 Yeah, bareft says to Armenian University and Armenia to who generously hosting this talk. Thank you very much. And also many greetings to Yerevan and Armenia, a place I like very much also because my partner's Armenian. I have been several times in your beautiful country. So thank you very much for hosting. So I'm talking, I'm Georg Nordhoff. I'm talking about an issue which touches our daily life more and more. Just recently a chat GPT has been introduced. My students actually often use it to improve their English and then to correct their supervisors English. And so my topic is do robots artificial intelligence have consciousness, a view from the brain. And I do think that there are some lessons we can learn from the brain. So that's why I asked basically you see here in the lower part of my talk. How does the brain do it lessons for artificial intelligence. So the $1,000 question is of course, what is conscious. Nobody knows. Everybody knows everybody experiences. But if you ask 10 people, you probably get 20 opinions. So I point out two features of consciousness here, which are key person perspective and point of view. What on earth are these strange terms. We have a first our brain in Dallas as with the first person perspective. So I look into the zoom, the zoom is part of a wider room and the room I can look out of the windows is part of the wider room. So I have a certain center a certain perspective, a certain person person perspective, which is unique to me. You have a different perspective. So we have a perspective view from on the world, but at the same time being located perspective view from within the world. How is that possible. So the key thing is philosophers have discussed that we have a point of view. The first is a point of view. Imagine, you're going to Yerevan onto the hills. Let's say, we're mother Aminia stance, and you have a point of view. Overall, Yerevan, the beautiful view you look into the valley, you see the city. Now, now you go down, you go down to Northern Avenue. Let's say to the opera and you have a completely different point of view. So now you see the hills up there, but maybe you see the mother Aminia, but not as clear, but you do not see the rest of the center. You have a point of view. And what is important that the point of view angers us in a wider context here in the world, your point of view is not within the brain itself. Your point of view is between you and the environmental context. Yeah. So point of view angers us and our body in the world, we become part of the world and can develop a perspective view on the world. That is key for consciousness. And I will tell you how that is generated. So the question is, the point of view provides our interface and our location within the world. But what makes this interface possible. That leads to the key topic of my talk. Basically, if you have no time and say, Oh, God, this note of this really boring, you can basically turn off and say, I got the message. It's just about time. However, in the following of my talk, I want to explain a little bit and give you examples why time what time scales are and why they are important. And as I like to always bring examples from nature. One example is the ice. The current neuroscience of the view of the brain and a lot of artificial intelligence to just operates at the surface of the iceberg. However, we all know that an iceberg has a deeper layer, a deeper layer beneath the water. That provides the interface and the stability of the iceberg. It locates us in the, it locates the iceberg in the ocean. But at the same time, it distinguishes the iceberg from the ocean. That's exactly the interface and these deeper layer is key for the brain key for consciousness. And my claim is that current artificial intelligence did not yet really exploit this deeper layer where the where we are where the artificial agent needs to be aligned to the environmental context in an adaptive. And the argument is that this deeper layer where you have this alignment between the iceberg and the water or between us and the world, slash the point of view that is about different timescales. Now, what do I mean by timescales. I will further elaborate in my talk. Here I just want to say that this timescales are very important in sort of matching between your brain body and the environment. I know this is a very complex picture, but what is important here you have a continuous matching. For instance, what do I mean by matching. You, I make certain movements, and maybe you try to match your movement with my move. So you align your movement to my movements. Bad example, I know, consider music, when you're dancing, then you try to match your inner rhythm with the rhythm of the music. So now you tap your, your legs and your arms, according to the rhythm of the music. So you match with the music. Excuse me, you can see the slides. We only we see the first one still. I don't know that I don't know. I don't know how this works. You can reshare again. I'm sorry, maybe you can reshare again this screen. So it works. Can you see the time slides now? Yes. Yeah. Okay, let me see when I push on the next one whether you see the next one. Yeah. So the matching is, so when you listen to music and you dance to music to continuously try to match your inner rhythm. To the rhythm of the music. And that's a pure statistical process. Stochastic matching. Bruce West, one of the key people of the dynamic system theory called complexity matching. And this is what I try to indicate here. And for that matching, the time scales are key. So you need to have, and as we all know in the environment, you have a large repertoire of timescales and that the brain tries to match its own repertoire of timescales. And I will elaborate. Now, let me know. Can you see the next slide now? Yes. Okay. So, as you know, I like to do analogous comparisons. The window. So look out the window. So here you see a window, a large window, and smaller windows. When you just look through one of this very small windows, you just see a little branch and maybe some leaves. So you don't really know it's just a tree is this a bush or something. So let's say you look at the six smaller windows together on the left. Ah, you see, you, you, you, you see the branches. And it's a trunk, but maybe you don't see all the leaves. So you don't really know whether it's winter. And the winter in Canada here can last very long as we all know six months, or whether it's spring with some emerging leaves, or it's summer, and so forth. And now you see, when you just see on the left, you see maybe that's a single single tree, but now when you take all the big window together you see, it's a whole forest. So meaning different window sizes allow one to see different features in the outside world. It's a continuous zoom in and zoom out. Actually, we are in zoom. So you can follow that. And basically now the time scales with the brain, the brain imagine the time scales like temple windows, you have presented up here, a repertoire of different time scales, temple windows, short and long. So it's a repertoire of different times. So to turn out the more time scales you have different time scales, the better you can align and interface with the environmental context. That means if you have more different window sizes in your living room, you can better zoom in and zoom out and you see more details of what is out there. And exactly the same with the brain. And interestingly, in the brain you see here in the lower, you have sort of a gradient from shorter to longer temple windows or timescales. So the shorter time scales the shorter temple windows allow you to zoom in to see the details to see the single finger maybe even if you had a very short window here even to see my finger nail. Now, if, and that's basically in the back of your brain like for instance in the visual regions of your brain. Then when you move forward, for the anterior part here where I