing because for instance, the amygdala doesn’t project very strongly parts of the, the, the, the basal lateral amygdala and parts of the amygdala that connect to prefrontal cortex do not project very strongly at all to, let’s say dorsal lateral prefrontal cortex. So that worked perfectly because in a, in a way, emotion is in the amygdala, cognition isn’t pre dorsal frontal cortex and they don’t even talk to each other very much. But to me, because of this kind of anatomical properties, why does it have to be a direct connection? Okay? A direct connection is very powerful. No one’s gonna deny that, but nonetheless, no one can drive from here to New York City in some straight line, and it gets me there a little bit quicker, but I can also go Philadelphia and do what have you and stop and, and, and, and, and, and, and do something, something different, uh, on the way than, and it still get to the same place.
Luiz 00:54:44 So there are these multiple ways of signals propagating that I think we get, have to become a little bit more familiar with this way of thinking that it’s not that everything has a chance to influence everything at the same time and equally well, but they’re ample opportunities of these signals to crisscross and mutually influence each other in many more ways than a direct inspection of the only strongest connections would indicate. So the first principle being this massive anatomical connectivity, but really having in mind another of the principles, which I think might be number two, I can’t remember the number, but it really is going into a functional space of relationships of these signals. You have these signals in the back of the brain, in the front of the brain, perhaps there’s no single one direct connection and there isn’t. But nonetheless, they’re functionally coupled in, in ways that have to do with the current behaviors that an animal is doing at that time.
Luiz 00:55:47 And I think that’s the space that we need to move to. And I think, I feel sometimes that people are a little too bound by the, by the anatomy and for instance, they, they have studies that showing, oh, functional connectivity in the resting state is really well predicted by the structural connectome. Okay, maybe that’s true, but, and not gonna deny that it, it’s going constrain and impact it. But I think that we have to understand how it goes away from that basic set of functional relationships that are just driven by the sort of the direct pathways, the most strong pathways or the, the ones that are in the more basic context, the more context independent ones that a person is there and, and, and, and, and not doing some active behavior in a scanner. Just to give one an arbitrary example, so I, I view these two principles as really highly related, and I like to emphasize the, the connectivity one because at least in, in, in, in, in my way of thinking has really influenced how I started to think and solidify some of these ideas was by taking the anatomy seriously. If the anatomy is like this and we’re revealing all these principles of connectivity, of, of anatomical connectivity, what are the functional consequences?
Paul 00:57:04 Yeah. So, so you tell the story of, uh, small world networks, um, and then relate that within a network, neuroscience pers per um, perspective, right? Uh, and, and you, it used to be thought, uh, well, because, you know, small world, oh, that’s optimal for communication. And it used to be thought you make the point, um, in the book that the brain is actually, um, connected, um, anatomically in a higher than small world, um, order of magnitude. Yeah. Uh, and then so what I want to ask you then, you know, going back to that, um, combinatorial anatomical connectivity is, you know, well, what does that do for us, right? And is that where we shift into the functional connectivity story? Is, is that an advantage of being more highly, um, more hi higher than small world network, more highly connected than small world network? Is, is that a, an, um, an easy way to shift into thinking about that, the importance of the functional connectivity?
Luiz 00:57:59 Yeah, I think, I think, I think that small world connectivity, the, the, the small world, world organization would already give a lot of those kinds of properties because I think that’s why it was so fundamental when Olaf’s Bornes and many others started suggesting this in the early two thousands. I think it was really fundamental because it really shows that even with a regular structure, with a few random long-range connections, you start getting properties that all of a sudden emerge in this system that you would not have. And you, you get it from relatively few long range random connections and which changes drastically the space of functional relationships that are possible in this kind of architecture. I think if you take that one step further, so I don’t have any problem with that, but I think that, again, it’s empirical, right? So that’s what we, we want it to be.
Luiz 00:58:53 We studied the brain and we see how does the architecture look? And in my view, my reading of the literature, it’s, there’s some, there’s some challenges and questions there because of the strengths of the connections and all sorts of nuances that we can’t ignore. But my reading of the literature is that it’s actually really more like a tiny world than a small world in the sense that it’s denser connectivity in cortex. There’s these highly structured large scale loops, basal ganglia, cerebellum, amygdala, mid-brain that compliment the, the cortical dense mesh of interconnections. So, I, I, what it, what I think it does is that it puts even more pressure, I think, on us to move to a space of functional relationships because the anatomy is such that you can integrate and you can, you can interact and integrate much more easily than we were. We, we thought because of these small world, or let’s say this is correct, and this tiny world hypothesis is, is flies.
Luiz 01:00:01 And, and so if, if it does, then, then we, we really need to, to, to shift spaces because anatomy supports many different kinds of functions because of the ways of communication. It’s highly efficient in ways that it wouldn’t if it was piecemeal. You know, I have to go to California via every single thing. It, it’s gonna, it’s a lot of steps that you have to take there. So it’s gonna take weeks, months, I don’t know, the, you know, when America went to the West, it was months, right? So it takes months. And if you have these shortcuts and these other ways of communicating, then it changes, changes things dramatically. Relationships can occur that would not be able to occur otherwise in space and time.
Paul 01:00:49 There’s time again, uh, principle three neurons as, uh, the main unit of the brain. Just kidding. Networks as functional units. <laugh>.
