#97: Predictive Processing & Ultimate Well-Being - Shamil Chandaria, PhD

0:00After spending a few years on the AI stuff, I started noticing that what we were doing in meditation deconstructing our experience, actually mapped onto the way computers and the brain are generating representations. The brain has to generate this internal representation and construct an idea of what's really going on. This whole experience, this whole conscious model is happening within us, But we think we are this avatar inside the model. Although it's a mathematical theory, it's, I would say, the closest thing that we have in neuroscience as a general theory of what's going on in the brain.

0:51What I am is this awareness. This whole experience is happening in the universe. That gives us a bridge to the fact that we are contacting reality. reality.

1:07Welcome to the Fitmind podcast where we explore the frontiers of human potential, mindfulness and neuroscience. I'm your new host Josh and I'll be filling in at least for now as my dear friend Liam dedicates more of his time to the practice. He'll still be super actively involved just a little bit more in the background. So today we've got an absolutely fascinating guest with us, Dr. Shamil Shandaria, a true polymath, a philosopher, scientist, long-term meditator and philanthropist who bridges the world between the worlds of contemplative practice and cutting edge neuroscience. With over 37 years of meditation experience, Dr.

1:46Shandaria's journey spans from transcendental meditation to non-dual practices, giving him a super profound experiential understanding of consciousness. Academically, his background is equally impressive with a PhD in mathematical modelling from the London School of Economics and a master's in philosophy from the University College London. Dr. Shandaria's work spans from computational neuroscience, machine learning and the philosophy of well-being. He's worked on artificial intelligence at Google's Deep Mind and now holds positions at Oxford University, Imperial College London and the University of London.

2:25In this episode we'll embark on a mind-bending journey into the nature of reality and consciousness. We'll explore the revolutionary Bayesian brain hypotheses and how our minds construct our perceived reality. We'll unpack the free energy principle and its profound implications for well-being and human flourishing. Dr. Shandaria will guide us through how meditation can help us deconstruct our experience, potentially leading to a radical shift in how we perceive both ourselves and the world around us. We'll also dive into the fascinating parallels between meditation and psychedelic experiences, examining how both can reshape our perception and sense of self.

3:10And we'll touch on cutting edge research about long term meditators that could revolutionize our understanding of consciousness. Now I should briefly mention that we had such an in depth conversation that we've decided to release our discussion about the jhanas profound meditative states as a separate video. So without further ado, let's dive into our conversation with Shamil Shandaria. Shamil, welcome to the podcast. Great to be here. With all this range, where do you currently feel that your main purpose lies? What is it that drives you at the minute and what gets you out of bed? What I'm really interested in is spiritual awakening.

3:52So that's really at the heart

3:58of thinking about human flourishing for me. Because we can think about human flourishing and making, you know, a kind of a good life and making your life look, you know, what you think your life is all about, making it look better or whatever. But that's not really getting to the nub of things. So for me, it's all about the awakening process and liberation. and I see that meditation is a very important part of that and psychedelics can be as well. So in simple terms, could you walk us through how the brain works, particularly in relation to the Bayesian brain model and predictive processing and then hopefully this will help us build a sort of a foundation of understanding how these practices can help us improve our well-being and overall happiness.

4:48Okay, so we are biological beings and we have this very interesting experience that we appear to be seeing the world. Outside my window are lots of lovely trees. And, you know, it looks like I am seeing the tree as it really is in itself. Now, how does that happen? if you've ever sort of considered the matter I mean what's coming you know most people have this idea that what they're seeing is what's coming through their eyes in fact most people we have this sort of naive realism idea that we are somehow behind our eyes and are looking out through windows in front of our face onto the world. But actually, that can't really be.

5:51And there's a very simple illustration, which is that if you put your hand out in front of you and look at your thumbnail, you see this very, very, very small portion of your visual field. And that's about all you can see in color and accurately. And yet you look around and there's this amazing, beautiful, colored, almost cartoonish definition, straight lines, a picture just appearing. How does that happen? Well, it's because the brain has to somehow generate this internal representation of what you're seeing. So basically, what any organism has to do is take their sensory data, their visual data, their auditory data, the somatosensory data of feeling and touch, body sensations, and construct an idea of what's really going on.