now do this, that's where you have longer timescales. Yeah, that's where you don't see the single finger nail anymore but you see the whole person, aha it's a hand, aha it's a hand of a person, aha it's a hand of Gio. So and that's so you have to imagine just like the situation in the living room. So basically through our brain and its timescales, we perceive the outside world in different features smaller feature larger features zoom in zoom out. And the brain has a larger repertoire of timescales quite amazing and we haven't really understood how the brain yields this large repertoire of timescales. It makes the brain so adaptive to different contexts, which again can be characterized by different timescales. So now you say, what are the timescales of the brain good for and I think I already indicated then and we introduced the concept of temple special alignment, which is sort of it's they allow aligning to your environmental environment. Remember, my example of music is exactly that. So you align your own inner rhythm to the rhythm of the music. And most important, that gives you a feeling of groove. It gives you a feeling of synchrony and you feel good. Yeah, so you extend your own self in a virtual way to the music, you are part of the music. And that's exactly what many artists tell me my partner is a composer, though we have a lot of musicians in our circles, and that's exactly what they tell me when they really become part of the music, then they feel the groove, and they play very well. And what I want to point out with this figure here, which was done by Philip klar who always one very clever guy in my group who does always beautiful figures. And you can see, I think that speaks for itself. You can see what is important that you the brain has an active mechanism. It does not just passively receive the input, but it tries to actively align and synchronize with the input. And I'm sure that when I do these kind of movements that maybe your brain tries to get in synchrony with that and by that it may be able to decipher the meaning of what I try to desperately convey. So this is probably that's why Philip made this arrow from the brain to the world. So the brain by actively aligning to the word becomes basically part of the world, and that provides your point of view, and that is key consciousness. Now, that's exactly what I want to point out now I give you some examples, how the time scales are important for distinct aspects consciousness. The Brian's time scales provides a temple windows for consciousness. Again, so this is an investigation which we did in patients who lost consciousness in different states, for instance in sleep, when you fall asleep. You gradually lose your consciousness. Now you want to know what happens with the time scale. If not of the right, then also your time scales should get changed when you lose consciousness during the gradual sleep stages. And that is exactly what we observed. So this was done by Federico Silvio in Italy. Actually, he's both a philosopher and an excellent neuroscientist, which is very nice. For instance, what you can see you measure the time scales more for the experts the engineers for instance here with what is called the autocorrelation window. And it's basically literally a temple window by means of which the brain can adapt or align to corresponding temple windows in the environment. So it's a matching process. Now, in the sleep, what Federico could observe this is that your time scales become abnormally long. So you lose the shorter time scales and only the long time scale. So you cannot zoom in anymore. So you do not see the details anymore. The fine grained the smaller temple windows completely disappeared. You can see this here, your time scales measured by the autocorrelation window becomes longer and longer during the different sleep stages. Interestingly, in the REM sleep, which is typical for dream, they sort of revert to someone a normal feature. That's maybe why we can have consciousness in dreams. But let me proceed. So important when you lose the faster time scales, or when your brain becomes abnormally slow, slower frequency shift or slower frequency, you're losing consciousness. And what does this mean that your time, your temple complexity becomes minimized, and you lose the different layers of the temple windows and when your consciousness also includes different time scales. Very fast, you can follow this one at the same time as is longer time scale because you perceive the continuity that here Georg not of the same person is sitting here and talking and talking and talking. So you have shorter and longer times. Now if you lose the shorter time scale, you will not be able to perceive these faster movements anymore. And when you become longer and longer, then also it becomes more and more static. So you will not be able to perceive these movements anymore. Yeah, and then ultimately, and this is indicated by here I know this is a complex figure this is this is done by Federico also very nice figure. So when you have, for instance here, he had this situation with the police with the thief and the policeman he has to see if he's a policeman, and the big police stop stop right now. And you need to reach me. So you see you need shorter time scale to pick up each word. You need also the slightly longer time scale to make the connection from police stop and stop right now, and you need very long time scale to put everything together. Now if you're losing the faster time scales, your picture becomes more blurry. Yeah, because you cannot distinguish the details anymore. Here's a speech. It's even longer. And now if you lose the faster completely you have only one long, everything is completely black. That's exactly what happens when you gradually lose consciousness. And I think that goes very well with your perception. Now, I give you another example how the time scales impact consciousness. The brain time scales following actively aligned to the time scales of the environment already mentioned the synchrony. And that's key. This alignment is key for consciousness. This, imagine here, you know, you know my preference for other examples already. Look, it's a surfer. So the surfer, when you ask the surfer, so I spoke with surface is very, very interesting. So before they go out and swim, let's say to the really high waves, more than 10 meters, that's a litmus, that's a litmus test for the professional surfer. They're more than 10 meter waves. They first, you see them, they observe, they don't go right into the waves, but first they sit there with a surfboard and look at the waves. And I'm sure that in their brain, they sort of encode and follow the dynamics of the wave and then they probably with their, for instance, their motor cortex, they simulate that dynamic of the wave. But now that they can basically tune their movements to follow the dynamic of the waves, you see this beautiful in this picture, the guy really adapts to the dynamics of the wave. And I, if I were doing a snapshot a millisecond later he would be in a different position, again in order to follow the dynamics of the wave in his own movements. He follows the dynamics of the wave in his own movements, the better the groove. That's what they all tell you, that's the groove. It's the same as I said with the musician, when you're completely into the rhythm of the piece, you become part of the piece and you have a fantastic feeling, a groove. The same here with the surfers, that's what they tell me. Now same with the brain. So this is an investigation, again by Philip Claer who does these beautiful figures. And you can see that here he presented external stimuli, auditory stimuli, every 20 to 30 seconds. And then he was saying, maybe your brain has in exactly the same frequency of 20 to 30 seconds, it has an increase in its activity slash its power spectrum. So in this 20 second, this is exactly this frequency range, you can see this