Luiz 01:01:00 Well, that’s a good point because not too long ago, neurons were the unit, and I think, yeah, I don’t know. I don’t know. I think that that one has probably died, but I’m not sure. Yeah. Yeah. But anyway, that’s discussion here today that I think that putting all these things together, I think that really speaks to our need now in neuroscience to focus on circuits. So the unit becomes a circuit, and what are the principles of the circuits that form, and why does, let’s say the fussy formm gyrus participate in certain circuits that support face recognition much more often than it participates in, perhaps, let’s say, than other circuits that support other kinds of things. So that we have to work those things out. That has to do with their place in this whole ana anatomical space and many other kinds of things. Obviously the, the sensory kinds of inputs that are more proximal and and whatnot. But that’s the kinds of ca questions that I think neuroscience should be focusing on.
Paul 01:02:03 Are, are you using circuits and networks synonymously in this, in this case?
Luiz 01:02:09 That is a problem that I, that’s a really good question because I’ve used that synonymously in the past, and a lot of people do not like it. And I think I probably should be a little bit more careful because I’ve interchange exchanged them, and a lot of people view networks almost like perhaps more traditional F M R I networks, F M R I, resting state networks.
Paul 01:02:36 Right? Right. I
Luiz 01:02:37 I never really thought of it that way because I, of course, it’s, it’s, it’s very commonly used
Paul 01:02:45 Circuits feels like closer to anatomical to me than networks.
Luiz 01:02:50 Right. Actually, in fact, someone recently in something that I submitted for publication, I think was saying that it was very confusing, and they, I think they said the same thing. Maybe you reviewed the paper that, that
Paul 01:03:03 Oh, yeah. They send them to me <laugh>.
Luiz 01:03:05 Because, because it, yeah. I think by using circuits, I think it semantically connected in this person’s mind, in this reviewer’s mind, perhaps something much more structural that I didn’t have in mind. So I, I don’t know what the best word is, then I just, I just don’t know because I was too casually using circuit very dynamically and very functionally, like these things forming dynamically as a circuit. Yeah. But if the readers, the listeners, everyone is, is interpreting this as something more, more anatomical and more rigid, then it’s a really bad way of de of, of, of referring to them and networks. Ha I have had this pushback too, because I think networks is just, people are thinking, oh, you’re thinking about, oh, and he also does some F M I research, so he’s always talking about these networks at these fmri resting state networks and, and some, right. Two thirds of neuroscience. Like who cares about FM r i research? And I, I, I didn’t mean it in that sense at all of fm r i networks of resting state networks. So I think my language there might have been really problematic. It’s throwing people completely off. But I do mean, I’ve been using networks and, and circuits more or less interchangeably and thinking of these highly dynamic modifiable, uh, overlapping entities, communities, um, coalitions, clusters,
Paul 01:04:44 Wonder if you could use motif
Luiz 01:04:46 Motif, perhaps. Yeah. Motif. Something like that. Yeah. So, but at motif, but it doesn’t even require something like a circuit motif. But I I if you say, um, circuit motif, people think that would almost, almost sound more structural to me, right. Circuit motif.
Paul 01:05:01 Well, I thought motif would give it a, give it more of a, um, functional flavor. Um, yeah. Yeah. Cause circuit sounds A connects to B, but then motif is like, well, there’s this family of them kind of or
Luiz 01:05:12 Something. Yeah. Yeah. I do. I I, I do think that maybe we need some ways of, of conveying that. That if these really, if there really are these things that form more dynamically, that, and this, this state, this, this in, in this time point the amygdala is, is, is functionally linked with several other regions, and that shifts dynamically in, in behavior, then I think we need to really use better words or, or, or, or qualify these words. And it could, that could, could be a little bit of explanation sometimes of the difficulty in communication and, and which happens in general with lots of these terms, right? Not to mention these complicated terms like complexity and emergence and what have you, <laugh>. So what exactly do we mean? Right? And, and even in this more basic sense, I think that you might have just spotted something that it can be confusing. Yeah. I should, I should be more careful in referring to these things, which I view as, as, as really very dynamic. Yeah. And, and malleable.
Paul 01:06:16 All right. Back on track, I suppose, principal, we have two more principles. Um, principle four, inter inter interactions via cortical subcortical loops. And, you know, this is harking back to the thalamus and cortico basal ganglia loops and the amygdala, uh, le loops. Yeah.
Luiz 01:06:29 So that, that’s really is, uh, uh, a principle that is, I could have ordered it slightly differently and linked it to the first principle, which is the, the backbone of, of, of connectivity. But I wanted to literally have a header there that a person would see separate in the piece of paper and, and, and be really drive home this need to be really thinking about the cortical subcortical, but even in a more general sense, not just the four brain subcortex, but everything below the cortex, mid-brain and hiin brain with cerebellum, that it’s really all this connectivity along the entire neuro axis that happens to be vertical in humans and, and horizontal and, and, and, and many other animals that we, we, that we really should be considering and not so much having the focus, which we tend to do, I think, especially in my part of the world, which is more human-oriented research of cortex, cortex, cortex, and, and, and in, in fact deriving what are the principles of, of the brain’s architecture, just from looking at cortical connectivity.
Luiz 01:07:42 I, I mean mm-hmm. <affirmative>, sometimes I mm-hmm. <affirmative>. I mean, I both admire people who do that. And also it drives me crazy because, well, we can’t isolate the cortex from all the other rest, the, the, all the rest. So how can we, how can we have even the courage of proposing that the brain architecture is, has these properties if we’re not taking into account all that other kind of structural arrangements? So I wanted to have that a separate heading, because I really think it helps organize our thinking and moving away from just maybe connectivity between any given level, just calling level here, you know, like a cortex, sub cortex, mid-brain, and just for convenience,
Paul 01:08:23 The eye roll <laugh>.
Luiz 01:08:25 Yeah. Right. So, so to break this, this, this, um, this, this tendency that we think mostly, you know, how does subcortical regions maybe are inter connected, which is a quite a lot of, of the animal work, right? The non-human animal research focuses a lot of on, on subcortical structures for historical reasons and many reasons. And then in human work tends to focus a lot on the, on the cortical regions. And, and so breaking a little bit this mode of thinking, I, I thought that I would have this as a separate principle.
