BI 231 Jaan Aru: Conscious AI? Not Even Close!

[00:00:03] Speaker A: Consciousness can be a computation, but it's a freaking complex computation that happens only in biological tissue that is so far ahead and away from anything that happens in large language models that the prior of basically having a consciousness in large language models is near zero. It's an inspiration for the younger generation that we don't have to be stuck with these old theories of consciousness. We can look into the brain, we can come up with more complex theories. Most people just, you know, they say whatever they think about consciousness and that's one of the frustrating things about consciousness. And that's one of the reasons why I sometimes have these off times because everybody's just getting their own stuff about consciousness and you know, ah, I want to do real science now. [00:01:15] Speaker B: This is brain inspired. Powered by the Transmitter Yann Aru is a co principal investigator of the Natural and Artificial Intelligence Lab at University of Tartar 2 in Estonia, where he's an associate professor. Jan's name has kept popping up on papers I've read over the past few years, sometimes alongside other guests that I've had on the podcast, like Matthew Larcomb and Max Schein. With Matthew and Mac and others, he has co authored papers exploring how some of the pesky biological details of brains might be important for our subjective conscious experience. Details like dendritic integration and loops between the cortex and the thalamus. Turns out, a recurring theme in his work is to connect lower level nitty gritty biological details with higher level cognitive functioning, and he has some thoughts about what that might mean for the prospects of consciousness in artificial systems. So. So we discuss many of those thoughts and we also touch on his more recent interest in understanding the brain basis of insight and creativity, connecting some of the more mundane kinds of insights during problem solving, for example, with some of the more profound kinds of insights during mystical and psychedelic experiences, for example. Find more in the show notes at BrainInspired Co podcast 231 where you can also find out how to support this podcast on Patreon to hear the full version of this and all episodes, or simply to express your appreciation for this here podcast. Thank you to my Patreons, as always. And now from Estonia. That'll make more sense if you stick around until the end of our conversation. Here's Jan. So yeah, I don't expect you to know this, but you've actually been on my list for a long time for someone to invite for an interview, and I don't remember exactly how I came across your work. And maybe it's just my interest in consciousness, but maybe it's because you've co authored papers with people like Max Schein and Matthew Larcombe who've been on the podcast before. Yeah, they are good people. That's one interesting thing, is like, they're really good people in academia, I find, which it's not necessarily has to be the case. I mean, of course, there's all varieties of people, but already I'm getting us off track here. So anyway, maybe we'll talk a little bit about your work with them along the way. But as I learn more about what you do, I kind of found it hard to nail down what drives you, what your interest is. Right. Because you have your fingers in a lot of different pots. And so maybe we could just start with what is that sort of the overarching drive of your work? I mean, there's creativity, there's intelligence, there's consciousness, there's what else? And I don't know, where does it come from? [00:04:22] Speaker A: Yeah, basically my agenda in science is that I'm free. Academia, for me is mental freedom. And whenever I feel that, oh, here is an interesting problem that I could attack, then I go for it. So, for example, I started off with consciousness, but at some time point it got a little boring for me. And then I was like, okay, what, what else? Big problems are there. And then I got really interested in insight, which is a really specific problem. Right. In problem solving, few people studied, but I really got really interested about it. And then I convinced some people to study it with me, and we wrote a couple of papers that are already useful for the community. So I do it a lot like that. I see, I feel this freedom. And whenever I see it, oh, there's an interesting problem, then I go to it and try to really get my head around it and do my best to figure things out. [00:05:28] Speaker B: Well, I've heard you mention that you at least often write papers out of frustration, that your papers are often born, the initial seed is from frustration. So maybe we'll get to that a little bit. But what do you mean? You feel freedom in academia? You feel freedom. That's an oxymoron, right? [00:05:45] Speaker A: Well, but for me, it is when I tell to young scientists about why to come to academia, then there are so many reasons why not to come. But I say the reason why to come is freedom, potential freedom. [00:05:59] Speaker B: Not everybody feels that freedom. [00:06:00] Speaker A: I don't understand how you. That's why I laugh. That's why I'm happy. I feel nobody can put the pressure on me that, Jan, you have to study this thing. Well, a Bit. A bit maybe. But you know, I don't like that. And I really like to study these difficult questions where most people would say yaan, why do I even bother? We will not figure these things out. But you know, I'm free, I want to study these things. And that's how I try to explain it also to my postdocs and PhD students that when you work with me, I will give you guidance, I will help to, you know, assist you in your way. But I want you to learn this freedom in academia. But I understand that it's not, not like that everywhere. [00:06:46] Speaker B: Yeah, well, it's really nice that I can tell there's kind of a joy from you in what you do, which is fairly rare. Fairly rare. Usually it's tenure that people seek to finally achieve that freedom. So you've gotten lucky and you have that disposition, it sounds like naturally, perhaps. [00:07:02] Speaker A: Yeah, I mean it's important for me and probably over these two hours you will see it many times that you know, when somebody who is already higher, in a higher position in science says something, then I never think like, oh, okay, it's like that. I always think, wait, do I actually agree? Because yeah, I don't have to agree with people, even those who are higher up in the hierarchy. Yeah, for me science is about, you know, being skeptical of everything and being free in the way you think. [00:07:36] Speaker B: So, so we're going to talk about consciousness, even though you find it very boring now apparently at times, I mean. [00:07:42] Speaker A: I have this on, off relationship. So during my PhD and before I was really, really high and dating consciousness basically every minute of my life. But then, you know, there was an off ramp after. Why? [00:07:55] Speaker B: Why the off ramp? [00:07:57] Speaker A: Well, because of course I went to the PhD that I did with Lucia Maloney and Wolf Singer in the Max Planck Institute of Brain Research in Frankfurt. I went with the idea that I will solve the problem of consciousness, which is a bit naive perhaps. But then during that PhD we published pretty well known paper that shows that you cannot solve consciousness the way people usually have tried to do it. So obviously it kind of led to this frustration that, you know, you go into the PhD with an idea of solving something and then you show you cannot do it. Actually that. And then I was frustrated and I searched for other topics after that a bit. [00:08:43] Speaker B: Yeah, yeah, all, all it does is produce new problems. Trying to solve, trying to solve a big question like that produces a lot of extra problems. Yeah, yeah, you've written a lot about consciousness, you've studied artificial intelligence and you know, there's this idea, oh, are artificial agents conscious? Can they be conscious? And a lot of what you've written about is studying this difference between biological and artificial systems with the pursuit of seeing, you know, finding differences, similarities and differences. That makes us out whether we should even believe that artificial systems could be conscious. But, but you framed it in, in a way that, like when people talk about artificial consciousness or artificial intelligence being conscious, that's, that's kind of upsetting to you. Why would that be upsetting to you? [00:09:33] Speaker A: I mean, I first give the short answer, short answer is that many people, even very, very clever people say that, you know, we don't know enough. So the centric, centristic position would be that, you know, maybe they are conscious, maybe they are not. We should have some kind of freedom there. But I think the prior, if you're a neuroscientist, then the prior that these systems are conscious is zero, basically, just nothing. It's zero. Right? [00:10:05] Speaker B: Yeah. [00:10:05] Speaker A: And then I feel this frustration. That's the short answer, I can give the long answer. [00:10:10] Speaker B: I share that with you too. But then you start like looking at all the issues, you start getting into the details and there are seeds of doubt, right? And even in your own papers you write, well, obviously it's not conscious. Could be one day, if we do this, we do that, we do this, we do that, right? And there's always that. And then AI folks can say, all right, well let's just implement that. And then boom, it'll be conscious. And so where is the end of that? Right? You could do that in infinite regress. So I don't know where the satisficing line would be right, to then say, ah, it's conscious. But, but I share that sort of zero prior intuition. But then I know I'm being a bad scientist because that is my bias, that is my prior. Right? [00:10:54] Speaker A: Yeah, well, it's not, it's not exactly zero. But you know, I always say, you know, if to me, all the others exaggerate in the wrong direction of all these might be conscious or, you know, so I exaggerate in the other extreme. But to your point, so as we also have written in some of our papers, right, there are these so called easy problems in the sense that, for example, we kind of think that a specific architecture that the brain shares has to be there for being conscious. We can discuss it later and something about thalamus, about the dendrites. But when we thought about these things, then I think a clear answer comes to mind indeed that these are easy problems because One could implement these things and then say, look conscious. So ever since 2022 I've been searching for, okay, what's a better answer? What's a more complex answer? And I think we have now at least two publications where we kind of delineate that. One is the 2023 paper in Trends in Neuros and that just came out in the end of 2025 in neuroscience Peer behavioral Reviews, where we basically say in very short terms, consciousness can be a computation, but it's a freaking complex computation that happens only in biological tissue that is so far ahead and away from anything that happens in large language mod that the prior of basically having a consciousness in large language models is near zero. That's a short way of answering that. [00:12:52] Speaker B: Yeah. And those two publications you just mentioned, one is short, the trends, one is short and then the more recent one that you referenced there actually goes into a lot of nitty gritty detail into how computers work and how the software hardware distinction and how memory is separate, et cetera. And it really goes into detail into why what you end up discussing partly why these integrated across scales kinds of computations and biological systems should be considered more