 good luck. Hello, I'm ready. Yes, we are in. It is January 23 2023. We're here in active inference guest stream number 34.1 with Avil within Karlu. We're going to have a one hour presentation, physics of creation, can the free energy principle ground the construction physical state spaces, followed by about an hour discussion. So if you're watching live, please feel free to add questions in the live chat and we'll also have some guests. So thank you, Avil for joining and off to you for the presentation. Hello. So this is a complete presentation. I will first introduce myself and the work. I am a regular Kalu. I have a program in physics and working with science in the version of Andy Clark. This work is essentially speaking the product of Kairos research, which is laboratory, which myself, Alexi Rosensky, who is present here today, and other people are funded to investigate issues of relevance to political through political evolution and community science. And the topic will be on the physics of creation. So creation is a typical term, let us say, it will be contextualized in presentation, but let us first define that it corresponds to the construction of physical possibilities, physical state spaces to talk in a more technical language. And the question that motivated its research is whether or not the free energy principle, a piece of math that's emerging in the cognitive science, which we will get to very sooner, is capable of representing this process of creation. So first, I have tried the first run of this without getting into the technicals. It did not work. So I will address the technicals first. We will see some concepts in physics, what they do, how they work, how they relate to what people typically do in physics. And first, I'd like to take a look at how physics explain. So physics explain with laws. They think that in nature you have laws, and those laws are basically formal statements that constrain the possible states of a system. And typically, those are entailing laws, that's from which you can derive logically what the system will do. An example of natural entailing law is Maxwell's equation of electromagnetic. It is a different Maxwell value, the one that is here. So what you see is four bits of vectorial calculus that will tell you how two fields that are collectively the electromagnetic field evolve with regard to another. So this is a formal constraint on what E and B can do, like the quality express constraints by definition. And those can be used to derive basically fully E and B given boundary conditions, which is how you do things in electromagnetism. But what is of reference here is that any of those law presupposes basically that can define E as an actual field. So it presupposes what is called a set space, which is mathematical space that represents the set of possible states. Before I state delta E and delta B is equal to this and this, we have to state that E exists and E can take such and such states. In that case it is a vectorial space in three dimensions. And this set space basically it has by construction to capture any difference that make a difference about the system that is targeted for any change in E to be basically computable by mastery. It has to be represented in the set space. So by construction I want the set space to categorify every variable of interest in my system given the scope and measurement capabilities of my theory. You have underlying notion that is basically implicitly underlying the notion of flows and set space that is the notion of symmetry. I cannot get into the very detail here but I will try to be as complete as possible. We typically understand physical theories outside specifically relativity theory and some weird quantum stuff to derive from symmetries. Symmetries are, I will say it out loud, algebraic group of transformation over the set space that do not affect the system's dynamics as granted as it's Laplacian. You don't have to understand what I just said. What you have to understand is that there are ways in which I can transform the system which does not change what it is. An example in symmetries, the sailing of the Lodfler-Moscow in its form, you see that you have basically the same pattern that is repeated again and again and again gradually. So if I make a small rotation in either direction it will not change a thing and if I repeat it and repeat it and repeat it and repeat it and repeat it and repeat it, I will still change my system like formally I will apply transformation but it does not change basically what observable properties the system have. And this is what a symmetry is. It is a set of transformation that is generated by a subset of smaller transformation and that does not change what the system observably is. And because of this, symmetries have a system that basically constrain what difference can make or difference some of the system. So implicitly when I formulate law or set space, I make a statement about the underlying symmetries. I make the statement that basically everything that is not in the system that is not represented as part of the system is something that is symmetric that the system is symmetric with regard to that on which any transformation will not affect the system trajectory at least not in a way that is relevant to my theory. It works for physics because it looks like in physics we have a few symmetries that are everywhere verified that are fundamental in some sense. It does not work so well in life in mind because we don't have that. To be specific living systems have a structure. I do not think it should be controversial and they have constitutive symmetries. They have changes that will not affect them. If I switch all of the carbon atoms in my body I scramble them between each other while keeping the relation to rest it did not change a thing. So it is very trivial in sense of symmetry but basically every structure that is me entails symmetries, entails transformations that are possible that do not affect the structure and other that will affect it. The structure it is not a given like symmetries are considered to be in physics. It is actively reconstructed by the activity of me. If I cut myself, if I fall, if my bone breaks they will regrow basically. You will have superficial tissue that will occur and that will make it okay again. It will change what structure is my body. Most importantly I was not born that way like big and with a bird and with 80 kilos of meat I grew and I grew from one single cell and this single cell just created an atomic structure apparently ex nihilo from basically a genetic code in the context and not only that but life in and of itself evolves in a way that builds structures that build symmetries and all of this means that you can't set a state space that is the state space of life in general of any living system. The symmetries that in the life system they change through time and they change by the own activity of this system. So any living system construct state spaces and to account for it we need a theory of state space construction. This is the motivation of the physics of creation line of work and this is basically the I did not from it as a question. It is the question that we the but you need to keep in mind in this discussion. So Daniel do we have any question yet? Anyone? Okay no technical issue that I should know about. Okay so what we will do is first look at basically what we say of creation in life in mind at a conceptual level what is the history of the concept and what Darwin to do. Then we look into the formal specifics of how we try to recreate to represent basically cognition as inference and what limits these attempts have and finally we will take the reflection further and see whether we can and how far we can represent physical creation, the construction of physical reality with the tentative formalism we have framed here. So let us look at creation in life in mind. Historically the term of creation it refers to autopoiesis that is that means literally self-creation and that is not done by an activist what became an activist so mainly the Maturana and Varela to ground the study of cognition the study of biology and the way well the link system work. They framed autopoetic theory as basically the idea that cognition is grounded in a way that link system construct itself like their metabolism allow them to reconstruct a structure and the sector does not have to be conserved by metabolism so you have self-creation and this is basically the concept that try to articulate to ground cognitive science and later Varela and colleagues try to basically concretize this by working on how the cognition was enacted which is it is implemented in the in the body and in the concrete activity of living systems and this basically was very determined in the sort of cognitive sense and it framed the RE approach so the extension of cognition beyond the brain let us say very very tentatively. The idea that they frame is that you have you can see the diagram that is taken from DiPaolo's work and that tries to explain the relation between self-constitution so let us say it is a weaker later more specified notion of apoptopiasis and agency. The idea is that you have self-constitution that is the fact the link system build themselves and it is basically afforded by the way link system couple with environments and agencies later higher order of regulation of this disability to define basically the terms of your coupling with the environments and the idea it generates is that living systems there are systems that have structural identity that they actively generate and sustain under precarious circumstances there is no benevolent universe that helps life that help life be life. Life has basically to generate environments where it is life and doing so it generates and it sustains what it is its identity and all of this is necessary for sense-making so let us say cognition to be something to occur and this is in my opinion and in to some extent in DiPaolo's own opinion somewhat wordy it is a lot of concepts a lot of Greek words and people have tried to frame this in a way that is more mathematical and formal and the question it generates is how to formalize this self-creative property of biological organization and cognition and in a way that is basically grounded in a material information flow in things that are concrete physical structure. We have a pretty good answer to that actually we have a pretty good work that made a point that you have related this somewhat high level concept to basic physical processes this work is the work of Moreno, Moschio and Montevihil mainly who work in the institute of answering the paybask in Spain and the idea is that living systems are constituted by a closure of constraints so the constraint is physical thing that shape how other physical thing work very simple if I have air in a box that is