Paul 01:08:53 So you, you describe, you know, the differences between open loops and closed loops, um, and give examples of these types of loops in the brain, which I like the idea of there being a closed loop as separate from an open loop, but everything also comes in degrees. And would it be correct to say that as we learn, the more we learn about anatomical connectivity, the more everything is actually an open loop?
Luiz 01:09:19 No, I don’t think necessarily. So, because I mean that, that could be the way that things turn out empirically. But I do think that with newer methods, right, we’re always evolving with methods, newer methods and new ways of doing these anatomical studies and just redoing them is, is incredibly important because we really have had, we haven’t had amazing breakthroughs in, in, in doing structural anatomy that have completely revolutionized things that compared to, to, uh, chemical injections and other kinds of more standard, um, approaches that have been used for many decades. And, and so we are really still learning an incredible amount that we forget sometimes that, uh, about that. But we’re still really learning a lot about the, the architecture itself, the anatomy itself. So it, it really could come to having certain kinds of loops that happen to be more closed, more in, in a closed loop type of, of, of, of organization in which the initial region projecting it’s the, is the region that eventually receives back the, the, the, the return, if you will, projections versus something that has a quite a, a little bit more, or quite a lot more of spread indicating that I, I would support different kinds of functional relationships, right?
Luiz 01:10:52 That, so that, that, that the segregation is clearly something that would be more functional. Segregation would be clearly favored by more parallel isolated circuits, parallel circuits than circuits that crosstalk. And because the emphasis in the past has been a little bit more about their relative independence in the loops, I’ve tried to emphasize in the book that we have been learning quite a bit about more, the more interactive and overlapping nature of these loops than it we thought perhaps a decade or two ago. And so appreciating that is, I think is very important. But I do think it’s, it’s something, it has to be the empirical outcome that we find out, okay, well, we thought was more distributed. It was because we had some markers that were diffusing in ways that we didn’t understand. And it’s not so open loop at, after all, it’s actually much more closed loop. Okay, great. So that’s, that’s the organization, and then we have to move forward with that, with the understanding that we have at the time.
Paul 01:11:59 But just because something is closed loop, so at first blush, you might think, well, that argues more strongly for a modular view of the brain. But, um, just because a loop is closed, it’s not so synonymous with modularity because there’s a lot of other interactions going on, um, that are not part of the loop.
Luiz 01:12:14 Agree. Cause yeah, that’s interesting thing is that there, there, if we, we have this habit of doing these basal ganglia loops and these kind of other, other loops, we have this habit, it’s a loop, right? But we, we have this habit of starting it with in the cortex, which is right
Paul 01:12:28 <laugh>,
Luiz 01:12:29 But it’s a loop, right? So we could have stayed, we start somewhere else, but let’s do the same thing and start at the cortex. And, and so if the, if the, if the loop was perfectly closed, right onto this exact same neuron, let’s say <laugh>, if the loop is perfectly closed, as you said, it doesn’t mean modular. Because if that same neuron, again, if we’re gonna stick to a single neuron, is actually connected, uh, in, in, in many ways it’s, it, it actually has the room for much more distributed processing and non modular processing. Even though these loops are closed, the closed loops tend to favor a little bit more segregated style of processing, I would say. But it doesn’t preclude, it constrains more, I think the kind of interaction of the signals. I do think that, I mean, I think that’s the overall way that I think about anatomy in enabling and constraining the types of relationships that are more natural.
Luiz 01:13:24 Because if I need to do something within a few seconds, it it, this, I can’t wait minutes for this thing to percolate and, and, and, and go 17 times around every single place in the brain. So there, obviously, there are temporal constraints that in behavior, thankfully, it’s not all the time that animals and humans need to engage in something that is so, such time pressure. But there are those occasions where there are these, these, these kinds of, uh, behavioral requirements in time that have to be executed quite rapidly. You step on a break and you have to stop immediately or catch something or what have you, that these are, are many behaviors, not to mention more survival related behaviors that we watch on, on YouTube or what have you, that, that are, that are incredibly time sensitive.
Paul 01:14:23 D does an open loop indicate higher complexity? Or does it indicate a higher decoupling between, uh, our senses and motor actions? Not necessarily. I suppose
Luiz 01:14:39 I, I would, I mean, again, at the first pass, I would have to think more carefully, but I think, I think I, I think that more open loops give more opportunities for more crosstalk, right? So for instance, if there’s a region that, for whatever reason, because of its receptor density, because of history of learning and other kinds of things, is, is very, uh, important for reward related mechanisms, and this is part of some loop that is, is is very open. It, it can be influenced by many things, and it can influence many other things. So it, it, it opens up the space of relationships. I think I, I, I, that’s my first inclination is to think that way. And so at least even in a rapid cycle of things, it has the ability to, to incorporate more things, right? Because it, it, it might be receiving signals from more diverse places that are more sens that are sensitive to different aspects of the external and in internal worlds.
Luiz 01:15:40 And, and, and thereby having, supporting more behavioral flexibility. Because if you’re just sensitive to a certain class of signal, then I’m more tied to that class of signal. But if I can have more signals that, let’s say just for example, a more internal world and a more external world, simultaneously, both worlds can, okay, do I go and fight well, but I was just injured, so it’s not a good idea, right? So I was just attacked and this pride of lions and, and your, you’ve been attacked and your, your leg is injured. The best behavior right now is to crouch in whatever do behaviorally that is better as opposed to striking back. And, and I think that these multiple elements of internal and external worlds, having the ability to influence behavior, the open loops, my first inclination is to think that they’d support the faster versions of these kinds of relationships and these behaviors.