important. Right. And that gives rise to more complexity. And what you just said is that the complexity. I'm paraphrasing what you just said. The complexity is so far removed, so far so much greater in biological systems that it's kind of a non starter. But is that the key is just complexity? What am I missing there? [00:13:53] Speaker A: Well, yeah, complexity sounds like very trivial, right? But I think that's already an important first step because when you come to consciousness science, then there are a lot of people, extremely smart people in cognitive science, cognitive neuroscience and psychology. But it turns out they have a really only a very basic understanding of the brain. So for them it seems that, you know, oh, the brain is very similar. You have neurons and you have synapses. But as you know, and as every neuroscientist would know, no, that's a very, very coarse abstraction of what the brain does. And in fact, the more we learn, the more we figure out that, you know, these nitty gritty details also matter. So the first step is to say that look, of course if you abstract on the network level, then it looks similar to AI systems. And if network level is the only level that is responsible for consciousness, perhaps. But the brain is way more complex in the sense that if you take the one unit, then if you take it from an artificial neural network, I mean, it's just a piece of code, it's nothing. It's. It's a piece of code, a very simple piece of code. But if, you know, if I would come open your brain, take a few neurons from your brain, put them into a petri dish, they can do things. They still are like living, doing something. And if you look into them, there are billions and billions of biochemical processes. So on the level of single unit, there is this enormous complexity that is not captured by artificial neural networks. And then to say, and some people have said it right, that artificial neural networks are nearing the complexity of the human brain or rat brain is just naive. It means that these people know about the brain, so that's about complexity. But now, the point we have made in these two papers, and much more thoroughly in the newer paper with Borian Milinkovic, is that, you know, it's not only the complexity, but the fact that the brain has these several scales of processing that are very different. The network level is just one slice of that. But the beauty of the brain comes from the fact that you might have a very tiny event, a molecular event, that then moves up the scales to network level, even affects the behavior, which in turn changes the molecular level. And that the real complexity that interests us happens at this intersection of these different levels and that these interactions. And the point I've been trying to make throughout these two papers is that we, as scientists, we don't understand the computations happening across these scales, these interactions. And if we would understand them, perhaps then we would also be a step closer to consciousness. But that's where the complexity. So it's not simply having the same thing. More, more, more, more. Scaling up is the word people use in AI. It's not about scaling up. [00:17:41] Speaker B: It's about they'll be scaling up at the same scale, right? [00:17:47] Speaker A: If you're scaling up the network level, that's not it. Because the brain is not only at the network level. And you can have discussions about how important the network level is. But having studied the brain and thinking about it, I would say that we actually often see that the more we study, the more we have to understand also the cellular level effects, right? I mean, think about memory, or working memory is a very good example where for a long time people thought it's the persistent activity of spiking, right? But now, over the last 10 years, 15 years, we have had the idea that now it's also, it can be like these silent memories that are not in spiking, right? That's just one example. And if you go to the broader memory research, then Again, for a long time people thought, oh, it could be just some spiking patterns. But now when we think about engrams, it's not only spiking patterns, it's a question of which molecular changes you have there. Right. And I think the same will happen eventually with consciousness. [00:18:58] Speaker B: For me, a cross level understanding is sort of one of the dreams. But a lot of people who are working within one level find, okay, well, we don't understand this at this single level. And it's sort of safe to stay in your own level and say, well, if we don't understand, we need to understand it at this level before we can start talking about this inter scale combination or integration or interaction between levels. And part of that frustration also is that we don't have the conceptual or mathematical tools, we don't have many to study across scales. So it's an enormously difficult problem. And I share with you that that is where the important part is, because that's where all the complexity happens. Right. And that's like the most difficult thing to like, well, how, if we, if we can't even understand it at one scale, how are we going to understand the interactions across scales? [00:19:58] Speaker A: But that's exactly science for me, right, that you don't go for the easy programs, you go for the hard ones. And I of course completely understand everybody who has a research program. They are successful getting papers in nature neuroscience, one after another doing the same thing. Right. I respect that. But yeah, my own thinking, as I already expressed in the beginning, is kind of different, that, you know, you have a very short life. I mean, not you, but you know, people in general, not me. And you know, you only have a couple of chances to do it. So ever since the start of my scientific career, I thought I just, you know, I pivot and I do something where I feel that it's, it's complex enough, but I can actually get something done. So, and I hope that, you know, more scientists also see it the way we do, that, you know, yeah, it is difficult. Yes, we are missing these tools, yes, we need new mathematics, etc. But that's why it's exciting and that's exactly why we have to do it right now. [00:21:08] Speaker B: Yeah. Okay. So I mean, lots of things to pivot off of, like what you were just saying, keeping with levels for a moment here. What you described is, okay, that's fine. You could say, okay, that's fine. Let's look at how the cellular level interacts with the neuronal population level. And we need to see how the population constrains the cellular activity. So that's just like across two scales. Right. But then there are some, you know, plenty of people think when you go up to the, when you jump across into the mental level, let's call it a level scale, whatever. Like when we're using psychological terms that because of things like emergent properties that you can, that there's some boundary and you can never really bridge the divide between the mental and the physical. Everything you were talking about was physical, right. Whether we're talking electrical fields, action potentials, cellular signaling at the lower spatial scale, for example, you know, neural populations interacting with each other, et cetera, et cetera. But, but then there's that physical to mental divide. So do you see this as a clean part of that scale integration or, or, or should they reside at different levels? I mean, like, like many people suggest. [00:22:32] Speaker A: Yeah, so first clarification. Also, some of my co authors are already probably mad at me that I didn't say it earlier, that you know, in some sense we would even go one step beyond and say that actually there are no clean levels in the brain. That it's, it's just our theory that puts it into these levels that in fact, if you, if you are a neuron, then you know, the signals you get are molecules. You don't know whether it's a network thing or you don't know whether it's a neurotransmitters or what it is. So. So in some sense the levels are much more tied in the brain than it, than it usually is taught, I think. Again. Well, if one operationalizes and you know, says that, you know, here we do these particular measurements, then it's fine to think about the scale. But in some sense in the brain it's very fluid. And sometimes he would say that, you know, you cannot really separated these scales. [00:23:35] Speaker B: Yeah, but we have to separate so we can name the thing that we're studying. You know, so the way that you guys phrase it in, or in the most recent paper, you and Bjorn, I think, and the way it's often phrased is that there's no privileged scale at which like computations happen. And what you just said is a little bit different. And you're saying that there's no, there's no level because it's. The scales are integrated. You have to think of it sort of as one fluid across scale process. Would that be the right way to phrase it? [00:24:05] Speaker A: Yeah, again, and it's fine if when people study it and they say that, you know, I study things on that scale. But when it comes to neural processing, that really happens then sometimes, you know, I think there's, you could say that these scales are not so clear cut. If you think about what a neuron is actually doing and sensing, et cetera, that it's much more interesting. Interesting then that you, I mean, I say that you have molecules, but you have also like electric fields from the other neurons. And well, in some sense electric field should be on a higher level than the molecules that are coming. Right. But are they really for the particular neuron? So it, and it's, it's, in short, it's just much more interesting probably. And I think the key idea here is that I think it broadens the horizon of thought. If you think, oh, maybe the scales are kind of an abstract thing we do to study these things, but in the brain it's not so clear cut. Right. [00:25:10] Speaker B: Yeah. In my current research, I often feel I sort of roll. So it's very exciting to me what I do. I have fun. I'm analyzing spike counts in populations of neurons. Right. But then very frequently I think, oh God, I'm just looking under the light for the keys like we all do. Like looking at action potentials, counting them, looking at their statistical patterns with each other in relation to each other. And that's such a, it's such a small, narrow view and it's the assumption that neuroscience has been going off of for many, many years now. Well, that's, oh, it's a spike. That's what we should look at. Right. But I know that what, what is more important, I'm with you rather, that what is more important is to look across scales and how there are interactions at all across all scales. And that neuron spiking perhaps, you know, is not the place. It's, it's one little place to look and that's where we're putting all of our resources. So maybe that's just commentary, less of a question. [00:26:15] Speaker A: No, but it's, but it's correct. And I, when, when somebody feels like strange about this, then just think like, well, how much of the fact that we study spikes is historical simply because people figured out the way to study spikes and then that was a scientific method you could use. But let's say in an alternative universe, people would have first figured out how to cleanly measure calcium activity or something like that, not spikes. [00:26:46] Speaker B: Right. [00:26:46] Speaker A: We will have a completely different theory about how the brain works. And, and, Right, right. [00:26:53] Speaker B: So it's like studying, you know, you go to like a rainforest and there's like one person in the rainforest, like standing there clapping. And that's what you can hear is the clapping. So you're like, ah, I shall pay attention to the clapping to understand this rainforest. Just