here is the blue bar. And you can see really here, this is what is called a power spectrum, this is a frequency, this is a power, you can see an increase in the power of the brain in exactly the frequency range of the task stimuli. This is called task periodicity. Now, what happens when you lose consciousness in anesthesia, you see it's completely free. Your brain doesn't follow anymore, the external environment, it's isolated. Yeah, it's isolated. And that's why you lose consciousness, because you cannot develop a point of view anymore. And I hope that your brain now follows my movements, so that you get the full meaning, the importance of this task periodicity. Yeah, and another thing, more for the experts, which might be interesting, you see that the power spectrum is completely flat. There's no difference anymore between slower frequency and faster frequency. This is called whitenose. And you see in the awake brain, you see the slower frequency have more power. Look at the seaside. The big waves are very powerful. And that's why they're good for the surface. When you go to Hawaii you see these big waves, amazing. Because the faster waves do not have as much power. So that is what is called pink nose. So you see really when you lose your pink nose becomes white nose, you lose the task periodicity, you lose consciousness. Yeah, so meaning you cannot be part of the environment, you cannot align and you cannot develop a point of view anymore. So this path to the sea, it's a groove of the brain, analogous to the groove of the self. Now, basically, what is important here, and that's a key point of the brain that we pointed out for the artificial intelligence later. It's a combination of neural waves and world waves. I just wrote a book that just came out called neural waves, and they need to interact with the world waves. And that interaction together gives you a point of view. Look at this guy here, the surfer. Due to his brain, his neural waves, he can adapt to the waves of the seaside, and that gives them a point of view, and that he can use the world's power of the sea wave to propel his circle. That gives them a feeling of groove. Now, if I'm right, then you would expect if the brain has a larger dynamic repertoire, more time scales should lead to more extensive interface with the environment. And again, there's an example for that meditation that gives you a form of an extended consciousness. I could make a one cheese sandwich. Please turn off your. Can you please turn off your mic. I think we heard some things you say in the background. So what happens in meditation. So this is an investigation done by Austin Cooper and now it tells that I'm muted. No, no, we'll listen to you. Okay, okay, good. So, so what happens in meditation. So here, Austin Cooper and Bianca Ventura from my group investigated naive versus proficient military. In key feature in using people you have in your consciousness, you have a dichotomy, you have a dualism between yourself and the world. You have to distinguish between yourself and the other. So that's what is called a duality and that duality in your consciousness stretches to your in the environment to the body to your own thoughts and everything. You always make a distinction between self and other. The proficient meditator. They don't experience this duality. They don't make a distinction of their self from the environmental context. They're completely aligned to the environmental context you see here the self as a distinction has vanished. So that's what is called the non duality. Interestingly, so exactly this sign of dual versus non dual organization of the mental side on the mental state you can also see in the brain. So usually the naive meditator was a common person has a dual organizational dual topography in the brain you distinguish between the more outer regions in the brain or called lateral regions and regions the blue ones more in the middle of the brain. Yeah, the middle regions of the brain are located to associated with your sense of self. The outer regions of the brain also called the central executive network are more associated with your with the other with your perception and cognition of the other and the outer environment. So, and this is called the default mode network so you have a clear dual what is called dual topography in your brain. However, when you look at the proficient meditators, there is no longer this principle distinction between these two regions, but they're very highly synchronized with each other. And there's basically what we described as a non dual topography. So you see what happens on the mental level shift from a dual mental state to non dual mental state also happens on the neuronal level, the shift from a dual neuronal topography organization to non dual topography. Now you might want to ask what happens to the time scale in these meditators. And what they found that the dynamic repertoire the repertoire of time scales, largely increased doing with increasing meditation proficiency. So your dynamic repertoire instead of five different different windows, you have 20 different temple wings. And that of course allows you to align much better with the environmental context in a much more fine grained way. So and then Austin added this beautiful figures which I always like. So here you have a non dual awareness and dual awareness, which basically distinction between yourself and the other, but here you completely immerse in a non dual way in the environment and that's exactly what these meditators proficient meditators tell you. So meaning you have an extension of your time scales, it's also an extension of your consciousness, and that's an extension of your interface with the environment, and that's exactly what these proficient meditators tell you. I finally come to the topic I should talk about, what artificial intelligence can and can't do. And I contrast that with, as you already know, what brains can and can't do. So, in human brain, the brains are remarkably adaptive. Absolutely. You can play go you can play chess you can listen to music you can make music. As a baby you can even drive the car, even if you're as young as this one. And it's not yet a self driving car. So the time scales make our brain highly adaptive because the more time scales you have, the better you can match with it, much larger time scales in the environment that makes the adaptive, and that also provides your point of view, consciousness. However, as we all know, we are not perfect. These different time scales make us prone to error. We are not error free. So we are never 100% that I call the error problem. So that can be contrasted with what is called the generalization problem in AI. And that is injury to so the current and I is good in one particular task. Just go chat GPT or text chess goes, goes clear, but they cannot do the other tasks, particularly they cannot do novel. So meaning they are not adaptive. Yeah, they're perfect and really good in one particular task. And when you look at these problems as bird for instance, natural language transformer has a very limited number of timeskills. Yeah, as bird when you look into the exact organization is basically three layers of implicit time scale, but not explicitly modeled as such. That's probably why they're so good, because they have just particular timescales designed for the task bingo. That's enough. So that makes some so good. But they're not confounded by the other timescales. Of course, and the more timescales, you might have a perceptual illusion if you use the wrong timescales in that moment. And that's exactly what happens in humans, because we have this large repertoire of timescales, sometimes we're using the wrong one. So we put we zoom in instead of zoom out or vice versa. Yeah, that for instance happens in mental disorders like schizophrenia and depression. As you know, I'm also psychiatrists. So we all know this. Nothing is for free. When