closed the air can't leave the box it is a constraint okay and the property that individuates living things is the fact that they are they are constituted by a set of constraints that canalize metabolism in a way that reconstruct the same way the same system of constraints and so from there you have self-constitution you have also structural let us say notion of biology that does not depend on the matter it is made of but on the way it is sustained that is I think pretty good if you want to have a physicalist notion of life and we have a very basic ontology in which constraints they shape metabolism and shape information flow as in the way living system react to things and that grounds the autonomy of living things the fact that they can recreate themselves basically but nothing gives a clear link between this and cognitive meaning or anything we could recognize as cognition so this is what we will try to articulate next so how do living systems as we just defined can create meaning the notion I would like to import is the notion of active inference which I will not get into the mathematical detail here but the idea is that the way the flow the dynamical flow literally that is entailed by an agent and its environment it individuates states that are meaningful basically an agent will predict in a relatively formal sense sensations that it will not perceive perceive but undergo and actions it will undertake and these predictions are basically what's underlie meaning for the agent it underlies what states will just swap into shape it's an action and be meaningful to them this can be framed by in a pretty simple or poetic move by stating that belgium cognitive systems they do not simply infer what is they bring about their own reality by actively inferring what is meaningful to them by actively looking for states that they think are really done for their own structure and then you start to have stronger relation between what a physical system of constraint is and how it individuates states that are meaningful but you still don't have a formal theory of basically how this metabolic flow builds learning or creation like builds a new state of possible a new space sorry of possible states that is not the one it began with I had I have a description of structure that is the set of constraints and this produces basically an activity of the system of flow of a destructor and this flow it somehow constraints or shape destructor at the stem takes a time t plus one which itself produce a new function a new flow etc etc etc so it is this dynamic of I don't want to call that self-referential unfolding whatever the the way a system of constraints produce a flow that reproduces some constraints that we want to formalize and the question we could reframe as such and we could reframe as how does cognition the ability to understand integrates information about the world it enacts within biological organization because cognition is as a kind of flow of political organization at least is entailed by it everything that I see is something that I but to say bring about that I look at because of the biological constraints I am and therefore the question of creation can be reframed as all the meaning I project onto the world is brought back to shape my my own structure and this is what we will try to frame mathematically in the framework of active inference Daniel any questions so the attempt that we make here is to see if we can how mathematically we can represent creation as a process of inference which is likely the most minimal highly recognizable way I could frame cognition as something that is distinct from simply its material in a presentation let us look at a piece of math that is called the furniture principle and that is the formal motivation or grounding for active inference it will be the hard part for people who do not do math or people who are not already familiar with the furniture principle I will work with slowly it is a piece of math that frames statistical inference from the chemical system theory chemical system theory is basically the idea that you have states and they move as a function of each other and what the FEP does is to prove that if I have something that is called Markov Blanket so a boundary that any information must travel through for two subsystems to communicate practically then you have a synchronization that emerge between states that are outside this boundary and states that are inside this boundary and and not only this but if I measure the basically every state that is on the inside of the boundary entails belief entails equal distribution over the external states and the flow of internal states is such that it minimizes the international distance between the actual external states well what I can observe and what I believe to be the case where I of course is the internal systems so that was abstract the basic informal meaning of this is that if I have a chemical system and I have a boundary that acts to mediate information exchange between two supports then both supports can basically represent each other and hold belief in each other in a way that is constructed by the systematic synchronization across the boundary it is quite abstract and quite minimal but it is found proof that we can have representation we can have cognitive meaning from a very very minimal set of hypothesis which is basically a consistent causal structure and it is a pretty good case that's a basic chemical system for example this type of constraints constraint will be more general but this kind of very basic non-committive things can entail permission it has been argued that it does underlie agency and creativity and similar concepts that would entail in my opinion creation as a prerequisite so we'll try to see whether the FEP actually can represent creation as the ability for agents to individuate states that are spinning to them and the thing is that the existence of Markov Blanket is demonstrated for any complex stochastic chemical system so chemical system that are random and this transmission relies as basically something that moves in function of itself plus random noise it is a very general formulation arguably anything that exists can be framed as such so we should be optimistic for the ability of this formalism to represent the cognitive the production of cognitive meaning my basic biological system and therefore creation it is not the case because when I frame something as x of t equal something something well I have to pre-defined x I have to define space of states that is x and I have defined creation as the ability to construct basically physical possibilities so by construction if you frame something within the chemical system theory you have to individuate states a priori what you have written down is not creation can be a ground creation can be interesting in the study of creation it is not creation is not the construction of physical spaces and it is likely not the construction of cognitive meaning at least in the sense that an activist and autopoietic theory people mean it it just in the right what become measurable for a given subsystem within the stochastic chemical system which is a lot but is not again creation let us think about how we should go we should think of cognitive creation um there will be more math here I'm afraid but if a system is to observe another system it will have to lose information about its broader environment it is uh pretty easy well it is a something can derive from first principle uh because basically um if you frame it as a quantum system uh observation will entail entanglement by definition I observe a system and it becomes uh something that has in the states because I become entangled with it so its states become basically synchronized with my states but I cannot synchronize with everything all the time so if I act so as to have more knowledge about the object I act so as to have less knowledge about my environment so there is no basically lost less observation it is something that can be derived in the quantum setting that can be reframed in other way in classical settings because of the way observation will either induce heatings information loss or cause myself to forget information you can pretty much derive it in a variety of case it is likely true in other cases and it means that observation as it is constrained by my own flow and the way I couple with the environment it does not simply entail the observation of pre-existing well-individuated physical states it brings about well-individuated well-defined observations for me to observe so the way how to say given this it is hard maybe not possible to frame a theory of observation from a mathematical theory that entails a theory well-individuated physical states if I frame such a theory of observation or measurement of cognitive meaning it will not be general and I pretty much have to work creation on the on the very fundamental scales at which my theory operates and this is something that is arguably done by reframing of the energy principle in for generic quantum systems by mainly Chris fields the idea is to use category theoretic equivalence between lots of things I will definitely not get into itself that's to reframe the synchronization dynamic that's in view that builds cognitive meaning according to the French principle as a series of binary measurements are equivalently symmetry breaking operators so the way a chemical system synchronize with the measurements under the FEP is reframed here as basically a series of binary questions that I ask my environment and that enforce my interaction with it in such a way that the answers to my questions become physical facts which you can derive from a rather general quantum information theory background but that gives basically no structure whatsoever to the measurement nothing in this theory says what kind of system can ask what kind of questions and this means that okay this drives creation maybe drives individuation of physical possibilities physical observables but it does not constrain at all what this creation is so it doesn't explain much and it doesn't help much to understand how a living system recreates itself so we still have somehow to recruit the cognitive and chemical formulation of the FEP that I discussed very tentatively earlier so let us take things from there I have turned during this formal discussion from the notion that we have to represent creation in the living and cognitive world to the notion that somehow it is necessary present creation of physical reality like the underlying states of physics this third part with this third part will serve mainly to discuss this claim and what it means and what we should do with it particularly first I want to draw basically that the claim that there is no a priori physical reality is a pretty basic claim it should not be treated as exceptional it is necessary property