a little crazy. [00:27:07] Speaker A: Exactly, exactly. That's a great analogy. And something like that happens. And he, again, it's okay to study spikes, but perhaps we should do less of that and perhaps we should just acknowledge that the fact that we study spikes and have historically studied spikes does not mean that they are the key currency of our intelligence or of consciousness. [00:27:33] Speaker B: Yeah, I guess. So what sent me on that little aside was the distinction between a privileged level and then what you were describing, that there are no levels, essentially, or there are no clean separations between scales because of the complex interactions between scales. I don't know how else to say that. But that original, then originally, before that, I had asked you about that distinction between the implementation sort of details, cells, populations, subcellular processes, and then what we consider the mental, psychological, mental phenomenon. Do you see that as part of, part of all the same scale if, if our mental world arises from, or is closely linked to, or however you want to phrase that, the, this across scale processing, you know, do you see a divide between there or is it just part of the continuum? [00:28:33] Speaker A: Yeah, yeah, no, it's, it's, it's a great question and one another clever, clever colleague also pointed out that, look, you know, of course you say that everything is tied up, but isn't consciousness like very cleanly separated from everything else in the brain? And I have to say that, you know, I have to take this point. I mean, I cannot say that, you know, consciousness is clearly intertwined with all these other things, at least as it looks, as it seems to us. So I don't have a good answer to that conundrum yet. [00:29:10] Speaker B: If we were three beers in and I, and I pinned you down and I said, do you think the brain generates consciousness or do you think consciousness is fundamental? What would you say? Or, or there, there, that those aren't the only two options. [00:29:25] Speaker A: But. Well, yeah, I mean, I, I, I tend to, Again, it, it segues a bit into, into what I have said also in the beginning that I try to be proud that I have a very open mind and I change the way I think. So I would, I would say that for, for years and for my earlier, earlier years, I would say the brain generates consciousness. [00:29:50] Speaker B: But that's what we're told, that's what we're told when we get into neuroscience. [00:29:53] Speaker A: Yeah, yeah, yeah, more, more like, yeah, yeah. But now I would agree that, you know, people like Tononi and Koch and others, they, they, they have a point that, you know, consciousness could also be a property of, a fundamental property of some kinds of processes. But I wouldn't perhaps agree with the rest of them, perhaps the rest of their arguments. I would say that if it is a fundamental property, perhaps it's a fundamental property of a system where different scales interact, for example. [00:30:25] Speaker B: Okay, another thing that I wanted to give, go back to. So you were talking about the level of like the network level that artificial systems are, that artificial systems implement digitally essentially. And one of the, and I completely agree with you, one of the surprising things about artificial systems is how quote unquote intelligently they can perform certain functions given that they have abstracted out all of those biological details, not just at the singular cell level, but the dynamics across populations, across scales, across levels. And so I don't know how you think about this, but you know, I don't know if you think that intelligence and consciousness are orthogonal or if they are related. My own sense is that they're completely orthogonal is my current sense. But anyway, it's impressive how much I can do function wise with so little with, with such bereft biological implementation as you were noting before. So how do you think about intelligence versus consciousness in that regard? [00:31:41] Speaker A: Yeah, I mean, exactly, exactly as you say. First, it's crazy really how, how many tasks these systems can. The thing we always have to keep in mind, and that I always show to my students is also that the energy demands are like millions of times, probably more than millions of times higher for these AI systems than our 30 or 40 year old brains. Right. But given all that, I think, yeah, at the moment I really also support the way Anil Seth has been conveying the message that look, as you said, they can be orthogonal. You know, we don't know if they are exactly orthogonal. They are perhaps not exactly orthogonal, but consciousness and intelligence, they are clearly not the same thing. And again, it's a problem if somebody thinks that oh, these systems get more and more and more intelligent at some time point, become cause of that they will be conscious. [00:32:45] Speaker B: No, so incredibly naive. It's so incredibly naive that. [00:32:49] Speaker A: Yeah, yeah, yeah. But I think the problem, I mentioned it briefly also the problem is that, you know, many of these people, they get paid high salary, they, people look up to them, so they never actually take the time to figure out how complex the brain is. And then, yeah, then they even cannot understand that this is naive. And that's sad. That's sad. So I really feel that at these times, you know, that the voice of neuroscience is not heard in the sense that most people just, you know, they say whatever they think about consciousness. And that's one of the frustrating things about consciousness. And that's one of the reasons why I sometimes have these of times because everybody's just. Just everybody has their own stuff about consciousness. And, you know, I want to do real science now that. [00:33:45] Speaker B: That is interesting. Like, when I first started this podcast, I asked everybody about consciousness and like, 70, 75% would say I just, I have too much to work on right now. It's, you know, why would I. Like, there's so many unknowns with consciousness. The problems I'm working on right now are interesting and important, and therefore I don't have anything to say about consciousness, which I respect, actually. You know, neuroscientists complain that, like, the computer science world and the AI world doesn't listen enough to us, but I. But I find, you know, they don't really need us right now, is my sense. [00:34:22] Speaker A: They don't need. Yeah, yeah. Although. Although there is like a new trend of people talking about AI welfare and AI consciousness and that this is an important topic. And I sometimes feel that this discussion is not really honest, that in some sense it's a hot topic. So you get invited to companies, you get high salaries for that. But actually it's not scientifically honest to talk so much about these topics. Again, of course, if you do not have this neuroscience background and you don't have this knowledge, then perhaps for you it is honest in the sense that you're not making up things up, but you're inside your own sandbox. It really seems that these systems can be conscious, and then it's a program on us neuroscientists to actually kind of try to convey this message that the brain is just way more interesting. [00:35:26] Speaker B: Also, at the same time, neuroscientists have a lot to contribute and yet should also be humble because it still is a mystery, the mind. [00:35:36] Speaker A: I think that's. That's the key. Everybody should be humble, you know, and. Sorry, sorry if I don't come across as humble. Actually, I, with respect that program, which I have studied, like, 22 years. I am very humble. [00:35:51] Speaker B: Yeah, yeah. Whenever someone says, I definitively know the answer to this, like, it is a red, red flag for me anyway. [00:35:58] Speaker A: Yeah. And. And it's. And it's the same thing, actually, about many things. Brain. Right. That there are many things we don't know. And I think. And that's what I always tell to young scientists also, just be humble, try to figure something out. But you know, don't believe anyone who tells you that they, you know, they really know that. Because in the brain everything is a mystery. Not everything, but, but, but most of the interesting things at least. And that's great, that's great for a young scientist and also still for me and you too. [00:36:27] Speaker B: So you're interested in studying across, studying the brain and the mind across scales. And one of your passions, I guess is that you've mentioned is connecting the, you know, the very low level details, subcellular processes, dendrites, et cetera, to these higher level cognitive abilities and capacities and functions. And we were just talking about how that's kind of the dream and how the, you know, we don't have the right tools yet necessarily, and that's why to do so and that's why it's exciting. So maybe, and you know, you've, you've studied with Matthew, Matthew Larcombe and dendritic integration theory, which, well, we can get into the details of that, but that's sort of low level, nitty gritty details that Matthew thinks are important. So maybe I could ask you where you're most comfortable or excited. Like what. It's an unfair question because what your real excitement is connecting is the bridging between the levels. But you know, go ahead. [00:37:31] Speaker A: But I think that that is the answer, that is the answer that, you know, I worked at this intersection of cognitive science, neuroscience and, and artificial intelligence. At each of these levels there are so many people who are way more clever than I am. But the large majority of these, they never think about preaching these scales, never, they don't care. But I'm one of these people that, you know, really, I, I like to preach things, I like to make these, put these things together. And sometimes other scientists tell me, you know, that's stupid and it's far fetched and we shouldn't do it. But so far I've been very, very fortunate also to find editors and reviewers who say that, yeah, it's a bit crazy, but it's, it's good enough in the sense of, of bringing science forward that you know, we try to do these speeches but because I think, and I think I hope it comes across, for me this is the goal of, of, of neuroscience and understanding the brain that should try to breach these things. And it might be a bit too early, but you have to start at some time point. And you know, for me it's just at least the most fun also. [00:38:52] Speaker B: Yeah, so There is like, for instance, in the AI world, right? There's this cottage industry of pointing out what AI can't do. And it's pretty easy to find problems in, in. In whatever field of study that, that you're in. It's. It's fair. And this goes back to destruction is easier than creation. Right. But at some point we have to, like, work on those problems. What. What do you think is the right balance for you to point out? Like, for example, what's missing in AI and why it would be important to integrate these things. So pointing those things out is one thing, and then actually working on the solutions is another thing. What's the right balance for you? [00:39:35] Speaker A: Yeah, I mean, in some sense I feel that I've been writing too much of these papers where we come up with some idea that's interesting to study, but then nobody studies it because everybody is busy doing their own thing. Right. So in that sense, I'm really, really happy to be working also with Borian Milinkovic, because we had this paper that we mentioned already in neuroscience and Pure Behavioral Reviews, but Orian, he's a serious guy who really wants to not only have this kind of conceptual paper, but really figure it out formally and then also try to implement it. So I'm very happy about that collaboration where we see that it's not only this