 that was just blues entry music. Hello and welcome everyone to Act In Flab live stream number 036.1. It's January 19th, 2022. Welcome to the Act In Flab. We are a participatory online lab that is communicating, learning, and practicing applied active inference. You can find us at the links here on this slide. This is a recorded and an archived live stream so please provide us with feedback so that we can improve our work. All backgrounds and perspectives are welcome and we'll be following good video etiquette for live streams. You can check out activeinference.org to learn about how to participate in any Act In Flab events which go beyond but also support the live streams and also the code of sight at the bottom will give you access to the table with metadata on the past and upcoming live streams. All right, today in Act In Flab number 36.1 we're going to be discussing and learning and seeing who joins for the paper. Modeling ourselves what the free energy principle reveals about our implicit notions of representation by Sims and Pazulo 2021. We went over some of the background in 36.0 and today we're just going to introduce ourselves, say hello, what we were excited about in the paper and then we wrote down some things to discuss but I'm sure both of you have other random things you wrote down and anything that people write in the live chat we can also look at. So I'm Daniel. I'm a researcher in California and I think I'm excited to just hear Dean's marginal comments and just see where it goes because there's just a lot in this paper and so it'll be a thick dot one. Dean? Well, I'm Dean. I'm here in Calgary and I really didn't know if I even had time to even look at this paper and then I listened in on Daniel and Blue's conversation in the zero and it piqued my interest and so I read the paper and I was really happy to read this paper because where Daniel and Blue kind of took a path to introducing the material, this is this is kind of familiar ground, familiar territory for me which is very unusual because most of the stuff that we look at I'm looking at it for the first time and so I'm really excited and I don't want to detour anything that you guys want to sort of look at at the point one maybe at the point two point of this process we can look at what the section five of this paper what can we learn from this debate because I think that section five of this is super interesting and not just trippy you guys use the word trippy which I thought was fantastic quite quite a good description but I think we can pull parts of this out and look at them and reassert them and reattach them in our own way and actually come up with something pretty pretty amazing. I'll pass it down to Blue. Hi everyone I'm Blue. I'm a researcher in New Mexico and this paper it was great like Dean was saying it was familiar to me but not because of my background just I thought it really explained and summarized a lot of like different points in the FEP and active inference and how we use it to model ourselves and it pulled out a lot of really important questions I think that I'm looking to looking forward to discussing with you guys. So we wrote down some things to discuss but maybe let's start with just a blank slate. Dean what do you think would be some approaches that we can take or some starting points because we already have other cards of our sleeve? Well I don't want to be I don't want to be a guy holding up a flag and detouring your conversation. One of the things we might get to near closer to the end hopefully is the authors speak to the possibility of you know when you set up your table Daniel one of the things that they potentially talked about in the section five of the paper is the idea that the lines are a little bit fuzzy here maybe they're broken and the idea of in some cases our ability to move from the non-representational into the representational it's kind of a big deal because you can have the debate that says here's the parts that would align with a non-representational view and here's the lines that would are parts that would align with a representational view but then they insert this part that says but maybe we slip into and out of these representations kind of like we slip into and out of the Q density and so maybe at near closer to the end today we can tap into that a bit. Let's start with that grid. Yeah because this is the structure of the paper and the elucidating contribution and also I think Dave left a youtube comment or something like there's other dimensions so this is kind of pointing away towards an interesting approach to making sense of some philosophical debates kind of laying them out in tables like this so how should we approach it there's the representational and the non-representational perspectives or stances or modes or archetypes that is definitely very blurry because of course it depends on what you mean by representation and that can have some fundamental ambiguity and then that representational versus not continuum or dialectic is then separated into a few domains. I think we could talk about what are the similarities and differences because some of it was a little nuanced and I wondered if there was other categories or they were organizational structural content related and functional so then this is your game board of eight lower dimensional projections that the thought can be in or it's something like one of these eight octants and then a lot of the papers going through some of the literature and rhetoric for and against each of these using the FEP as kind of the interpretant what is being interpreted to speak for one of those eight cells whether either of you want to continue it there. Well on the action domains on that sensory motor stuff and Yelly Broomberg talks a good argument around the idea that for example I could be in a flow state and there is no model that I'm or plan that I'm necessarily following and that's true if we're talking about extension and embedded as two of those four E's you don't necessarily have to move to that place that requires something to work off of you could just be flowing and working in something right that that that rate of change is is you extended so there are there are times when you literally don't have to be working off of a plan you could also be moving from a blueprint to some kind of outcome that you want to achieve and then that it's pretty obvious then that the representational piece has to exist