you look at biology at evolution brain and the same applies to artificial intelligence. Nothing is for free. A large repertoire of timescales is excellent is good for adaptation and having consciousness, but bad for error free perfection. And you have mental disorders with people who really suffer from this lack of error free perfection, for instance, obsessive compulsive disorder. Yeah, so there you see nothing is for free. So we need to find a balance. Now, let me go beyond the humans. And then I come back to artificial intelligence. And then I will come to my conclusion. It's not non human species. Non human species are for instance, plants. Do plants have movements. You would say, well, the plant is just standing there. It doesn't have movement. That's not true. The plants move. But they move so slow that we cannot perceive it because we don't have that time scale. The plant is in 24 as a plant move. But we cannot perceive it because our brains timescales do not allow for the perceiving those movements of the plants. Another beautiful example, much discussed in philosophy of mind, Thomas Nagel, 1974, what is it like to be a bet. The bet is a particular example, which can perceive ultra sonar waves, which we can't. And he says that endows the bed with a different point of view within and towards the world when compared to the human point of view. And he's right. We cannot perceive ultra sonar waves. So meaning different species, different timescales, different degrees of interface with the world. And the degree of overlap between human and non human timescales probably decides how much we can communicate. So here, this is a schematic figure done by one of my former students, Meschat Golozoki, who actually was an engineer, a very clever engineer. Unfortunately, now he works for the stock market industry, which I find very sad as a brain scientist, but he wanted to go that way. So here he drew that different species have different timescales. And that provides you with different interfaces with the environment. And the degree to which the timescales between different species overlap might decide upon the degree to which we can communicate. Now, the key question, can we build timescales into artificial intelligence? In principle, yes. U.S. engineers might say, oh, not of your justice psychiatrist for loss of a neuroscience, you have no idea how complicated that is. But I think that's an area where neuroscience and AI and engineers need to work together. And I'm very open to that and you will see why, because for me, this has an important implication. So now, if we know the timescales of the brain, I showed you that the timescales of the brain are key for aligning, synchronizing us and the brain with the body and the environmental context. That's our interface. Like, remember the living room example, the windows of the living room, our interface with the external environment when you're in the house. Same the timescales of the brain. They are the interface, our interface through which we are connected with the outside world. Now, if we want to make that connection more specific or more focused, we might want to change. We might want to build artificial agents with specific timescales. Yeah, and we indicated that, for instance, here, let's say you can build an, I know we cannot build an artificial brain. The idea of a blue brain by Henry Markram didn't go so far as we all know in neuroscience. But you can add additional timescales, or you can pronounce certain timescales. You make certain timescales more specific, more fine-grained or enlarged. So that, for instance, indicated here. And I think that's where beautiful, for instance, the deep learning networks, the different layers. I could imagine that into these layers, I would love to build in some timescales to also to use them for better prediction of timescale data from the brain, and then, of course, also for better adaptive agents. Now, this is what I'm, the question is, what exactly do you want your agent adapt? And then you need to say, okay, these are the timescales I need for that feature in the environment. I want my artificial agent adapt. And then it's exactly remember this picture of it I showed you earlier, the complexity matching between environment and brain. Then you have, again, a complexity matching of the stochastic structure of the timescales of the spontaneous activity of the agent and the stochastic structure of the environment. So it's really a complexity matching process. That's why it's really important to include the environmental context in the modeling of your agent, because that gives you the goal, the purpose with which it should interface. And I would deconstruct, of course, no wonder you already know all that environmental context in terms of timescales. Now, what are the timescales in artificial agents? Good for why on earth not of do you want to build artificial agents with timescales? Is it just scientific playing? Or is it just curiosity? Or does it have a specific purpose? The letter. And I tell you why. So, when you have here, this is a good example, you have a biological agent, you have an artificial agent, and this is quite well illustrated, you have a certain larger array of timescales. That allows you a more broader perception or more specific perception and a more specific decision. Now, why do I want that? Because as a psychiatrist, I'm dealing with mental disorder, like schizophrenia, depression, anxiety, bipolar, post-traumatic stresses, all of that I can continue following. And we currently do not really have biomarkers, meaning diagnostic markers. Diagnosis is sort of guessing based on some lists, scales, but we don't really have markers like in other biomarkers like in other medical disciplines. And that also means that therapy is guided by trial and error. You have to try out maybe this drug work, maybe cognitive behavioral therapy works. So we do not really have a clear biomarker, nor really precision therapy. So the question is, how could we change that? So that's a real issue and the mental health problems are increasing. So and most of these psychiatric disorders show specific temporal disturbance. I give you the example of depression. In depression, people experience themselves as too slow, they're lagging behind. And the whole brain showed in various studies is indeed literally too slow. So that's why you experience yourself as too slow relative to the faster environment. Your visual perception is too slow. Your movements are too slow, psychomotor retardation. Your thoughts are too slow. You can all show this. So the speed. So depression is primarily a speed disorder. So then the question is, and I remember I tell you very typically one of my patients was a young girl, 16 year old, she came with her mother, her mother was speaking normal speed. And the young girl didn't speak at all. Nothing. She was completely as we say mute. So I said, what is this? So it turned out to be a pure depression, no neurological disorder. And later I asked her, why didn't you speak at all? She said, my mother, I knew that my mother was speaking normal, but for me it was too fast. I couldn't follow that by a shutter. Because inner time was too slow relative to the outer time. And that's what you literally see in these patients. So how can you treat them? Keep that in mind. Now I show you an example, and then I will come back. So the idea, what would be great to do and maybe we can even do it together in collaboration with the Armenian University of Armenia, if you're interested in my institute here. But we build a particular robot with different timescales. Let me show you this one. So this is a robot, obviously, and this is a real robot, one of my former students actually build it, a real robot. And what you want you build in different timescales into the same. So when you listen to music, you have different timescales. You have fast timescales, very short timescales, longer