of a coherent physics it is necessary property actually of the formulation or description of a world that is beyond what we call metaphysics let us be more specific in the argumentation I call this argument diagram from ontological consistency an ontology that is naturally stick that serves to present nature if it is consistent cannot call onto excellent object that would not be natural to explain the properties of its own objects for example if I tell you this thing is black because it reflects the light in such and such way and I can also explain the properties of light it is self-consistent if I tell you this thing is black because God will need and I cannot explain you what God is physically it is not self-consistent there is no other property that is framed by Alexi Rosensky who is present here as the property of self-functional consistency that is that the ontology should basically include itself the discourse we have on physics is unless we have very radical position about the relation between human and the other world it is something that exists in nature and that is something that we need to account as an element of nature this entails two things one thing actually it entails that scientific presentations the models and theory we have and the property those models and theory have they must be accounted for as natural entities in their own rights and there they must be at least in principle explainables by the tools of physics and science and this entails in turn that I cannot have such a thing as a law a symmetry or a set space that pre-exist physics any statement I make about physics must itself be explained by physical processes I will elaborate on what it means the view of physicists is typically that we have theories and those theories are representation of the world that just so happen to account for fundamental structure of the world for example we have the Maxwell law of electromagnetics electromagnetism and those laws are representations of a force that exists and that is generated by SU1 I think by a symmetry in the Hamiltonian sorry a symmetry in the quantum property of systems but this symmetry pre-exists basically any attempt to represent and any physical process it is a fundamental property what I say is that you cannot have fundamental property you cannot have something that predates physics and that cannot ultimately be explained and if you admit such a thing your ontology is not consistent because you have to admit some things that are beyond that predate nature to explain things in nature which is the opposite of what self-consistency is so if I want to frame a world where you do not have explanation that call on two things that are beyond physics you have to have laws symmetries and spaces that are constructed by physical processes and it is not a pragmatic view to have at least epistemologically it is necessary let us discuss what it means concretely we have cognitive systems dot cognitive systems they perceive and this perception brings about cognitively relevant states which the cognitive agent experience has real we have physical observers these physical observers bring about what we call quantum states that are well-individuated states that are visible and those states constitute physical reality at least as any given observer can experience it what I say here is that this is not epistemic notions the quantum states and the cognitively relevant states that perception and the cognitively relevant states under perception by agents they are real and they are individuated as real by their observation and if we frame it properly the qfep at least what we could do with the qfep it represents how proto-mines at least particles agents that have properties that we classically understand as mindful they individuates physical observable which means that they create spaces which means that this is a theory of creation not simply a theory of how access reality but a theory of how we get reality in the first place and to do anything like grounding properly this claim it will take a lot more details and I will try to work by showing what the such a theory would do because obviously I do not have such a theory that is framed formally from start to finish out of no and if I did I will not present it now let us look at the way constraints basically creates the relative experience so this is a diagram of how as humans in the world the social world basically understand things like constraints over what we do and can do and what it entails I understand that if I talk to you in proper English you will understand if I talk to you in French or Swedish you will not understand understand that if I undress it will be inappropriate and if I hold myself in a certain way it will lead you to believe me more or less understand that if I start speaking very loud it will disturb people and they will act for me to stop talking very loud I understand very basic facts about how this world work and I understand how this constrain the action I can take and the consequences of that action what experience is basically a sociocultural escape which exists independently from me and that I act within and what I claim is that this sociocultural escape evolves in a way that is creative that is basically abuse new norms exist that did not exist 10 years ago and that is even more true for one millennium ago or before humans were a thing and the way I can represent how I don't try to understand this social landscape and possibly fail our erode norms and the landscape is such way it is equivalent to a presentation of the landscape itself and the constraints that constitute this landscape and the way the flow of landscape reconfigures the landscape so what I did here is to basically frame a theory of creative evolution this of a set of constraints well of a sociocultural landscape on the way individual agents act understand and act within this landscape I have sense in which constant observation and cognition about physical here a social reality but still physical sense that it exists in a physical world how this landscape can be created by something that is cognitive and this basic ontology if I am correct here is something we can automatize I hope so and that we can frame as a basic theory of well the individuation of new physical possibilities creation i.e. not here create to this graph is given to me and what is of importance here is the duality between two representation two perspective the perspective of agents that basically enact sociocultural constraints that the experience has a physical fact as a reality that cannot go beyond or escape or maybe affect in brute force that's possibility but not escape and the perspective of a set of constraints that endlessly unfolds in a way that is creative and that is a formal intuition I plan to build on and I can say something a bit more specific about how it works by analogy to the fantasy principle and the holographic principle basically a creation operator in my opinion I could argue for that I will not do this in detail right now is something that goes for a space of it's not actually a space of physical possibilities it's a space of potentialities it does not have possible states and impossible states it is just the statement that something is possible it is very close to what the Greeks call chaos so you have such a space of unindivided possibilities and from this space unfolds a pair of an individual observable which is called E for environment and an agent that is capable to observe this observable and the agent is also capable to maintain a boundary between itself and the observable or basically it's available to stay existing as observable to be observed and from this I think that some mathematical operator that has this basic flow is capable to describe the construction of physical set spaces as they are the unfolding strictly strictly no metaphorically a physical structure as basically the yes the emergence the co-construction the individuation as you will of well individuated observable states and observer that is capable to see the state and bias in this state maintaining whatever worked as an individuated space of possibilities before as a well individuated boundary between itself and the system let us think for example of a piece of paper I an agent can basically write things on it and by writing things on it I create new physical observable which is the content of what I've written and the fact I write on it and or read it makes a piece of paper not a piece of paper but a boundary between myself and a semiotic space where the writing occurs this is the basic dynamic that's must be framed for us to understand cognitive meaning and that if we framed in such way could work to describe the construction of state spaces X Nilo because literally you do not need physical possibilities to be a thing or be to exist it is the point of this presentation you don't need a space of possibilities to pre-exist the individuation of the agent and environment this is why it is relevant and this is why it is hard to formulate and given an operator with this property we can pretty much work it back at the foundation of physics and see what happens so have a physics of creation let us tie the threads of this presentation I will formulate three take on messages first we have something that is called the FEP that is not framed that is most mainly framed as a Bayesian mechanics as informational enrichment of chemical system theory but that can also be framed as something in quantum information theory it says that agents actively infer states of being through their engagement with the environment in a way that maintains the structure this gives us a formal background the quantum framing of it gives us a very very tentative and how to say I would not say ill-specified but very few specified it does not have a lot of detail so it gives us a very not detailed background for the study of how cognitive meaning underwrites the construction of physical state spaces but because it is very abstract and not very specified we have to relate it in a much stronger way to the classical formulation of the FEP where the internal flow of the agents and the synchronization manifolds that emerge from the interaction have a very specific topology of a specific very specific geometry and can be equipped with meaning about the underlying physical structure second we have to admit that pass battery realism is a thing we have to admit that there is something wrong in how we view science let us quote Einstein if without in any way disturbing the system we can predict with certainty the value of a physical quantity then there exists an element of reality corresponding to that quantity by this account nothing is real which is a very big problem a very bad position to hold and we have in some sense to account for the role observation in creating well-invisiating physical