so maybe the argument isn't that both aren't at work it's how we transition from a density out of a density and then back into a density something to something to at least have a conversation about because I think blue was touching on this last time too she said I just want somebody to be able to show me how how these lines are in this this sort of structured relationship exist and I when I when I heard her say that I was kind of like no I think that's a that's a fair question because those those lines seem to be blurred so something that came up yesterday is the idea of free will and I just wonder how that might play into these different grid boxes like like is there free will in all of these or in none of them or is that even related okay let's write it down one interesting thing about this is it has a little bit of that visual illusion with the black grid and the white dots at the junctions so it's kind of a a perceptive artifact it's just sort of messing with our visual representation and visual representations are are uh you know when you see it you believe it that kind of focus system in a lot of the active literature and vision okay so let's let's save free will and the the blurriness for later and then come to some of the things we wrote down for the previous list okay so in what sense can or are computational models or any other kind of models be fully observable can I can I give maybe a possible real-life example that maybe speaks starts to speak to that sure okay so in the paper he the authors say however these differences matter if one considers structural aspects and the degree of resemblance between hidden variables and environmental dynamics so I was thinking about your question and I so I wrote this down if dynamics then the material i.e the content the structure and the organization is going to have a different change rate than the function so the first the first three columns are going to behave differently than the fourth for example the screwdriver as tool remains very stable hopefully relative to the spiral effect or function then it can generate on the screw to the one driving the screw there are now two rates to juggle the assumption that the tool applying the function as a rule is stable and consistent and the effect sought will be a form change meaning action orientation where a new representation outcome happens you quite literally start and end with a dance between the stabilities and the dynamics leveraging off of each other so back and forth so we need to keep open the possibility that the that the stable or the representational and the dynamic which as Yelly points out can be non-representational as well are both in play and then I wrote here because I've used the expression when in doubt zoom in zoom out when markoff blanket zoom in zoom out so in terms of in terms of the computation there's actually two computations one around around what we think should remain relatively consistent and one which we then apply that consistency to something that we actually want to see a rate change expedited so there's actually I think there's two calculations here to the person that's trying to figure out how they're going to see what's on the other side of the blanket very interesting Lorenzo Laplace so I like this idea of zooming in and out with a markoff blanket and I think it speaks to the the way that agents are nested or or anything is nested I don't want to necessarily ascribe agency or free will to different levels of of organization but I think when markoff blanket zoom in different markoff blanket zoom out different markoff blanket so for me like it speaks to multi-scale systems and how applying markoff blankets is kind of scale-friendly or like imputed on to the the system in question one extension or maybe example of that we've probably talked before like the table it's in the process ontology it's material that's tabling but then that material is also process based so sort of this multi-scale process ontology and the identification of a given system of interest at stationarity like an object that can be tracked with but there's objects that are too fast to be tracked and there's objects that are too slow to be tracked below the just not noticeable difference and so within a given measurement or active inference regime a given entity can have a attention on a given system of interest it can't have attention on things that are slower or faster like out of its frequency band but that's discarding 99 of the system to not consider outside of the frequency band it's like the observables are just one they're they're very bounded and so that the slower things are the larger things and the faster things are the smaller things so that's kind of a nice and and and let me just piggyback on that and if we're talking about specificity or scale-friendly versus cosmic and scale-free that's also part of that zoom in zoom out exercise because there there are quite literally elements of this that are still scale-free that if again as they said modeling ourselves we can we can we can start from that position of scale-free and then find the specifics or we can find the specifics and ask ourselves how far out can we extend our confidence that the stability will hold so that's again piggyback on what you just said Daniel there's all there's all kinds of ways of interpreting zoom in and zoom out and I think that's maybe that's one of the insights that we get when we don't get ourselves hung up on being a really radical inactivist even though I'm of the camp that says we start out in the non-representational and activist leaning side and then we the plans and the and the representations and the effects of that are representational all right I'll think of an example that a wet lab example so I'm thinking of doing PCR or a wet lab technique where you're not observing perhaps the outcomes of what's happening for multiple steps and even then it might be a color change or a number on a screen and so then you get limited observables those are sometimes your degrees of manipulation like in a thermocycler that's changing the temperature the only thing you're measuring during the reaction is the temperature sometimes you might also look at like the optical property but usually just changing the temperature and then it's very representation like to have a mental or a cognitive representation which can include extended phenotype like the notebook and the computer but that's