 Okay, so today we have a professor, Takashi Kigami from University of Tokyo, so he is kind of a leading figure in the field, the research field called Artificial Life, so, and, okay, so his background is statistical physics, and he got PhD in the University of Tokyo physics department, then he moved to the Kyoto University and Kobe University, then he came back to the University of Tokyo as a full professor, so today he will talk about Artificial Life and Corrective Intelligence, and also he's quite famous artist, international famous artist, he has collaborated, okay, so he's quite famous artist in noise arts like things, I never saw international, but anyway, so today I don't know whether he is going to perform today or not, but yeah, please enjoy his talk, yeah, please. Okay, thank you very much, Jun. For me and Jun for being with us for more than 20 years now, I think, yeah, almost 30 years. Okay, thank you very much for a nice introduction, and I'm very happy to be here as a visiting researcher here in noise, you know, I was expecting like a very sunny and nice ocean view here, and actually it was, you know, I haven't been here since the 8th of August, so it's already two months, but I haven't seen so many, many people here like this room, so it's my first time, and I'm very excited to give a talk in front of many audience, so today I'm going to talk about Artificial Life, the introduction is Artificial Life, and then I'll switch to what my current interest is, try to explain it in connection with what's the people in Artificial Life been thinking, and then what's the concept about Artificial Life. So I got many questions, what is Artificial Life is, and my main question to answer to the question is, Artificial Life is about the theory of brain and evolution, right? Well, usually what we answer is that computer is a good metaphor of the brain, and then our answer is, well, may or may not, right? And then do we have a CPU, do we have a RAM, or do we have a global clock? Of course not. So what is the good metaphor? So that's one of the questions that we usually ask in this community, and what is individuality is another big question in this field. So the idea of evolution assumes the object of selection is individual, right? But what's the definition of individuality is? And I don't think we have a definite answer to this, and we don't have any definition of this. So this research is about, you know, how we can understand individuality, and then what is the evolution of individuality? And this question is one of my big obsessions since I was in Stanford Institute, like 30 years ago now, that a little bus and water fountain are asking this question, that what is the evolution of individuality is. And then why this one is a big question is because the first question is because I think everybody knows John Conway's Game of Life. You don't know if anyone knows, right? Of course, I hope so. But I can even show some of this, so it's gone. So I cannot show this. Wait a minute. Yeah, for example, like the Game of Life is the grid, the game, right? It's black and white. The white is the life state, then black is the dead state, right? So when the white cell is surrounded by exactly the three states of a living cell, then the black cell becomes white. And if the white cell is surrounded by two or three living states, and it's continuously in the same state, right? But otherwise, it's going to be the dead state. That's what the very simple living systems are, the game, right? Then if you simulate this, this way, some crowded things, but you can see there, right? I mean, of course, you know this, this is what we call Glider, our favorite creature, right? So this one too, and this one too. So what are these? These are the little Glider, or we can call it a soliton for some, you know, moving things. So this is where the artificial life actually started, because people are fascinated with this idea of Glider, and how you can find it, and how you can play around with that. Then actually, but the John Conway doesn't like it, you know. So he said, don't ask me about the game of life, right? So we invited him to as a keynote speaker in 2014 in New York, and he said, okay, guys, don't ask me about the game of life, right? But that's what the people want to ask about, right? So that was interesting, but he doesn't like it because the game of life is very deterministic, right? The game of life, it doesn't stochastic or deterministic. It's very much deterministic, but he also says, don't use stochastic equations, because it's not also, you cannot describe life. In order to describe life, you have to not use deterministic rules, not use stochastic rules, but you have to use indeterministic ones. But what is it? You know, we don't know it, right? So that the people are very much, you know, actually, I asked him, so what do you think of the game of life? And he was so pissed off against me. I said, I don't know. Anyway, so that's what we started, you know, playing around game of life. But the glider is whether the concept of gliders came up from his John Conway's game of life. But of course, you know, it's not easy to translate the game of life into chemical reactions. So there's a way to do this. And then one of the things that we're doing is like the grayscot model, the very simple two components chemical system. But as you can see here, right, they can replicate and they can move around. So the grayscot model in the 2G space that they can create a glider, which I mean, actually, we have to put the third components to this. And then it can start to move around, which the tone progress here in OIST, we wrote a paper on this with the Nathaniel. So this grayscot model can be this continuous version of game of life, and then the glider, we think. But also, there are some interesting recently, that we have very nice simulation of the continuous version. Maybe I should make it in a real space, right? The first one was a very discrete space and time and state cell automata that you can see the glider. The second one is the continuous state, continuous time. But I'm going to spend