Paul 01:16:43 Okay. Louise, principle five, and then we’ll move on to, uh, some other topics here. Um, and I’m curious, you know, whether you felt, if this one was just tacked on, but I know that this is, this is near and dear to your heart also. Okay. Principle five, connectivity with the body,
Luiz 01:17:00 Right? So connectivity with the body. I think that one of the weaknesses, to be honest of the book is literally that I didn’t develop as much the body into the book as I, I in, and I, looking back, ideally, I, I could have, I should have, because I do think that brains being embodied is fundamental. Like a lot of people have been talking about this for decades and even more so in the past five, 10 years. So obviously there’s no originality whatsoever there. I’m just saying the same thing that everyone is saying that a lot of people have been saying. So, but I do think it’s really a fundamental principle. And if later we get to talk about AI and, and certain things that AI needs to do, I, I would bring back the body and, and, but that’s, that’s for later maybe mm-hmm. <affirmative>. But I do think that this, this close link of brain and body is really fundamental.
Luiz 01:17:55 And I think that because I wanted to have a book short, but I also, the book was for a general audience, and I have professionally to get grants and papers and whatnot, I have to put an end to the project. I think I, you know, in a perfect world, I would have had a lot more on the body in really highlighting this aspect, which I think is fundamental. So in that sense, it might have felt, like you said, was it tacked on? So in that sense, it might have been appearing that way because of the lack of depth and integration of these ideas throughout the book. But I do think it’s really fundamental. Well, I really, I, I, I really think that we are missing a lot by, by considering this as a disembodied organs brain that is just, um, a brain and that is not closely connected in communicating with, with a, with an acting body and in, in this embodiment.
Paul 01:18:55 Yeah. Yeah. When you talk about interception, but that also just links with behavior, which you stress, you know, throughout the book as, you know, that’s, that’s what it’s all about anyways, connecting brain and behavior.
Luiz 01:19:05 Yeah. Right. Exactly. Um,
Paul 01:19:09 Alright. We are, we got through the five principles. Oh, so I want to, um, just ask, um, the way that we’ve been talking is in terms of which area, which regions are connected to, which other regions and the, and the, the functional motifs, Sam, now I’ve gotta use language and it’s awful, right? Um, but, uh, but then, but this is all like with neuronal, um, connections and areas. Um, you know, do you, when you think about these things, do you also think in terms of neuromodulators and con, you know, within the context sensitivity astrocytes and, and those sorts of system level properties in, in combination with these organ, uh, principles of organization, it’s, and so on.
Luiz 01:19:53 Yeah, I, I, I, I do think in term in those terms, I think it reflects my own lack of sufficient amount of knowledge that I don’t have a more comprehensive treatment in, in terms broader terms, including astrocytes and, and, and more in-depth discussions of, of neurotransmitter systems and their interactions and their combinations. And the way that, that, that, that’s something that I, that I could have explored and, and included in the book. But I, I do think that it’s, it’s, it’s, it’s really, it is really critical and, and integral to understanding these systems is that these systems exist in an architecture that is made even more flexible, even more powerful given this, these modulations that these, these these systems afford, provide the system in ways that can engage large, large, large parts of the brain can be more focal sometimes. And, and, and so it’s really in incredibly central to the, to the story. I think. I, I think that that’s something that if I, you know, some, uh, someone, I can’t not envision this ever happening, but if they’re an entangled brain version two, uh, then that’s the direction that it should grow and develop and, and, and work out and in what ways these, these, these aspects need to be integrated. I think that’s, that’s really important.
Paul 01:21:28 So I, I don’t want to make us go down a long path here, but you are very interested in connecting brain with behavior, and you argue about the complexity, uh, of behavior and needing to track that. Um, and when I got into neuroscience, I was really interested in consciousness, whatever mm-hmm. <affirmative> the hell that is, right? Mm-hmm. <affirmative>. Um, and, and, you know, thinking and our mental world es essentially our subjective mental world. And I’m wondering if you feel like the, you know, these principles of organization and these principles of, um, uh, complexity and so on, uh, can we extend those, you know, well, well, here’s the question. Has this changed or, um, made you think about our mental world in, in any different way? Or do you feel like these principles can, um, apply to the mental and we can start to get a grasp on things that aren’t quote unquote behavior?
Luiz 01:22:24 Yeah, I do, I do think I, I, I think that’s clearly a part of the space that I don’t explore in my thinking in general. It’s something that I, that I hear and that I read sometimes. And, and a lot of people are interested, a lot of people bring like language and other kinds of, other kinds of, um, yeah. Human capabilities, at least with this level of how language is developed, uh, the how developed language is and, and whatnot. Clearly, clearly, clearly is clearly the case. It’s something that I don’t bring into, into focus in my thinking or, or my, my focus of research. But, but it’s something that I do think it, it, it’s closely related. It has to be closely related to, to the aspects that we were discussing before, essentially in a very loose sense and, and and kind of hand wavy way is, is, is there are all these, these higher order interactions that, that, that, that could afford some mental world and event and, and, and space that, that i, that needs to be supported in some way, cuz it, it exists.
Luiz 01:23:40 And so it need to make charges along, along that direction. I often, like you said, in many of papers, I, I bring these examples from threat processing and other kinds of more basic forms of behavior, even if they’re highly, highly complex, might involve certain kinds of, at least in the scenarios like foraging scenarios and how they predator and prey and whatnot. In some of the discussions that I’ve, that I’ve, that I’ve written, but, and doesn’t approximate the type of many of the behaviors that many investigators and researchers are really interested in. Right? So that, that gets to, to your question where these, these mental world of imagination, of planning of these levels upon levels of further decoupling of sensory stimulation and, and current state of the body and, and the, the mental, the content of, of, of some mental augmentation. And, and so it’s, it’s something that I, that I don’t have, uh, that I haven’t investigated or thought deeply enough. Um, my means.