timescales. Your timescales are very intricately built into each other. That makes the complexity of music. Listen to Bach, listen to Kacchatoyan. I don't need to tell you as Armenians. Kacchatoyan has very complex layers of different timescales in his music. So now, now imagine, so you expose this robot. Your robot has only one timescale to the complexity of the music. So if the robot has only fast timescales, longer timescales, here meaning only the fast timescales, the robot will always dance too fast to the music. Now, if your robot has only the longer timescales, but not the shorter ones, it will always be too slow. So you want to build a robot with a variety of different timescales, which it can match with the timescales of the music that will dance to the rhythm. Remember my rhythm example to the music. Remember the surfer. If the surfer has more timescales, it can better align to the different timescales of the way. Same here. Now, if you're depressed, sort of your timescales are more shifted towards the longer end. It doesn't look like this, but the faster timescales you can literally see this in the brain are a little bit decreased so you cannot react to fast stimuli. Your sensitivity to react to faster stimuli is hugely decreased in depression in your perception, as well as in your brain. We could demonstrate this. So now what you want in depression, you want an agent, an artificial agent or robot, which can dance with the patient in a timescale which is slightly faster than the timescale of the depressed patient, but can still, whether the depressed patient can still follow. So ideally you want, as I like to speak of a temporally augmenting agent, you want a temporally augmenting agent with particular shifts the depressed, very slow timescale towards slightly faster timescales and then gradually towards faster. We, with our vast array of different timescales, will have problems of specifically targeting that timescale, to which the depressed patient is still responsive to, but which also pushes him away from the very slow state. So my hope is that such temporally augmenting agent can really enhance in a very individualized precision based ways of timescales in, for instance, depressed patient in, for instance, music therapy and dance therapy. So here, because if you can build in specific timescales, you can really specify and zoom into exactly that timescale which is needed for instance the depressed patient. So for instance, one application and I would be very happy if we could work on this together. So now, of course, the million dollar question, can robots and AI exhibit consciousness. For those, you probably know, some of you may know, when in Soviet times, they were in the West, the radio Yerevan jokes, and in principle, yes, there were all this question, can you get a car in principle, yes, but you have to wait for 10 or 20 years. So now here, I have the same approach, but disclaimer is not a joke, what I'm saying. So can robots and I have consciousness in principle, yes. However, that will take time to build in the various timescales. Look at the human brain. It took a long evolution to develop all these timescales. Yesterday, I had a discussion with one of my students. And I think it justified you said, it's not just the timescales, it's the physiology behind the timescales because these timescales, you need energy. The brain is a hugely energy demanding organ. It, it's only 2% of the weight of the body, but it consumes 20% of the energy. And as we know, particularly in these days, it's important. So, and then this 20% the energy consumed of the body energy is 95% is invested into its spontaneous activity. So when you have tasks and when you see me and do all these movements, your brain will just incrementally increase in energy consumption by 5%. So what on earth does the brain with all these metabolic energy, it sustains its timescales and it was repertory. So you see there's a very intricate interface between physiological metabolic energy and the timescales and we haven't really fully understand that connection yet. So that's really, really important. But in principle, I would say, I hope that we can build an agent with specific timescales, whether they have consciousness or not is a second question. So with that, I'm closing. So I hope I could give some ideas how consciousness, what consciousness is how consciousness is yielded by the brain and how that provides some lesson and maybe ideas for artificial agents. As I said, ideally, I would love to have a temporarily augmenting agent. And that really can be subdued under the umbrella of what we developed here in our group, the temple special theory of consciousness. And as you already know, waves are important. That's why she related my title, my last book, which is came out neuro waves, Brian time and consciousness and as in physics, time is key. There's nothing special about consciousness. We don't need to invoke a mind some special facilities of functions in the brain is a basic dynamic process look into the ways of the world, and you understand what happens in the brain and consciousness. Thank you very much. I have a question. Sorry, sorry, sorry. I can barely understand you. Maybe you can come closer or something it's, I didn't really capture what you said. Okay, let me start again. Yeah, I think a little better yeah. It seems like those entries will have a major to build stuff. And not attractive for us that are in the human brain or human work. It seems like we have been studying how the human brain works for centuries and now we are getting all that knowledge to build the road that communication of us. It's really difficult for me to capture. I'm here and talk and he will hear. Yeah, yeah, yeah, I can understand you much better. Yeah, great. Thank you. Just one second. Okay, okay. So longer seconds. Hello. Hello. Yeah, I can hear you very well. Okay, so I'm repeating the question for the third time. Sorry. So, you know, when we look back in our history, we have been building stuff that are imitations of nature. But now we're trying to create robots that are the imitation of us and we've been studying the how the human brain works for centuries. And now we're using that knowledge to imitate ourselves I want to know why you have that desire. Okay. So we capture very nice points learning from nature. And we haven't fully understood nature biology and dynamics and dynamics is a key feature of nature. Absolutely. So, whether I want to imitate the brain. I don't think so. And by the way, nature imitates itself. And by imitation, you get mutations. And by mutations, you get evolution. So imitation is a very basic process of nature. And it's also in our psychological repertoire. We imitate the little baby imitates the mother and by that it learns, and then it integrates into its own spontaneous activity and by that it develops as an individual. So imitation is a basic natural process. And my key is, of course, I'm a fully aware of the dangers of artificial intelligence. And actually, we're just together with Taiwan and Germany, we submitted a bigger grant on the cultural implications and relevance of artificial intelligence. And I think the cultural perception might be different in the east slash in Taiwan, and in the West, let's particularly in the Anglo-American world. Yeah, so there's a definitely a cultural context to it of course because we are all environmentally context dependent. And you see that's also in my plea for agent. And also from the European Union, which just yesterday was in the news that they want to restrict, make clear guidelines for productive use of AI versus destructive use. Yeah, so I'm completely go with that. One should not throw out the baby with a bathwater. And that means taking out the disadvantages. And I put out one potential area. I indicated that very clearly, I think in my psychiatric example at the end, that we can use these robots also in a very