some time and then also state. But it's not still in a real world. So that I try to make the living glider, the real chemical reactions, that can generate glider. That's what I started working so much in Hanzik. He's now in Trento, of the University of Trento, Italy. So this is the oleic anhydride put in high pH water. Then looking through the microscope, and about 0.1 micrometers, millimeters, that you can see this droplet is formulated, and it starts to move around. So I was so surprised and I was fascinated with this idea that, okay, this chemical reaction is not always happening in global space, but it's localized and they start to move around. And actually, this droplet is fenced to the environmental pH gradient that they can climb up and then sometimes they fall back because they understand the gradient of the pH. So there are three different ways to model gliders. And the question is whether this is life or not. But we can say that, okay, okay, we already made a life. So we don't have to do anything more. But still, I don't think this one is not life. But also, recently, I did this simple chemical reaction. This is a saturated with ACS. So it's different components. Still, you can see a different droplet. It's moving around and interacting with each other. And different initial configurations, which you put this way, or if you put that way, and then you can make a different droplet. So the droplet is not only one kind, but there are the many different kinds of droplet that you can generate even in a chemical reaction. So there's a bunch of different types of gliders like in the game of life. So it's more or less this one is the corresponding to the game of life in the real world. However, of course, the biode is never satisfied with this kind of experiment, right? They say, well, it's just a chemical reaction. So why you call it life? And then actually, my friend stood by that in context and he said, it's not life. We have to call it life, right? So different sparing. So that's where the artificial life is currently in the state. But then I remember in 2009, adaptive behavior, it's a journal that there's a bubble web. She studies the small insects and then here that she wrote animal births, animal paper. Why not model the real one? Why are you stuff, you know, playing around with some strange toy model? That's what she said. So there's a bunch of paper that we wrote and I also have paper there. I'm happy, I'm happy taking biology as a natural history that's my answer to bubble paper. So it's so difficult to convince biologists, I don't have to, but it's so difficult to convince biologists to understand. You don't have to really think cell and DNA, but you really can study something else, but still can understand the living system, I think. And then I remember that one of the conference that the big professor in biochemistry came into the conference and he said, okay, I totally understand what's happening in the cell, microchemical state, and then things. So don't ask me what is life or what is consciousness. I understand that these neurons are connected with each other, right? So if you don't ask me what is consciousness, I'm very happy. So think there's no life, there's no consciousness, then everybody's so happy. That's exactly what he said in the conference. I mean, well, actually, you know, studying a neural network is fine, but the studying consciousness is different. It's very difficult, right? And then also, what is life is a big question that people cannot answer to it, but people can study biochemical processes. So it's a way different and how to bridge the gap is a big question. That's what the obsession about artificial life, of course. And then I remember in 2002 that Rook, he's a famous guy that we met at Roomba, I have two Roomba in my house, and Roomba is quite popular, especially in Spain and Japan. I don't know why in Spain it's so popular, but the Roomba, and when he said he made a Roomba, then we say, so this is the first application for artificial life ideas. So now that life, artificial life comes in the real world. But he said, and then I discussed with him that, you know, but it's not still life, he said, but this is not life yet, right? The robot never becomes life or your computer simulation never becomes life. So what's missing there? There's something that is totally missing in our computer simulations or making robots. And he said, maybe the parameters are wrong, because we have tons of parameters of the computer simulation. These days, I don't know, more than 1,000 or 2,000 parameters. So, but we really have to adjust the parameters. That's the trick, that's what he said. Well, maybe the complexity threshold is lower enough, right? Because at that time, we don't have deep learning. And then the models are so simple. And he said, okay, the complexity is still low, that we have to put more complexity, more complex things, then we can have the real given system. Well, maybe the lacking of computer power. Computing power now is, I think it is getting better and better, right? Everybody is using GPU and then supercomputers. So that maybe this can be overcome. But of course, maybe we're missing something fundamental, like Tom's eruption theory, I don't know. Some missing part is that we need to understand what is life. So we call, this is Brooks, by the way. So I invited him to the artificial life conference in 2018. Then I asked him, why don't you talk about your missing piece? And then we call it Brooks juice. So the question is, what is Brooks juice? That's the artificial life challenge. We have to find the Brooks juice. And once we found it, and then we can put it in our robot, then the robot becomes alive. So the question is, what is Brooks juice is the thing that we have to understand. But then, for me, we've been discussing what is Brooks juice. Actually, he didn't know we called it Brooks juice. So he was surprised. But anyway, so the juice is a bit difficult to find by just looking at the computer simulations and then just swimming in the ocean. So we really have to do something else. So that in considering the