7:03Because, of course, the brain is just enclosed in this bony skull. And the only things going into the brain are time series data carried mostly through nerves. and somehow in this dark bony cavity we are generating some sort of conscious experience somehow so how does that happen well there is a there is a technically mathematically optimal way of taking data and trying to figure out what's going on and that's called bayesian inference And the way it works is that you are looking out the window, and you have lots of sensory data. And what you're trying to do is to figure out what's the probability that this is a tree given my sensory data.

8:06So that's called the posterior. That's what we're trying to figure out. And it turns out that you use Bayes' theorem for that. And Bayes' theorem, roughly speaking, is to calculate the probability that that's a tree given my sensory data. The way you do it is you have to kind of go backwards. You have to say that's equal to the probability that my sensory... What's the probability that I got this particular sensory data if it was indeed a tree? times the prior probability that there are trees in general, which is reasonably good. And then that whole thing is divided by a normalizing term, which is what's in general the likelihood, the probability that I get this kind of sensory data.

8:58Now, that Bayesian inference turns out to be what's called computationally explosive. So we can't actually calculate it, Because to calculate it, you would have to figure out every single hypothesis of where the data is coming from, including, you know, it's a tree, it's an alien spacecraft, it's a snake, you know, everything. And you can't do that. So both in AI and in biology, in evolution, we have to do something called approximate Bayesian inference. And here was the clever hack that was figured out by evolution. And that is exactly what we're doing in machine learning and AI. The hack is, OK, so let me guess at what's going on.

9:54Let me guess at what's going on. And, you know, you start off guessing as a baby and the guesses aren't that great. But as you go on, your guesses get better and better. And what you do is you sort of have this neural network, you know, coming in from the edges of your brain where the object nerve, the visual processing, early visual processing is happening, for example. And you feed forward and say, oh, looks like it's a tree. Okay. this is the basic idea. But in fact, what's really going on is that you've got this idea that it could be a tree, but now the brain says, okay, let's imagine, let's actually imagine that it is a tree.

10:41Now the brain fills in the entire picture and guesses exactly what's going on using its prior probabilities using its best estimates. And then to make sure that we're not just making it up, what we then do is compare the little bit of sensory data that I got, you know, this ridiculously small amount of the visual field with few isochords, and compare that sensory data with the output of your simulation of this guess model, right? And you compare your simulation with this few pieces of sensory data. And then what you do is you adjust your simulation, your internal mental model of what's going on.

11:26You adjust the parameters, the dials and knobs and switches of that simulation to minimize the prediction error, to minimize the difference between the output of your generative model, that's the simulation, the generative model, and your sensory data. And that's called the prediction error. and as you minimize this what you end up with is a generative model an internal simulation if you like a hallucination but it's a hallucination which is a special hallucination which is it's a hallucination which is trying to minimize the prediction error with the actual sensory data in other words it's a controlled hallucination okay so and then we get this very important hypothesis which i call the neuro phenomenological hypothesis big word but very important in the spada neuro phenomenological basically being that the that hypothesis is that our conscious experience is in fact our internal model, or some part of our internal model.

12:46Some parts we may not be conscious of, but we are experiencing our internal model. In fact, we are not experiencing the world as it is in itself. That's a philosophical conundrum. We've never actually seen the world as in itself. We are living in our own internal simulation, like a video game, if you think about it, right? And in this internal simulation, we have to represent ourselves as well. So this representation is in fact an avatar, if you like, within my own internal simulation, within my own internal genital model. And the idea of me is all, and my narrative self and my social self and my bodily feelings, these are all combined to make sense of an internal model.

13:49it's very important to have an internal what's called a phenomenal phenomenal self model means a kind of internal conscious self model so it's very important to have that because from an evolutionary perspective you know if there's a tiger running towards me there I need to know where I am and that the tiger is over there and then I can jump up that tree over there so it's a good data structure within my generative model to make sure I have a model of the agent itself. And funnily enough, even the idea that we are acting is a part of this agent model, an acting model. And so there are other parts of this generative model, which is that, in fact, as another data structure, a philosopher called Immanuel Kant, a couple of hundred years ago, actually figured out that we also have to have the concept of space and time and causality to organize our sensory data.

15:06So we're bringing all of this to our internal genital model. So we are simulating this world with data structures, or they're called hyperpriors, you know, these top-level beliefs, using space, time, causality, an internal phenomenal genital model. now so that's a very strange situation and we are therefore there's a couple of things to note there the first thing is that we've got a kind of a weird um sort of error or an illusion going on which is that, in fact, this whole experience, this whole conscious model is happening within us. But we think we are this avatar inside the model. So we've taken ourselves to be this little mini-me inside of our awareness.