idea, because this really frustrates me and I, when we put out Trends in Neurosciences paper with Max Schein and Matthew Larcombe, then I think we already said many key ideas there, but indeed, how do you study? How do you do it? But now, at least currently, if Borian can keep it up, then it might be a very crazy thing that we try to do. Everybody's welcome to join us, but at least we try to do it and we try to do the next step. Because I think here, you know, for me, one of the key ideas is that in consciousness science we need theories that are like the iit, but are not iit. We need theories that have the mathematical grounding and formalism, like iit, but that come from a different angle because currently everyone who is interested in consciousness and mathematics will go mostly to iit, because that's just where you can. Then it's the only game in town. Yeah, that's the only game in town. And that's a problem. That's a problem for science. So we need to develop another theory or many other theories also that do this and where also people can study these mathematical properties. Because I strongly believe that science needs this diversity of Ideas. And in the end it might be that IIT is correct. But we won't get there if we only have IIT, right? We need other theories. It's extremely hard because IIT has basically head start of 20 or something more years, right? Even more. But yeah, that's the only way. [00:42:27] Speaker B: Yeah, no, I agree and I should correct myself because I posited two sort of dichotomous things. Pointing out problems and implementing solutions to those problems. But really what gives me joy, where I'm happiest is in formulating what the problem space even is, which is where no, very few people spend time. And that's exactly where you're spending time. It's not that you're just like pointing out problems. It's like, what is the right question? Which not many people actually spend a lot of time. [00:42:57] Speaker A: What are even the possible angles? Or exactly as you say, what's the problem space? Because that happens often in our lives and that's related to insight also that, you know, we get stuck on one solution and we don't even see that there are other possibilities. And then we need the other scientist, another person to say, hey, actually you could look at it like that too. And this might open up a third possibility, a fourth possibility. That's how science should work. At least that's how I think. And that's also how insight works and our problem solving. [00:43:31] Speaker B: Oh, I want to talk more about insight. But yeah, I mean, sticking on this, defining the problem space and how that is like the problem that is like one of the biggest challenges. Like for example, so many of us, I will say neuroscientists, and I have been guilty of this also. I don't want to say guilty as if it's a super negative thing, but just take on the assumption that, yeah, okay, the brain performs computations and carries out algorithms. Like if I mention, like I don't think to a colleague, random colleague, yeah, I don't think the brain actually implements algorithms. It's a non starter for them because everyone is so enmeshed. The assumption that has been made so long ago that the brain performs computations, the neural population implements this computation via this algorithm. For example, if we're talking like Mars Level's approach that's so ingrained in people that they can't even think otherwise, it is a problem. [00:44:35] Speaker A: But yeah, when I was thinking about consciousness in that respect and whether consciousness is a computation, then I realized that one big thing always that we have to kind of differentiate is also that there are many different types of computations and mostly when people talk about these things, then they are of course, pretty naive. And then they think that, you know, digital computation is exactly like neural computation. It's the same thing, especially if you talk about spikes, etc. Right. However, you could also think, and that's what we also say, you know, you could think that still the biological tissue also performs some types of computation, but it's very different because you cannot cleanly separate it, you cannot dissociate it from that substrate, because the substrate is the algorithm, partly, at least. And if it's to a large extent the algorithm, then, you know, there's no way of cleanly taking it apart and putting it to a machine or doing something like that. So I think when we talk about computation, it is very useful to think whether it's this naive computation or actually it's like this biological computation. And I think more and more people are seeing it like that. [00:45:58] Speaker B: Can you elaborate a little bit on the phrase the substrate is the algorithm? Because I agree with it and I'm not sure how to articulate it myself. [00:46:09] Speaker A: No idea how to articulate it? No. I mean, I think one possibility is to also think about another example which we know very well, like is the genes and, and DNA replication, right, that of course you can write it down as, you know, letter changes to another letter, but that doesn't mean you now have, you can build a cell from that. Because in the biological issue in the cell, of course, this algorithm of changing or having the complementary nucleotides, it also comes from the fact that how they are put together, how this thing is constructed. So, yeah, I probably didn't explain myself very good, but the point is that part of the algorithm is already in the, in the substrate. And when we go to the, to the brain and think about neural computations, then we probably can also find these examples rather, rather easily. [00:47:32] Speaker B: I kind of think of algorithms as this platonic ideal, right? Or like a mathematical structure, let's say, or mathematics happening very separate from the substrate. And what you were just describing is like, when you're actually going to implement it in a substrate, then the substrate has to necessarily almost be part of the processing part of the, if you want to call it an algorithm, you. [00:48:00] Speaker A: Could think of it like this also, Paul and there are other thinkers who have also said it that, you know, for example, if you have a brain, then the way these, the way the molecules are, the way it spikes, the way you have electric fields, it constrains what types of algorithms you can even have. And in the End it becomes inseparable from some of these algorithms because these algorithms are so tuned to what actually is even possible to do in the brain, because you only have a certain amount of things. And so that's the reason why I think you cannot cleanly separate things from the implementation you mentioned. MAR was incredibly smart in proposing these ideas back then, but at some time point, we scientists, we also have to move on and say, this is a great idea. It makes sense in many contexts, but in other contexts it doesn't make sense. And we should not always think that you can cleanly separate the algorithm from what the brain is really doing. [00:49:16] Speaker B: Well, you mentioned a magic word to me currently in my own thinking, which is constraint, which is related to the organization of biological processes, which I've come to greatly appreciate. When you take a constraint away, like let's say for an engine, right, if you take the walls away from the piston function and compressing the gasoline, et cetera, compressing the air to drive the engine, I just gave a really poor summary of how pistons work. But you take that wall away and there's no pressure, there's no work, right? And the constraint is completely necessary. And so I just gave a machine analogy. But the same thing is happening at all scales, at all times, in cells, in brains, in wet biological systems that the. So I've come to think the constraints in these complex organized systems, it's actually the constraints that have a vastly larger causal effect on the system than what we think of as the causal billiard ball going through it. In fact, the constraints, there's some sort of energy going through it, and the constraints are all the main thing that drive how that energy flows through the system. I don't know. So I don't know how you think about that. I know that you reference some of the biological organization literature in theoretical biology, etc. So maybe that wasn't such a question, but maybe I could ask how you think about. [00:50:49] Speaker A: I think it's fine if you also explain things. I mean, you're clever enough, you don't only have to ask questions. Perhaps I start asking questions from you now. [00:51:00] Speaker B: Actually, no, no, no. So that's, that's my fear, right? So I, I fear often I go on this diatribe and then I'm left and I'm like, oh God, what's my question? And then there's just blankness, right? So, but, but I guess, I guess. [00:51:13] Speaker A: You do this thing, you do this podcast to figure things out also for you to. And this is not only asking questions, but it's also this realization that. Oh, no, it relates back to this idea that I have that constraints are really important. And indeed, we can think that, you know, again, the implementation so heavily constrains the algorithms that in the end, it becomes inseparable. [00:51:40] Speaker B: Yeah, well, that's what I was getting at is whether that's how you think about. Just because you mentioned the word constraint and automatically I thought, oh, well, when you're implementing a quote unquote algorithm in a biological substrate, it's all constraints. And so that's how maybe I could think of the substrate being the algorithm in that respect. Maybe I was trying to make that connection between constraints and organization and the substrate being the algorithm. Part of the problem is that we just don't have the vocabulary and conceptual. [00:52:08] Speaker A: Yeah. And the mathematics and everything. But as we said before, that is the reason why we should try to develop it. It's, of course, hard. It's hard to get grants for that. But in the end, that is science, and that is what is currently missing. Right. And we need to do that. [00:52:27] Speaker B: Right. Don't you feel safe, though, knowing that the cortex is where intelligence and consciousness is? You feel safe in that, right? No, no, That's a hard pivot. [00:52:42] Speaker A: Oh, that's. That's cortical chauvinism there. That's cortical. [00:52:46] Speaker B: Is that what that is? Oh, I didn't know. Oh, I didn't understand. [00:52:49] Speaker A: What is cortical chauvinism? Shame on you. Shame. [00:52:55] Speaker B: So you think that the thalamus has something to do with things? Is that the thing? So what. One of. One of the things that you point to in terms of what. What could be important for consciousness. And just, you know, the way that we understand how brains give rise to intelligence and mental phenomenon are these thalamocortical loops. And not just with the thalamus, but other subcortical structures. So talk to me a little bit about what you've. What. Why the thalamus. The thalamocortical loops are important and perhaps why they're neglected, largely. [00:53:27] Speaker A: Yeah, yeah. In some sense. Of course, it's. It's not that they are fully neglected. If you. If you look, of course, there's this fantastic work of Rodolfo Linas, for example, who very early on said, that's the key for consciousness. Really? Early Dononi, also very early. The first papers on IIT are about these thalamocortical systems and how it is different from the cerebellum. Right. So people kind of know it, but now, especially based on the research Also that that happened in the. In the lab of Matthew Larcomb, who has been on this show. We kind of tried to update this a bit, try to explain it a bit more in more detail about what is it about these talumn