states not only for quantum physics not only for cognitive science but for everything in between which is in some sense entailed by quantum physics in some sense entailed by cognitive science so I don't think it is very big deal to say that but I think it will be taken as such and the statement I want to make is that not only doing a way with this brand of realism is compatible with naturalistic demarch with science but it's also necessary because for naturalistic science to be a thing I have to represent myself representing as there is something metaphysical that comes to play is there as taking myself as a detached non-physical uh describer of the world or by postulating some kind of godlike pre-existing structure that I just saw point to stumble upon you will have something that is beyond physics in some sense if you don't do away with physical with this brand of sense of realism and it has been said to me that it is not a new view no it is not it has been framed as a criticism of the notion of atman thinking itself by a early Buddhist philosopher about three million years ago and the fact it is not new is not really a problem in my opinion we should go with things that seem true whether or not they are exciting and novel and the fact that we can have a cosmology without pre-existing nature and other people do not find any problem with it should be a sign that maybe our insistence that no their things in themselves is not well motivated and finally um what I propose here um beyond physical considerations and a discussion of what some piece of math do is a very very very very basic not draft but drawing of physics creation that holds in two movements the first is abstractly the individuation of something from nothing of order of chaos given a space of undefined possibilities be you are the duration of an observable that is well defined and of an agent whose observation of e also individuates a boundary so we have this kind of framing that could work as a handle on how to write the space construction it does have a very strong category theoretical field from the way basically the content of b is not what's at stake this diagram is pretty much agnostic concerning what b is and is not and it must be because it underwrites the becoming of b um and there is a lot of work if I want the physics of question to be a thing in framing properly what kind of mathematical operator basically follow the flow of this diagram and what those mathematical operators do and the kind of physics like kind of lie well for this I have uh considering case I have uh the case of how humans collaborate to build and recreate social norms because it is a case of creation like it is pretty much the single case of creation I could convince anyone is a thing uh there is money because we believe there is money and it works as we believe it works because we believe it works this way it is a ritual statement to say that norms constrain us uh because we believe in norms and the duality it affords between the representation of single agents trying to understand and act within a social world and the flow the interesting flow that is created by the social constraints and that recreate social constraints it is pretty specific it is much more specific than uh the vague operator on the left and I know this operator must represent that so it gives basically some level of grip at this entity level on how the physics of creation would work and uh they would work by describing the unfolding of physical structure as the construction within physics of meaning and or constraints over what is observable by given adjective observers and uh yeah pretty much we have to work from there because um I do not think there is a there is no self consistent in the sense I introduced earlier there is no self consistent theory of physics if we don't do that so let's do that um this work again is uh to some extent my work it is to some extent the collective work of Keros which is basically um it is perspective laboratory that works towards uh an active sense an active naturalism a sense that is naturalism in the sense that it plays the observer within the world and works within well-defined causal ontology rather than less well-defined laws and abstract principles and it tries to do so in a way that enables people as in humans to understand the reality to basically use this knowledge and how the world works to constrain their own behavior and uh navigate the world especially given well the recommendation of structural constraints that will uh we expect and be at play in the coming decades so I strongly suggest you take a look to our website and our general line of work and consider if you want to talk to us or join us the QR code is I believe a link to our website so thanks to Maxwell thanks to Daniel thanks to Alexi thanks to Yonah thanks to pretty much everyone who helped me write this and sync it through this big work uh so I will run through this slide if someone wants to read them in the video and now we can go uh toward the discussion excellent thank you thank you Alexi and Maxwell welcome back if you want to join and please either of you provide a first statement or any set of questions however you want to continue just go for it you know what give me five minutes I will uh try to um destroy my head so chat but don't expect me to okay well do you want to share your perspective on the project things that Aval would be familiar with while you're here uh yeah I can begin I guess um yeah so I worked on the metaphysical or ontological parts and yeah to me Aval gave it a good good presentation yeah it was really nice to see it framed into the whole way of talking about everything like linking it to FEP and then to other stuff so yeah um so the point of the if I can share a bit more about uh the metaphysics of the ontology the idea was to create an ontology that is naturalist and as FEP seems to be really naturalist then I wanted like to make it work together at first it was only in my own uh in my own field that I developed this ontology and then Aval said to me that it was really linked together so that's what what we tried to do here yeah what does that look like how are you evaluating what a metaphysic is or how you know when you found the right one or an incompatible one yeah so the basic question was about the question of what exists more generally so that was my uh my anger of attack if it's a thing in English yeah so um there are many discussions about what there is in philosophy uh mainly in the 20th century due to quite uh I think it is popular in general in the culture of science um so I followed these debates on what there is and pretty much there are a lot of views about what there is and what there is not so I was a bit how to say that disturbed by the fact that there are things that are not um and I wanted to give it a try to try to talk about what seems not to be and to make it exist in some way and so um I'm going to say that better like generally what doesn't exist is linked to consciousness like um Pegasus uh or let's say stupid sentences such as uh the round square copula of the clay college um or something else such as I don't know some weird non-really mathematical entities and I was like okay so I admit that they do not work but at least we are talking about them so can't we just say they exist in some way and then I went into social ontology uh and I found that uh we can call them social artifacts and then I was a bit relieved with this view so I tried to to make it work with also phenomenality or cognitive or personal experience human experience because being a social artifact doesn't make it all so yeah uh then I I just tried to make it work between ontology social ontology and cognitive ontology and yeah that's it and also yeah so uh something else Avel talked about the self-referential inconsistency so it's a debate uh uh I've found uh by by working on ontology and I was really into it the fact that uh but the theory is inconsistent if it doesn't applies to itself and and I found some articles about it uh which were against naturalism but one kind of naturalism only so it was mostly an enemy enemy nativism and I saw that the people who were working on that they didn't know about a lot of naturalism and as I I work with Kairos and in my study uh I found that there are many kinds of naturalism so I saw that the the argument doesn't work with other kinds of naturalism such as the the one that is developed here yeah uh I hope it's clear yep I'll ask one question on this topic from the chat and then feel free to take it in any direction or Maxwell so Adam Rostowski in the chat asked thanks Avel question maybe for Alexi what if state space is considered theoretical ideology rather than causal ontology its presupposition would then not be a threat to ontological consistency so what if state space is considered theoretical ideology rather than causal ontology maybe just unpack the question and then give a thought yeah I will write it now in order to reflect on it perhaps Avel just a direct question what are the state spaces that you're concerned with how would you characterize what they are by construction state spaces there are spaces of postulated physical possibilities you know that practically I do not understand what the notion of ideology how it works here you have it's pretty trivial that what we call ideology will define what a person or political system or institution is capable slash interested in seeing it is it is a topic of scott seeing like a states uh foundational let's say work of other case anthropology that looks at basically how whenever states try to do things uh overtly usually for population their main concern in effect is to build the capability of control and usually they do so by cutting through complexity aggressively uh very basic instance of that is scientific forestry that is I can say it is indeed forestry um the idea was that uh freshly did uh war boats loads of them they also needed to be able to predict how many war boats war boats they would build on a given time frame so I need to know how many war boats compatible trees they had but you don't have war boat compatible trees and not war boat multiple trees you have trees of different sizes from ages um different essence so if you just counted trees um you found with a well-defined measure you would not really know how many trees you have so they just raised forest and planted a multitude of spots of a given specie at a given time with a given soil so they knew that when they knew they had 64 uh I don't know what trees there are I actually know what trees there are I don't know it's told in English so when they are the I will it's not oak but I will say oak to say something when they knew they had 64 30 years oak well that's 64 30 I did say 30 30 30 is oaks and that's it they can cut the trees and they know exactly what they can