#922 Stephen Grossberg - Conscious Mind, Resonant Brain: How Each Brain Makes a Mind

00:00 Ricardo Lopes: Hello, everybody. Welcome to a new episode of the Decent. I'm your host, Ricard Loops. And today I'm joined by Doctor Steven Grossberg. He is Wang, professor of Cognitive and neuro Systems, Professor Emeritus of Mathematics and Statistics, psychological and brain sciences and biomedical engineering at Boston University. And today we're focusing on one of his books, the latest one. And I guess that Steve would agree with me when I say that it's probably his magnum opus conscious mind, resonant brain. How each brain makes a mind. So, Steve, welcome to the show. It's an honor to everyone.

00:41 Stephen Grossberg: Well, it's a delight to be here and it's one of the wonders of modern zoom that you're in Portugal and I'm in Newton Massachusetts. So there we go.

00:53 Ricardo Lopes: Yeah. There we go. It's modern technology, right?

00:58 Stephen Grossberg: Yeah. Which got really amplified greatly when COVID happened. That's when zoom really took off.

01:07 Ricardo Lopes: Yeah, that, that's totally right. So to get into the topics here, then I, in the first few sections of your book, you get into how we tackle basically the brain, how we understand the brain, the mind and the relationship between them and how you can do it through several different perspectives, mostly you focus on physics and psychology. So to introduce the topic, could you tell us perhaps and also to add some historical background here for the audience? Could you tell us how uh physics and psychology approach approach this issue? And historically, when and how did they split in their approaches?

01:58 Stephen Grossberg: Yeah. Well, at the um in the late 18 hundreds, uh three very great physicists Helmholtz, I'm sorry, Hermann, Von Helmholtz in Germany, Clerk Maxwell in England and Ernst Mach in Austria made seminal contributions to psychology and neuroscience. Uh At that time, if you were interested in the structure of physical space and time, you would just as likely try to study psychological space and time. But um so they were uh interested in understanding how the observed world which is physics developed side by side their analysis of the observer, which is us. But by the time we got to the young Albert Einstein, oh, he didn't do psychology anymore. He just did physics. And I wanna quote just a little thing that he wrote to his friend, Queen Elizabeth of Belgium in 1933 which was much later. But this was his stance throughout life that quote, most of us prefer to look outside rather than inside ourselves. For in the latter case we see but a dark hole which means nothing at all. So in the late 18 hundreds of the greatest physicists did psychology and physics and some neuroscience. And then Einstein who's at least is great and did only physics. And the question is why. And one main reason is due to the properties of the experimental data that Helmholtz Maxwell and Mach uh discovered. And that data has characteristics that I like saying a nonlinear non stationary and non local. Whereas traditional physics use models that are linear stationary and local. So what do those words mean? Well, nonlinear means uh in the simplest form, the whole isn't the sum of its parts. You know, you chop us up into parts and life is no longer supported. But nonlinearity more technically means that you multiply variables. You don't just add them linear means you just add stuff non stationary means we're always evolving from being babies. We develop quickly and then we go on through lifelong learning and non local in the sense that I'm using it means that there are long range interactions between many, many neurons in our brains. And that is because these interactions enable our brains to understand the context and the objects in which we experience the world is not just pixel by pixel in a painting that our brains look at the whole pattern which requires interactions. Uh And uh the physicists just didn't have the concepts or the math to do it. And in fact, Helmholz, who started out deeply interested in psychology and physiology ended his life only doing physics. So maybe that's enough to get started.

05:41 Ricardo Lopes: And of course, when I asked you here about how people from physics and psychology, for example, think about or understand the brain, the mind and the relationship. I was also at least to some extent alluding to the mind body problem. So, could you tell us what exactly is the mind body problem and why has it been so difficult to tackle?

06:08 Stephen Grossberg: Ok. Well, when we think of our brain, which is a central feature in our body and the one that's relevant to our discussion today, we may think of the little gray cells of Hercule Poirot. I don't know if you like mysteries, but I always loved thinking about the little gray cells, but we think of our mind, we may think about a beautiful sunset, a feeling of being in love, uh you know, deep thoughts or poetry. And there's a huge gap between those two. And the question is, how do you bridge the gap? And that's what my magnum opus uh tries to do as reflected by its title Conscious Mind Resonant Brain. How each brain makes a mind? How do we bridge the gap? And my book therefore, tries to have a comprehensive and unified description of the main processes whereby our brains make our minds. And that can be done because there are vast databases that were accumulated in psychology and neuroscience uh since the time of Helm Helmholtz Maxwell and Max. So the problem wasn't um what am I supposed to explain the problem was how can I find any explanation of hundreds of experiments and which experiments cohere together? So in order to even even begin this, you have to have an intuition in which data uh are probing the same underlying processes. And what are those processes? So? Well, we start with perception and at least two main ones, visual and auditory. I won't even talk about tactile and that goes on to cognition and emotion and action. And we can't just be satisfied with discussing how that goes on in uh healthy or typical individuals. It's important to also understand well, what happens when they break down which they do and can lead to many clinical disorders. And in the course of my work, I noticed that if there's certain imbalances or breakdowns in the neural network processes that I used to model lots of uh data about typical individuals in a principled and unifying way. Then al popped uh symptoms of diseases as varied as Alzheimer's disease, autism, medial temporal amnesia, schizophrenia, A DH D PTSD, auditory and visual neglect and agnosia. Even what goes wrong with slow wave sleep things which I never dreamed of. And so one of the main things to think about is, you know, how are you forced into thinking about things you never dreamed you would know anything about. And more than that um our brains operate in cycles as a perception, cognition, emotion action cycle where we're continually interacting with and adapting to a rapidly changing environment. So with this background, let's come down back to the mind body problem. Why are they so intimately linked? And one theme in my book and, and think about mind body problem is brain evolution needs to achieve behavioral success. So the uh the meat, the little gray cells, uh the body part of it is intimately linked through Darwinian selection to uh the mind part of it. Uh They, they can't be separated. Um So the book explains how multiple brain circuits and multiple brain regions interact to generate the behaviors that control our day to day uh activities and mind is what you can most efficiently call an emergent property, an interactive property of these brain mechanisms. And even Von Neumann knew that that the intelligence isn't just in the local interaction, it's how many of them interact. So why my book is unusual is that most books are about mind or brain, but not about each brain makes a mind. And in this regard, I was lucky because I started my life's work when I was 17 years old, I was a freshman taking introductory psychology. And there was a lot of quantitative data about how we learn a classical data, Hovland Hall. Many other Great American psychologists were into learning. And so my brain exploded and I was driven into what became neural networks uh including the main laws of short term memory, medium term memory, long term memory that people use today to explain how brains make minds. So because I started in psychology and was forced into brain, I was able to make the link and then build and build and build with many gifted collaborators for 67 years. So it was a lucky start. But why my brain exploded, I can't explain. And um so um and these are not traditional concepts from computer science, engineering or traditional symbolic A I, although now a lot of people call me the father of A I. Because over the years when symbolic A I failed, the people in IA I wanted to take over neural networks to be A I. Marvin Minsky was one of them. And um anyway, so I'm called the father of A I because I introduced the modern paradigm in 1957. Um And along the way, you know, with all the discoveries that just poured out uh over these years, uh it's not just classical stuff because I had to introduce new computational paradigms to link brain to mind. And I call two of them complementary computing and laminar computing and complementary computing explains the nature of brain specialization. You know, we don't have independent modules like you don't first compute uh edges and then shading and then texture and then depth rather. When we look at an object, all of these properties can be overlaid, we're looking at a texture depth, full object with contours boundaries. And so there's a lot of multiplexing of um properties. And so um complementary computing was a discovery I gradually made. I didn't wake up one day and say, oh let's let's assume the brain as complementary computing. Now I over decades, I we try to explain more and more data and in every one of them, there were complementary pairs of processes and a kind of Yin Yang relationship computationally complementary until I realized it was a global design principle for the brain. Like I think we might come to it in a while that visual boundaries and surfaces obey computationally complementary laws. The ventral watt stream that um does perception and recognition in our brain and the dorsal we stream that the facial representation and action obey computational complementary laws. And one of them the what and where is on a huge scale of atomical organization and the boundaries and surfaces are much smaller. So you have a multilevel um uh representation of complimentary complementarity in all brain processes. And I mentioned lam in a computer and laminar computing explains how all uh higher forms of biological intelligence are um explained by variations of a single canonical Lain a circuit. I mean some listeners may know that if you look at uh the Neocortex that supports vision, speech language cognition, there are six characteristic layers and characteristic sub lamina with characteristic feedback interactions and horizontal interactions. So bottom up, top down horizontal in all of these circuits. And one of the marvels of evolution is that specializations of the same canonical circuit can do vision, speech language and cognition. And that's very important when you think, how do we develop a sense of self cause our whole cortex shares a design and all the parts can seamlessly fit together into an emergent uh architecture that can be resonating uh as needed to do, you know everything that we do in our conscious experience. So I've always felt it was remarkable that some such seemingly different behavioral competences emerge out of variations of this ship canonical circuit. And no less remarkable is that circuit supports different kinds of conscious awareness. And because all those circuits obey a similar kind of input output scheme, so they can all be interacted, feed forward feedback, et cetera that supports consciousness. Um But maybe that's all I should say about that you have a lot of questions you want me to answer.