16:20That's the first thing that is kind of strange. And the second thing that's kind of strange is that we believe it's all real. This is the real me. That's the real tree. This is the real bottle, right? It's really a bottle. It's not a kind of a representation of a bottle or, you know, but it's real. And the philosophical word that we use for real is that we have reified this object. We believe it's the real object. Now, what happens when we start to meditate is that, if I can take this step by step through meditation and then come to a... What happens when we start to meditate is that we come to see, first of all, that, ha well first of all we come to see that i'm not my thoughts because we can come to sort of observe our thoughts arising and body sensations arising and emotions arising and if we're if we're observing them arise that you know they're not us and we also um come to notice that those are kind of as we really tune into them we can come to notice that those are made of fine vibratory sensations we can start to deconstruct them we can start to see that these body sensations and these thoughts are fleeting and and and impermanent so not only are they not us, but they're fleeting and impermanent and shimmering and diaphanous and almost like fog or have a dreamlike quality.

18:19And then we start to, you know, that's, there's a technical Buddhist word called emptiness, a very good word, actually. Some people don't like it. I actually like it. I think that it's a wonderful word. And we come to see that all the phenomena are empty, they're impermanent, they're diaphanous, they're almost translucent and transparent. What we come to see in meditation is that, in fact, what we take to be real is not really real. This is, in neuroscience and philosophical terms, we tend to call that we have dereified the objects, so they no longer are real. And of course, seeing these objects as empty or reified is very important, but the most shockingly important part of the meditation process is to see ourselves, all the various components of ourselves, of the phenomenal self-model, as empty.

19:26that they are, they reified. They're also not real. They're just constructions. Doesn't mean they're not there. Doesn't mean that they disappear or that thoughts disappear or their body sensations or emotions disappear. It's just that they become empty. And the whole notion of being identified with them starts breaking down. And at that stage, something pretty amazing can happen, which is that we can come to see that all of this experience, all this arising, these thoughts, these body sensations, this sound, this visual field, are all arising within a space of awareness or a space of perception, and that it's all constructed, and that who we take ourselves to be is kind of empty as well.

20:43and then we can wake up like becoming lucid within a dream, that's an analogy, we can become lucid within our own generative model, within our own internal simulation. We can actually become lucid and wake up and see it as a construction, see it as empty, and start this slow process of awakening and in fact ultimately becoming the master of the genital model so that's that's that's the awakening dimension we can come to the liberation dimension later but so any questions wonderful wonderful masterfully described yeah yeah beautiful you touched on it briefly i'd love to talk a little bit about the history and if you could describe the free energy principle, why its significance in neuroscience, and how the brain works to minimize free energy and its correlation with well-being and happiness.

21:52Right. Okay, good. So good, good spot, because we haven't yet talked about suffering, and we haven't talked about well-being. And so let me talk about this buzzword, the free energy principle. You know, I sort of said that we are seeking to minimize our prediction error. We're seeking to have an internal simulation, a generative model, that minimizes the prediction error with the sensory data. Now, there's also another thing that it minimizes. It also minimizes the difference that the you know whatever i'm seeing i'm also trying to minimize uh the the the distance between um uh my priors and what i'm seeing in other words i'm trying to have a picture of the world that not only minimizes the prediction area but also minimize is what I would normally expect.

23:00So if I saw this tree as a spacecraft, that would have a high free energy because generally, one's prior of seeing a spacecraft would be quite low. And so interpreting it as a spacecraft would be very surprising, in a sense. So that's minimizing free energy, minimizing prediction error minimizing deviations from our price now before I come on to action and then ultimately happiness and suffering it's probably worth talking about the fact that this model is actually built in layers and this is very very important I mean I do think it's extremely important that your listeners kind of understand that how is this magic trick pulled off you know it's too unbelievable but it turns out that the magic trick is pulled off because it's done in layers and layers and layers and layers and each layer runs on beliefs from the layer above and eventually just through smoke and mirrors and layers and layers and layers, we end up with this real belief.