life itself, maybe that the evolution must be taken seriously. Life is always with evolution, evolution, evolutionary processes. So that we decided to device a minimal experiment, which is still sufficiently complex enough. But we can discuss several productions and can discuss individualities and evolutionary processes. And then what is it? That's the thing that I want to discuss in the rest of the time. Okay. So my choice is a tetraheminal. I don't know whether this one is the best material or not, but I want to discuss, I want to study this tetraheminal and then thanks to current technology that we can identify each individual and then tracking them for many, many hours. And that's the current technology that I can use. So even though there are thousands of biological agents that we can track, identify tracking with a unit or some other deep neural network that we can use. And then I also, I remember one of the students that I, one of the books that I read is that saying that conditional learning is possible with a, with a protozoa is a, you know, relative of a tetraheminal. That if this protozoa is conditioned with 500 health, it's followed by electric shock and it's controlled by 100 health is followed by nothing. But then he immediately or he or she immediately understand that once you get 500 health, that you are expecting electric shock is coming. So that they start to tremble and then moving first. But when there's 100 health, he can be relieved. He doesn't have any electric shock. So this is already possible even with the, you know, conditional stimulus. And I think, I don't know why, but maybe the chemical network inside the protozoa is also like a neural network can do reinforcement learning and it can learn something, can memorize something. So which I thought was fascinating, you know, even like this small little thing with a chemical network inside, then they can learn and memorize things. And actually, what tetraheminal is sensing is that chemical substance is released from the other, other individual so that he knows where the other one is. That's one thing. And then also there's a two cycles, right? Actually, the tetraheminal has a six, six, six different six. We have only two, but they have six, they have seven. So we just used one type, so there's no sexual cycle, but we can, we're only looking at the viscidative cycle, so that there is no sexual recombination, but there's only self dividing by self, right? That's the experiment that we're going to look at. And this is the experimental setup. Of course, I'm not a biologist so that I cannot do this, but I have a collaborator in Hibosaki University. He's Professor Kashiwagi that he set up this one, the microscope, and then put EDMS on the heating up plate so that you can observe and you can record all those tetraheminal moving around under the microscope. So it's a very simple experiment, but it's very difficult to do it more precisely, right? And then you say that this EDMS that we put up 0.4 micrometer diameter poles so that we can put oxygen and also food from those poles so that the tetraheminal is not starving, they can get oxygen, fresh oxygen and fresh food. This one is a good example. So these are the moving around, right? So we have to, we can identify them and then we can track them for many hours. And then we use a Baxter algorithm and then some package from NIH so that we can identify which one is which, how they move around so that we can calculate kinetic energy and then, you know, angular momentum and all these things that we want to understand from the physical point of view, the physics point of view. The one of the things that I'm interested in is there are the two exactly the same genotype and then the same numbers of the individuals, but this one and this one behaves differently, right? We pick up from the same gene pool and then put in the same different well, a tetradish and then observe. But this one is the, it's not so much gink, right? But this one is more active, right? So what makes this one is, is not so active and this one is very ginky, right? So we have to understand what the gink is coming from. It's from, from interruptions. It is from something else. So that's one of the, actually many biologists, my friend said, okay, I am doing exactly the same conditions, but even though this tetradition, this tetradition is different and that's the common sense that people know it. So I was wondering, you know, why this one happened and also even the same genotype makes phenotype. Of course, everybody knows it right now. There's one too many mapping, even though there's only one genotype, but still there are many phenotypes that can exist. The question is, what is the phenotype of tetrahymine? The question of tetrahymine and one of the things is that because this behavior is different, right? So how we can take like a phenotype as a kinetic energy distribution because my background is in the statistical physics. So one of the things that I can do is, okay, why don't you just look at the kind of, you know, energy distribution of the, of the tetrahymine. So that's what we did. So this is taking a logarithmic scale and this is the energy. So if this is a Maxwell Boltzmann distribution, then it should be straight line, right? Because it's a logarithmic scale and this is the energy. So if this is a linear, then it's a, okay, tetrahymine is just a Boltzmann particle. Maybe then it's gonna be so much interesting. But fortunately, there are different types, right? It doesn't look like Maxwell Boltzmann distribution, right? So, and also when you have many cells, like eight cells in the same tetradition, sometimes they are synchronized and they're going to the same distribution, but it's not linear form. So maybe just something else, but it's also some of the cells, individuals of the different forms and many, everybody has a different form. So, so it also depends on the case. If you like equations, by the way, I don't like equations, right? Usually that people say, okay, you know, these are small microbes, just random Brownian motions, right? So you can describe it with a