Paul 01:24:53 So let, let’s move on to talk about, and this is kind of related, um, emotion. Um, although you, you know, you are have been interested in the, um, the connection between cognition and emotion, but I, I suppose, you know, you mainly focus on the, is related to emotion. Let’s talk about emotion and, uh, and then I wanna bring in AI before we get to some extra time for Patreon listeners. Um, I, I just want your take on how, um, the concept of emotion has sort of historically evolved, right? Um, a as we’ve learned more, you know, so it used to be emotion is in the amygdala or subcortex, right? Or something like that. And now we have a much more nuanced view of it. So how have we come to appreciate emotion in the modern sense based on its historical context?
Luiz 01:25:40 Yeah, so I think that I, I think that’s a, that’s, that’s a really important, important question or theme, which is historically, so one of the central ideas in, in, in emotion research pertaining to the brain is, is the, the, the existence that the suggestion of an existence of a, a specialized circuit for emotion and even more so than a specialized circuit circuit, uh, let’s say a more structural circuit, a more structure called define and constrained circuit for emotion. Not only that was, was a major theme, but the presence, or the, the existence I should say, of a specific brain region, let’s say the hypothalamus as the centerpiece of this little structural, uh, circuit for emotion. And so historically, that has been something that has always been emphasized. And I think to a good extent, it’s still informs a lot of the debate that there is specificity in the structural circuit, a set of regions that are involved.
Luiz 01:26:46 And there are some really heavy players, hypothalamus amygdala and, and so on, peria, ductal gray and some other favorites of, uh, depending on the focus of a research group. But so in the animal world, there, there’s still, in the, in the, in the animal research, non-human animal research, there’s still a lot of emphasis on, on on these players. But I think in the past two decades or so with research in humans, F M R I to a good extent, but I think more broadly even conceived, it really has informed perhaps strongly, a strongly initial push was, uh, Antonio Damasio’s, uh, decartes error to, to really suggest that these systems at the very least are, are interacting with other things like perception and cognition in really important ways that, for instance, decision making is a cognitive emotional process. Like let’s say somatic mar somatic marker hypothesis by the maan colleagues and so on.
Luiz 01:27:51 So, so I think that really broadened the scope of thinking about emotion in the brain, the emotional brain in the past two decades. And I think that that is permeating even a little bit into the rodent literature. For instance, a lot of investigators have been, now, some investigators I should say are now, uh, looking at the role of the anterior insular and right, I should say that in, in, in the, in the, in the, in the rodent, the roles of the medial prefrontal cortex, that part of cortex was, has been there for quite some time, but not a broader, uh, view of, of, of more players as perhaps in the case of the human research. So I guess, I guess in, in, in, in summary, I think what, what we’re, I think we are observing is that it’s becoming more natural to, for people to think of at the very least, more interactions between these cognitive emotional systems, cognitive action systems, cognitive perceptual systems.
Luiz 01:28:57 I mean, as you know, my push is to kind of dissolve these boundaries or heavily blur at the very, very least heavily blur these boundaries. But I think that there, there, there, there’s a, a, a wider appreciation of this, of the, at the very least, of the interactions between these systems. So I think emotion research has really evolved tremendously in the past 30 years or 25 years in that direction that the circuits are becoming really hard to, to, to, to you. You look at these reviews and the circuits now have a dozen or two dozen regions. And if you, if we look at it in, if we, if we actually be interesting, the number of regions per year, per decade, you know, plot in a, in a graph, and people were pro probably looking at two or three regions and, and now they’re, they’re, they’re showing reviews with, with really complex, really fascinating set of regions interacting in complex ways for fear learning, fear extinction and, and their relationships and whatnot. Uh, Steve, me has a, a review paper is, is really fascinating and it’s like a savvy page, uh, uh, review. And it has these very large, these growing circuits now that are becoming more mainstream in, in, in neuroscience. So I think that it’s, I think it’s, it’s really important that, that this direction of it’s not under it is isolated units anymore. It’s, it’s, um, the opposite of isolating, right? It’s, it is bringing together with the rest. And I think that it’s, it’s,
Paul 01:30:31 We, we still, we still draw it though as, as kind of units connected to units, you know? And, uh, is there a way around that? I mean, I, we do, so you, you’ve written a paper where you, you have kind of, you, you’ve diagrammed out a functional emotional, uh, integrative kind of circuit, and that’s not so much the boxing arrow, um, version.
Luiz 01:30:48 So yeah, I think that, I think another thing that you were speaking about learning how to use language in different ways or learning to do science in a way that it’s, it’s more or less it’s, it’s less context dependent. I think another informed direction for us as a field and perhaps other fields is to learn to do science that is less boxes and arrows, right? <laugh>. So somehow if we could, cuz we have to draw these and I, if we let go draw, so we wish to draw things, I think that would be a huge advance already. And, and not only we have to draw things, but sometimes we have to put labels because we’re just trying to summarize, we’re trying to do dimensional dimensionality reduction on the paper, right? And so we put one word there, but then if we do that one word, we ruined everything, right?
Luiz 01:31:36 Because okay, that’s not what we meant. Or maybe we did mean that it was just that word, but is that what you really meant? Or you meant, I don’t know. What did you mean exactly? So, but I think that dimensionality reduction is so extreme that how do we go, I mean, we can’t have a plot with a cloud of a, a word cloud around surrounding and rotating and spinning, I dunno, but, so yeah. But is is that a problem? I sometimes I do think so that I, I just don’t want to even do a figure because it, it is just going to ruin the whole idea.