constructive way. And that's exactly what I try to indicate that you have to work on both sides. For instance, chat GPT was not allowed in England, in Italy, my Italian students are officially not officially not allowed to use it. Yeah, I think that's a boundary we need to delineate and demarcate. I hope that provides an answer to your question in some way. Thank you. We have here in question. It says, would you briefly explain once more the nature of depression in the viewpoint of time lapses. It's in the chat. The question. Yeah. Okay. So when you ask these patients, they're too slow. They feel too slow. Their movements are too slow. They cannot get up in the morning. Their mood. Nothing changes. Yeah, their thoughts don't change. It's always the same thought. It's called rumination. The perception is very static, very dynamic, and you see exactly this kind of slowness also in the brain. The brain is literally too slow. Where is that coming from? We currently do not know. Probably it is related to some normal for the experts for some biochemical modulations by for instance, serotonin, which comes from subcortical regions to cortical. And it's not just serotonin. It's a whole balance. But it is clear. There's strong findings that the whole brain is too slow. It shifted thoughts these midline regions where your sense of self is remember in the meditation. I explained that these regions in the middle of the brain are too strong, too slow. Your self is too slow. You always focus on the self and everything is negative. The exact causes of that remain yet unclean. I hope that provides an answer to the question as it is wanted. Hi, can I ask a question? Yeah. Hi. I was wondering whether this misalignment between timescales is also observing other psychedelic disorder like for instance and hyperactive person is it like a faster than the environment or other kinds of psychedelic disorder like addiction or hyper anxiety and so on. Thank you. Thank you for that question. So you open a can of worms because this is my one of my favorites. We call the special temple psychiatry. So you have in different psychiatric disorders, different temple disorder. So I indicated some depression is a speed disorder. Another disorder which we investigate a lot of schizophrenia. It's a temple imprecision. You have a millisecond. For instance, a recent study by ours and confirmed by many others really shows that they cannot synchronize with external inputs. Yeah, the rhythm of their brain doesn't really can align to the rhythm of the external stimuli. Now for the expert, it's done in auditory oddball paradigms. So you see decrease what is called intertri-face coherence. And this is all because the face, the activity shifts in a temple imprecise and milliseconds, which was shown in a recent paper by one of my other students, Stefan Lechner in Vienna, he invented a very clever new methodology to measure the spontaneous millimetre shifts in the activity of the brain. And that's highly increased in schizophrenia. So, and I could go on so different disorders have different kind of temporal disturbances. And that reverberates throughout the whole brain and that difference. If you want to know more about it, read our stuff on special temple psychopathology, as we call it. Thank you. Hello, Professor. Hello. Hi. Thank you for the presentation and have a question. So, basically, from what, from my understanding of AI, the simplest way to explain how it works is, and correct me if I'm wrong, you gave it, you give it an environment of data and then what it does it just kind of studies the relationships between, you know, inputs outputs and then that way, if you give it your own input, you can just get you another output based on the, on the, on the environment you gave it. And if I, if I try to make the analogy with how the brain works, I think it's missing two things, which is making its own environment, which is the equivalent of the experiences that you know a human goes through, you just make your own, you know, your own experience, which is make your own environment. And the second thing is, is the resting state, the brains, the brains resting state. So the question, the questions are, do we understand the resting state enough to reproduce it in an AI, you know, model, you know, that basic, that basic intrinsic activity. And then if we add to it the ability to create their own environment, which is have their own experience. And then what the AI already has is to just go further. And then basically maybe we could have a conscious AI so is that possible. Yeah, good question highlights some of the key issues. I think AI is very good at capturing some of what goes inside in the brain what is called cognition. However, the brain lacks exactly what you described and what I tried to point out the alignment. And let me specify that. So the AI passively receives the input. The AI has no chance of shaping the input and selecting the input it wants to receive maybe selecting but it cannot shape the input. So it is this active shaping of the input, which is key in the brain. For instance, if I have only slow time scales, I will not be able to process fast inputs, different from slow input. So my time scales, actively shape and select the input I can receive and process on. And this active process is philosophically this goes against a human like view of the brain, which is very still prevalent in current neuroscience. Yeah, there's one line of research, which is active sensing Schroeder and Lakatos wrote fantastic papers on. And that goes, it's this active selection, this active shaping of the input. Yeah, and that seems to be also abnormal in mental disorders. And for that the brain seems to use a time scales and it's spontaneous activity. So what he described what we measure as a resting state spontaneous or resting state activity is basically when you don't perform a specific task like watching me or listening to me or talking. Yeah, but in a way the brain is never rest this always something going on. And, and that, indeed, I would argue that this resting state spontaneous activity is a key feature of this as a brain. For instance, we do now a lot of computational modeling. And we often discuss, and the border is always that these computational models do not have a spontaneous activity. So that we often come to that pilot and then we can't really model it because the spontaneous activity in the model is not there. So there you see that support. And indeed I would argue that your agents need to have a spontaneous activity independent of externally. That's the key feature. I didn't go into that spontaneous activity business so much. But that's exactly the point because this spontaneous activity has all these different times get it's a repertoire of times. The actual task is just, okay, tapping into one of this repertoire. It's like say your living room has 20 different window sizes. And now you decide to look specifically through one window size and see the tree. Yeah, now, but then you can also decide. If you look through all windows then you see the whole force. But if you have only two windows, then you don't have a much of a selection. That's what your spontaneous activity from my point of view is good for me. And that repertoire I think is missing that dynamic what is called dynamic repertoire is missing the spontaneous dynamic repertoire is missing in the agent. Thank you for the illuminating. I want to give you an opportunity to expand on an idea. I know you have many things to say. And let me give you a bit of context. And of course, ask is AI conscious can I do this that a human brain can, and we end up comparing our consciousness with AI's mental abilities artificial mental abilities. However, we know that there is a big difference between the neural constitution and functioning that the brain has and the underlying architecture of the AI. The question would be, what's specific about the brain's