17:57 Ricardo Lopes: Yeah. Yeah. Yeah. Yeah. There's a lot there to unpack and it relates to also to some of the questions that I have

18:04 Stephen Grossberg: also say for listeners, I think, you know, think of this interview as listening to his story, everything in the story can be thought about and studied on multiple levels. But first get the gist of the story and, and maybe as we go along, uh you'll ask me to flesh out things more but, but your questions really asked me to flesh things out more. So I, I think we should follow the questions and then see what made the needs to be added. Of

18:38 Ricardo Lopes: course. And so uh you've mentioned words like and computation there. And I would like to perhaps clarify that a little bit more. So, are our brains like computers? I mean, when people talk about com brains as computers, when they use that metaphor, is it accurate or not?

19:01 Stephen Grossberg: Well emphatically not. Um If I start, I just noted we need new computational paradigms. Uh LIKE complementary computing, laminate computing. Uh Moreover, traditional computers aren't conscious. Um So my work provides or off as a scientific solution, the mind body problem by showing how where in our brains and even why from a deep computational perspective, humans can consciously see here, feel and know things about the world. And we don't just get conscious to serenely contemplate the beauty of the world rather we use our conscious states for reasons I hope will clarify as we go along because they are competent to uh um control uh effective planning and action to realize valued goals. And so consciousness mediates between mind and the world of action. Um And so to do this, our brains are self organizing, meaning we can learn by ourselves autonomously throughout life. Their analog, which means that the neurons can have values going from minimum to max and everything in between, not namely shades of gray, much as you might see an achromatic image in shades of gray and everything operates in parallel because everything about our intelligence is organized in terms of spatially distributed patterns of activation. So self-organizing analog parallel, but present day computers are not self-organizing and they're digital and serial. So the architectures are fundamentally different. Um My own belief is that as uh computers evolve and they are, you know, they get even more powerful, like, you know, my iphone is a supercomputer. Um We have systems that on the one hand, uh maybe dedicated VL SI chips that embody more and more of the kind of intelligence that humans embody are complemented by a lot of things humans don't do well, which digital computers do like adding a million numbers in a fraction of a second. And that will be a very powerful tool when when and if it happens.

22:00 Ricardo Lopes: And uh you've already mentioned their uh conscious awareness. How do we experience conscious moments of seeing, hearing, feeling, knowing and why do things like perception and awareness uh matter here?

22:21 Stephen Grossberg: Well, I wanna emphasize that although I've led the development of what I think is arguably the most advanced neural theory of consciousness in the sense that it provides principal and unifying explanations of the most psychological and or biological data about conscious experiences. I never tried to directly understand consciousness. Uh You don't wake up in the morning and say, hey, you know, let's understand consciousness today. Uh But instead all the deepest insights about how our brains work a concern. One of the things I just mentioned how they self organize uh through childhood development. And do you have any kids?

23:16 Ricardo Lopes: Uh No, not yet. At

23:17 Stephen Grossberg: least. Well, we have one daughter. Well, anyone who has observed little kids learn language that's fast self organization and lifelong learning. We continue learning throughout life. Remarkably. Uh My models of learning led me forced me kicking and screaming into insights about consciousness. And uh so my work on learning led to the most advanced, currently, the most advanced cognitive and neural the about how brains learn to pay attention, recognize and predict objects and events in a changing world emphatically. One that's filled with unexpected events because that's what we got to do to deal with evolutionary challenges. And this is the theory I call adaptive resonance theory for a reason. I'll make clear in a moment and all of the foundational hypotheses have uh from which I could derive art have been supported by subsequent psychological and neurobiological data. And art has also provided principled and unifying explanations of many additional experiments. But for me, the bottom line is really that I was able to derive art. I say art A RT instead of adaptive Resonance theory in 1980 in an off cited article in the journal Psychological Ro review from a thought experiment. And this thought experiment is about how any learning system can autonomously learn to correct predictive errors. So this is all about autonomous learning and the hypotheses of the thought experiment are just a few facts that are familiar to us all. And they're familiar because they represent ubiquitous environmental pressures on the evolution of our brain. So they're not something. Oh, well, why should I believe that? You believe it because you've been living with it your whole life. And moreover, during the thought experiment, I never used the word mind or brain. Mhm. So, the thought experiment and the principles and mechanisms that come out of the wash are thus the universal solution to the learning problem. Wherever those hypotheses are used to derive their logical consequences, you'll get a system like art and the learning problem, which I didn't say one crucial fact about lifelong learning implies that when I learn something, I won't just forget it right away, which is called catastrophic forgetting. For example, Ricardo, I never uh saw your face before and now I'll remember it for the rest of my life. So I learn it in a second. I'll remember it for years. So I call this the stability plasticity dilemma. How any system can learn without catastrophe, forgetting plasticity of because we experience incredibly fast learning often and then stability. The fact that our memory of that fast learning can persist for a very long time. And let me just say, for those of you who know some popular current A I like deep learning or it experiences catastrophic forgetting big time in multiple ways. Uh If we maybe we will talk about a little later. So I think therefore, it's remarkable. But that these results about learning without catastrophic forgetting is what led me to rigorous neural models of how we consciously see, hear and feel and know things about the world. So I derive the models I was explaining with simulation, parametric quantitative computer simulations with my many students and postdoc all this quantitative data about seeing, hearing, feeling and knowing. And then I realized all those parametric properties or properties of conscious experiences. So consciousness sort of came in through the back door. Uh And the thing that all of these experiences had in common are resonances which occur when there's excitatory feedback bottom up, top down excitatory feedback between pairs or greater numbers of brain regions when they match their signals well enough to cause the active cells to synchronize boom, boom, boom, boom, boom, boom, boom, boom. Um And that's one reason why we have brain rhythms, but different brain rhythms support different processes and sustain. They're firing long enough to trigger not only learning but consciousness. In other words, the resonances and the requirement for matching or what dynamically buffer a learning process against catastrophic forgetting. And the remarkable thing is that the contents of these resonances match properties of the conscious experiences that they're learning to represent. That's why I call my theory adaptive resonance theory because of how resonances trigger learning that doesn't experience catastrophic forgetting. So the proposed solution on mind body problem occurs due to a computational analysis of how humans and other animals autonomously learn in a changing world without catastrophic forgetting. And because the hypotheses and the thought experiments that lead to cognitive models like adaptive resonance theory, as well as cognitive emotional models, like what I call my cog and model, the cognitive emotional motor model never mentioned the word mind and brain. Um My work and my book show how these models provide a blueprint for designing autonomous adaptive intelligent applications to engineering technology and A I. So if you want autonomous adaptive intelligence, we already have the foundations of a blueprint for it for the future. Um In particular, my books spend a lot of time uh explaining how our brains consciously see the world. And that's not surprising because more than half of various mammalian brains are devoted to vision from excuse me, early sensory processing up to visually based cognitive planning and navigation. Um In other words, we need perception and consciousness to be able to act upon the world by looking, reaching and navigating uh in order to survive environmental challenges. So this link between consciousness and action I think is very interesting mediated by learning that solves the stability plasticity dilemma. Sorry about this. I have had asthma for over 70 years.

31:41 Ricardo Lopes: Uh So uh another thing that I would like to ask you about and it's something that you also tackle in the book is Visual Illusion. So how do they work and what can we learn about? I guess not only the visual system but information processing by understanding how visual illusions work.