24:24And the way the layers work is that the early layers, the low-level layers, are like low-level perception, like edges and corners and lines and various things like that. And then you go up and you have mid-level, like geometrical shapes, and then you have higher levels, like faces and motion. And then ultimately you get up to concepts and and and and language right and most of the time we spend our time in these top level layers we're just totally caught up in our thoughts and our concepts and all that um and of course one of the things meditation helps us to do is go down and and really start um putting more in the jargon it's called more precision weighting on these lower level layers more attention on these lower level layers.

25:20And that takes you out of your thoughts. And that's part of the process of getting freed up. Now, so that's the hierarchical generative model. So the free energy principle says nature has to minimize free energy. The brain is trying to minimize free energy. Free energy is, roughly speaking, the prediction errors and also deviation from your priors. And these priors are coming from these high-level layers, and the priors of the mid-level layers are coming from the layer above, and the priors of the low-level layers are coming from the layer above that. Okay. And there may be sort of half a dozen layers or something like that.

26:03now where does action and suffering and happiness all fit into this because so far we just talked about perception and but ultimately spiritual teachers like the buddha talked about the alleviation from suffering right so what do we have in mind there well um this is the really amazing thing about active inference and the free energy principle. So active inference is this idea that action, the active means the action part, action is combined with the inference, the perception part of it. What happens is that we're trying to minimize the deviation from the prize right so we're trying to to perceive the world in a way that's close to the prize now if you bring action into it what you then do is you say you're the reward states that you want um like getting an apple or or having a drink of water um those reward states are your prize You expect to get the water.

27:23You're thirsty. You expect to get the water. And in fact, right now I'm thirsty and I expect to get the water. Right. So how do I minimize the deviation from like I'm not getting the water, but I expect to get it. so I want to minimize the deviation from my prior having the water and the situation right now, I want to minimize the free energy, and so the way I do it is by taking an action to drink water and so what's happening is that in this framework minimizing free energy then entails getting what we want and not getting what we don't want okay so i don't want danger i don't want pain um i want pleasure i want uh food and and drink um and so in this amazing synthesis that has been developed in the last uh decade or or so what we now have is a unified theory of perception and action.

28:46And this is an amazing thing. I mean, one of the main architects is someone called Carl Friston. And although it's a mathematical theory, it's, I would say, the closest thing that we have in neuroscience as a theoretical model of a general theory of what's going on in the brain. So, I mean, this is very exciting stuff. And, you know, because it's so technical, it hasn't yet filtered into the public domain. But it's very important for people to understand. Now, let's bring it back to meditation, well-being and awakening.

29:33so this is very interesting so now we have this whole action framework and there's a bit of a problem with this action framework because it's designed to make us successful from an evolutionary perspective right we run away from the tiger we get the apple we get the water and the environment is constantly changing so we're trying to minimize free energy okay so if we imagine that there are many environmental situations and we're a little ball on this curve and we're trying to roll down the landscape so this you know very large landscape and we're constantly trying to roll the ball get lower and lower in this free energy landscape which is closer to what we want, closer to homeostasis, right?

30:25So having enough energy, having enough hydration, being safe, etc. So homeostasis, or in general, something called allostasis, which is a kind of a dynamic version of homeostasis, is minimizing free energy. So we're in a good physiological state, we're happy, we have what we want, we don't get what we don't want. and signals like pleasure and pain give us rewards you know good feeling when we get closer to the homeostasis when we actually start minimizing free energy now the world constantly buffets us around it's a it's kind of an impermanent world you know oh the bus is late and i'm gonna like be late for work like it's raining and i haven't got my umbrella and I'm gonna get soaked and I got a meeting and whatever so what happens is that the ball gets constantly moved around okay because of the environment and so what you notice is that you're constantly in a state of not being at the minimum of free energy you're not in a kind of a homeostatic state you're not in a state of where you want to be you're constantly up and down this curve, I'm doing it as a two-dimensional curve, but I mean, this is massively high-dimensional space.

31:53You're constantly moving away, and you're at a disequilibrium. You're not at the lowest point in the valley. You're at a disequilibrium constantly. And when you're away from that minimum, it's unsatisfactory. It's not nice. I don't like that. But the problem is the world's constantly buffeting you around. And this is the way, this is the sort of carrot and stick way that evolution gets us to maintain ourselves in a homeostatic situation in a way that we want. Of course, I'm talking about physiology, but eventually it turns out to be all this social stuff and status games and all of that sort of wraps, all these sort of, these social dimensions also wrap into this framework.