Fokker-Prunk equation. And if it obeys a Fokker-Prunk equation, then the distribution becomes like a Maxwell Boltzmann role. Okay, so I was very much afraid of, you know, this can be applicable to my little creatures, but fortunately, no, right? So it's something else. So one of the, one of the techniques that I use, very simple technique, and I think everybody use it right now, but it's called the Kalbach-Lieber Divergence. Well, you can see me to write these channels and channel divergences, that you have two different distributions. And whether they are similar or not can be can be measured by this simple formula, like taking a logarithm of p over q, then you can measure quantitatively how much different this distribution on the other distribution. So if it's exactly the same, then p, Kalbach-Lieber or Jensen-Chanon divergences before zero, and it's very different, and it becomes 0.6 or something. So that this difference is, can be measured by computing this one, which is the Kalbach-Lieber distribution. Okay, so first of all, I'm looking into, these are the two different, this is the reference one, this is one, just one individual in the one petri dish. And then computing the distances, I mean the differences between the kinetic energy distribution with the two different individuals, which is of course the independence because it's in different dish, different dish, but still you can compare with this one and this one. And then if you take a swan plot, swan plot is just plotting all these Kalbach-Lieber points, then it's distributed like this. But once you have just two, sorry, I made a mistake here, but there's two individuals in the one petri dish, three individuals in the one petri dish, or individuals in the one petri dish. Then if you increase it, then the JST goes up, means that more individuals in the well, in the petri dish, more diversity that you can see. So however, comparing with this one, which is totally independent, it's different. So maybe there's some kind of interaction that exists. So we are afraid of that, even though we put different individuals in the same petri dish, and then I don't know whether they are interacting with each other or not. But it looks like there are some correlations. We measure like a special correlation, the two-point correlation, and there is some positive value. So that there is a correlation, so that there is an interaction in there. But as each petri dish, they are converging into some point, so that each petri dish has a different variety, but the individual has more diversity. I think it's difficult to say, but this is converging because of the central limit theorem, I think, because it's divided by those numbers by eight, so that it's averaging over by the number becomes more of stabilizing this behavior. But the message from here is that there are the interactions, and more individuals in the same petri dish, there is a more distribution, different distribution of the kind of energy that we can observe. So now we can come to the replications. In a repleting case, that we can also see this one is interacting now, so that this is here in this timeline. Once they, so this guy is a little bit inactive, right? Then start to replicate. So when they replicate, they start, they are slowing down and they almost stop, then they replicate, and then start to move around again. Still they are sort of hanging on each other, but maybe they are just going on, and they start to interact freely. So we look into this, this is all only the x-coordinate, because we have a 2D space, and then also we are only taking x-coordinate, and then this one is the first individual, and then dividing into two, and this is the time x-coordinate of one cell, and then second one, and this one is dividing into two cells for each other, right? And then this one goes to the final there, the eight individuals that are exist. Then we put the time series of x-coordinates, and well, you cannot say anything, but this is about 14 hours that we can have this one. However, when you look at the kinetic energy distribution, this is the first one, first generation, this is the second generation, the third generation, this is the fourth generation. I don't know, but this one and this one is similar, right? Well, it looks like the same, right? This one is also... Maybe this kinetic energy distribution is inherited to the off-screen, that's from our observation. But when you see this one, the different individuals, this is the first generation, this is the second, this is the third generation, this is the fourth generation. See, the energy distribution is similar, right? There's the DNA, there's no... We calculated, but there's no communication happens, so DNA is exactly the same. Still, the phenotype, there's something is inherited from the parent to the off-screen. And again, so this is the third example that this is the first one, and then you see, it's similar. So there is some inherited from the parent to the off-screen to the third generation, fourth generation. So maybe this kinetic energy distribution can be the candidate of the phenotype of this tetrahymn. And so this one, this one, this one. So the question can be, is the kinetic energy distribution the phenotype? Are they called mutations? Yes. Actually, there are the mutations. It means that this distribution is changing. As you can see here, this has more flux here, but it's more smooth here. And then it's different, right? And then also it's different from whether this one is coming from this guy or this guy is coming from this one. So these are the nephews, right? And these are the brothers, but these are the nephews. And then if the distance is increasing in terms of generation, then also the phenotype variation can be generated, I think. So again, of course, we can measure them by the Habak library again from Shannon, right? Then we can see like this diagram. So this is the, so this one is the same. So this first one is the first generation, the second generation, third generation, fourth generation. Fourth generation has eight individuals, third generation four individuals. And second