32:03 Stephen Grossberg: Yeah, I wish we were able to figure out how to show my powerpoint slides. I'll try to use the words uh about what the pictures would have shown more clearly. So first let me sketch why we experience visual illusions, which will also start to explain how they work. Um So let's start with the retina, the photosensitive retina, that's where light is registered before light signals all over the retina are bundled through axons as the pathways uh into the optic nerve to set pattern light from the whole retina up to the brain because of the need to form the optic nerve to carry the light signals to the brain. The retina has no photo detectors on the part of the retina where the optic nerve forms, that part of the retina is blind. And um you might say, well, big deal, I don't see it, it hasn't caused any trouble. But if you look down on the retina and I had a pretty picture to show you that. And if you get my book, um it has all these pictures, it has over 600 color pictures. So everything I say is illustrated, another part of the retina not far from the blind spot is called the fovea Fove A. That's where all of our high acuity vision occurs. And it's because we need the phia to really see things clearly that we move our eyes several times every second with psychotic eye movements to point them to the things we really wanna see clearly. Things in the periphery get fuzzier and fuzzier and the blind spot is as big as the phobia. And now you might say, but then why don't I see it? And um, before getting into, why think about it, we couldn't look at or reach for an object that happened to re registered where the blind spot is. If we didn't somehow do something to make that problem go away and that could lead to disaster, you know, a leopard is leaping at you in the place where the blind spot would have seen it tough. So why aren't we aware of the missing visual signals due to the blind spot in particular, how do you complete the retina representation so that we can look at and reach for things which otherwise might have been uh invisible. And one thing you have to deal with right away is our eyes are rapidly jiggling in the orbit. They're sort of micro the cards and they're happening so fast and they're so small, you don't notice it, but it's functionally very important because there may be a stationary object in the world and you are stationary with respect to it. But your retina is jiggling back and forth around every edge of that object and it's creating transient signals. Um And these transient signals, refresh visual signals um which otherwise would fade. I like to mention Jurassic Park here. And did you see Jurassic Park? Yes, I do. You know that famous scene where um this guy is standing out of the car and T Rex comes looking very ravenous and they scream, stand still. He can only see relative motion and then T Rex moves his head. Yes, relative motion and scoops them up in one bite. Anyway, the blind spot is stabilized on the retina. So it doesn't create transients and so it fades. So it's not gonna cause a problem in itself, but you still have uh an incomplete retina limit with a big hole in it. And so how do you complete that? And that is comp accomplished by multiple stages of completing boundaries across the hall and fill in with surface brightness and color. And it takes several stages to do it. Um And I call processes whereby incomplete representations are completed by multiple stages of processing a hierarchical resolution of uncertainty because when you're done, you're going to get completed boundaries and completed surface representations. So you don't know there was a hole there. Mhm But then that raises a deep question. How does the brain know at what stage the cortical representation say of a surface representing your face is complete enough uh to um control looking at you or maybe touching your nose? How do you know what stage? Because it took a lot of stages. And you don't want to use a stage that has an incomplete representation because then the leopard could leap at you through your blind spot. And the answer that I propose is that a resonance with a completed surface representation. And the next processing stage. Whoa whoa whoa whoa whoa whoa whoa surface representation. Next stage lights up the surface representation makes it consciously visible. And then we use the conscious representation to look and reach. And so the hierarchical resolution of uncertainty created the need to have some active process light up the completed representation. And I call this a surface shroud resonance because spatial attention focusing on the surface is called a intentional shroud. And I predict it occurs between a pre stride visual cortex V four. In other words, there's a hierarchy of cortical stages. It took several days stages to complete the surface and the posterior parietal cortex which is known to support spatial attention. So V four P PC I predict supports the surface route resonance and there's a lot of data supporting that. So getting back to the original question, what's a visual illusion? If the boundaries that are completed do not contain expected surface brightness and colors, then we say we're looking at a visual illusion. The truth is that almost everything we see is a visual illusion in the sense that we're completing boundaries and surfaces not only over the blind spot also over the plexus of retinal veins that covers the photo detectors. In addition, where compensating for variable illumination, we not seeing the illumination gradient. So the final representation of objects and scenes are visual illusions, but we only say, oh that's a visual illusion. If the color isn't what it we think it should be based on our experience with that kind of boundary. So everything basically in vision is an illusion and the ones that look funny, we call illusions.

41:22 Ricardo Lopes: OK? But about those ones that look funny and we call illusions. How does understanding how we experience those connect to our experience of art?

41:37 Stephen Grossberg: Well, here, I would have shown you several pictures that would have been uh let people listening. We weren't able to figure out how to link my powerpoint to the zoom. I've done it in the past, but I think the software changed or something. I don't know. First, I want to say a little bit more about boundaries and surfaces because they are the general functional units of how we see. That's an early prediction of mine that various people and supported psychophysics. So, um there's a wonderful uh illusion that it's painful to me not to be able to show it to you because it's gonna be so hard to explain is a certain kind of Kinnier Square, Gaetano Kita famously introduced lots of visual illusions. He's very great uh Italian perceptual psychologist who died some years ago in this illusion. He takes um circles and he'll make part of the circle black and part of it light blue in a lot of concentric circles so that the black blue interface is horizontal. But I'll make a Pacman out of it where the black part has a Pacman shape. These are all black lines and then there's a little quarter of it, which is the blue parts which have less contrast and he'll do four of them, 12 parallel to each other, like oriented three. It is hard to do, but it's backwards. So we'll have pairs in a row and then pairs one on top of the other. And so if you look at where the black and blue uh interfaces occur, they all line up to form an illusory square and my book has it. I don't know if you can see it.

44:08 Ricardo Lopes: Yes. Yes, I can. Yes.

44:10 Stephen Grossberg: So you see the point is that because of the bound, the strong black boundaries inhibiting the weak blue boundaries, that's enough to make the square form. And it's also enough to let surface color pour out of the broken blue boundary to fill in the entire square. So what you see is a blue square, neither the blue nor the square are in the image. But what this illustrates and that's why it's such a lovely example. Boundaries are completed inwardly between pairs or greater numbers of inducers in an oriented way. That's how you complete the square and the blue color that spills out of the little blue uh broken boundaries spreads in a uh un oriented way um to fill in the square. So um so the color spreads outwardly and un oriented and the boundaries are completed inwardly and oriented. And so you could see from this outward, inward oriented, un oriented, a sense of what I mean, that boundaries and surfaces are complementary and, and Yang, because you can't be both oriented and un oriented and you can't be both outward and inward and without both of them interacting though you don't see anything. And so you have these parallel cortical streams that are doing complementary computing and then interacting to overcome their complementary deficiencies. And there's one more big property of boundaries and surfaces. For starters, we can consciously see surface brightness and color we see while the, but we can't consciously see boundaries except when they separate surfaces with different brightness and color boundaries are invisible. It's called the offset grading. Uh Yes. OK. So all I've drawn is horizontal black lines, but usually you can recognize that there's a vertical boundary of some kind between the ends of those lines. Can you recognize that? So, but the boundary is invisible? OK. All boundaries are invisible. The only reason why you'll see color is because if you create uh let me see if you can see this. Whoops. Uh

47:37 Ricardo Lopes: uh Yes, I can, I can. This

47:39 Stephen Grossberg: is called the Ehrenstein disk. All I drew were black lines. It completed a roughly circular boundary, perpendicular lines. And then at the line ends, there were little black white contrasts that were little brightness buttons that filled in within the boundary, the circular boundary and makes it look brighter. Do you see? It looks brighter?

48:06 Ricardo Lopes: Yes, I see.

48:08 Stephen Grossberg: OK. So that's a case where because the color can't escape, it makes for a visible brightness difference, but here the color can escape because it can flow all out of there and spread uniformly on both sides of the vertical boundary. And so it's invisible. So I let me show you, I think I wanna show you a very famous picture. That's a lot of fun. Let me see. This is a famous picture

48:47 Ricardo Lopes: uh perhaps just please move it a little bit away from the camera because OK, like that like that,

48:55 Stephen Grossberg: I don't know if you can see, I can see first, it looks just like splotches of black on white. But if you look at it for a while, you can see a Dalmatian dog in snow. It's hard to do the image all the way to my right is the kita square. Mhm I don't know if you can see. Yes, I can see and right next to it horizontally, what's called a reverse contrast kita square? Now the Kinnier square just like uh when I made that circle with the lines, it looks brighter because the Pac Man, the four pac men create square boundaries and then they put bright contrast inside which fill in the square and look brighter. But if you look at the one right next to it, it has two black and two white inducers on gray background. And there you can recognize that there is a square but the gray inside and outside of the square is the same. So it's visible. So in the case of the reverse contrast, Ken it a square, I'm so sorry, I should have just made.