Paul 01:32:12 It limits the concept conveyed. Yeah.
Luiz 01:32:14 Yeah. And we, that’s the, the concept that we’re conveying. And obviously we’re not gonna, perhaps there are other ways that are even available now, but for a regular small lab completely beyond anyone’s means to generate a movie to explain some of these ideas in some complex computer graphics manipulation,
Paul 01:32:35 3D glasses. Yeah. The IMAX movies for every paper. Yeah.
Luiz 01:32:39 Some visualization that is super fancy, perhaps that even exists. But the, we don’t, that science is not like that yet, perhaps. Yeah. It needs to be a little bit more,
Paul 01:32:49 Maybe when we develop, uh, agi, they can just, we can just tell them to do that for us. Right. But will they have, uh, emotions? So you’ve written, you’ve written that there’s a need for, uh, artificial intelligence to incorporate emotions. How is that for a segue by the way? That’s pretty good. That is,
Luiz 01:33:06 Yeah, that is a good one. That’s a perfect one. That’s, I I do think that if, if we want to, of course we can have, we can have a computer science, we can have, um, artificial intelligence that is completely different from how humans do things. It’s, it’s, it’s fine, right? So we can, we do, um, obviously complex, um, mathematics that is, is not just how we, or, or logic or what have you. It’s not the way we reason. It’s, it’s, it’s, it’s, it’s completely fine. It’s, it’s not, there’s nothing inconsistent with, with, uh, trying ways of formalizing knowledge systems or, or, or artificial artificial intelligence in ways that are completely different. But if we want to solve problems that are similar to what embodied animals solve in intelligent ways, I think we need to solve a couple of problems that bring us to emotion. And, and here’s my way of thinking. My way of thinking is that like, let’s say we think to make this concrete, and this how it helps me is to think about a robot, right? So a brain, a robot has a brain and it moves around. And so if it does that, it needs to, we would like to send to perhaps Mars is not a good analogy anymore because
Paul 01:34:39 <laugh>
Luiz 01:34:39 Of things that I don’t even want to get into, but we send, we wanna send some, some expedition, some, some, uh, spacecraft to another planet or some, some other
Paul 01:34:52 Something. Yeah.
Luiz 01:34:53 Right? So we need to solve the, it needs to be an autonomous agent, right? It needs to be, it needs to survive to be autonomous. So, so I think the problem to be solved for the kinds of intelligence that we engage in and other animals engage in, there might be other kinds of intelligence that are fascinating as well. But these kinds of intelligence really requires us to solve the, the, the behavioral autonomy problem, which is how do you build a system that is completely autonomous and can behave flexibly and intelligently in complex, unpredictable worlds so that the, the rules are completely, uh, they’re flexible enough that they, they pose, uh, sufficient challenges to, to an agent. So in, in that setting, I think that we really need to have agents that can have both motivation and emotion in the sense of learning what is of biological significance in that case.
Luiz 01:36:06 Obviously not biological so, but what we call biological significance. In that case for agent, it’s, it’s, it’s significant to the agent that it needs to conserve some kind of energy because it’s battery will run out and it needs to find some source of what have you, some source of some nutrient or I don’t know what. And, and it needs to navigate in certain ways. So certain things have value in, in that, in that way because of its the integrity of the agent that will not be maintained if it does not consider those factors. So I think emotion and motivation come very naturally to organize behavior. Even behavior that is taught to be historically has been taught to be independent of these factors, not necessarily perception planning. And, and, and in general, problem solving have to incorporate these factors. And I think these factors are incorporated in ways that we call emotion and, and a emotion and motivation in, in biological systems.
Luiz 01:37:15 And so that’s, well, you, yeah, fine label, you can call it whatever we want. And it’s, it’s, it’s a label and it’s, I think it’s a, it’s a really way of summarizing the existential and survivals in the, in the survival sense of agents that exist in a, in a, in a flexible un unpredictable world of limited resources, right? So you can’t just, um, turn yourself off and beams some new energy source and, and then you turn yourself back on, it turns yourself, it turns you, it turns you on back on. It’s something that in the, in these kinds of scenarios, I think it has to be solved perhaps to navigate worlds that are not limited by factors such as space and time in ways that biology faces and limited resources, perhaps the games kind of change. And so maybe we, we can’t, we can’t get by with many kind of other architectures, and maybe it’s will surprise us that we can get a lot more done by just brute force, right?
Luiz 01:38:24 We have such astronomical computational power now compared to what we had 50 years ago. I mean, what was gonna happen in 50 years. So if we have, a lot of the problems that we’ve been solving in many ways are largely driven by computational power, right? It’s like brute force, brute force in the sense that we have, I don’t know, have a trillion parameters and we have unlimited energy and we have, and we have time in our hands, but, and so maybe that’s gonna be available. And if that’s available, then we don’t need a solution that is, if functions on whatever the, the, I have no idea if it’s, it’s, it’s a, a reliable estimate of whatever number of watts the brain consumes per unit of time.