with it um so it is a way in which what something we could account as ideology so the notion that we should be able to control blah blah blah it's um basically construct physical reality in very concrete sense because they cut trees that plant other trees that fit their expectation it is a form of niche construction it is a form of um um how do you call that cognitive incision in the sense that the loads they would need to account for how trees are they just uh afloat it onto the environment by making trees actually all the same and it is a form of meaning projection they have a stamp of meaning which is that they need to count trees that are both compatible and they project it onto like our actual forests are organized but and this is uh these those are physical possibilities as understood by the state those are state spaces we have a state space that was very complex with a lot of possible trees possible spaces possible uh size that is uh um brought down to how many trees I have and it is uh could you repeat the question word for word please because it was not clear to me how the ocean of ideology uh intervenes what if state space is considered theoretical ideology rather than causal ontology its presupposition would then not be a threat to ontological consistency the causal yes okay I understand the causal ontology precedes a state space it is the ontology of creation slash constraints that they have waived that um somewhat impressively at the moment uh state space are are I think I think I will be able to demonstrate it but I can't at the moment they express the point of view of an agent so I imagine that this is what is meant to be traduced by uh a illogical theory theoretical ideology and um I do not think state space are a threat to ontology I think ontology is not consider state spaces as fundamental are flawed ontology is fine thank you avu olexi feel free to chime in and then max well if you want to add anything yeah um so I think I will join Availa under in the commentary um also I would like to say that um if we take coin's definition of ideology if I will not explain this now but well yeah so the idea I think is that the concepts are or tool and not fundamental stuff that's what I think it is implied by ideology here so what I'd say is that um you give the value of being an ideology uh or consul ontology depending of your your goal and your criteria um but at least minimally your this stuff you're talking about it is that's the fundamental stuff uh that I would say yeah so even though it is ideological even though it is instrumental let's say minimally it is something and it has consequences on the world and that's what I think availa said with uh niche construction uh of social time it is didn't really say that in these words but yeah even though your ideas are let's say stupid or absurd and there and people appear to hear you then your ideas they are in the world and they do something yeah so I hope I insert well if not then I can simply take questions thank you max well do you want to add something or I don't currently have anything burning to say uh availa I think this is the uh well this is the end iteration of our discussions around this I mean as always I think uh you did a great job presenting what I think is um you know a genuine issue uh and I mean I've just been rehearsing the same um you know appeals to multi-scalarity um I mean I find it interesting uh that you uh would speak uh an answer to the these uh issues in quantum f.e.p um precisely because of its scale free nature so you know you can't appeal to a spacetime background when speaking in uh the quantum f.e.p formulation so I think uh that in my mind that makes things um potentially more difficult I see why you would want to do that because of the focus on um you know the the definition of an observation as a physical interaction I guess um Maya my kind of uh overarching question is whether the multi-scalarity of classical f.e.p helps at all uh in the sense that we've described uh before so you have a set of observables some of them change too fast to really like disclose or allow us to track reliably some phenomenon behind the blanket others do uh in the latter case we talk about states in the former we talk about random fluctuations um so I suspect that you know the the kind of definitional uh discussions like uh by which I mean like you know definitions of states like these kind of foundational issues that arise uh in classical f.e.p might be helpful here about I also suspect that you'll disagree with me so I'm looking forward to hearing your thoughts so um basically uh you have a the distinction is clear between us at least as you frame the problem you seem to be a realist about the physical possibilities about state spaces which I'm not I have an agreement uh for non-normalism which is a self-constitutional what ontological consistency and determinants of self-constituency and self-constitutional consistency that brings us to say that you can have primitive state spaces in which uh to frame in which to classify have physical observables that somehow pre-exists physics nature the process that are within it so that is one physical argument the question is then can I demonstrate conclusively that you do not have physical variables before observation but state spaces do in fact try to use the perspective of server and basically if I can demonstrate it the uh the debate is over um until then you have you can basically uh hold you can hold structural realism but you have to account uh of how things exist if they are not fundamental I think uh multi-scale FEP can do that so it's not a critical problem for you everything holds on the uh formal demonstration that we do not in fact have uh state spaces as a primitive physical effect well just I would push a little bit back against your realism versus anti-realism uh distinction here I'm really just talking about uh I'm really just talking about like at the level of the formalism you start with some of observables which seems to be what you are also uh staying and that then you uh can talk about the way that those observables are related to uh some other things which may be latent all of that is a formal statement as far as I'm concerned and like it is useful whether or not it maps in some kind of fundamental realist ontology sense I guess like if there is a there is a kind of transcendental over tone to what I'm saying in the sense that I you know we've argued in print in this on the map territory fallacy fallacy paper that the the FEP theoretic maths and you know the the related cluster of principles unitarity maximum entropy and so forth provide ultimate constraints on any model building exercise just full stop so again that that that doesn't presuppose any realism about what maps map onto it's really just a statement about like the construction and maybe the process of constructing um you're correct it is I wasn't I was uh uh attributing I don't know to say it's English it was not accurate to frame you as a realist about set spaces it was correct however to frame you as an instrumentalist which I refuse to be uh because of naturalism basically I consider it necessary to frame causal ontologies that map onto reality properly as fine, grandly as we can and to do that we have to include basically the translation relation and this is when everything becomes uh complicated because you can't say if there are such physical possibilities then this occurs you can't use that language because by doing that you frame physical possibilities as something that is abstracted from underlying physical reality you um decouple the formal from the actual and I think whatever ontology we use must be what reality is but must correspond structurally to what drives reality so I have uh it is the contrary it is uh I who is the realist but the constraint I was going to suggest I mean maybe uh maybe actually the so I guess you know basically I go for ontology I go for ontology that is consistent in the sense I friend and the constraint of that are much much much much much stronger than the constraint over uh instrumentalists ontology that just yeah we say interesting things which is not actually easy it's just much more easy but suppose we started from the assumption that we don't have to have a strong position on the reality of state spaces uh does that affect your argument at all so you know if I were to say like I I just as the practitioner of science start from my observables you can be a radical phenomenologist from this point of view if you want you're just starting from your observables and then working out like what resists your manipulations or interactions yeah does that in your view kind of allow us to move around asking the questions about the situation you can move around with oh about the fix of creation no um move around yes understand some things with cognitive meaning yes you have a f.e.p. grounds cognitive meaning in the micro system theory it's a huge deal um what it um what it does not is afford ontological consistent theory of physics and the difference between us is not what we hold the f.e.p. to do I think we're both instrumentalists um we love f.e.p. grounds from you as a realist um but the difference is that I see an affordance to build a much stronger self-consistent physical theory that draws elements and a general philosophy from the f.e.p. the philosophy of active inference basically and so I push in this direction because I think the goal is variable it is ultimately an ethical call if you do not agree that the goal is variable you should not agree that it is useful to push it I do not have ultimately an issue with that so maybe um there is a kind of Kantian take um here because I mean I think in the at least in the history of philosophy this split between realism and anti-realism or instrumentalism I mean uh like I think it's perhaps overly simplistic as a description of the potential options uh there you know there is a whole Kantian line of thinking uh that talks about you know something intermediate in some sense uh so you know yeah I don't know uh just a thought so um I could elaborate is uh Adam in the room Adam if you're in the room uh use the chat but in a sense I agree that not in the room but I can read another question whenever you're done I'll read another question from the chat yeah but I was going to say something which is that I agree that the basically the distinction between instrumentalism and and realism must be dissolved because both none of them are ontologically consistent both suppose a representation relation that is outside the world and the question is whether the words as a formal construct correspond to the reality as a structural thing and it is not a constant question to ask the all of the things are in the world we have to have a theory of how things are true or not that affords for the statements we have