50:29 Ricardo Lopes: No, no, no worries. I can edit this.

50:33 Stephen Grossberg: Yeah. So you can consciously recognize a a square boundary that you don't see. Um Anyway, many artists knew uh instinctively about these things, including matisse who I love and matisse used to like writing that all bound. Well, he didn't write all boundaries are invisible, but he painted with color blotches and knowing that the boundaries would be completed to create a scene. And one of my favorites is the self care and

51:28 Ricardo Lopes: I leave it, leave it like that, please. Yeah.

51:30 Stephen Grossberg: Yeah. He painted that in uh 1905. And if you see it uh in my book or on the web or wherever you'll see that there are a lot of color splotches, but they're arranged in a way that invisible boundaries join them and create surface representations of objects. So let me see if I can, they get the schematic of that.

52:02 Ricardo Lopes: Yes, I can see it now. Yeah,

52:05 Stephen Grossberg: you see the splotches create a modal boundary groupings which then capture the splotches into surface color. So, OK, anyway, so um that is one link between art and um illusions. And I discussed the work of 14 different artists showing how they all instinctively exploited different properties of how we see. I mean, just consider like uh the impressionists like Monet and Point Allistic painting sera. Um YOU know how these little points of color arranged uh in an appropriate way, generated emergent boundaries and surfaces that you can interpret as a beautiful scene. So uh my book spends a lot of time showing how different artists and artistic movements um uh generated uh painterly precepts that we understand often as scenes and objects even, you know, and like how you get effects of three dimensional curvature in a two D painting. And one of the big things my book explains is how the way our visual system adapted to the three dimensional world, those same laws automatically enable us to understand paintings. Um ZOOM screen movies as 3d objects. You know, I'm not seeing you like a pancake, I'm seeing a frontal view of a three dimensional face.

54:13 Ricardo Lopes: And so we've been talking here a lot about the visual system, but how do we go from seeing to recognizing and also then to predicting the world?

54:28 Stephen Grossberg: Um So first let's distinguish, seeing and recognizing. So seeing, we open our eyes and if there's enough light we'll see, it could be unfamiliar or familiar. Uh So seeing is a general purpose process, but we only recognize familiar things. So recognition requires prior learning of recognition categories for objects and scenes. Uh SO that these categories would selectively fire after learning in response to the objects and scenes that we've experienced before. And that raises the question, how do you learn a recognition category? And I already noted that Adaptive Resonance Theory or art explains how we learn to recognize stuff without experiencing catastrophic forgetting. And one thing I didn't say about learning in what I said, it enables us to pay attention. But what it's what we're learning as we're learning categories that can be used for prediction is what are the critical feature combinations that are going to predict uh successful action? How do we suppress the irrelevant outliers? And when you study adaptive residency, you see how over learning trials outlier, irrelevant features get actively suppressed and you gradually converge it can be very fast. It depends on the complexity of the problem on a set of critical features in the activation pattern. Uh THAT you know, the top down expectation is lighting up and it's so critical features that go into the bottom up adaptive filters that choose the category and in the top down expectation that the category reads out to focus attention. And so the critical features enable art categories to be associated with predictive consequences. Um Yeah. So and so and different resonances then support seeing and recognizing I mentioned mhm A surface shroud resonance supports conscious seeing and uh the shroud is in posterior parietal cortex. Um THE surfaces in V four, when that resonance occurs, it propagates top down all the way down to uh lateral chani it all the way up to prefrontal cortex. And because of what's called the art matching room, namely how we pay attention. It selects the critical features that uh we should be looking at. But then the feature category resonance um which goes into the watch stream, recognition and perception stream that supports conscious knowing. So that when a surface shroud resonance and a feature category resonance synchronize, when I look at a familiar person's face, I both see them and know who they are. Um And I had a nice, nice picture of that. But if, if you're interested anyway about my book, if you are interested, I I should say this because, you know, you might think such a big book. Uh If you buy it on Kindle, it's 15 bucks. And if you buy it from Amazon, it's about 30 it would have cost over 100. But I worked very hard on the book and I want the people who are interested to be able to afford it. So I I asked Oxford University Press how much of my own money I had to give them to bring the prices down. So if you get it, it's cheap because I subsidized it.

59:03 Ricardo Lopes: Well, thank you so much for that. That's very generous. I guess that people will

59:08 Stephen Grossberg: be. How much did it cost in Portugal? Yeah, I don't know why the, the you, you use euros in Portugal.

59:17 Ricardo Lopes: Uh, Y yes, I, I mean, I, I think it cost around, uh, I don't know, €30 I guess I um order it from Amazon Spain. So,

59:29 Stephen Grossberg: yeah, so that's not that different from $30.

59:33 Ricardo Lopes: Yeah. Yeah. Yeah, that's right. That's

59:35 Stephen Grossberg: right. Well, thank you for buying it. So um should we, should we move on? Uh

59:40 Ricardo Lopes: Yes. Uh And I was going to ask you now, uh actually earlier, you've mentioned for example, your uh cognition, emotion motor model. But how do you look at the relationship between emotion and cognition

59:58 Stephen Grossberg: and brave cognition about is what we know about the world and emotion embodies, how we feel about what we know. And it's emotion that gives value to what we know. And here another resonance occurs between our cognitive alike representations and our emotional representations. So the alike ones are in places like infra temporal and prefrontal cortex. The emotional ones are in places like the Amygdala and hypothalamus. And when there's a resonance between cognitive and emotional uh representations that combines what we know with how we feel about it. And the feedback from the emotional representation is called incentive motivation. Not only to pay sustained attention to something that we currently value, but it also gives the motivation to release actions to realize goals about things we value. So, you know, if I see my wife's face a little way from here, I'll feel all the good things I feel to her and the motivation to approach her. So a cognitive emotional resonance sends positive feedback between the cognitive and emotional representations. So just as a current object or event can trigger an emotion, an active emotion can select and focus attention on objects and events that are compatible with that emotion. And that can help us achieve, you know, um results that are compatible with it where emotion could be hunger and we just wanna go have lunch. So um how to notice desire attend and eat that ice cream cone. It's all wrapped up into cognitive emotional resonances in their output pathways. Um So the cognitive emotional resonance is a different kind of consciousness because it links, for example, um to surface shroud resonance. Well, I think we're gonna talk about that in a moment.

1:02:52 Ricardo Lopes: So, so I I mean, these different kinds of awareness that we have and we've talked about some of them, how, how are they then integrated into what we experience as unified moments of conscious awareness?

1:03:09 Stephen Grossberg: Yeah. So for that multiple adaptive resonance is synchronized. I mentioned two of them that could. But so example, during a conversation with a friend, we might be synchronizing surface shroud residents to consciously see them stream shroud resonances that consciously hear them feature category resonances to consciously recognize what we're hearing. Cognitive emotional resonance is to consciously have feelings about what we're discussing and then what's called item list resonances to consciously understand the conversation because they're supporting things like working memory of sequential sounds that form language phrases and sentences, and all of them come together to enable us to act appropriately during the conversation. So it's not just a single resonance, there's a continually evolving series of resident states as stuff keeps happening.

1:04:25 Ricardo Lopes: And so uh we've already talked here about the relationship between cognition and emotion. Uh I would like to ask you now specifically about the prefrontal cortex. So, in all of what we've been talking about here, does the prefrontal cortex play a special role?

1:04:44 Stephen Grossberg: Um Well, I'm gonna want to show you a figure of that for sure. But, but first, let me say it carries out or controls many of the higher cognitive, emotional and decision making processes that define human intelligence as it's also controlling release of actions uh and aimed at achieving value gold. And as I just mentioned, one of the things prefrontal cortex has it has um working memories that can uh generate uh what I call um list chunks or list categories like words, sentences, phrases, thoughts that can sequentially determine what's appropriate to do in a given situation. So it's not just one thing and it enabled us to generate behaviors that are flexible and adaptive notably in novel situations and to suppress ones that aren't by using this contextual knowledge. Um Can you see the seven regions on top in red and green?

1:06:09 Ricardo Lopes: Uh uh Yeah, I, I mean the green, it's not perceptible here. But yeah, I see the, the several

1:06:17 Stephen Grossberg: regions on top, they're all different regions of prefrontal cortex. And basically, when you put together all the models into this larger neural architecture, it explains how the prefrontal cortex interacts with uh cognitive emotional decision making processes to make contextually based decisions about what to do or think next. So it's uh so you got to understand how working memories work and a whole series of things we won't discuss much today.