pretty representational to have counterfactuals about what molecules would do that have never been visually observed um and so science becomes representational and I think how far does that reach back and it kind of blurs off into increasingly non-representational forms not saying that science is the most representational thing but what are the most representational things and how we move back and forth on that is what I'm seeing with a two feet in two feet out you but you just sit on a really important point because you said science becomes so so what makes up the becomes like what what exists in that that space prior to it us all agreeing that it's now science and I think that's what this I think that's what this modeling as opposed to us as model asks us to consider that's a really that's a really you touched on the the big the big deal here abductively so what makes up that what what what has to be present for us to be able to say okay now you've crossed a threshold and now we're talking science yep it reminds me of the paper of Brunberg anticipating brain is not a scientist and we can just take the the always prescient titles and content of these papers to be targeting attention point like is the Markov blanket a valid or useful construct more recently but back several years ago identifying some similarities or at least a area of interaction with the anticipatory and cybernetics of systems which is pretty system agnostic can apply to very abstract representations as well as perhaps real systems and then the physical brain hypotheses and the inactivist and the ecological perspectives on cognition with this activity called science and the role of a scientist and so a few ways to parse science one is as we heard from Majid Benny recently with model-based science so that's one like relevant way that we learned about and talked about this recently like science being the application of models but then other systems are perhaps doing something different well what is science anyone in the live chat should definitely respond what is science and also I would like to hear what both of you think and how that relates to anything we're talking about here or ever so I always think about science as the application of the scientific method to anything right like so observations hypothesis experiment etc and because I think of science like like this science itself is just a model also right like so so we use science it like we build a model of whatever phenomenon it is that we're trying to replicate or observe or test or whatever and we do we experiment with our model like essentially so so the scientific method itself is a model for doing science and then the process of doing it is creating a model of of the world so it's like nested models all the way down in science in my world so blue can ask you a quick question to what do you think precedes the engagement with what you're describing is or is because the authors always say assumptions matter what what assumptions do you think goes into a person saying okay well now I'm going to use scientific method are there are there things that that priors that we should sort of pull out and examine that are consistent when a person gets to a place where they say okay well I'm gonna I'm gonna do that yeah I mean so so there's a lot of stuff that's in the scientific when I say the scientific method it's like you I mean it encompasses things like scientific bias and positive control and a negative control and and so all of these things are kind of this in this like umbrella category when I say scientific method an experimental design like it is not always simple but but definitely like what precedes any science happening is simply observation right like so you have to like be observant and to have observation you have to have some instrument with which to measure or gauge the phenomena which you're trying to examine right so even if it's just your brain like which is probably the most powerful instrument of all of the instruments because it's fundamental to any test it testable instrument that tests anything in the world so I think yeah like that so that's a lot of priors I think so so is it fair to say then that to get to a place of of adopting a scientific method one might first possess before they can actually get into the formalities of that a field seeking curiosity and a way finding bias and those are two different things right one is I can be a spectator the other one is I've got to get in the boat I've got to get my skin in the game do you think that those two things are necessary to adopt scientific method or to commit to scientific method so I'd like for you to unpack those a little bit more so um yeah can you just describe what you mean about like the those two the way finding and the the curiosity and yeah the yeah the field seeking curiosity well curiosity I mean it's got it's kind of got four parts right there's there's an epistemic piece there's a there's a sensing piece I can't remember what the other two are right off the top of my head but but bottom line is is that curiosity doesn't just it's not sort of a monolith um and and part of that determination because you don't know but you you still you're you're you're prepared to put in some time and effort uh to get past not knowing means that you can take one of two essential paths you can find out right you can go to a source that already knows where you can figure out based on sort of collecting and and curating evidence so that that's the that's the field seeking piece and then the wait wait wait what's the difference between finding out and figuring out like to me those are the same so like whether you're collecting evidence like from google or from that's epistemic right but to to figure out is to is to not have a source already not a single source of truth that would be the difference so one you can find you can find a truth source and then either believe it or not and the other is you have to you have to develop your own truth through it takes a lot more time because you have to go around and gather a whole bunch of other bits of information that aren't necessarily prepackaged for you does that make sense yeah and and also like what is truth well right right i mean you ultimately have to be the one that decides that right whether to trust and have confidence in something or not cool lot to um yeah really go for it no go ahead Daniel um so let's pull one level back from or two