properties, neural properties and functioning in time that create I'm scarce and a certain number of timescares, because then, as you, and I suppose you see you mentioned the thing about the, the energy constraints. There will be certain trade offs in devising agents, right, as number of units a volume of connectivity transmission speed of information and be that all key factors in constructing the things that timescales that the brain does. However, in AI it's a different story as the in CEO constitution enables different physical properties and evolution in time. Could you expand a bit on this difference in constructing timescales. That's, that's a good question, and a very difficult one, because indeed so energy is indeed required. And Philip pointed that out so when we discussed that yesterday in a meeting, you need energy and you need biology and you need physiology, and I think he's right. So that's the question. What on earth is the brain using all this energy the metabolic energy and the physiology. Yes, for timescales, but how does it construct the timescales. And that's where you see is coming in, and Samira who do the computational modeling in my group. We simply don't know as far as I know now we're trying to investigate. So yesterday I just saw data where we really show that the excitatory to excitatory connection, having huge impact on your timescales. Yeah, that's of course computational modeling you can do that you cannot do this in imaging. So indeed, probably, and that's a very good taken point here. I think that indeed the subcellular cellular level and how that yields timescales, because there must be let's say the brain has excitatory cells, pyramidal cells and inhibitory cells, interneurons. And that balance is key if you change that balance you probably change your timescales, but we are not really clear about that. How that relation goes from the excitation inhibition balance as it calls to the timescales. Then of course now as an engineer who tries okay not of tells us to construct some timescales now we put in some timescales. Then of course the question arises, do we really need this balance between excitatory and inhibitory neurons excitation and the ability to construct timescales. Is that a necessary feature to yield timescales or not because no doubt excitation inhibition balance is an old species not just a human brain thing. It's a basic feature that nature invented that I cannot really answer that question. But that's exactly the direction as I said yesterday we just raised when I saw those data. This timescale constituted they are an emergent phenomenon from the cellular level of course I would agree, Yassir would smile because I contradicted yesterday. But we really don't know. Yeah, but that could probably be the bridge between physiology and the timescale stuff. How, and that question will be key for implementing timescales and robot. I have one more question in the chat. One second we have two questions actually one of them is as human develops their consciousness adapts the timescales by noticing their breed, following the audio instruction how do you think I will develop their consciousness. Okay, can can you read again, or maybe I can I should be able to see it myself here because yes sure I can read it again. Okay, do you see it. Okay, as human develops a conscious adapted timescale by noting their breath. Is that what you mean, following the audio instruction how do you think I will develop the time scales. That's the question yeah, I think that's okay. Yeah, interesting point. So indeed, breathing. Of course, breathing has certain timescales and the breathing timescales connect with the brain. And actually, we're just doing. This is done by Josh go ahead in my group. A larger study on breathing and anxiety. And the main, what we seek so far in the pilot data is indeed that there is a certain information dynamic information shared between the breathing timescales and the brain timescales. And the conclusion here is really that the more the breathing and the body and the brain timescales are shared, the less anxiety. So maybe anxiety. That's sort of what we also see in the data is about the decent colonization of the timescales of the brain from the body from the breathing. And that's uncertainty, because you cannot predict the next breathing thing and anxiety is exactly that template uncertainty. You do not know what happens in the next time point and you get nervous and anxious. That sort of and the breathing therapy astoundingly helps very well in almost every single subject we have seen so far. So, indeed, here, what I said for the environment in the brain also applies to the body and the brain. Similar thing heart and brain. There's a lot of studies on that synchronization too. Thank you. And we have one last question here also mentioned please. Do you see it or I can read it for you. Yeah, I see I see one hand up here by ties a start and I see one question in the chat. Yeah, I can. What shall I start with. Okay, I read in the chat. Okay, so hello. There's quite a lot of literature and why humans perceive Tamil of a particular sound as a whole and how they dissect it into partial with their oral abilities, it obviously has to do with zooming in and out. And on why is it possible to perceive all of the different polyrhythm in an episode of music as an organic whole. So my question is how do the timescales align to the polyrhythm I think this is a very good question. And remember, so thank you for the question and my partner as a musician, the composer so as an under for use from Armenia. And so we often discuss these issues. Indeed so imagine music and polyrhythm what is polyrhythm you have different timescales and remember man. Remember my example with the thief and the police guy. Yeah, that there you have you need fine grained timescales to see where the fine grained changes in shorter time windows. But you also need larger timescales longer timescales to put some of that together. So in the polyrhythm, I expect that you probably exploit a lot of your repertoire of different timescales. And if you do not have the longer timescales, or they're very weak in you, you have probably difficulties of putting all these polyrhythm together into one organic whole, meaning you will perceive the piece in a different way. So ideally, I would like to do an experiment here. Basically, I look at the your polyrhythmic music piece. And I look at the different rhythm, and I can deconstruct that in terms of timescales. That's the first step, then I go to my subjects, investigate the brain in the resting state without listening to the polyrhythmic piece here. And then I look, whether the timescales of the piece, let's say, and you can measure that in terms of autocorrelation window in the piece itself, autocorrelation. This is more for the experts now. And then you can measure the same timescales with the same measure slash autocorrelation window in the brain of the individual subject. And ideally, if I'm right, I should basically from the timescales in my brain, I should predict how they will perceive in what gradient they perceive the different rhythms in your polyrhythmic piece, based on the gradients of the corresponding timescales in the brain, whether they perceive the faster rhythm of the polyrhythmic stronger relative to the slower, or they perceive the slower, stronger to the fast, because that should be related to corresponding timescales in the brain if you perceive the slower, rhythmic pieces of your polyrhythmic pieces stronger, then I expect that you have more stronger, longer timescales in the brain corresponding to faster and of course vice versa. But basically have a readout. And that's very important then also to construct my individual agent. Yeah, so then I will try to, if I have a readout of my individual subject timescales, and that predicts the perception of that subject. Let's say of the music piece remember my