1:07:04 Ricardo Lopes: So earlier, you've mentioned learning and I want to ask you a little bit more about it because actually in the book, you also explain why we need to understand learning to then properly understand consciousness. And you distinguish between two different types of learning, perceiving and knowing versus moving and navigating. How do they differ from one another?

1:07:31 Stephen Grossberg: Right. Well, so I briefly mentioned before that perception and cognition are represented in the w or ventral uh cortical stream and then a spatial organization in action including navigation of representing the where or dorsal cortical stream. And um uh they obey computationally complementary laws. I'm gonna just say in words what they are. And then it may be that one of the sly, one of the um copies I was trying to make in my powerpoint was of that. And I can actually show it. I'll run that to my other study. So perception and cognition, use exci toy matching and match based learning. What does that mean? So if I read out a top down expectation, um a prime the system for getting ready to respond more vigorously to an expected event, oh, you're here, I recognize your face faster. So that's excitatory matching and match based learning means that you only go into adaptive resonance if what you're expecting and what you're learning about uh c close enough together. So you can suppress stuff and learn a better and better critical feature pattern. So the learning is ex excited to it's match based and the matching is side to it. But navigation and action use inhibitory matching and mismatch based learning. What is an inhibitory match, let's say within our movement, I have a representation of where my hand is now and I wanted to go here. So this is my present position. This is my target position. Then I have a volitional go signal from the basal ganglia that makes my arm move from here to here. But when I get here, I am where I wanna be. So that present target minus present is zero, that's inhibitory matching. And it turns out you use that mismatch in an appropriate way to tune the circuit so that when you are, where you wanna be the difference vector is zero. So first here, I can't see what I'm showing you

1:10:28 Ricardo Lopes: I

1:10:29 Stephen Grossberg: can see it, here's a slide of adaptive resonance and you can see the light green down here. Yes, that's the critical feature pattern so that you have a top down match on the critical feature pattern. You go through an excitatory resonance to try to learn that. And the way that works if we're gonna discuss it is via the art matching rule. Uh MOVE

1:11:02 Ricardo Lopes: it just a little bit to your left, please. Yes, like that. Yes.

1:11:08 Stephen Grossberg: So the what the art matching rule shows, you see, there's, if I have bottom up excited Tory input, it can fire a feature selective cells. If nothing else is happening, if I activate a category and I try to read out in those spreading green adapted pathways, uh AAA critical feature pattern or prototype, I can't fully do it because I'm also activating this inhibitory of surround which approximately balances the excitatory on center. That's how I prime. When I read out that, that expectation, there might be some subthreshold activation of the cells that got activated and a deep trough of inhibition around it. But then if I have a matching bottom up input, it's two against one, too excited to one bottom up, one top down one inhibitory, you can start to fire and that can lead to resonance. And I don't know if we discuss, we're gonna discuss that. Uh What do you do if something unexpected happens? There's a complimentary attentional system and orienting system where the learning goes on the attentional system. And if you get a big enough mismatch or novelty or unexpectedness, it activates the oriented system which automatically discovers a better matching category or a new category to learn the new stuff. So, OK, that was a little a field. So sorry, let's get back to your question.

1:13:01 Ricardo Lopes: Uh So then, I mean, why is it, why do we need to understand learning then to properly understand consciousness?

1:13:13 Stephen Grossberg: Well, to summarize my work shows that learning without catastrophic forgetting is solved by adaptive resonances. And that properties of the adaptive resonances just after years of modeling happen to support multiple conscious states with different functions in different parts of the brain. We reviewed some of them, all of them are supported by lots of explanations of modeling simulations and explanations of data. So, um you know, science is never complete but um uh I'm I'm sort of baffled in a way, you know, we live in the world of Google, you can presumably search for anything in no time at all. And yet I've explained data from hundreds of experiments. Never mentioning the word quantum no quantum. Um There were quantum effects at the photoreceptors at the when you know noise hits the the ear at the periphery right away, our brains are designed to suppress quantum noise and to deal with it to become deterministic. And yet today, there are still people who are promoting that, you know, consciousness must be quantum because they don't understand it and must be in microtubules. So how do you recite Keats out of a microtubule? That's beyond my understanding. So there's overwhelming evidence that what I'm saying is at least on the road to the truth and the thought experiments cannot be denied because the hypotheses are familiar facts. Um When I grew up learning science, I thought you studied the theories that explain the most data in a principled way and nothing comes close. So I hope some people are interested in studying it. And I wrote my book in particular to be self-contained and non technical. But that being said, we are talking about the human mind, which is perhaps the most complex system in nature that we can hope to understand scientifically. So uh it requires thoughtful reading. I have friends who've read it uh parts of it, you know, who know no science at all. Um A rabbi, a pastor, a, a visual artist, a gallery owner, a lawyer, the social worker. Um AND they would read the parts that interest them and, and that's possible because I wrote the chapter. So after you read the preface and introductory chapter, one, each chapter can be read independently of the others. So you don't have to try to read the whole thing who has that much time or interest. But you know, if you're interested in about feelings, you can jump to the chapter about feelings and so on.

1:17:05 Ricardo Lopes: So let me ask you now a little bit about the evolution of the brain. And actually earlier, you've already touched a little bit on the idea or the proposal that some people have of modularity. But what do you think about that idea more specifically about the idea of modularity of the brain slash

1:17:28 Stephen Grossberg: mind? Well, brain anatomy is definitely specialized. There are multiple brain regions that carry out different functions like vision, audition, cognition, emotion, and so on. But the processes aren't carried out by independent modules. In fact, if you look at f the famous Feldman and Van Essen macro circuit, which I had a picture of um they all interact. Um I can't, you see, I can't see it when, when you can see it.

1:18:14 Ricardo Lopes: Yes, but I can see it.

1:18:16 Stephen Grossberg: Yes. OK. So on the one hand in the lower right, that's the anatomical diagram by Feldman and Van Een of the interactions in the visual system dense bottom up, top down multiple stages. Remember, hierarchical processing of uncertainty and the bigger picture if you could read it gives the functions carried out by different parts of visual cortex along with what region it is. And there are bottom up, top down horizontal interactions and and as the figure says, the bottom up horizontal and top down interactions overcome complementary processing deficiencies. So if it weren't for them, those interactions nothing could be completed effectively. And moreover, each of the processes do what one's called multiplexing information. So a visual boundary, it's not just an edge. If you know, I the the famous example of a shaded ellipse, if all it was was a bounding edge at the instead of a boundary, then the color inside would just fill in and you'd see a flat gray ellipse, but boundaries respond to edges, texture shading and depth. And so, um in addition to hierarchical resolution of uncertainty with all these interactions, each of the computations are multiplexing, multiple properties. And as I said before, otherwise, you couldn't, I couldn't ever learn to recognize your face. You know, you have texture, you have shading, you have boundaries. Um It's all overlaid. I have to be able to deal with it holistically. Did that help? No.

1:20:31 Ricardo Lopes: Yes, that helped. And uh earlier or just a minute ago, actually, you mentioned how uh many theories of consciousness out there, people, apparently people who propose them apparently do not even look at uh scientific data. So, but why do you think there are so many theories of consciousness out there? And uh would you have a solution for how we should actually approach it that

1:21:01 Stephen Grossberg: we all have a conscious experience? So maybe we feel entitled, but we also experience cognitive impenetrability, which means that in our daily experience, if you don't study neuroscience, you don't even know you have a brain. And for hundreds of years, people thought that our experience was in our pancreas or our liver or you know, so um you can't introspect the answer. Mhm And so I don't believe that a theory which discusses consciousness is a theory of consciousness. Unless it can show how interacting brain processes give rise to conscious psychological experiences, you have to make the link and without a linking hypothesis between brain and mind, brain processes have no functional role and psychological processes have no mechanistic explanation. So you're high and dry. Now, as I mentioned, I'm lucky, I started passionately studying psychology as a 17 year old. And that drove me because I always think in real time into studying brain dynamics, I I didn't mention but I was driven into neural networks to explain data about verbal learning, serial, verbal learning, how we learn lists of anything. And I didn't know they were neural networks. I had never studied, I was a freshman after all cell bodies of neurons, axons, synapses, synoptic knobs, transmitters, action potentials. They were all in my model from a psychological derivation. And so only when I talked to pre-med, friends who were studying it in their pre-med courses that I realized the things that they told me they're learning, I already knew. So you have to make that link. And from a very early age, I discovered a way to penetrate cognitive and penetrability and people who want to know how I've done it and others have tried to do it. Um They should read my book and then if they get really interested. Um If you go to my web page, which is just sites, si Tes dot Bu for Boston university.edu/, Steve Gsteveg sites W to use flash Steve G, you'll see maybe 570 archival articles and quite a few videos of self-contained lectures and interviews to learn more about it in detail. But my book is one is a good, you know, one place to go shopping for this information. There are 17 chapters on different stuff and it covers a lot of stuff.