32:43And so we basically have rediscovered the first noble truth of the Buddha, which was that life is full of suffering. And so because we're constantly being buffeted around and not at the local equilibrium, local minimum. so that is a fundamental issue about human life and that many great sages have figured out but now there is something that we can learn from meditation which is that everything was construction and actually not only was all of perception also construction but the pleasures and pains and and dissatisfaction is also a part of that construction and it can also be day reified. So I have a pain in my knee and this is unacceptable I don't want to feel a pain in my knee I like I must move I must do something to get rid of it but as I tune into that pain in my knee and And I noticed that it's just this buzzing, burbling, fleeting sensations, impermanent.

34:22And I also noticed that coupled into that tangled complex is also this feeling that there is these body sensations. And then separately, there's this, I don't want that. And I can notice that I don't want that is also a kind of construction. and when you notice that it's also derefied it loses its potency it loses its stickiness this whole idea that like i don't want it actually what really makes the unsuffering is this idea that like it's there and i don't want it but if i relax around the i don't want it and And if that scene is as empty, then the suffering goes away. The whole nature of that stickiness and pushing and pulling of those thoughts starts to grip less.

35:31You can see it still arising, but it actually no longer needs to be connected to action. And that's almost, I mean, this is a very subtle thing, but it's almost because we develop a prior that these things aren't derefied, that they are impermanent. and therefore in some sense we have actually hacked the free energy principle because we already said that oh i expect it to be uncertain so therefore at a higher level that uncertainty has been now explained away that those deviations have been explained away so this is a very subtle process and and what we end up doing is ending we end up having a sense of equanimity a sense of everything is okay even when the weather isn't great or there's stuff going on Oh, look, there's this kind of pain arising and there's this discomfort here.

36:44Oh, look, I just missed the bus. Oh, gosh. Well, I better walk a bit faster. You know, I mean, it doesn't mean that you become a blob or like just not motivated. You know, one doesn't lose data, but we get more room between stimulus and response. We start getting less buffeted around. So this is this whole process of starting to grip less because things are more empty. And so this is a wonderful part of the benefits of meditation and it's part of the awakening journey. And it's therefore deeply linked to human well-being. Wonderful. Yeah, absolutely. It does. And it just reminds me, I'll often have, say, there'll be like a general feeling of anxiety or some sort of worry.

37:42And then if I look at that and I just relax and I open to it, then quite often it will actually become excitement or what feels actually quite pleasurable. not necessarily labelling it as excitement, but it's not what it was. Immediately, it's not what it was, at least what I thought it was. I think it's Mingya Rinpoche that talks about the sort of backing off and just noticing there was just a subtle pushing down, even that I didn't notice. It's very subtle, didn't realise that I was just gently pushing down this feeling of what becomes excitement which previously felt like anxiety and he actually he he grew up in obviously a culture of meditation he had a great a very famous dad he's actually some one of sam harris's teachers and he grew up with panic attacks he talks about I think from when he was around seven until 11, he was like seemingly this prodigy.

38:52He had this famous father who was an incredible meditation teacher. And there was an expectation that he would follow in his footsteps. And then he somehow managed to convince his mom to go on a three-year retreat. Actually, he got his mom to convince his dad to put him on a three-year retreat. I think he was the youngest person to go on three or four years, actually, to become a llama. and yeah he eventually spoke to his dad and it was becoming so intense this feeling of panic he ended up locking himself in a room and just sat with the panic he talks about making friends with the panic and inviting it in for tea and this is something that his dad had told him to do so yeah i wonder how how you would from your perspective look at that the the sort of making friends with the panic.

39:42He actually says after three days, he was very sad because his friend panic doesn't come around anymore. So this is, of course, a very, very important part of the meditation journey. I think that, that, as you say, that, especially in the non-dual tradition, it is, you know, the idea is actually to fully experience these emotions and body sensations. And if we, in general, this is the sort of slightly more advanced version, but if in general we can experience them with a state of acceptance, loving acceptance in a state of awareness when we are in a state of wide awareness sort of a vast awareness and an awake awareness ultimately and these emotions which are as you just described they're ultimately just energies and we interpret them as, for example, anxiety provoking, but in the end they can just be interpreted as energies and that's why you sort of intimated that those energies can be reinterpreted as excitement or actually they can just be empty.