Paul 01:39:14 It’s exactly 20. Yeah,
Luiz 01:39:15 Exactly 20, right? So it is not gonna work on, on that kind of setting, but who cares if that’s not the setting in which the things are operating. But I think that in many problems that involve autonomy, that will have to be part of the solution, or, or at the very least, that would be really useful, a, a really useful part of the solution, I believe, at least my, my my my, my take on this,
Paul 01:39:48 One of the things that you just mentioned a moment ago is, you know, it doesn’t really matter what you call it, it’s a label, right? Um, and let’s say the word emotion, maybe what we should have said, sort of, uh, you know, up ahead is we’ve been talking about how the modular view of the brain should sort of dissipate and eventually, um, allow this other way of thinking that we’ve been, um, discussing here. But, you know, are we still u comfortable using the word emotion? Because that word itself has a conceptual space that historically, at least in western culture is associated with things like weakness or, uh, frivolity or non-use usefulness. But as we learn more about, you know, let’s say the emotional circuit motifs and the integration of emotion with cognition and its importance, does that change our conception of emotion enough? Um, you know, or do we, you know, those boundaries of concepts, do they need to be kind of blurred as
Luiz 01:40:43 Well? I think, I think that’s a really great point. I really like that because I think it has been changing, right? So I think that, you know, when you started saying this, I, it took me a, a fraction of a second to up update things. And so yeah, of course, you’re right. Cuz historically that’s how emotion has been conceptualized in we literature and, and society. And, but I do think that in the past 30, 40 years, I guess it’s let’s say 30 years with, uh, uh, uh, books by, by Joel Ledou and by the Maio and by many others, I do think that it’s actually going in it, it could be going, I, I’m, I’m not sure. That’s a great question actually. I think it could be going in a way that is, is changing the way that we see it and, and, and see it in a way that, for instance, if we take some of the, let’s say, just say that the masu ideas that is integral to decision making, so it, it becomes the connotation becomes very different, right?
Luiz 01:41:39 And so we start thinking of it as, as an intelligent thing to do. And so if you were, if you were to eliminate that, that factor that is so important, determining, am I gonna go all the way to the West Boston to do a PhD and I’m gonna leave all my family, all my friends here back in this other country. I mean, the flight is super long as I am a graduate student, have no money. That’s not gonna, I’m maximum one time a year. I’m gonna be able to see, I mean, how do you, the decision there is is humans, at least it’s highly complex, right? It’s not just, oh, I’m gonna increase my knowledge of artificial neuro networks in computational neuroscience. It, it’s, so that way of deciding, insofar with as we think that it is used, it, it, it’s improved some kinds of decision making. Obviously some kinds of decision making are not improved
Paul 01:42:32 By Right? Right.
Luiz 01:42:34 Humans, how humans fail miserably, I think is pretty obvious in, in ways of using types of decision making that, that are influencing by, by factors that are clearly detrimental. At least if we think of, you know, the greater, some greater good of all the, the beings on the planet as opposed to a sub, a small subgroup. But, but if we think that that is beneficial in some context, then I, I do think it sh it shifts the boundaries of the concepts themselves. And we start thinking of, of, of start valuing people who have I’m, I don’t know, call it emo, I don’t even know exactly how it’s conceptualize, but call it emotional intelligence or what have you. I don’t, you know, so I
Paul 01:43:14 I What about emotional train wrecks? What should we do with those people? <laugh>?
Luiz 01:43:19 Yeah, well, yeah, exactly. So it’s, it’s, it’s messy
Paul 01:43:25 <laugh>, we talked about emotion and, um, it’s, um, usefulness in artificial systems. But what about the other kind of principles that we’ve been talking about that you write about in your book? You know, these, do you think that the ar ai world, let’s say, is appreciating these kinds of principles that are being more appreciated and more discovered in our brains? You know, the interactions between systems, the interactions, the loops, and the openness and closedness of loops? I mean, are, do these principles need to be built into ai? And do you think they’re being appreciated enough?
Luiz 01:44:01 I do. I, I, I don’t think they’re appreciated enough. And I, I think it’s fine. I, I mean, well, I, I should, I should put it this way, I think, I mean, we have finite resources, and so if you want to stick to one style of computation, if you will, it’s fine. Mm-hmm. <affirmative>, let’s build things that have these abstract parameterizations and transformers and whatnot that we can build things and, and understand the underlying mathematics eventually and, and, and do things or what have you. You can, you can have, you can have one style of doing things. It, it’s, it’s, I think it’s fine. But I think that what I find very interesting is that some groups are starting to pay attention. You know, one group that always comes to mind is a large group. So I’m not, don’t need to name any specific, it’s a, in Montreal to me, comes, pops to mind as a, a hub of people who kind of integrating these ideas much more strongly than perhaps, I mean, as an outsider.
Luiz 01:45:01 So I could be wrong, but it feels to me that I’m always hearing something from, from those groups, so many, many groups that becomes a hub for thinking about how we can, how can neuroscience kind of give us some architectural ideas or themes or motifs that that can, can be interesting. So, okay, so I don’t have just one recurrent neural network, one R n N, but I have multiple R and ns and okay, now I have to work out how they’re gonna interact. And, and so, so I think it opens the space of architectures in ways that I as a completely, uh, a fictional about the brain find fascinating, because I wanna understand in what ways maybe some of these principles could, could open up ways of thinking about distributed processing computation more generally. And I think that it, it could turn out that, that, that they turn out to be very beneficial for ai for general, for let’s say general intelligence.
Luiz 01:46:06 And it, it could, right? So I, I just don’t know because if you, it could be the system, the system could benefit from some of these things, but only modestly, unless we take it to a much more right. Further elements of the architecture and perhaps the embodiment and, and even the fact that that animals have a, a long de mammals have a really long development phase. They spend years, uh, many, it’s not just humans that they can’t, they wouldn’t be able to survive on their own by any means. So it’s, it’s a really, ty a really different kind of solution to this problem, right? Have a, create something that depends on supervision, not in the supervision supervised learning sense, but on sup on, on someone literally taking care of it for, for years. So what if that is part of a, a, an a really important type of solution that, that that evolved.