1:24:49 Ricardo Lopes: So yeah,

1:24:52 Stephen Grossberg: yeah. So what makes me frustrated with some very well known scientists? I won't mention like the quantum guys if they could explain even a tiny fraction of the amount of data I've explained by that. I mean, how brain dynamics give rise to parametric psychological experiences in data. I would think they were making a serious contribution but they don't even try and, and they never try to compare their contribution with mine, which I just feel is unscientific. It's not how I was brought up beyond bad manners. It's unscientific, which is the severe criticism in my lexicon.

1:25:49 Ricardo Lopes: So let's move ahead here. I I think that earlier we've already touched on what consciousness is good for and you've also explained what resonance states and resonance mean. So I mean, another thing that people many times bring to the table when talking about consciousness, particularly the philosophers. And this is something that was coined, I guess by uh David Chalmers in the nineties is the hard problem of consciousness. Do you think that it can be solved scientifically?

1:26:22 Stephen Grossberg: Well, let me, um of course, it depends how you define it. Let me just read first Wikipedia says, and I'll quote, the hard problem of consciousness is the problem of explaining how and why we have quality or phenomenal experiences, um how sensations acquire characteristics such as colors and taste. So to a certain extent, and I will come back to this moment. All my theories do just that. Like my theory of vision is trying to explain how we see, you know, boundaries surf, you know, shading texture, color, all these properties of surfaces, paintings and so on. But let me just give another quote from Nagel in 74. He wrote in part when we see, for example, we experience visual sensations, the felt quality of redness, the experience of dark and light, the quality of depth in a visual field. Other experiences go along with perception and different modalities, the sound of a clarinet, the smell of mothballs. Well, um one can give a scientific explanation of what's going on in your brain when you see red, dark and light depth, um sounds of instruments, but no theory in science that's defined by mathematical equations as all scientific theories to the present have can experience the color red or any other conscious qualia equations don't have qualia. But in other sciences um like I I love this uh quote after you've read Lamb's huge treatise on hydrodynamics, you still don't know that water is wet, but you know, just about everything you'd ever wanna know about water and other fluids, including air um to design airplanes that fly boats that float on the surface of marines. You have an effective understanding of water and the fact that you don't know it's wet matters less and less because you can give a scientific explanation of what is happening. First, what characterizes the state of a liquid versus a gas versus solid through the molecular theory of gasses and this and that and the other thing. So it just doesn't matter after a while. And so in my life, given what I know that I know um the fact that I'm seeing beautiful colors around me, I can understand a lot of how that is happening and it is outside what an equation I can give you representations of the objects and the colors and their forms and their brightnesses. But I can't say why it looks red. I can't because an equation can't do that. And the the, but the answer is an equation can't do that in any science. You know, there's this famous um uh example, I hope I'm remembering it right? You know, when um astronauts were coming down and landing and they landed somewhere in, I don't know, Arizona or wherever or wherever, there was a flat arid surface. And the first thing I guess maybe these were the first guys to the moon. I don't remember. But the, the, the anecdote is the first thing they said when they got out was the bump, was there? Meaning that the computation of their trajectory coming in to the atmosphere predicted there would be a little bump and it was there. But that's not the same thing as the qualia uh f experiencing the bump. Uh So science can be effective and useful, but I do think it's outside the province of equations to say why water feels wet, you know, we can say why a fluid has certain properties but why it feels wet. I, I don't think equations can do and if someone can, I, I will study them passionately. I don't see how it can be done in principle though. Mhm

1:32:09 Ricardo Lopes: So uh let's talk here a little bit about some of the applications of the way you approach consciousness in your book. Before, toward the end, we talk a little bit about uh principles of brain design and some other general questions. So do you think that your approach to consciousness would have any implications for the way we understand mental disorders?

1:32:38 Stephen Grossberg: Well, um this is one of the things I call an acquired taste, just like consciousness was an acquire taste. It's not something I ever dreamed I'd have anything to offer. Um But so all of the neural models start by studying the brain dynamics and mental functions of normal or typical individuals because there are vast databases on that, you know, many hundreds thousands of scientists. Like if you go to the annual meeting of the Society for Neuroscience, I think a recent meeting at 26,000 scientists, they're all doing incredible work, huge databases. But at a certain point, I noticed that if some of the processes get imbalanced or damaged in specific ways, then I could see formal behavioral symptoms of mental disorders emerged. And, you know, I just had to notice that it was in the model. Uh And so it's in this spirit, I've so far explained behavioral symptoms based on precisely defined imbalances. Remember, complementary computing is a balancing act or, or actual lesions in Alzheimer's disease, autism, uh Fragile X syndrome, schizophrenia, medial temporal amnesia, visual and auditory amnesia and neglect and problems of slow wave sleep. So that came in through the back door just like all of it is. So the moral of the story for young scientist is uh don't look for shortcuts, uh try to understand the underlying principles and designs. Uh And then try to model it in an authentic and serious way. It's hard work. Uh Sometimes I wasn't ready to even write down anything about a problem for 10 or 20 years until I'd accumulated enough uh information to be able to proceed. But then uh I like saying it's the gift that keeps on giving consciousness was part of the gift that kept on giving, understanding a little bit of mental disorder is part of the gift that kept on giving. So if you try to get to the heart of the matter, it's richness will reward you many times over. But I've been lucky because I started very young. And even though it's controversial for quite a few years before neural networks were even a thing, there were always enough people in Washington who had faith in me and the work to give us enough funding to do the work. And it's not like the kind of funding where you need 50 million to get a fancy machine. All the money that supported me paid for uh graduate fellows ships and postdocs uh for people to learn from me and learn to do science with me. So it's pennies on the dollar was a good investment anyway. Um Did that say enough about that? Mhm. Uh

1:36:38 Ricardo Lopes: And another topic that you also cover in your book is irrational decisions. What is, what is an irrational decision exactly? And why do we make them?

1:36:54 Stephen Grossberg: Well, you know, there are several different kinds and uh let me just make a comment before I focus on one kind in general. If a decision fails to achieve one's goal, when another decision that would achieve the goals readily available, it wasn't chosen. We may want to say they were being irrational, you know, and just like in cognitive emotional interactions, you might have feelings about stuff and you just can't bring yourself to do it. Uh And there are many factors that can cause it, including an experience with this kind of decision making or the need to make a quick decision around the stress. You know, there are speed accuracy, trade-off problems. But the most famous psychological database concerns what's called decision making under risk. That is decisions in situations that are uncertain because the outcomes are defined by probabilities. You know, if you do this and that the probability of getting this payoff is that if you do that and this the probability of getting that pay off is that and which do people choose? And um Danny Conman and Amos Tversky won the Nobel Prize in Economics in 2002 for their experiments in an algebraic model. They published to fit that data in 1979 explaining paradoxical properties of irrational decision making in probabilistic environments and their algebraic model, which is, you know, phenomenological model that tries to compress the description of the data. And it is very useful, it's called prospect theory. But then in 1987 I think my student, Bill Gutowski and I uh went beyond it in a uh well known article in psychological review. Uh BECAUSE my article with Bill introduced a real time neural network model, we gave it a name because I started realizing you got to name everything you don't need to prevent other people from plagiarizing it. I won't stop them. But at least you could say that's just uh affective balance theory is what we called it that quantitatively explains and simulated the Danny Kamman and Amos Tversky data as emergent properties. And all we did was we applied my already known cognitive emotional model to that data. So we didn't do any more modeling. We showed that their data fell out of the wash when the model was exposed to the kind of risky environments like those in prospect theory. So that I thought had an important lesson that Darwinian selection does shape our mind to be adaptive, but it it fails in certain environments. Um You know, it can't anticipate every kind of risky environment. And in the same spirit in chapter 17, in my book, I summarize other examples of irrational behaviors including superstitious and self punitive behaviors in a similar way how in certain environments, you know, and you know, um conspiratorial thinking, sadly how certain social environments can support conspire uh conspiracy kind of ideas which we see all too much in the United States these days to say. So my abstract with Bill I, I summarize a little here. These explanations illustrate that the data concerning decision making under risk may now be related to data concerning the dynamics of condition and cognition and emotion as consequences of single psychophysiological theory. So in other words, you didn't have to build a new theory and it really, and we explained some risky decision making data that common and Tversky could not like pref what's called preference reversals. So our theory is more powerful both in the sense of um uh explaining their data as the merchant properties of known brain on the motional dynamics, but also explaining more data because Bill, you know, didn't experiment uh on the uh preference reversals. But you know, our paper was published in 87 uh Danny and Amos was in 79 They got the Nobel Prize in 2002. We didn't share it. So go figure.