cortical loops. And as Matthew also probably explained on your show. Right. What the Mototaka Suzuki in his lab figured out is that you have these dendrites. You basically have a switch that you can decouple the dendrites. But because this big layer, five neurons, are the key component in both the thalamocortical and also corticocortical loops, then by doing this decoupling on the dendritic level, actually decouple the whole thalamocortical system. So it's very easy then from that switch to switch on and off, for example, consciousness, or also have a more nuanced control of whole processing. Right. And once you think about that and you realize that the thalamus is in a very interesting position where it is part of the loop, but also it can control, because that's what the research of Moto Daka and Matthew shows, that the thalamus, other connections from thalamus, control this dendritic integration. Is it very strange if I show it like that? [00:55:30] Speaker B: It's weird, but it's cool. [00:55:33] Speaker A: But, you know, the thalamus also controls this interaction. And then you realize, well, yeah, thalamus. Thalamus is in this very unique position in the brain where it gets all of these subcortical information. It gets all of these cortical information. It is in a process of actually changing what happens in the. In the cortex. And then you understand that, well, the cortex is not actually on the driver's seat here, but it's fully. Like the thalamus can fully control everything that goes on in the cortex. And once we started to think about that, then we realized that, of course, it's anatomically very clear that every area of the cortex sends direct input to thalamus, and thalamus can control every area of the cortex. And then we said, look, why should we think that the brain is deep, rather with respect to the thalamus? It's very shallow because the thalamus looks at the cortex as basically one sheet all across the brain. That's why we then wrote the paper how deep is the brain? Although I tried to convince my co authors that it should be how deep is your brain? To an old song by Take that. Right. But anyway, what song is that? I won't sing it, but it. But the Song is from Take that. How deep is your love. Oh, yeah, yeah. And it's okay if you don't know it. I also shouldn't know it. And probably nobody of the listeners knows that song. It's you. I mean, but it's a classic. And that's. That's why I said that. Let's have the title, how deep is your brain. But in the end, we have it almost like that. [00:57:24] Speaker B: Well, it's also just better pr. Right, because then people think, oh, my brain. But, okay, so you're alluding to that paper which you discuss. You just discussed some of the details of why you call it the shallow brain hypothesis. And I know that maybe you think maybe that wasn't the best way, Best phrase to call it in retrospect. But so the idea is that. So you have the thalamus. Okay, I'm going to step back just a little bit and restate what you said with a little bit more layman's kind of details. So in the cortex. So the cortex is made up of a bunch of what are called cortical columns that have variation across the cortex but are uniform enough from column to column that we call it a column, right. At, like a unit of processing. Within those columns are made up of different layers. Then when within layer 5, you get these big pyramidal neurons, which are called pyramidal neurons because they're big and they have this pyramidal kind of shape. And these neurons project down into subcortical areas. They project to the spinal cord. They're like sort of the major output of the cortical column. And they also receive. They have these, what are called apical dendrites that reach up into the more superficial layers of the cortex that are receiving feedback from these thalamocortical loops, among other things, feedback from other cortical columns and from some subcortical areas, et cetera. So they project down to the thalamus. The thalamus projects back eventually, and it constitutes a loop. Right. So these neurons are getting information from all sorts of different inputs from different contexts, et cetera. So that was a very basic, like, sort of biology of it. But the point with the shallow brain. You're smiling. [00:59:15] Speaker A: Why are you. [00:59:16] Speaker B: Why are you. What did I miss that I should have? [00:59:19] Speaker A: Just thinking that whether your explanation was really so much simpler than what I. [00:59:24] Speaker B: No, of course it wasn't. You start to do it and it's like, oh, shoot, I'm getting in trouble now. The details. Anyway, there's these big cells that project that are important. Yeah, but. But the point with the shallow brain is that all across the. You have. All across the cortex, you have these repeating motifs, these coracle columns, and you can think of that as sort of one layer. Even though we think of the brain as a hierarchical structure and that hierarchy, even within the cortex, we think of, like, early sensory processing to later. More abstract. More abstract. And then you get in the prefrontal cortex and everything is like, super abstract. And we think of this as like a very deep architecture, like you think of in deep neural networks. And in that sense, it's deep. But your point is you can also consider the cortex as sort of this one layer of all projecting to this one area, the thalamus. And that thalamus, which you alluded to, can be thought of as a controller and can switch off whether a cortical column continues that loop or whether there's a break in the loop. And so in this sense, the brain is shallow. There you go. How was that for a higher level description? [01:00:31] Speaker A: Yeah, excellent, excellent. Yeah. So that is the basic idea. And I think. Yeah, what's the kind of. The problem of the title is in some sense that it's the shallow brain hypothesis, but in some sense it's no hypothesis because it's just anatomy. There's no speculation that this. It's like that. It's. It's just anatomy. Every cortical column in V1 in prefrontal cortex sends direct input to the. To the iro. The thalamus. So it is like that simply. You could still, of course, wonder whether the best word is shallow, but in some sense it's it. That's what we try to say in the papers. It's not. It's not speculation or. It's not some theory, it's the facts, man. [01:01:20] Speaker B: That's like. You should. You could. I like the. Your brain is shallow sort of approach. It's like the mind is flat by Nick. [01:01:28] Speaker A: Yeah, Nick Schaeter. Yeah, yeah, yeah, yeah. [01:01:31] Speaker B: So just more PR stuff. But. So do you think of the thalamus, then, as. Or at least higher order thalamus, which is an old, evolutionarily older subcortical structure. Do you think it's all control? What do you think of the thalamus, then? [01:01:47] Speaker A: Yeah, I mean, you don't want to put the homunculus to the thalamus. Right. So rather you would like to say that of course, the brain is a complex dynamical system, but the thalamus is like a central control hall where you can actually root things and you can more easily in one place, basically. Do some of the control. Right. As it happens in a dynamical system. Right. Because it is simply at this, as I said, it's not only that the cortex sends input to the thalamus, but also all subcortical areas. And if you think everything that comes here, it ends up in the thalamus one way or another. So it is a very interesting central actor, perhaps. And that's why I would like to think it's a conductor in some sense. But again, in the sense of in the dynamical system without being a homunculus. But it's one place where you can actually conduct and make sure that everybody's on the same page in the brain. [01:02:56] Speaker B: And one way to contrast this with the classical view of the thalamus. However, whatever words you used to describe what the thalamus is doing, it's not simply a relay station in the brain, which is how it has been thought of in sort of large swaths of neuroscience for a long time. Yeah, the wiring just has to go through the thalamus to get to the next part of the brain. And that's what it's doing, is relaying that information to the next part of the brain. But it has a much more active role and part of the goal is to figure out what that role is. Of course. [01:03:31] Speaker A: Yeah, exactly. So, yeah, we will try to do here also experiments. My friend Mototaka is doing these experiments. Others doing great experiments. But yeah, so we, we think that there is still a lot to figure out about how the higher the thalamus kind of has a role in, in rooting things in and controlling things in the cortex. And again, that, you know, we should really reduce this cortical chauvinism and we should really not think that, you know, that there's this king cortex that, you know, does everything. Rather, it's just a very convenient computational space that is used by these other actors. [01:04:18] Speaker B: What are your assessments of the prospects for artificial consciousness? Do we just need to implement thalamocortical loops and then we're there? Do we really need to figure out this massive across scale integration and then we're there. What are your current thoughts about the possibilities of implementing artificial consciousness and if so, how to go about doing it? What would be the first step toward that from where we are now? [01:04:48] Speaker A: Yeah, well, now I, I have to reveal that once I got to understand large language models and modern AI systems, then I understood that also we are cooked in the sense that, you know, we cannot maintain a long time this position that, you know, it's just Dendritic integration and thalamocortical integration, because that is something you can fundamentally, like, digitally implement, something like that. But rather, as I also alluded to earlier, then I immediately. And that was like, early 2022, something like that. I started to think about, okay, what's more fundamental? What are the more fundamental things that we are missing? And that led to all of this research and ideas about how actually it's not only about the dendrites, it's not only about thalamocortical loops, but it's about these interactions between the scales and how the processes that happen between them might be relevant for consciousness. And again, it's speculation, but I think we have done the first steps toward describing conceptually what we think. And now I said there are some great young minds trying to figure out whether we can also formalize it a bit better. So fundamentally, I would say that it's not worth. Maybe. It's not worth defending this idea that definitely it has to be these dendritic integration and thalamocortical loops. There are probably deeper reasons why or how consciousness arises from the brain, and we have to go and figure that one out. And some people who listen to that think. Jan, again, you're crazy because you had this theory, this dendritic integration theory, in 2020, right now, you're already pivoting. What is that? But I always say that, for me, the dendritic integration theory, the key idea of that is, first, we have to have serious neurobiological theories. We have too many of the cognitive theories, computational theories. We have to have a serious neurobiological theory. And second, it is not and never was for me, a definite theory, but rather it's an inspiration for the younger generation that we don't have to be stuck with these old theories of consciousness. We can look into the brain. We can come up with more complex theories. Dendritic integration theory was like an example of that, and I hope a good example of how you can do it. But we need to move also on. We need to go to the next step and next step and continue moving. [01:08:02] Speaker B: Well, so what is