last example, then I can tailor the robot timescales accordingly for the therapeutic use. We have a question here just one second please. Hello professor thank you very much for the wonderful lecture. I'm a computer science student and I have maybe a naive question on the concept of timescales. So I believe for some time in the field of neuroscience there is this ongoing search for some kind of quality quantitative measure or benchmark of consciousness, like how much more conscious is a person asleep then a person under anesthesia or something like that. And then, like you illustrated in one of the slides, can the amount of how well the environment timescales and brain timescales are aligned. Can that serve as some kind of reliable qualitative measure. And can we come up with a similar measure for AI, given a specific architecture is implemented. Good question so if you want to include the timescales in your devices, I'm happy to work with you together so my idea would be really remember my example of the polyrhythmic music. So basically because I assume that you have a correspondence between the timescales of your environmental input. The timescales of your device slash agent slash brain and the timescales of your perception. Common currency. Yeah, so that's shared and the more you share the timescales of your brain with the environment the more you will perceive the timescales in the environment. So, for me the lacking piece because basically remember, that's why I want to have basically a read out from the timescales of the brain for your consciousness for your perception. And that is my job. And we're working on that but if I can do that, then I can come back to you and say built in those and those timescales in your device and then I can predict how that agent will be able to interact with my specific person and its perception. I hope that addresses the question. I'm really serious. I wouldn't know, because I'm not an engineer, I'm not a computer scientist, how to build in these timescales. I mean there are measures like auto correlation window, permutation entropy we currently just Andrea, which allowed you for my good in Italy. I think it's that we're also linking the excitation inhibition balance to timescales Samira is doing that in my group. So we try to address some of those gaps which I think could be relevant also for you as a computer scientist. Yeah. Thank you very much. Yeah, so if somebody of you guys are interested in that I'm very open. Yeah. I think we can also have her. Hello, Dr. North. Yeah, hello everyone. Yeah. Yes. Okay, thank you. I just I want, I would like to stop on our particular point, because, okay, we are speaking from biological and cellular human perspective about the consciousness. Without our other another type of consciousness which we can't detect we can't understand, or we can't assess. There is potential for artificial intelligence to go out of our understanding for consciousness and our rules, because even our, we are even with our understanding to our consciousness we are still in the beginning. What about another type, another state, another that another another method of processing and another rules of consciousness, which you can reach out of our abilities, our for assessment or understanding. Yeah, I mean you saw that I tried to go beyond the human spaces that I included some other spaces like plants and also other biological. biological brains. I would not bring sorry biological back from, we are just reflecting ourselves in different states or in different layers. But I mean what about totally different type of consciousness was different which out of our, we are not reflecting different consciousness. If you want to go beyond biology, you will go to the spiritual realm. It sounds like a metaphysical something which we can't assess with our tools like our mental or our consciousness. Yeah, I mean I mentioned for instance the example of plants, which we cannot perceive the it was an initial example of that, but of course you seem to be much more radical. So I think I cannot say anything at this point in time because we have no evidence for that. But I can't exclude it either because of our limitations. So and I think I clearly pointed out the limitations because of course, the dog has more fine grained time scales for smell than us. Perceives the world in a much more fine grained smelling away than we can. Yeah, and that's absolutely okay and I think it's a gradual point. I think the point, what is behind that and I agree with that, that we should not fall back into an anthropocentrism and saying and then also is what I call the argument of specialness that our consciousness and brain is special. And I think I pointed out that is not it's just basic dynamical principles of nature we just haven't understood that yet. And what is the metaphysics of today might be the science of tomorrow. So I definitely agree with that and I think we can learn as well. The question of rules, which you addressed. It's of course a difficult one when you try to go into computational modeling. You need rules, you know, mathematical formalisms. And I would argue that and you saw that we really go strongly rely on dynamic system theory and dynamic system theory doesn't come from neuroscience. It doesn't. It comes from biology. It comes even from physics and chemistry. Yeah, so, and we do a lot of try to really look into these fields into physics biology engineering. These are sort of the key fields where we look, and of course mathematics, where we get a lot of inspiration so we try really to conceive the human brain and the brain and Janet in this wider context. So I think that's extremely important also to really avoid an anthropocentrism and much of the consciousness debate and current neuroscience is clearly pre Copernican. I mean, you see in some books of mine I speak for the need of a Copernican turn or post Copernican view or Copernican revolution, if you want to say so. And that's clearly true and that's a problem in neuroscience. It's very pre Copernican, very central, very anthropocentric even if they go to the animals because they still take the humans as a reference. I agree with that. Yeah, this is, yeah, this is a point which we are just taking what we know as reference, but even as artificial intelligence can use physics which we didn't understand or we didn't know yet. Right, right. Definitely. Yesterday I was confronted with quantum computing. So maybe the quantum computing allows you a sick a kind of signal processing, which we cannot do with the current computers. I said wonderful. If it allows to process the signal in a more complex way, maybe we find something in the brain. Yeah, complexity background activity which is now considered noise and cancelled out and we cannot capture. I'm all for it. I'm way too much of a scientist and too curious also to close the doors for the opposite. Yeah, and that's the way I also see AI. Yeah, so I see that as a deeper layer to better understand the deeper layer for instance if we have a deep learning model. Yeah, we have a deep learning model. And now let's say we have a deep learning model with time scales and one deep learning model without time scales. Now, I fit in my raw data for my brain imaging data into these two models. Now, if my deep learning model with time scales predict my brain data better than the deep learning model without time scales. Yeah, bingo. Yeah, there you see that would give me a retrospective support. Okay, these time scales are important in my data because they're better predicted by the deep learning model with time scales. I would really like to see that implemented by computer scientists or engineers. So that's why I'm very open to collaboration. So that's the way I see it the same way where I see this template augmenting agent. Yeah, to highlight specific time scales and use it for the therapy in my psychology patients. Okay, Dr. Thank you very much.