1:42:41 Ricardo Lopes: Well, uh so we've already also touched on the or addressed the problem of cognitive impenetrability. So let's talk about something that we, you've alluded to in the beginning of our conversation, that is A I system. So do you think that having a better understanding of our brains give rise to minds would also be important for designing A I systems?

1:43:09 Stephen Grossberg: Well, you know, today, when people say the word A I, they usually mean neural networks, which wasn't the case when A I was symbolic, Marvin Minsky realized he ought to get on the gravy train. In fact, there's a funny anecdote about that. Uh Marvin Minsky and Seymour Pappert when they were doing symbolic A I, they were at MIT and I was an assistant professor and they really attacked me brutally because I was doing neural networks. And Marvin when he was a young man tried to do neural networks and he failed. And since Marvin was the smartest man in the world. How can this little dipshit kid be allowed to do neural networks? They attacked me. Anyway, years later, after I'd found it and became first president of the International Neural Network Society and the Archival journal neural Networks. And I was veteran chief for many years, I had a lot of influence over the organization of international conferences on neural networks. And I said, let's invite Marvin Minsky because he had just written his book Society of Mind, which basically was just plagiarism. But hey, that was nothing new for the People in Tech Square. And I figured, let's hear what Marvin has to say. Sadly, Marvin ever prepared. Um But he got up and the first thing he said was there are two things I regret in my life. One was I underestimated gross Burt. And then he, the rest of his presentation was telling jokes, he wasn't prepared poor Marvin. He was such an egotist amazing. And you know, he did some useful earlier work and then he just, you know, shoveled in the guilt anyway. Uh What can I say? So my neural networks illustrate autonomous adaptive intelligence and not all neural networks do. So I was trying to say now when people say neural networks or a I, they usually mean deep learning. And uh I published a paper a few years ago because I was invited to where I explain why deep learning is untrustworthy because it's not explainable. Mm If you make a prediction, using deep learning and let's say it's a medical prediction and someone dies as a result. And the surviving family are grief stricken and they say, why did you tell the doctor to do that? And your answer would be, I don't know, you're gonna be sued out for everything you're worth. So it's untrustworthy because is unexplainable and it's unreliable because it can experience catastrophic forgetting. In fact, in 1988 I wrote a off sighed paper that I review in my magnum opus with 17 prop problems that back propagation which deep learning uses as its learning engine still has that art never had and yet they still choose not to use it. So deep learning is deeply flawed and then chat GP t you know, a generative A I, I mean, you know, they've thrown into it immense databases, they scrub the internet. And so perhaps not surprisingly, it can generate a lot of phrases and sentences that sound intelligent. But basically, you know, it starts with a sentential context and has a probabilistic algorithm for trying to guess what to say next. The foundational problems are, it has no goals, it has no meaning and it literally does not know what it's talking about literally. And in fact, um on the connections website now, I don't know if you follow it. There's been this very heated discussion of this uh attack uh on A I of this kind led by Gary Marcus who's a professor at NYU and everything they're saying against it is correct. But I believe the future of humanity will benefit from an A I that is increasingly autonomous, adaptive and intelligent in its algorithms, machines and mobile robots, especially ones that embody brain inspired designs because these are the ones that can seamlessly interact in brain machine uh um co-operative ventures. And I'm the guy who's led that and we're well on the way to it in terms of meaning, you know, I don't want to presume that everything's understood. In fact, just last year I published an article on my web page about how Children learn in real time from their parents and other teachers to learn effective and perceptual language meanings. It's not in my book. I wasn't ready to write that paper till last year and that's just the beginning of what has to be a huge a research program that I clearly will not live to see completed. But at least now there's that foundation you can't say as people in a, I used to do all the time with the best game in town. They've never been the best game in town, but they spend so much time marketing with people like me, spend so much time thinking and writing and discovering I, I can't do that. That's not why I wanna stay alive. Um Yeah. So should I say, uh maybe I'll say, you know, the best example of autonomous adaptive intelligence is the human mind. We're autonomous because we can act on our own. We're adaptive because we can learn throughout life because we solve the stability plasticity dilemma. And we're intelligent. Hm, at least in the sense that we can as a species make wise enough decisions and plans to create civilizations, even though we can be superstitious, conspiratorial, terribly cruel and violent as well. And the end of my book discusses, you know, why they and you know, there's a kind of Yin Yang in the human mind and good and evil are built right into us. But I also try to clarify why good has a a leg up. Um BECAUSE bad things usually pub punished and either suppressed or avoided where you can, good things can be sustained and you can build beautiful things, important, useful things on them. So there's a broken symmetry and evolution between good and evil and our lives depend on it.

1:51:57 Ricardo Lopes: Uh So talking about good and evil and you also mentioned a little bit earlier. Superstition. Uh DOES the approach that you present in the book will also have implications for how we understand things like morality, religion, creativity and the human condition where I guess many of these topics are traditionally uh approached in philosophy.

1:52:24 Stephen Grossberg: Yeah. Well, um remember that Newton thought he was doing natural philosophy when you do deep science, it gets philosophical, deep philosophical. Um Well, I wanna do a quote chapter 17 in my book, likely discusses some of these, the things they could be discussed much more, more extensively on page 621 of the header of this section in question is more or less what I was just talking about symmetry breaking of positive and negative in cognition and emotion, coal and a reason for hope. I think hope is an important word and especially in our crazy world. Let me just quote it. It's not too long as with all predictive behaviors. There are cognitive and emotional reasons for morality that are built into the most basic properties of our brains in both cognition and emotion. There is, I'll argue a broken symmetry between positive and negative processes with the positive emphasized because it is just a broken symmetry. Much as every source of light can cast a shadow, both good and evil deeds are possible. Although I would claim there's a built in bias toward the good so that values aren't purely relative. And as I was just indicating, uh an example of the broken symmetry is that bad deeds are punished and punishment suppresses behavior uh which causes us to avoid them. Uh And if you don't avoid them, you end up in jail or worse, whereas good deeds are rewarded and thus can be sustained and built upon. Um And I believe in that very much, you know, it's a broken symmetry that's not very romantic, but I think understanding it, you know, you don't have to understand much more than on the level I said is very clarifying a lot of our development and evolution due to broken symmetries. Um YOU know, and just look at us, we're bilaterally symmetric but with broken, you know, we have a dominant hemisphere in our cortex, uh we can be right handed or left handed. So we are broken but we are symmetric and is built into everything. And I believe that a lot of uh the origin of the universe, although I shouldn't talk about things I know nothing about. I, well, not, not that I know nothing but I think a lot of broken symmetries drove evolution. So,

1:55:50 Ricardo Lopes: so uh let's uh we've already also talked earlier about the brain design principles that you uh uh go through in the book Complementarity and Certainty and Resonance. So perhaps let, let's go back to the very beginning of our conversation when I asked you about when psychology slash neuroscience and physics diverged in terms of their approach to the brain and the mind. Do you think that the paradigm you present in your book can help unify them again?