Luiz 01:47:07 And so we don’t know, but I think that, yeah, to me, as a basic scientist and my, with my origins in computer science, I, I, I’d be fascinating if we had more space to investigate those things too. That would be, that I find extremely attractive that that type of scenario, just to understand, not necessarily, I, I’m, I’m a little more skeptical about, I’m, to put it mildly, sometimes I’m, I’m more skeptical about the ways in which these architectures will inform us about the brain, but I view them more as abstract architectures that we would gain a lot to understand, even from a computer science or call it whatever discipline. I mean, again, names of disciplines are, that’s one thing that they’ve Sure. Yep. <laugh> useless, dissolved everything. Yeah. Yeah, exactly. I mean, I, I was never, I was in the computer science department and I liked biology.
Luiz 01:48:03 I mean, I’m psychology department. I didn’t have a biology a a psychology class. And so yeah, it’s, for me, it doesn’t work. But, but, but I, I think it would be fa it would be so fascinating to, to, to, to, to study these kinds of things. And if they can inform the brain, because inform our understanding of the brain so much, the better. I think that that, that I’m open to that possibility. Yep. As well, as long as it’s a real collaboration, right? People are really interested in, in, in, in really understanding these systems and real collab, really real interdisciplinary collaboration, multidisciplinary. So I, I really want to understand the biology and the behavior, and I always, I I also want to understand the computational principles, distributed processing, parallel distributed processing types of architectures and how these things communicate, interact with each other at a, at, at, at, at a, at a, at a, at a high level, at a, at, um, at a broader level, conceptual level, but with the details of the disciplines, right?
Luiz 01:49:11 It’s, it’s, it’s the same that if, if someone gets into this and, and doesn’t know underlying, um, computer architecture and distributed systems, doesn’t know that a little separate from ai, I’m just talking more, even more broadly now. Yep. So I think that, that these two collaborations, I think would be, would be fascinating. So I think there’s a lot of room for, for those. And I think that, so I mean, just literally yesterday I bumped into a paper of saying what happens to, I don’t even know what it exactly means, I have to read it. It’s like, what happens with, um, what happens to a spatially constrained, I don’t know if you saw it, it was literally a spatially constrained r and n. And so what emerges, what, let’s use another word, what happens to a, a spatially constrained r and n recurrent neural network when it literally is spatially constrained. So it’s insights, it’s not just an abstract. Oh, so I, I literally have, I I, I was like, okay, I’m, I bookmarked it and I want to read it. I wanna see what it is. So if you start putting a physical constraint to the number of connections or the size, or, I, I don’t even know what they exactly did, but they, they, they had a, a thread on Mastodon, and it seemed fascinating because they were, they were saying that we’re finding all sorts of things that when you spatially constrained, constrained an R N N.
Paul 01:50:31 Hmm. Okay. Yeah.
Luiz 01:50:33 I mean,
Paul 01:50:33 I, I vaguely remember this, but, uh, maybe some, maybe it was like a year ago or something, or
Luiz 01:50:38 Okay, maybe there was another version, or, you know, an earlier, maybe it got published now and it was like different
Paul 01:50:43 Street print. I don’t know.
Luiz 01:50:44 And I, I had, I noticed just literally just yesterday, I, I think it was still a pre-print yesterday, but, uh, in fact, I, I’ll I’ll just send you the link, but
Paul 01:50:52 Please do
Luiz 01:50:53 Just, I, just as an example, the kinds of things, right? So that’s why I find it so fascinating. I wanna put brains, I mean, if we could clone ourselves, right, and have multiple careers at the same time, it’s like working in robotics and put a physical, what if we build this thing and this thing has to navigate in the world, right? Oh, so what, I mean, what do we learn from, from, from doing that, from, so my suggestion is that cognitive emotional interactions have to be integral to this architecture. That cognitive emotion integration has to be integral to this architecture and not something that is kind of a, you have a traditional cognitive architecture and you add a layer that you call emotion because it has some kind of
Paul 01:51:42 Right
Luiz 01:51:43 Neuromodulatory low pass thing that tracks, you know, how many positive and how negative, how many negative things happen in its past one hour. And it modulates a little bit how, I don’t know how fast it moves or how fast it, it, it, it, it answers a question. I don’t, I don’t think that that’s, that’s the way to go. I think that we need to, has to be much more integrated. And I think that that offers a lot more, uh, room for diversity of, of processing that than, than the, the segregated view. But again, that’s obviously something that, uh, it’s, it’s, it’s, it’s not, in this case, it’s not empirical. You have to construct and compare the, you know, com compete the two types of architectures and c right. End up doing.
Paul 01:52:33 So, believe me, uh, I would rather spend, uh, another half hour with you than go do what I’m about to do, which is, uh, try to teach children, uh, things a guitar and such. Uh, but anyway, I’ve gotta go be a parent, and I’ve taken you long enough as well. But, um, Louise, I, I really, uh, I didn’t say this upfront, but I really enjoyed the book, and I think it’s a great book, uh, especially for, uh, a more general, um, uh, audience. More, more general readership, because many of the books that I read, um, you know, that people I, I have on the podcast is a little more specialized. So I think you, you really hit the nail on the head there. So congratulations on that. Looking forward to entangled part two when that comes out. And, uh, I really enjoyed the conversation, so thanks, thanks
Luiz 01:53:14 So much. That means a lot. Thanks so much. I had great fun. Thank you.
Paul 01:53:32 I alone produce Brainin inspired. If you value this podcast, consider supporting it through Patreon to access full versions of all the episodes and to join our Discord community. Or if you wanna learn more about the intersection of neuroscience and ai, consider signing up for my online course, neuro Ai, the quest to explain intelligence. Go to brand inspired.co. To learn more, to get in touch with me, email Paul brand inspired.co. You’re hearing music by the new year. Find email@example.com. Thank you. Thank you for your support. See you next time.