about the world to be in the world because they are um and so from the moment we're here from the moment you agree that uh ontological consistency is something to have a lot of the distinction we have in western philosophy will be dissolved which makes communication complicated of course Maxwell you want a closing thought on that or I'll ask the next question um you can ask the next the next question I think that that that brings us uh nicely to uh yeah this is from Alex Kiefer so Alex wrote I love the positive story in this paper in so far as I understand it which I'm still working on and the idea of treating even such things as state spaces and scientific reps as amenable to physical descriptions but to bounce off what Maxwell is saying I'm not sure I see the deep critique of thing in itself realism here probably depends what one means by that I tend to think of the strictly incoherent conscient idea of something unknowable behind phenomena so why is treating state space creation as part of phenomena at odds with that and it's also pasted in the chat that's uh just to pick up on that I think um you know we you don't need to be a strong realist uh you can just believe that you know you are more or less getting it right and the the semantics that we've proposed uh you know that uses active inference and the free energy principle is a kind of uh formal basis but precisely uses the fact that you know you have a certain perspective you have certain beliefs and your beliefs are going to be more or less wrong uh and that the the prediction error or the free energy is precisely what like what you need to uh the kind of minimal ingredient that you need to hook you into a situation enough to you know be doing something you know pragmatic that you might call truth telling I think the yeah that that's really all you need from a formal point of view is just the ability to be wrong and to change your mind in light of new evidence so my claim is not that everything in in physics is false and um formal statement about postulated possibilities of no epistemic value or trivial epistemic value my claim is that those representation as any representation are um relational they relate to objects and the truth value or value in a sense that we have to give the meaning we have to give to that uh if we are to give anything at all must be grounded in the study of physical dynamics that's on the lighter margins if we are to have again an ontologic consistent theory we could just say that sense is discourse that is not entirely false and say things that are uh not entirely false we could do that uh again the notion that series um physical discourse must be ontologically consistent is ultimately a sequel like it is a segment on what you should and should not do I can say that it is better because xyz you can say that xyz are not meaningful targets to have and that whatever we do is okay essentially or maybe it's better than xyz but the notion that I am I do not critique set spaces but at least it's an open point I have to frame it as a critique to afford to resonate but the point is that set spaces are rational property and that to study this uh rationality it reframes the entirety of physics it's epistemology it's ontology it's cosmology and if you look at the flow of the theory it's not a pure abstraction it underlies social evolution so evolution is agent projecting meaning as norms and then the norms when they think it is something that is very basic uh for us to understand how through system evolve and yet there is no such theory there is no theory of social evolution that do not presupposes variation and abstract away any form of structure and metabolism it simply does not exist so the non-existence of a naturalistic theory of social evolution is in my view a pretty big problem for a purely predictive uh pragmatic value uh standpoint from a more abstracting point we don't have a current cosmology we don't have an explanation for why why there are things and why the things that are look like what they are physically we just have to postulate that states by space time is such and such for physics to make sense it is a problem we can't have a physical theory that is that stands only on the fact that its statement does not give rise to obvious problems we have to have something beyond the statement we have to have some way to account for how this statement became true and it is a pretty basic point to make in psychology biology sociology we can't just say things are such and such we have to say how they become such and such and why they become such and such it is a nothing in evolution nothing biology makes like nothing in biology makes sense except in light evolution and it is because those are systems that are i don't like telegeco let's say teleonomical those are some that are organized that are coherent that work in a specific way and so we have to cannot just say they do x we have to say how and why they do x and this epistemology has not has not permeated physics physically because of the belief of physicists in a platonic truth of symmetries or formalism or symmetry i have a question for you abel so i'm thinking about this in terms of the direction of travel of our research on the free energy principle which i mean there are two main components to that we could discuss that are relevant here the first is a move from state-based to path-based or path-integral formulations where what what's at stake is precisely probability density functions not over individual states but over trajectories entire trajectories of the system where like what what your generative model or lagrangian is describing is the probability of some specific history so this kind of yeah intrinsically historical formalization and on the other hand as a kind of second main direction of travel really formalizing these kind of three nested scales of inference where you have inference over state factors and individual states like you know color and red green blue and so on inference over the parameters of the model that characterize the relations between states and configurations of states and finally the structure of the model which is really like the the actual you know relations and functionality of like what is connected to what so one of the things that we're you know working on is getting to the mathematical foundations of these basically mutually constraining scales of inference so suppose you know you had access to a path based description of the free energy you know formalism supplemented by a clear you know a clear formulation of inference nested within parameter optimization nested within structure optimization you know in other words like assuming that we solve structure learning which is a big assuming but you know if you granted those two things would there be still much of a problem you think from a formal point of view if you could just like move in the space of all possible model structures and paths through such spaces would you not basically address what you have in mind here um yeah so um three things first uh path integral it solves the issue of um presets meaning or logical systems and the issue of whether you solve that already but it is not something like cold on here so it would not help with the issue I discussed second is structure learning the question is how basically letting things project structure if we have that we would progress in a big way or a theory of cognitive meaning and agency but you will still operate within a given space of possibilities you will have an attraction slash contraction of the state of possibles but by the by virtue of what the FEP is it is it builds a manifold so the manifold like nothing will evolve on on the manifold where it will evolve um so some of this space will become trivial some of the dimension will become trivial um but and some things that would not that do not really appear per scene the set space will become relevant so they will become part of the state space as experienced by the agent but the hard part is to account for how this test space um as experienced by the agent becomes this space per se how you have matrix of six per six that serves as a boundary let us say and how agents will basically collaboratively build systems meaning where they have alphabet and can communicate long sentences using this this grid and this is not in and of itself a question of structure learning although solving structure learning wasn't able to to have a lot of computational experiments that would help greatly with looking at the fine grain structure the question is how you use how you use mathematics in a way um well in a way that describes creation this was the the construction of set spaces that's what i'm suggesting within the space that's what i'm suggesting when um so the you know what what i'm talking about is if if you could figure out when to add a new node or connect or uh edge to your generative model like precisely when is it appropriate to add a new parameter when is it appropriate to say oh these two parameters are connected to that stage in these proportions if you have that worked out i i could see a kind of straightforward resolution to what we're just talking about i'm saying like you know there is the the system just unfolds dynamically and occasionally you know it's it learns new structure or parameters which basically means like you know new degrees of freedom are added to the system we were again talking at the epistemic level you can describe a lot of things by basically looking at the dynamics and being smart about it you don't you don't even need advanced mathematics to do that this statement we make is different let us say that you have people that agree that shells are an appropriate means of exchange for x y and z uh then money becomes a thing it is a thing that exists individually because uh it has structural properties and these properties drive observables through dynamics and equally inevitably did not exist in one before and the question we ask is how we account for an evolution within a space that builds another space and we cannot just look at what's going in the space and declare how it has changed it is not the problem we're uh posing problem is how it wasn't mathematically something that occurs within space that proves another space that is the same but not the same that is structurally um similar that inherits from what existed before but is different in qualitative aspect because this one thing became so what we have to have is the theory of becoming basically and um we can be very smart about modeling things without having a theory of becoming we can be very precise in chemical inference and uh very efficient at um seeing uh regime changes that does not mean we have a self consistent theory of becoming creation space construction call it what we like those are two different problems although one may help the other regrettably