one has two individuals. The first one is the first one, just one individual. And then the color means that whether they are similar or not, this kind of technology. And then the dark one is exactly the same, and the light one is very different, right? So when you see this, that for this one, first generation, the second generation is very similar, right? However, second generation, third and fourth is very different, right? So they're inherited from first one to second one, but there's no inheritance happens from the second one to third one. However, from the third one to fourth one, again, there's a very big blue square means that from the third one to fourth one, the inheritance still emerges again, right? However, when you see this, first one and fourth one is very different, right? So even though this one is inheriting something, this one is inheriting something, but what is inherited is different from this form, from this form. And you see here, so the collective within the same generation, they have a similar distribution, right? However, there is no inheritance. So it's quite complicated in a way that sometimes it is inherited, but sometimes it is not. Also, there's a collective effect so that in the same generation, they seem to have, sometimes they seem to have the same kind of energy distribution, but there's no inheritance from the previous generation. And an interesting point is, for example, like this one, it's from the second to third to fourth, they tend to have a similar one, but this nephew one, the first one, the kind of energy of the first generation is not inherited to the later generation, but this one is quite nicely inherited. So one individual from the second generation, its kind of technology is inherited to the third one and fourth generation. However, the second individual of the second generation, its kind of technology is not similar to the third and fourth generation. So even though it's from the same second generation, but the inheritance, the strength of inheritance is different from individual to individual. So the final one is the comparison of this, the individuals starting from the same, the common ancestor, and then same genotype but randomly sampled from the two, and then compare with them. I mean, this one is the same from the single cell, this one is, and this one is the same from the same population, but they don't have any common parents. So this one is from the same ancestor at the beginning, but this one is somewhere, but most of them have the same genotype. So the question is whether those behaviors are the same or not. And interestingly, if you look carefully, they're different. And then actually, my collaborator, Kashiwami sensei, said, okay, when you look at, I thought it was something wrong, but this one is about to rotate at the same time. I don't know what kind of synchronization effect this one makes possible. And then she, many times she saw this one, only from the population where the single ancestor can generate all of them. And then they have this kind of interesting behavior. So maybe there's some strange correlation in the population where there is only one, the same ancestor that they have. So there's another technique, which is also, which is also quite popular, but it is called multidimensional scaling. I think I hope everybody knows this. But basically, what you can do is pick up some distance, this one, and this one, as measured by distance, you can have the distance between distance. Then, but this point and this point is closer, but this point and this point is much further on. So measure this matrix of distance, and then projecting onto a 2D space with maintaining the distance, that's possible. So if this one is like a euclidean distance, then it's more like a PCA analysis. But you can do it, you can generalize this one to the multidimensional scaling. So we can use, I used a distance by the Kalberg library distances between two individuals. So that the individuals can be mapped onto a 2D space, but also preserving the distance between the two measured by the Kalberg library distances. Okay. Of each time scale, I mean, that's, I should be careful. So each time, each, oh, each two minutes that I calculate a Kalberg library distribution, and then calculate the JST with the other time series. So time series is segmented into two minutes pieces, then calculating a current energy distribution, and then also angular momentum distribution over this segment, right? Then comparing with the other one, then if it's, it's very similar and it's zero. It's almost come to the same point, but if it's very different, they are plotting at a very different point. So mapping into the multidimensional scaling, then you can see that if it's very, if it's stopping, almost stopping because of the preparing for replicating case, then it's coming to here, right? But if it's just moving straight and fast movement, they go to the different edge. And then in the middle, it's slow in the middle space, and then they're not so much rotating, but then it comes to in this area. So I computed, so this one is a three, three examples that I saw, I have demonstrated. This is the first one and dividing into two, dividing into four, and then this is a fourth generation. So we have here that we have eight trajectories, we have four trajectories, we have two trajectories, there's one trajectory, then they are moving in the empty space like this way. So comparing with this, I don't know how they are different. So it's not a kinetic energy distribution, but a little bit different, but we don't see so much difference between those three. However, when you're comparing with randomly sample individuals, this one has eight individuals, this one has four individuals, this one has four individuals, five, six, and eight, then when you're picking up from the randomly from the pool, they stick to here. They just in the middle, they don't stop, they don't move fast, but they're just making a little point on here. However, when the ensemble is from the same ancestor, they are moving quite globally, their behavior is more rich. And then I