the. I'm not asking what your goal is here, but when I'm thinking. So to me, like, I had this base intuition, right, that, okay, well, artificial systems aren't conscious, and I think they can't be. But then there's a thing in artificial intelligence that I said earlier, right? Well, okay, artificial system doesn't have this. All right, well, we'll just add it as an algorithm, and now it has it and you can do that step on in ad infinitum, potentially. Right? So there's the extreme of biological naturalism, which I think Anil Seth maybe has landed on, where really it is like something fundamental in principle about the biological substrate that negates any possibility of consciousness in an artificial system. I may have that subtly wrong or completely wrong, but I think that that's the position of biological naturalism, and that's kind of at the extreme. At the other end of the extreme is like functional computationalism, essentially, where if you just replicate the computation, you get consciousness for free. But what you're talking about, so what I said in the beginning is like, well, what's the goal? Is the goal to like show that there's an imprint divide where, where machines can't be conscious? Or is the goal to find the differences between machines and biological systems and produce and fill in those gaps and then. And then eventually we'll have synthetic systems that are conscious. So maybe not what's your goal? [01:09:35] Speaker A: But. [01:09:37] Speaker B: From which you're speaking, what is the goal? You know, what is the background from which we're coming from? [01:09:44] Speaker A: The goal is simple. We have this naive functional computationalism, we have biological naturalism. Our goal is to show that there is a middle way that is actually the most plausible, the most reasonable, which is that unlike naive computationalism, we say computations in the brain are extremely complex. We don't understand them because we don't understand how exactly the substrate constrains the computations. But they are still computations, very complex computations, we don't know, but still computations. Hence also biological naturalism is wrong because fundamentally you can at some time point build machines that are conscious, but they are not like large language models, because large language models, they only do everything digitally. There's nothing that constrains them the way brains are constrained. So it has to be a bit more brain. Like it has to be neuromorphic in that sense, but not simple neuromorphic. But probably, you know, you have to take into account several aspects of what are the key process in the brain. So in short, we want to say that there is a middle way and actually that this is a quite a reasonable way that people could take if they think about consciousness. Did that make sense? Was it clear? [01:11:22] Speaker B: Does, and I want to refine a little bit and have you elaborate. So there's no in principle reason why a synthetic machine could not be conscious. And in fact, like the Ship of Thesis, you know, you could replace every molecule of a cell and as long as they're sort of functionally organized and dynamically organized the same way and operate the same way. You could then replace the whole brain or whatever and have a conscious machine essentially with machine parts. But you almost wouldn't consider it a machine in that respect because it is, it's about the organization and the integration across scales. Something that is, it's just way, way far away from our being able to understand and or implement it. So you, that's where you are is it's not in principle impossible. We're just really still far away from. [01:12:13] Speaker A: It is that we will not get there with large language models which are great and which do all these interesting intelligent tasks, but we will not get to consciousness like that. [01:12:28] Speaker B: Yeah, you want to get to consciousness in a machine. [01:12:34] Speaker A: I mean. Yeah, I mean I can say that I think it's so far off as of today that I seriously haven't thought about it. But I know, I mean Anil Seth would say that, you know, is this even something we want to achieve? I would need some thinking time here or I would need to call a friend perhaps. But yeah, I haven't thought very carefully about it because I do think that it's currently it's so far off that you know, we don't have not to be worried about. [01:13:09] Speaker B: Yeah, yeah. I mean my main goal also is to really selfishly just to understand it. And I suppose if we understand it we can in principle implement it, but maybe not. Maybe we can't implement. Maybe there's a level of understanding I would be satisfied with in terms of these organizational across scale principles that would satisfy my curiosity and not necessarily allow me to build something in the Feynman sense. What I, what I cannot create, I do not understand. [01:13:39] Speaker A: Yeah, yeah. And something you said before is also very important that you know, again, many as I have tried to explain many of these details of the brain will matter because the computations that run in the brain can only run on these kinds of substrate. Which means that on some aspects at least you might end up almost with the brain if you want to have synthetic consciousness. But probably it's not that you that it's the every aspect of the brain, but there might be some aspects that are basically exactly as they are in the brain. Like also physically implemented exactly as they are in the brain. But as of today we don't know which aspects. We don't know almost anything about that. But that's why we have to study it and that's why you have to try to figure it out in, in. [01:14:35] Speaker B: Terms of the what I think we're up to about 22437 Theories of consciousness now. Something, something like that. Do you throw them all away? Are any of them more or less intriguing to you? [01:14:50] Speaker A: Well, I, I think, you know, I mean, I've done my research. I said I've been in the game for 22 years. I read probably hundreds of theories of consciousness. But at some time point, yeah, I feel that, you know, the problem is that it's very easy to do verbal theories of consciousness. [01:15:10] Speaker B: Yeah. [01:15:10] Speaker A: And talking to theories of consciousness, very easy to come up with something. But at some time point, especially when I was in the lab of Matthew Larcombe from 2018 on, then I thought, you know, now, you know, only, only thing that makes sense is we really constrain our thinking about consciousness in terms of neurobiology. And I would say that until this day, actually there are not so many people who are willing to do that because I think partly also our brains are lazy. And when they are, then, you know, they are faced with all of these facts about the brain, then it's easier to say, ah, probably that doesn't matter, it's just the computation anyway. Right. And of course I say that that's the problem of science also that, you know, if you have figured out the way how you can get the papers and publications, then basically you don't need to go deeper. Right. You don't need to open up that Pandora's box. So what the point of this very long and emotional response is, is that we have a lot of theories, but actually we do not have so many good neurobiological theories. So even if I would agree that we have too many theories, I would still invite young researchers to come up with new neurobiological theories of consciousness. [01:16:37] Speaker B: I think part of the, my own hesitancy you mentioned how lazy brains are, is that it's really all right thinking about privileged level and how there's no privileged level. Well, when that's the case, okay, you still have to think about what's happening at any given level at any time. And then you're supposed to hold this in mind while you're moving the other parts in relation to that particular part. And then if you want to do some research, you're going to have to measure some stuff at a level and then you're going to, your focus is going to have to be on that level and you just can't hold all of the other levels in mind. And so there's a real hesitancy then to, for one to like try to keep this galaxy brain, you know, keep all the pieces and how they interact in mind. But also there's a hesitancy for me then to dive into one level because you have to go real deep. You know, you're going to run into all these issues and spend all your time thinking about that one little level when what you're really interested in is the across level story. For example. [01:17:41] Speaker A: Yeah, I think the problem, and again, in science, we should stimulate this kind of research that people go for these really complex problems. But partly, I think that's a problem of science. We haven't done that very, very, very well over the last years. And for example, I always say that, look, if you want to figure out consciousness, we should invest into young people who do some experiments, for example, in basic neuroscience. But instead the money often goes to big projects that study these classic theories. I think that's wrong and I think there is a very good case to be made because I've been in that situation in Matthew's lab, right. The first time I meet with Moto Taka, he says, okay, I want to show you something. And he shows me this data that was unpublished back then about this decoupling. And even if many people would say, oh, that tells nothing about consciousness, or it might not be a piece of a big solution, here is a fundamental piece of evidence that we didn't have and we couldn't have if Mototaka wouldn't have done this research and we would have no idea about it. And now imagine we, instead of only Mototaka, we would have 10 labs or 20 labs doing these kinds of research. Of course, some of that would not end up with anything new, but we would get these discoveries, some aspects that we even didn't imagine that didn't fit into our mental box before. And then we could think of consciousness in a new way. So that's what I think that, you know, we should try to, if you want to solve consciousness, then we should, should not be happy with what we have right now, but we should try to do these kind of novel experiments and try to do, use the most recent neuroscientific techniques. And we should really invest into also doing that. [01:19:49] Speaker B: Yeah, but I think, you know, it takes, you said like basic neurobiology research, right? And that's. That is exactly the point I was worrying about is that when you start to do basic neuroscience research, neurobiological. Neurobiological research, you can easily lose the big picture, right? And you get into the details and it takes people like you, like Matthew Larcombe. [01:20:16] Speaker A: It's a very good example, because I think, I hope that it was partly me who kind of helped to make Mototaka's research known in consciousness science. Because they, of course, wrote a very good paper. In the end, it ended up in Cell, which a very good journal. But most of the people on the cognitive side, or cognitive neuroscience, would never read that paper because it's also like anesthesia and stuff. So my goal was, okay, I come in. Whoa, I see this research. [01:20:46] Speaker B: Am I. That was you who ushered that in. That's not something you should know. [01:20:51] Speaker A: No, no. I mean, in the sense that they had the result, right? Yeah, most of the time they showed me the result. But then I went to Matthew's lab, actually to study predictive coding. But at that moment I understood. Oh, sorry, sorry, sorry for that word. But basically it was like that, that, you know, oh, okay, consciousness. Here we go again. Okay, we're getting back together. And basically I said to Matthew, and what a talk. Okay, you have to write this paper. And then it ended up