both may help the other greatly because um the thing we are looking at is the same we're looking at two very different aspects of the thing Max will want to add something or Alexi on that no I'm going to think about that um thank you Avil awesome yeah from the chat on that topic Tom wrote a similar question Avil could you elaborate more on how active could be expanded to accommodate state space creation assuming active is normative about free energy minimization when should an agent expand the state space and Alex Kieffer also kind of uh added a note it's appropriate to add a new parameter when the model complexity accuracy trade-off says so more seriously isn't free energy measurable across models or different state spaces so it's almost like there's the structure learning meta cognition layer and it almost is like can the box be wrapped within the box can it be possible to have a state space that within its state space represents itself and or to what extent do these physical intermediates and their associated physics and metaphysics need to come into play for different systems and at what level of generality basically about this I'd subscribe to a variant of consistent history and interpretation of quantum physics but it's not quantum physics it's just something that has the kind of physical quantities quantum physics has um well I did commit to quantum information theory earlier so okay let us say it is a in-nature quantum which is a bit trivial because yes it has pieces in it which is what being quantum is but whatever um um the this is the most specific uh mathematically consistent way to present it that I found you still see those slides yep eba eba bay um and this is basically category theory because um this diagram as a property which is that following the arrows basically change what the space is b begins as a neutral interface so a space of possibilities that in principle are for interaction but it is not semantically loaded and the fact that you have the individuation of e the observables and a I framed it as an agent but it would rather be agentive capabilities or constraints that enable the observation of e huh so um this does not conserve the property of b because b becomes equipped with a semantics so real theories that look at spaces without looking at what is in spaces those are category theories and those are the most abstract uh field of mathematics as far as I know and they do not afford uh semantically loaded things like free energy energy is a measure of not only it entails states but it also entails uh measure of states I think uh the basic mathematical structure we need to represent that is much more basic than any other things and any country measure will break down much before we get there but what we are that is more semantically loaded is this the duality between the experience of uh struggle to landscape by agents and the um engineer's flaw that is directed by the landscape and reconstruct the landscape so um the question was what was the question could you um repeat it word for word Daniel it was just about the uh normativity around model selection and whether modeling state spaces could or would require okay the actual artifact the normativity um is uh there it is in this representation of the problem I have a small agent that represents uh circle landscape and acts within it and I can just claim okay there is a set space there is a front instead space I can defend energy this agents mean my strategy and then this is where structural learning intervenes basically because those agents will try to understand the one they do not experience the world directly they will infer the world in a way that is robust and competent but they can make up new things they can fail to understand they can cheat they can erode the structure of the landscape that is collectively enacted in many ways and this um again is due all to something that is the uh dynamical it's not dynamical actually I do not know what it is it doesn't have a word the unfolding of the structure of the landscape under its own sound flow that again is not dynamical because it's not what dynamical sensory says so we described what you're talking about in terms of niche construction yeah so uh in human systems I took yeah we had a Carl and I had an interesting discussion about this a few weeks ago and yeah I think what what really distinguishes human systems from non-human systems is the degree to which the active inference loop is mediated by what would otherwise be internal states but that end up having this kind of ambiguous function of yeah they're environmental but they they encode the material aspects of some normative practice or some they store some information about what you're doing like a like a blackboard for instance or or these slides that we're all looking at and and you know in some sense encode the the thought or aspects of the thought that you're trying to communicate um so the yeah maybe a lot of the problem here can be finessed by noting that it's not just about agents interacting with each other there is the kind of common store of you know or traces you know like a kind of material thing that we all have in common exactly this yeah and but if you know you have a common uh collective inaction of a given structural landscape because human are such that then they synchronize in a robust and some kind of loaded way and that's structural landscape it does not exist it it is not something that pre-exists the fact that agents do that agents infer properties and then enact them and this makes it an open landscape and the normativity of the FEP affords probably a computational experiment design that help us see how such landscape unfolds and look at mathematical theory where the agent is absent where it is landscape that produce the landscape and once we have that we have a theory of creation it is what a theory of creation is so so what you're saying is if we had a theory where the agent was able to like expand its state space through structure and parameter learning in particular then we would have what you're looking for am i misunderstanding uh yes but i'm not sure that we agree that we're aligned on the constraints that this segment entails not sure because the the goal is not to um understand how the agent engage with a personal environment it is and to abstract if we get uh to the target we can abstract entirely the agent and what they think and just look at constraints and how they unfold i went light on the constraint ontology because we do not have it it was not the place for that but we need to look at specifically constraint over what is observable and how they are individuated and how they produce other constraints etc etc we have to get back at this constraints constraints are dual to priors i mean you're talking about the same thing so but the formulation that is taken in physics often by belief is dst it entails a primitive space and this is what needs to be um to address the problem i want to address this is why to have to be overcome and this is likely possible given a constraint centric ontology so ontology what you not have possibilities then constraints over the possibilities but constraints then possibilities are folded by constraints and i do not know what i'm saying is that those two stories are the same story uh that's what basian mechanics says that that yeah it is from within dst uh but they're both from dst what the i mean so a free energy principle theoretic model of the belief updating of a particle in a system is in the strong sense of like is dual to if i have one i have the other right is a maximum entropy dynamical systems model of the whole system as in the whole system of particles and environment to which it's coupled and all of that sensory motor like that that's okay that's the cool thing it's that the the the f e p is the cononically optimal dynamical systems model uh it's maximum entropy in the sense that you know it uh is the it makes the least assumptions given your current state of knowledge about the system you know it's it's jane's jane's as in you know j a y n e s it's jane's optimal in the parlance of belton shaftie vaudeville uh yes so if you have a constraints model you also have an f e p model and vice versa if you work it out from the perspective of the f e p you can map back onto a constraints but definitionally if you can uh frame a constraint model with no reference whatsoever to the michael system theory or pre-existing spaces yes i agree you have solved the problem but the paper which we refer to frames it within the st so i have that's interesting which is those that we need to get beyond to solve this specific problem again i do not mean to say basically you have uh you have uh framed this equivalence i think up to details let us start up to details i would need to think about it actually um but uh with hypotheses that are precisely those that i want to use the equivalence to overcome so it is not it is not the same question i guess i'm just saying that the you know constraints don't get you out of the problem that you were posing originally either like both no in you just you have another kind of layer of you know pre-given static is not we have to characterize constraints it is that we have to have an ontology of constraints so much medical constructs that use this constraint with no reference to anything else just constraints what it means i don't know but if we have that in the sense while conserving the sense that is given by monsieur moreno and while leveraging the equivalence between constraints and belief that what meaning belief whatever that you derived we have one but it is not something that has been done because f e p is framed within uh well vanilla f e p v f e p is uh framed within d s t which principally states and q f e p is one does not specify what build up survivability just specify that there are survivables that we can make survivables but we can't solve this within d s t or within uh pure abstract quantum information theory it is not afforded by the structure of these theories one supposed states the other prescriptions observers we have to show how those are become integrated and the only way forward i see is a constraint ontology so right but what i'm saying is just mathematically speaking if you're talking about constraints then there is a in the end it's probably functorial i mean you know there is a rigorous and systematic mapping from what you're talking about to what the f e p is already saying um that's that's really yeah i mean that that's that's really what the duality is saying like yeah uh maximizing your entropy with respect to some set of constraints is dual to minimizing your free energy with respect to some model it's the same thing if you have one you always have the other if uh how would you frame the notion of a constraint in the sense of monsieur Montevil and moreno