still don't know why, but the one reason is, so I said, population created by the sequential division of a single individual has a greater diversity of movement patterns than the population of randomly selected individuals, perhaps due to the disparate cell cycles or the summer character. So once they are replicating, they are cell cycles in the same phase, but once you are randomly picking up from the gene pool, they tend to have different cell cycles. That's our guess. We try to understand whether that guess is true or not by using RNA-seq seconds, single cell RNA-seq, that maybe there we can see why this is more rich and why this is not. But anyway, so this petri dish starting from the single cell or petri dish randomly picking up from other pool, even though they have the same genotype, but their behavior is very different. So the plasticity, one of the plasticity is not a single individual's effect, but also is the collective behavior. The plasticity, I mean, no, no, the phenotype is not created by the individual only, but the phenotype of the individual is also affected by the collective of other phenotypes. So I think it is very, that point is interesting. And also, finally, again, this is the Khabibak library. And this one is the Khabibak library of the unrelated individual. And this one is the related because it's from two to four to eight. And then they are more similar, right? Their kind of technology is more similar, but you see here that kind of technology is different. So this is the point one, this is the point one. So their behavior is much more similar to each other and almost exactly the same because my guess is that in case of a population which is created from a single cell, maybe there is no strong interaction because the chemical secreted from this individual is the same as the chemical secreted from other individuals because they are, they are in the same, they are the relative and also they are replicating in pace so that maybe the cell cycle is sort of synchronized, I think. But if you're picking up from the random pool, maybe the cell cycles are different so that their behavior is strongly interacting, strongly, sometimes it's correlated, but sometimes they're very not correlated because of the interaction is not just for synchronization, but the interaction between the cells is maybe try not to synchronize so their behavior becomes more sort of different. So the diversity is coming from also from the unrelated individual. That's so far, so I cannot conclude that behavior, that observation, but the more discussion is that the first one is maybe the kinetic distribution is a good as a phenotype. And then also when I, when we started this work, I noticed that the library in Princeton of the United States that he also already did a similar experiment, but without any interaction between the individual, that's the difference. And then he said that kinetic energy, kinetic speed and also angular momentum he thinks can be the candidate for the phenotype, but we think that the kinetic energy distribution is one of the characteristics of a phenotype, phenotype of the individual. Also there's a phenotype in the variation which is suppressed by the sensor interactions in both Homo and hetero populations. So there's a phenotype variation that exists, but maybe because of the population interaction between individual, sometimes phenotype is different. So the mutation is not caused by DNA with replacing with some other sequences, but this phenotype variation is coming from the interaction in the macro level. And the third one is that this is coherent, but the rich dynamics in Homo are comparing with hetero. So when it is created from the same ancestor, then their population becomes more coherent, but as you see in the MDS space, they are moving around quite widely. So the dynamics become rich in the sense that the population is more low dimensional dynamical system, I'm not sure, but in hetero population, hetero in the sense that hetero over cell cycle with the same gene, maybe because of this different cycles that they behavior is more, it's not so much rich, but this point is, we really have to think again and again, we tested things, but we still cannot come to the different conclusions. But again, it's interesting point is that the stability and the viability is quite correlated, right? And without thinking DNA or without thinking any actual mutation, but still we have a, when it can be hetero, it can happen, and how the viability emerges by looking into the interaction between the individuals. So this is actually what the So I'm pretty much interested in identifying individuals and tracking everything, and maybe it can change the way that we think what is evolution, and then maybe if we are lucky enough that we can update the Darwinian evolutionary theory. And what I wrote here is both that people want to know, but you said the collective intelligence, what is collective intelligence, right? So I think well, in order to understand what is the collective intelligence of the collectiveness, maybe we have to open up the environment, and then open environment is very important for understanding the collective behavior, I think. The collective intelligence can be only understood by coupling with the open environment. That's what we are doing. One of the things, so I don't go into the detail, but this is the largest scale web services, and it's a social tagging services. Then maybe the tag becomes like an individual that is replicating, but it's also mutating to others, and then interacting with each other. So this is the one example, and this is unbehavior. We are using a unit that's identifying and tracking. Then we are making a food portion for connecting to this part. So you can see interesting interactions, and then role differentiation can happen only when you are coupling. This nest is coupling with the open environment. Some other thing is that we are tracking honeybees. This time, we are putting QR code on its back. Thousands of bees in the honeybees hive, then there is only one layer and put in front of the video camera. Then after