1:56:25 Stephen Grossberg: Well, let me quote again from my book and then if uh maybe I'll make some other comments on pages 635 and 636. Again, chapter 17, as I explained in chapter five, brains can also bind into coherent moments of conscious experience within a unified self measurements taken from multiple physical sources light sound pressure, temperature, bodily biochemical processes and the like. Um SO like when someone asks, how do you feel today? You can tell them all sorts of stuff. And it's not just like, you know, I have a light meter, a sound meter, a pressure meter, it's all unified into a coherent moment of conscious awareness. So the brain is also a universal. Remember I talk a little about that self organizing measurement system of the world and in the world. Um CHAPTER one explained that due to this universal measurement process, brains may play a role that'll be compared with the role played by played in physics by Max Planck's law of black body radiation. Cause by describing a radiation spectrum that's universal across all matter planks, lushes and quantum mechanics. Whereby to explain all subatomic phenomena, the brain's universal measurement process can be expected to have a comparable impact on future science. So, you know, we don't, our brains don't function when it comes to cognition, emotion, et cetera on a quantum level. Um YOU know, even if there are uh noisy fluctuations in individual cells, often cell populations have mean activities uh that are much larger than the variants. So we can have a a kind of um uh deterministic or mean field dynamic equations or operating dynamics that's not operating uh with quantum fluctuations uh needed to be considered. But we have this universality in terms of all these physical fields as a measurement device that's self organizing. And I um also in chapter 17, talk about how some of these principles um like complementarity uncertainty and resonance operate even in single non neural cells, a a very primitive uh cell types that have existed long before humans. So there is a universal developmental code in which art adaptive resonance is part of that larger code. It is one example of the kind of resonance that you find uh even in morphogenesis of very simple species. Um So, you know, I talk about the kind of resonance that occurs when a blastula becomes a gastro, I talk about. Um uh WELL, you know, read the book, I'm getting a little worn out of my search and my brain isn't. So there is a universal developmental code that um that includes the things that we experience when our brains start actually enabling us to remember what we know about the world. And resonance goes way way back in morphosis. And you know, I've always thought, but you know, maybe if someone makes this precise or shows its wrong, I've always thought of um the double helix as an example of complementary computing. And you know, Barbara mcclintock got the Nobel Prize for showing it's very dynamic. Um So, you know, if that is properly thought of as an example of complementary computing, uh There are many examples of this kind of complementarity uh throughout all of life, um et cetera, et cetera, et cetera. Uh And as I say, you know, when, um so for example, some of the competitive dynamics I discuss like mass action recurrent on center off around networks competition uh is universal, not only when you think of it on the species level, a a Darwin, but it also occurs at every level of cellular organization. Uh So the aggregation of slime molds is an example of a competitive dynamic with an on center of surround recurrent network. Uh So there were these universal computational principles that are not only about us that are also about all living things. And you know, when I read theoretical physics, which I've been doing since I was a boy, you know, I see what I view as physical versions of on center of surround networks galaxy as it were, you know, it's a balance balancing act, but I can't say anything useful about that.

2:03:31 Ricardo Lopes: So, uh let me just ask you one last question then and I want to ask you this because I also, I actually found it very interesting that you talk about this in the book and whenever people in science talk about laws many times, uh I mean, people get upset about talking about laws that are not laws of physics, I mean laws of biology, psychology or whatever. I mean, people are not usually very fond of that. But do you think that we can actually talk about laws of cell biology?

2:04:08 Stephen Grossberg: Well, cell biology is vast and these days he biology usually means genetics. And uh I'm not a geneticist, although I did just make a very speculative comment about the double helix. Um So I'm not sure what context to put the question, you know, it is part of cell biology. What I said about um uh slime mold aggregation, their cells, their biology, the growth of hydrous heads, their cells. It's biology in that sense, there are examples of shared laws, um you know, gastro relation, it's cell biology, it's not in a brain, there is no brain yet. So um I feel, you know, even my work has shown that what I'm doing is a special case of more general laws that you can see reflected in examples of cells in biology. But um if you, if what you mean is genetics, then I'm not a geneticist, but it does strike me that the double helix. Uh AND for all I know it's being done, maybe the stuff that Barbara mcclintock got started is maybe being studied very actively and I just don't read about it in terms of the dynamics of those complimentary strands. Um YOU know, complementary computing generally gives you a form of stability. You know, the Yin and the Yang alone aren't enough but together and bound by the right kind of feedback. They can give you stable organization just like with the boundaries and surfaces, for example, like with the water and West streams, one of the great triumphs of the watch stream, which I try to explain through models is how you develop invariant recognition categories, size and variant view and variant position variant. So I could recognize you from any view, distance angle, you know. But the problem with being position and variant is at the place where you're computing this and the anterior of infra temporal cortex, I have no idea where your face is in space. And that's what the, so that's in the W stream, the West stream gives a spatial representation. So you have to be able to go from an invariant representation into the watch stream into a variant position representation in the West stream. And I explain how that happens. Um uh But by themselves, you know, you'd have positions without objects and objects, you didn't know where they are. So each of them is useful and important, but it's not enough his survival. So complementary computing on every level of our brains is necessary for survival. And it may very well be uh at every level of cell biology. But I'm not a person who studied non brainin cell systems for long enough to make a strong comment. I'd love to know more about it. Maybe I'll do more of that. Uh In my remaining years. I don't know.

2:08:34 Ricardo Lopes: Well, uh there are many other topics that of course, we could cover here because the book is massive and very detailed. And again, it is the conscious mind resonant brain how, which brain makes a mind and I'm leaving a link in the description to it and I hope that my audience runs and buys it. It's a very interesting book. And uh Steve, apart from the book, would you like to tell people where they can find you and your work on the internet?

2:09:06 Stephen Grossberg: Oh, yes. So I had mentioned once, but let me say again, I have a personal web page sites, si tes dot bu.edu. So sites Boston University Education slash Steve Gsteveg. Um uh All my articles can be downloaded from there and quite a few interviews and uh some videos of keynote lectures on different topics. Um And if anyone wants to write me with a question about my work, my email is simple. It's just Steve Steve at bu.edu. Steve at bu.edu. So I love to discuss science. And um um what can I say? I started 67 years ago and I'm still at it. And um and that's a testimony and I only need to my good luck but to really respect the foundations of what I was studying and not go for shortcuts and they've thrust me forward. So I never hit a brick wall. And uh I've loved every minute of it even though it's been really hard work.

2:10:42 Ricardo Lopes: Well. And uh thank you so much for that work and it's been an immense pleasure to have you on the show. Thank you so much for accepting the

2:10:52 Stephen Grossberg: invitation. Thank you for inviting me and I'll look forward to getting the URL for it so I can uh hope I don't cringe too much.

2:11:06 Ricardo Lopes: No, it was great. It was great. So, thank you again.

2:11:10 Stephen Grossberg: My pleasure.

2:11:11 Ricardo Lopes: Hi guys. Thank you for watching this interview. Until the end. If you liked it, please share it. Leave a like and hit the subscription button. The show is brought to you by N Lights Learning and development. Then differently check the website at N lights.com and also please consider supporting the show on Patreon or paypal. I would also like to give a huge thank you to my main patrons and paypal supporters, Perera Larson, Jerry Muller and Frederick Suno, Bernard Seche O of Alex Adam, Castle Matthew Whitting B no Wolf, Tim Ho Erica LJ Connors, Philip Forrest Connolly. Then the Met Robert Wine in NAI Z Mark Nevs calling in Holbrook Field. Governor Mikel Stormer Samuel Andre Francis for Agns Ferger, Ken Herz J and Lain Jung Y and the Samuel K Hes Mark Smith J. Tom Hummel Sran, David Sloan Wilson Yasa, Dear Roman Roach Diego, Jan Punter Romani Charlotte Bli Nicole Barba Adam Hunt Pavlo Stass, Nale Me, Gary G Alman, Samos, Ari and YPJ Barboza Julian Price Edward Hall, Eden Broner Douglas Fry Franka Gilon Cortez or Solis Scott Zachary. Ftw Daniel Friedman, William Buckner, Paul Giorgino, Luke Loki, Georgio, Theophano Chris Williams and Peter Wo David Williams Di A Costa Anton Erickson Charles Murray, Alex Chao Marie Martinez. Coralie Chevalier, Bangalore Fist, Larry Dey Junior, Old Einon Starry Michael Bailey. Then Spur by Robert Grassy Zorn. Jeff mcmahon, Jake Zul Barnabas Radis Mark Temple, Thomas Dvor Luke Neeson, Chris Tory Kimberley Johnson, Benjamin Gilbert Jessica. No, Linda Brendan Nicholas Carlson, Ismael Bensley Man, George Katis Valentine Steinman, Perlis Kate Van Goler, Alexander Abert Liam Dan Biar Masoud Ali Mohammadi Perpendicular J Ner Urla. Good enough Gregory Hastings David Pins of Sean Nelson, Mike Levin and Jos Net. A special thanks to my producers is our web, Jim Frank Luca Stuffin, Tom Veg and Bernard N Cortes Dixon, Bendik Muller Thomas Trumble, Catherine and Patrick Tobin, John Carlman, Negro, Nick Ortiz and Nick Golden. And to my executive producers, Matthew Lavender, Si, Adrian Bogdan Knits and Rosie. Thank you for all.