as the dendritic integration theory. So I had my tiny role in that equation. [01:21:30] Speaker B: That's not tiny. [01:21:32] Speaker A: I think the point of that is also that you have to have in the lab these people who can do these amazing, precise, complex experiments like Mototox. But in the best case scenario, you also have some people like me with the same haircut also who recognize that, oh, this is relevant for these people and we can actually write it together. And this is exciting for a lot of people. So you need to have a right composition of people. And Matthew definitely had these people back then. [01:22:08] Speaker B: Yeah. That's great. Well, I'm sorry, I didn't know that that was. I mean, that's not a tiny role. You called it a tiny role. That's quite a large role. I mean, those connections are not the rule. Right. I think that that is the exception when you have people who can connect what's happening at one level with concepts at other levels. [01:22:27] Speaker A: So, yeah, that's the only thing I can. Besides, that's the thing too. [01:22:31] Speaker B: Right, right, right. Yeah. So that sometimes I think that's my only skill too. And I don't know if how good I am even at that skill. Like, I'm a poor neurophysiology, I'm poor at everything I do, but something that I'm slightly better at. And I. That. But that's one of the things I admire in other people. So there you go. I admire you, Jan. [01:22:48] Speaker A: Thank you. I. It's the same feeling Paul, I admire you, too. [01:22:53] Speaker B: Thanks. All right. [01:22:54] Speaker A: Not only because of your haircut, but there are. [01:22:57] Speaker B: I try. I gave a fresh shave just to look a little bit closer to one of my, One of my heroes here. But you. So, so. But part of that, connecting those ideas takes a lot of creativity. So here we go. Here's a pivot. Why are you interested in creativity and insight? [01:23:16] Speaker A: I think partly it is because of that, because I really. My mind goes all over the place. And, you know, my wife also hates it that, you know, when she expects a normal answer, then I give kind of different answer. [01:23:30] Speaker B: Right. [01:23:32] Speaker A: I don't do normal. I like different. I like to think always about that. [01:23:35] Speaker B: Can be very frustrating in my household, too. [01:23:39] Speaker A: Exactly. Well, yeah, we are alike. Right? And then, you know, again, this was a time where I was kind of a bit having this off time from consciousness, and then I don't remember, I probably, I had some kind of idea and then I started to think, okay, but where did that idea come from? And then I was reminded of insight. Then I started to read about the insight, and then I started to think, oh, we don't understand it, and we can do so much better in understanding that. And you know, what bothers me when it comes to insight is that many people think, and it's also in the textbooks like that, that it's a really tiny thing in problem solving. It's a really marginal thing. Right. But then in our papers, I mean, we haven't been maybe super successful, but we tried, we try to make the case that, you know, of course, the laboratory insight of figuring out a picture, it's a very tiny thing. But there are these moments in lives of people where suddenly they understand something about themselves, their own life. And these are not tiny things. These are the most important things in the lives of these people. We have to try to understand what the mechanism is behind these transformations. So that's how I, how I got into it. And then in the beginning, there was just ideas, and I started reading and collecting examples. But then when I came back to Estonia in 2020, then I, I got a postdoc, and fortunately I, I managed to convince her that this is the thing we have to study. And now she's, she's, she's doing great research on, on, on, on, on that thing. And I, I had some PhD students coming in. I don't really. I, I'm very hard at convincing my PhD students at what they have to do because I want to have their freedom, right? But I convinced at least some of them and Then we studied these things. [01:25:58] Speaker B: So what do you want to share about insight? I've actually had a few people over the years studying creativity and aha. Moments and people I'm sure that you're aware of. John Cunhos has been on the podcast and so if you're. There just aren't that many people studying it. Right. So yeah, that's another thing. [01:26:23] Speaker A: And as I said, but I think it should be actually much more interesting. And then the, you know, the first thing we did with this, with this postdoc was to, to. I gave her almost an impossible mission. I said look, I think that there's this core phenomenon of insight. Of course it started in, in problem solving. But let's try to write. [01:26:46] Speaker B: What do you mean core. What do you mean core phenomena. [01:26:51] Speaker A: That, you know, people come to a solution in a very broad sense. They have this, oh, now I see. And of course it started in problem solving. But I think, and because I had read many examples, I thought that that's just a very tiny piece of this thing. But let's write a review where we take problem solving, but then we also go over the literature and study insight on the psychedelics. We look at insight in meditation, we look at insight in psychotherapy and we look at insight as it happens in delusions. Because these were examples where I had already found some examples. I knew there has to be something. And then imagine you give your postdoc this kind of a task. You have to go to five completely different literatures and try to find this phenomenon from all of them. Never has been done before like that. Right. But she managed and indeed we found that there are these common threads about this insight phenomenon and this means that it's not such a tiny marginal thing, but it's something that happens here and there and that is worth actually figuring out. [01:28:12] Speaker B: So maybe this is not the right place to ask this, I guess, but there are different sizes of insights, different impacts that insights can have. And you talked about the various ways of looking at it, the various sources from which to draw insightful experiences. And one of them is one experience that people who experience this draw massive insight from our near death experiences. And I forget, did you have you looked into near death experiences as well? [01:28:50] Speaker A: We didn't. I mean I have to tell my postdoc that, sorry, we have to redo it because Paul said we have to. No, you're, you are right, you're right that people, of course they come up, come with, with life changing experiences. [01:29:01] Speaker B: But, but it, but that's a life changing Experience, but I don't. Is that an insight necessarily. So maybe the border between insights. Fuzzy. [01:29:10] Speaker A: I would need to. Need to read a couple of them. That's again, that's how we proceeded. Right. It's not that we say, look, there are insights in meditation. That's it. But you know, we went into the papers, we looked through the. How they are described, so. Exactly. That's how I. How I would also need to. Need to proceed here. That. Because inside it's kind of. The one component is also that it's kind of sudden. So it's not that it kind of slowly. There is this aha component. But yeah, I would need to look that up. [01:29:47] Speaker B: Yeah. [01:29:48] Speaker A: So yeah, we didn't cover everything, as you can see. [01:29:50] Speaker B: Yeah, yeah. Well, there is the aha component of it. But often when I have those, I can trace it back to a slow buildup of, you know, various aspects of knowledge that I can you know, sort of retrospectively appreciate as having led to that aha moment. And it reminds me of. And I've used this example before, but a lot of musicians, for example, talk about when a song sort of just comes to them, you know, and that there's this insight or insightful burst of creativity. This aha moment. I'm thinking of people like. Like Tom Petty. I mean, there's a lot of people who have. Who have talked about this. However, it's always preceded by lots and lots of work and noodling around on the guitar for hours. And then finally it feels like something drops into their lap. And it. You have this experience of the aha moment, but it's embedded within this. You've been surrounded by all the things that will lead up to that aha moment for intensely. For some. For some time. Is that a common theme in your studies? [01:30:54] Speaker A: Yeah, yeah, yeah, yeah. I mean, it is. When I give. Talks about it, then I always also say that, you know, it's not that it comes out of the blue in the sense. In that sense, but really you usually. You work hard on it and you could think about it that, you know, you set up algorithms in your brain which kind of continue fiddling around with the program. Right. And they do it and you have learned it to do it and they continue and continue. But nevertheless, the interesting aspect then to study is also this moment where you actually experience it. Because at that moment, although there might be this slow buildup of this unconscious processing, you get this sudden change of consciousness and sometimes this sudden change of behavior of the sudden change of the person. Right. And of course As a neuroscientist, you want to understand, okay, how can it happen that, you know, you basically reconfigure the brain at one step? Basically, even if there is slow buildup, the reconfiguration has to happen there and it has to have its signatures. But of course the problem is that, you know, if you put people into the scanner, they won't have life changing moments. Right. You can do tiny things. But yeah, as we say, and you said it also nicely, there's a continuum of, you know, tiny things. But you know, we eventually want to understand these big things. So again, it's in that sense people are have also said, yeah, that's crazy. I mean, you never can really figure out how these big things happen. But you know, we take it step by step and we'll see. [01:32:44] Speaker B: So, so what is the, how would you summarize your current thinking about insight and how it arises and the neurobiology, neurobiological aspect of it? [01:32:54] Speaker A: Well, yeah, especially in terms of neurobiological aspect. We really, we have been thinking a lot about psychedelics because that's, you know, a clear manipulation that you can do on the brain where we know some of the correlates and there you can try to make this link. And I think we have discussed theories of consciousness, but I also would say that, you know, theories of psychedelic effects also, they could also use some refreshments of what we have learned about the brain. And yeah, it comes as a surprise to no one when I say that I think that the dendritic mechanisms and thalamogortical effects are also really important. So essentially what we think that happens during insights is that you have a reconfiguration of these thalamocortical complexes and this, that the main driver of it are again on these apical dendrites. Again of course I am speculating, but it's a very interesting anatomical fact that a lot of these are five H2A receptors are where. On the apical dendrites. Exactly there where Mototaka found that you can control the coupling and decoupling exactly there. [01:34:25] Speaker B: Right. [01:34:26] Speaker A: And so you can control these aspects there through psychedelics. And I think it's a very neat, a non coincidental aspect of our brain. So that's where our theory is going and that's how we also try to link it to insights. [01:34:49] Speaker B: So but the phenomenal experience of insight and the phenomenal experience of subjective awareness and of dreaming and of wakefulness, they're all subtly different. So