two days, open up the door so the bees can go out and come back. So this is done by Gene Robinson and his group in Auburn Champagne in the United States, and then we are collaborating with them to see what's happened there. And one of the interesting things that we found again is kinetic energy. But when we compute the kinetic energy of each individual, before the door opens, so this one is a closed room, then there is a bursting behavior happens. But after you open the door, this bursting behavior becomes more periodic. And this green one is the entropy that I calculated, and the entropy is related to the role differentiation of the bee hive. And then differentiation happens when you open the door. So my guess is that if you open the door, and then there's open environment, which you cannot make a good model of it, then the information is coming from the environment to the hive. This information is the cause of role differentiation in honey hive. So that says it's kind of hypothesis, but as you know that the information thermodynamics, I don't know whether that's connected to this one or not. But the information thermodynamics tells us that the correlation between the system and the environment is very important. And we can measure it by the mutual information, and then it can be added as a third term to the first of the thermodynamics. So maybe this coupling with the open environment can drive us to understand what is the collective intelligence in this system. But again, I said, you know, we already have a very interesting phenomena without having any DNA or cell, but still you know, it's an interesting heredity, an interesting variation can happen in a smaller time scale so that they can add something to the evolutionary theory so that maybe we can understand the details of Darwinian theory. Because at the time, Darwin didn't know this micro-evolutionary process, but now that we know it. Thank you very much. Yeah, thank you very much. Are there any questions, discussion? Yeah, thank you very much. So what is your definition of open environment? So the like a petrodisha looks to me like a closed environment. So this one doesn't have any open environment. So in order to understand the actual collective intelligence, you really have to couple with a couple with the open environment. So like there's a flower, there are complex something. So is the earth like open environment or is it closed? Which aren't? Earth? One planet. Yeah, that's a big question. So I think it's this earth, the earth itself is a throw system, right? Because it's in the universe, there's not happening with others. So in this sense, the complexity or the intelligence is limited. Yeah, it's too much to say about this, but yeah, you're right. So the earth itself, the only the energy is coming from the sun and it's not normal flow, the information flow is coming from others. Well, maybe the alien is coming without the information flow. Otherwise, I don't think there's maybe it's not just the open and closed degree of openness. Thank you. So also thinking like on this planetary scale, I'm really interested when I hear at the Santa Fe Institute, Eric Smith is talking about metabolism and the origin of life. And he says that the Krebs cycle is like a hurricane. And it's like the hurricane is a dissipative process that like moves energy along the strong energy gradient, the strong difference. And so I hear him saying that like life, maybe this is not the Brooks juice, but maybe this is a component of it. It's that it's this process that efficiently, you know, kind of like dissipates energy across this gradient. I wonder if you if you consider metabolism. Yeah, well, for tetraheminar, there's a metal metabolic cycle inside already. So we don't have to think about the metabolic cycle. But for the chemical reaction that we do have a very simple chemical reaction. This one, too. So the simple chemical reaction is there. So I don't know whether you can call it metabolic cycle. But it's taking up the chemical from outside, having chemicals and making a membrane. And because of the membrane that this one can move. Then again, they can take chemicals from outside. So there's a feedback going on between the cell and the environment. So that can be the minimal metabolic cycle, I think. Yeah, yeah, because like Eric Smith and Moravec, they discussed like a reverse TCA cycle can be the first metabolic cycle that we can find in life. So it's possible that we can replace this simple grayscot model with a TCA reverse TCA cycle and put into this one. However, do you think you can understand what it's like? Or do you think that becomes life? My fear is I don't think so. Maybe this, even if we replace grayscot reaction with a reverse TCA cycle, but still we're going to have this kind of behavior. But we don't think that this one is the life itself. It's something else I think it's missing here. So that's the one thing is that maybe the infinite complexity of the environment coupling with it, there is an information flow. And that might explain how the life is. But I'm not sure. Thanks for the talk. I'm a biologist. And so in biology, we usually define life like these tick boxes. And we say is it check off if you can check off all the boxes, it's alive. And so are those boxes different in different fields? Like in your field, how do you define what life is and does that differ from how other fields define what life is? I mean, so what's the definition of life in artificial life? Like how do you, in the beginning, you were saying, is this alive? Like is the Roomba alive? I would say, of course not. It can't reproduce. You can't check off that box. But I mean, do other fields have different definitions of life in order to define what's alive and what's not? But from my experience, the biologists never care whether it does life or not. I asked many researchers studying biology, and if I asked, what do you think is life is? And they said, what? So here it is. So they don't really want to understand what they want to define what life is, because that's exactly what they're working with, right? If one day somebody