with no reference to either d s t or in related states well i mean in these reaction networks that they're mapping they are implicitly using some kind of same reaction networks it is a another line of work right but like i'm just saying that there is a similar kind of like individuation of things that's going on when you map out like some kind of constrained uh schematic in in that sense like yeah the the point that you're making sort of like grandfathers both here which is okay but how do you you know write down a state space model at all and how do they how do you grow them in particular um but i i guess i'm just saying like yeah constraints are just as subject to this as to the extent that the f e p is subject to your criticism the constraint stuff is also for the same or dual reasons like if you just inverted all of the arrows in your reasoning you would also be able to apply this to the constraint stuff equally like that's not an exit to what you're pointing out but i guess my my complementary point is if you think there is a if you think moving to a constraint ontology helps us then that's interesting because any like constraint mapping is also dual to an f e p theoretic model uh so maybe the f e p can be used like very directly just based on the way that you're like if you solve the problem in terms of constraints you could flip it back it doesn't really uh and i don't know if you're stuck along the way then you can flip it at that point and think okay well does looking at does looking at it from this like do i want to look at self-organization from the perspective of the selves that are self-organizing or from the perspective of organization and you know the maximization and entropy and all that stuff i guess so if you're agreeing that uh the f e p treats another differently problem naturally different problem this is not a view i hold it is a view i oppose specifically i do agree that physics of and by belief derives within the st rigorous relation between basically what are the two sides of this actually tells um what i am saying is framing as a goal the fact to leverage that as a way to build a mathematical theory of state space construction and i also claim that this would be essentially equivalent to build a self-consistent and ontologically consistent constraint ontology for for for that for this diagram and i'm not sure what you are going to get because it is an ambiguous that your research as of now is framed within the st which presupposes states which the constraint ontology will not and it's ambiguous that is very relevant because it's uh as you said entails uh one duality at least in one specific case which is bits of dst that is uh the key problem to address so i think you wait you may be combining two issues in this worry i mean so the there's the worry of like new states arising and there's the worry of using states at all so i mean i think the life is very okay so yeah so it's the problem is not really that you know dynamical systems theory assumes states it's just the static nest of the or the you know the predeterminateness or the prespecified nest okay so yeah so specifically you're worried that there is a very fruitful way forward i think to address a lot of them my my uh my proposal is a state space construction not a state space negationism to to get uh uh right i need b to begin with i need physical things to be there well to have a actual self-construction theory you would need this to work from uh x nihilo to get something from nothing uh but it is not the issue at hand here the issue at hand here is how existing things can unfold create whatever meaning states so it is not the notion of states that is the problem the problem is uh the fact that we do not have a theory of how states are created and we cannot frame it exclusively from a mathematical theory that by construction presupposes a space of possibilities we can however leverage a result from this theory to uh generalize it to the space construction which is what i'm proposing so what about the renormalization group so if you start really from the idea that you just have you know observations and they uh you know have a scale of noise associated with them um yeah you can i think i mean there are principled you know physics the kind of ways of bootstrapping something kind of saying well you know some stuff changes so fast relative to my reference scale that it basically averages out and just contributes kind of random fluctuations to what i'm observing in uh whereas other states of possibilities and you look at basically why to renormalize it to look at relevant scales but maybe it would be afforded to have a past realism but it does not entail and does not model past realism it is a model of how complex things complex enemy can fold so it is relevant it is not what i'm uh pointing at right as we come near the end for each of us yep maybe you can spend the most time at avil but for each of us what are our next steps what cognitive counterfactuals and physical embodiments are going to be shaped by this work in discussion maybe we can have non avil go first and then you can have the last word sound good okay well personally i'm very excited to work on the category theory this year learning about the recent uh dissertation of uh toby smith also i think it's interesting to bring in that niche modification active inference variational ecology a lot of the work that as max well brought up earlier had been framed within a multi-scale setting bringing some of these new developments over the last one or two years like the path-based formalization renormalization group environmental process uh generative process generative model symmetries weak markoff blankets there's just many new concepts and toolkits on the table today in 2023 then i think there were one or two or three years ago so it's been a super interesting discussion and it brings a lot of topics to our attention that i expect will be able to reduce our uncertainty about a lot in the coming year and i'll pass to alexi then max well with an athlete the opposite it would be more consistent because alexi and i work directly together in caros okay max well um well i mean daniel i would i would echo your overall assessment i share your enthusiasm for where like everything is at right now i think we're poised for some very exciting work and a lot of very uh interesting stuff has happened over the short term in terms of what avel has presented i think uh you know uh in addition to what you were describing one of the things i want to uh like work out this year is really um the philosophy of the free energy principle or Bayesian mechanics i think uh up until recently what what most everyone has been doing myself included was just coming at this thing from whatever you know disciplinary background we had access to and trying to just make sense of it but uh i think that has made us kind of blind to some of the and everyone has kind of their blind spots you know all of these so-called interpretations of the fvp signal to me that there's there's actually yeah there's there's philosophical work to be done to kind of understand uh jules hedges put it nicely on twitter i think he said the he's become convinced of two things first the fvp is true in the sense that it's quote unquote the right way to do it and second that it's a lot more like boring than all the hype suggests in the sense that it's yeah it just follows from the way that a bunch of physics wicks up it's very very fundamental and important but like yeah i mean um there are a it it's it's part of the kind of core package now and the there are new exciting horizons of like unknown but you know more more to the the point is that there's there's like a kind of core philosophy to the fvp that remains to be uh developed so for example the duality between the principle of maximum entropy which you know implements a form of like octimus raiser in some sense and the free energy principle i think is kind of deep and but we've only started to think about you know the implications that had and chris fields his group has been suggesting that there's a similar duality between the fvp and uh unitarity or the conservation of classical information in yeah quantum information theory so uh that's very exciting sorry if you're talking so long awesome exciting times indeed all right alexi and aval with the final words yeah so uh i will be short so i'm really excited about how these debates will move the lines in theoretical but also practical philosophy and on our direct perception of stuff in life in general to be a bit broad uh also i hope that all of this will make like i hope this will make the difference in metaphysics and ontology in philosophy and meta ontology or some and naturalism of course as it seems to do in active inference lab and chaos and aval so there are three things we need to do to settle the dust on the argument i generated there the first is to formalize much better of the elemental ontological consistency and past pluralism both need to be very clear in their reach and entailments which is something we'll do with alexi mainly and possibly other group members something that we need to do is to put a lot of work in working out what these basic things mean i do not know i i can tell you that it is essentially correct i'm pretty confident that it's actually correct but i have no idea whatsoever what it means mathematically which is a pretty big problem as far as i'm concerned so there will be a lot of work for me and other members mainly yulab renak and the vaccine trimule to work out the specifics and of this of this mathematical thing that allows order to unfold from chaos and most pressingly in my opinion at least it's it's what i'm supposed to focus on in terms of my spectral flow it is to work out what this means how social constraints are enacted and how they evolve because ultimately the rest is obstruction this is what has content this is what allows us to say okay this is our what exists these are the social constraints that shape things and this is what is their flow it is critical from a scientific perspective because it is what gives me to listen to her accounts it is critical from a well pragmatic perspective because it is what tells us what world we are in and what is its flow which is the goal of kairos something we'll do with um uh mainly my non-job and renas canzola and um so basically we will work around the specific of this and we will uh if someone is thirsty to collaborate please write us we will be very interested uh because there is a lot of all of these problems are big problems all we say to a lot of firepower so yeah basically anyone who wants to um we're interested in working all the specifics with us is very much invited to reach out excellent thank you alexi avel and max well for joining thanks everybody for watching till 34.2 it's over thanks fellas um