it's just. And if you relate them all to dendritic decoupling. There are, there have. There must be some subtle differences. What? Now why are you laughing? [01:35:11] Speaker A: No, because of course if you say it like that, then everything I say sounds so stupid and silly. Everything is in the dendrites. It's in the dentist. [01:35:20] Speaker B: That's not what I mean. [01:35:21] Speaker A: No, no, but, but fundamentally you are right. I mean, in the, in, in some. [01:35:26] Speaker B: Very coarse level, dude, it's not all dendrites. Okay. No, I'm just kidding. [01:35:31] Speaker A: But, but yeah, if you also think about, for example, dreaming versus psychedelics, then indeed both of them affect dendrites. We have papers on both of them. But the effects are different. Right. During REM sleep, you have a lot of acetylcholine, for example, Right. Which has, of course, different receptors than that of. Of serotonin. They are not exactly the same things, so. Exactly. Some of the neural effects will be very different, some of them slightly different between dreaming and, for example, psychedelics. So details matter, and that's what we try to also convey in our papers. But on some more abstract level, all these effects really, for some very strange reason, they converge on these dendrites because again, these metapotropic cholinergic receptors are also on these apical dendrites, or to be more precise, on oblique dendrites. I think it's so late into the podcast that nobody is listening to us who doesn't care about the brain anyway. So we can go into this. Metapotropic acetylcholine receptors on the oblique dendrites. Right. [01:36:56] Speaker B: Okay. [01:36:56] Speaker A: The point is that, you know, all of these mechanisms, both serotonergic mechanism and this mechanism to act acetylcholine, they affect these dendritic coupling could be a coincidence, but perhaps not. Right. And then it means that, you know, dendrites are underappreciated and they really have an important role in actually figuring out these things. [01:37:26] Speaker B: A lot of people who start studying, like psychedelics, Right. Then they shift everything into psychedelics. Is that going to happen with you? I doubt it. Because you're. You're going to spend a little time there and then move on. [01:37:37] Speaker A: No, yeah, I mean, I, Yeah, I, I like my freedom and I, I really am passionate about also being fresh and changing topics because what I see is also scientists, we also, we get very much into our box when we study the same thing all the time. So I try to keep myself fresh, changing things, changing pace. So far it has worked, I hope, but of course it's always also very difficult. Now you take a new topic. Let's Say education, and now you try to write papers on it. Then you know, if you, and that's the problem, you send out, send a paper to the journal and let's see, let's say that it's a good paper, just, it's simpler. Right? But now the editor looks at, who's this guy? I have never heard of him. [01:38:30] Speaker B: Yeah, right. [01:38:31] Speaker A: It's not someone who works in education. And then, you know, they, why should they even consider your paper? Even if it, let's say for the example, it is a good paper. So that's, that's difficult. But nevertheless, I think I will keep on doing fish. [01:38:46] Speaker B: That's not how it's supposed to work. But that, that is how it still works. [01:38:50] Speaker A: Right? [01:38:50] Speaker B: You have to. Oh, I've never. Where did this person come from? That shouldn't be how science works, but it is. [01:38:55] Speaker A: Yeah, yeah. And you know, maybe, maybe also here, in the end, it's worth saying and worth giving a shout out all people, all researchers in Eastern Europe and also all across the southern areas, Africa and South America, etc. That, you know, there are so many great minds out there. So many great minds. But then let's say in this example again, you send a paper to a journal, of course it should be objective, but in the end, the editors are humans, of course. So when they see that you are affiliated to University of Tartu, Estonia, and they don't know you because you're now on a new field, then they simply, they have so many papers, so many things to do, it's a split second decision where they say, okay, seems nice, but you know, I cannot trust it. And now maybe it's also written by an AI. So I think it's a, it's a really, it's something I see a program and we have had this discussion of the diversity in science, but I think we really, they really haven't solved that particular issue. There are editors who are really good and who are more objective, but often when you get a desk rejection, then you think, okay, was it because of the paper or was it because of my affiliation? [01:40:21] Speaker B: Yeah, I mean, I had already invited you when I learned that you were in Estonia. Had I known that, I wouldn't have never have invited you. [01:40:30] Speaker A: Because that's why nobody's listening. I mean, nobody's listening at this point, this time point, because it's some random dude from Estonia. [01:40:38] Speaker B: Is he Estonian? So what is it? [01:40:42] Speaker A: What is Estonia actually? What, what is Estonia? [01:40:46] Speaker B: Did you grow up, did you grow up in Estonia? [01:40:49] Speaker A: No. Yeah, I mean, I'm I'm. Yeah, I'm Estonian. That's why I work here also. But I, Yeah, after, when I finished school, then I went to do my undergrad in Berlin and then I did my PhD in Frankfurt. But yeah, this is also a funny story, you know that I saw a good friend, old friend, see a professor position, get the professor position in a university and I wrote to him and congratulated him and then he writes back, no worries, Jan, you will also get it one day. And I'm like here in Estonia, I'm not applying anywhere, I'm not trying to escape Estonia, I want to be here. But obviously he was a good friend and he hadn't understood it. He thought that I'm constantly sending applications to get to Germany or UK or something. Oh yeah, I don't. [01:41:48] Speaker B: The science world is a little different, right? With open source stuff and so you have access to more things. I mean, does that change the landscape? I guess what I'm wondering is what. How you, upon reflection, having grown up in Estonia, gone away and studied things elsewhere and then having chosen to go back to Estonia and I don't have any, I don't. I could care less about where someone's from or to me, it doesn't say anything. If someone's from some Ivy League school. I despise it when someone's being introduced and they're from Estonia, then they might not even mention where they're from. But if they're from an Ivy League school, they will say they, I'm from Stanford. This person, you know, and it means absolutely nothing, means very, very, very little. It about their intellect, it means absolutely nothing. But so, you know, maybe just reflecting on it, like what, what does it mean to you that like, where does Estonia. Not where does Estonia sit? I guess I'm just asking about your reflections on what it's like working in. In. In that. What is that kind of environment and what is it like working in that kind of environment? [01:42:55] Speaker A: As I said, sometimes it's very difficult and frustrating. So let's say again, I send an email to a colleague whom I don't know, let's say in Harvard or Stanford. I usually, I send brief notes, very like polite questions about something. But let's say I don't get a reply, then I'm always, always thinking, oh. [01:43:19] Speaker B: Is it because I'm. [01:43:22] Speaker A: Is it because he's super busy? Which is valid. [01:43:25] Speaker B: Yes. [01:43:25] Speaker A: But what if I would have an email account from Cambridge, let's say, and sent the same email? Would he or she also not have replied to Me, is it partly because of that? That's frustrating. But on the other hand, I do think that the key currency in science is how well the people can think, how well your postdocs and PhD students can think. And here in Estonia, we have really excellent brains. Yeah, I think we are, we are at a very good level. And that's why I also take pride in, you know, cultivating that and helping these young scientists get out from Estonia, come back. But, you know, saying that, you know, we can do top things here also. Yeah. [01:44:15] Speaker B: All right, so we'll end on. I know you're going to attend a 75th birthday party. And in fact, I, I don't know if I said this earlier, but I appreciate you coming on because I know you're working on like educational tools right now. And, and to come back to these topics has been a real task switching effort on your part, which you've handled extremely well. So I really appreciate that. [01:44:34] Speaker A: Well, let's see. Let our listeners, let the listeners. Great. [01:44:39] Speaker B: What you were saying earlier about the way that you go about things. Let me see if this resonates with you. So I was listening to some, to the, to the, to how standup comedians work. And one of the things that stand up comedians focus on, like developing the next hour. Right. Of their, like, for a show. And so they, they go out and they try things and they, over a period of time, they see what works, what doesn't. And then eventually that gets put into like an hour. [01:45:08] Speaker A: Right. [01:45:09] Speaker B: And then they deliver that hour of their standup comedy and then they just have to start totally new after that on a new hour. And I was thinking, man, what a great way to go about like a scientific trajectory in a career. And I, that is sort of the way I would like to go about doing it. And that's kind of the way that you're going about doing it. So kudos to you that, that's really impressive and fun. [01:45:36] Speaker A: Well, let's see. I mean, yeah, it's, it has its upsides and downsides. [01:45:43] Speaker B: Sure. [01:45:43] Speaker A: But yeah, now when I, for example, when I work on education and this thing is really getting over my head, so much bureaucracy and so many things that we need to solve here, then I say to my colleagues, guys, now I need to work on something simple. Now I will go and work on. [01:46:02] Speaker B: Consciousness, the boring, simple thing. Yeah. [01:46:07] Speaker A: So, yeah, that's just simple things. [01:46:09] Speaker B: Yeah. [01:46:10] Speaker A: But yeah, yeah, I wish really that, you know, if somebody still is listening, that they take this, also take this podcast and think about scientific freedom, because I think that should be the key reason why anyone stays in science, is that you have these, this mental freedom to study things and you shouldn't follow the patterns of your supervisors and what has worked well in the past. You know, let's do real science. Let's do mental breakthroughs. [01:46:47] Speaker B: Thank you Jan for your time and I'll let you get back to what you're whatever you're currently working on these days and. [01:46:52] Speaker A: Yeah, what am I working on? I have no idea. [01:46:55] Speaker B: Now you have to task switch back. So anyway, thanks for being here. [01:46:59] Speaker A: Thank you. [01:47:07] Speaker B: Brain Inspired is powered by the Transmitter, an online publication that aims to deliver useful information, insights and tools to build bridges across neuroscience and advance research. Visit thetransmitter.org to explore the latest neuroscience news and perspectives written by journalists and scientists. Scientists. If you value Brain Inspired, support it through Patreon to access full length episodes, join our Discord community and even influence who I invite to the podcast. Go to Brain Inspired co to learn more. The music you hear is a little slow jazzy blues performed by my friend Kyle Donovan. Thank you for your support. See you next time. [01:47:50] Speaker A: Sam.