comes to the researcher and says, okay, what you're doing is just a poem or nothing to do, right? Then do you think he's going to give up his research? I mean, so I think what is the life kind of question is only for physicists or computer scientists, because they are very much embarrassed with this idea. But for the biologists, what life is, just they can touch, they can see in the test tube, they can do some experiments. That's kind of my bias. But what do you think? I mean, in intro biology textbooks, the first page tells you what life is. And it tells you it must have these particular qualities. And if it doesn't have all of them, like viruses don't have one or two of them. So it's not so clear like whether viruses are alive or not. There's a debate. Some people would say there's no debate. Some people would say there is a debate. And some people would say some of those checkboxes don't belong. But I mean, they do exist. There is a list that changes over time, as it probably should. But I think biologists often think about what's alive and what's not like things that just like replication in and of itself does not mean life. I mean crystals replicate. So I mean, you have to have a number of different categories that you can check off in order to call something alive. I think what those are, I don't think are decided at all. Well, I can take your point. That's why I said artificial life is about the evolution of the brain. If you ask people what is consciousness, I don't think brain scientists can answer to that question. Because they understand the neurons and they understand how the neurons connect to the other. But if I ask, so what is consciousness, I mean, they say, well, you know, may or may not exist. So the emergence is something that we don't understand yet. Life itself is also the same thing. We can understand the biochemical reactions, connections and networks. But if somebody asks, what is life, then, you know, there's no simple answer to that, right? So yeah, you're right. So I, there's a bunch of this, the properties, that's one of the definitions of life. But still, you know, because my background is physics, so that we are still thinking that maybe there's a missing fundamental principle that creates life. Thank you. Thank you for your talk. I mentioned about the distribution of kinetic energy is some sort of phenotype. And remember in your examples, sometimes it looks like inheritable, but in other cases, it's not. For me, it looks like the kinetic energy distribution is like a continuous trace of what variation is influenced by both genetic and the non-genetic factors. And I'm wondering if you have any ideas like how inheritable the kinetic energy distribution is. Yeah, that's a good question. I cannot answer properly to that question, but I think the cell system can allocate energy in his lifespan. Then the kinetic energy is the result of allocation, scheduling of this energy. So that's why they are very much similar for this one, and this one, and this one. They are different from each other, right? Even though after forced generation, this one is similar, but this one is totally different from this one, right? So the scheduling of energy allocation is written in, I don't know, maybe just a way to, you know, protein, I mean messenger RNA activation or something. And that is inherited to, so that except DNA parts, you know, the rest of the cell can be inherited. Maybe it's an initial state or maybe that's the initial configuration of the chemical components surrounding the DNA can be the media for non-DNA, I think. But I'm not sure, you know, that's why we are looking into this RNA system to see whether there is some media that can carry non-DNA in the transportable. Thank you very much. Thank you. Okay, I have a question also regarding the kinetic energy distribution. So it's more of a curiosity thing, and I don't know if you did this in the lab or not, but you showed the difference between the vitreta dish which had the individuals from the same ancestor and the other one which had the individuals from different ancestors, right? And you showed how the kinetic energy is different in both vitreta dishes. But then, like, I would also think that maybe there's just something in the vitreta, in this vitreta dish that's different from the vitreta dish. So, like, have you done the experiment where you tried to allocate the individuals from one vitreta dish to the vitreta dish of another group? So it's like the whole eight in another, so that to see if the difference in kinetic energy is actually due to the environment itself or is it due to the genetics? For the moment, we don't have the technique to pick up one of those cells to bring it to the vitreta dish. Well, yeah, definitely that's something that we have to do. Okay, yeah, because I also thought maybe it's like, I'm not a microbiologist, so I don't know how to do it. But then when you mentioned that there were different individuals from different ancestors, I thought maybe there is a way to do it. Just always recording everything. And then by the unit is that they can identify each individual can track everything so that we can calculate kinetic energy, but physically picking up one individual and then putting into another vitreta dish is something else, right? Okay, so yeah, but your question is very much like a special one. Yeah. Okay, thank you so much for the talk. If you know any good way to do that. Oh, I wish I'm not a microbiologist at all, so I have no idea. Thank you. Thank you very much. In this experiment, there's a small guys when you like increase the number of the small creatures, the size of the dish was the same. Yeah, so like the density also was changing. So did you like try to take a big dish or something like this? So maybe the size, the size matters, but yeah, it is 800 micrometers, and then the micro is smaller than this. But it's like relatively small. Yeah, thank you. Oh, it's about 40 micrometers. So yeah, may or may matter, I don't know. Yeah, but you're right. So we also have to check with the larger size vitreta dish. Yeah. Okay. Thank you very much.