from one and a half fractal dimensions from the curiosity fourfold distinction there was like the personal factors so the question was what proceeds the type of engagement that might be called scientific right and there's some personal factors they're basically psychodynamic or like situational behavioral archetypes but they're things about individual uh people humans what other kind of factors even if the individual is the nexus of agency what other factors come into play and like what precedes the kind of engagement that might be called scientific if the person is our system of interest what do we how do we zoom in and out from there well that that's my that's my question my question isn't i don't have an answer necessarily but what i have is so you know before the big bang kind of thinking we don't know but is it possible that there actually is something that we can put our finger on that leads someone to have confidence in a scientific method and not just sort of remain in a kind of in a place of stasis and and not be curious right because there's lots and lots of people who are kind of quite comfortable just going going along with whatever is happening around them but then there's others that appear to have um a potential for a deeper commitment so and so for example when when we were talking prior to magie coming with us on with us um blue raised the specter that genes don't necessarily have to have time in order to be considered prediction matter expertise because it has both the cause and the effect included in the in the structure so that got me curious right like some for some people just making that statement or that or asking that question blow right by them they wouldn't even give it a second thought but then for some people just making that claim arrests them slows them down and gets them asking questions on their own now so that's why i'm asking this in the context of so why would somebody now go from i don't care i don't give a rat's backside about science to i i'm very passionate about science i'm very interested in in learning this method so i think um science really is seen by many as a kind of truth finding endeavor um because whether it's through mining previous knowledge or generating new information maybe knowledge um either one of those is definitely like like a a a thinking of the truth um and in that way like aren't we all just looking to like minimize our uncertainty reduce our uncertainty so so i i'd look at science like like it is an uncertainty reduction task i mean that's the the point of it perhaps so i don't know what factors um you know other than the fact that like maybe we're in a simulation like like we could go there um because that that would like precede the scientific endeavors that we undertake in our truth finding efforts um but but really like the knowledge of science as a way to determine truth um i think might be part of it i mean i mean and maybe not everyone has taught that um i think about like you know in western culture we're all like indoctrinated with the scientific method from i don't know fourth grade on or something like that right and then they give you the phd at the end when you're fully indoctrinated but but here's a thing blue and this is a would you bring up a really cool point and you know fourth grade indoctrinated and i think one of the things you and i have kind kind of gone revolved back and forth is so yes science and so somebody has has said this is a this has got some practical things and some truth finding things but i've always come back with great science plan these these are material things my question has always been when and you bring up the well we've decided when it's grade four but i'm not sure that some someone who's afraid for let me see that be about nine ten years of age is saying is the same sciences for somebody who's 29 30 or 39 40 right so we can we can say plan but i'm what i'm suggesting is is that before the the the plan or the science there's something before that some really important things that we're just assuming or willfully ignoring because now we're on to the plan right because that's what has been reified and and given the the the holy blessing sort of thing so i think you're you're hitting something that that's critical and it's why i'm disagreeing with you so i don't think that the science that you're doing when you're nine is that different from the science that you're doing when you're 39 um and the perfect example of that is um a fine man like i was reading i don't remember maybe it's in in the meaning of it all like maybe it's in that book but um you know he talks about how like for him he just plays right and so a lot of his science is like oh i get to like make this toy and play with it and try to look at it and i keep playing with it and i keep playing with it and so this is something i think that's like innate in us like to just the figuring out part like the truth finding part so even as a baby like from the time you can sit up like what is this block like what is this shape what is this slinky right like it's finding out and figuring out like what are the affordances of the objects right right and um some of them are imposed by the outside like don't stand on the chair um because kids will use chairs as tents and step stools and you know like so it's some of it like that we impose some structure on you can't use this toy in that way but but really like kids will they don't know they'll use anything anyway like they'll they do all kinds of crazy things like like they'll you know slide down a mountain on you know a trash bag or something right like so so they don't know right how to what what the limits are um and so they'll explore new limits and i think that scientists um at least the ones that really love to do science uh play a lot also so so again the the the trial and error piece can be like the triangles piece it can be scale free and i would never push back on that piece the epistemic and the specific is what you just described the other two parts of curiosity are the distributed and the sensing and that's kind of the sort of the more act active part where you have to kind of go out and do the skin in the game you have to get in the boat and cross the south pacific you can be doing the epistemic and the specific as well because you can be taking samples but you can get you can kind of see that there's another side to this so my my question around the when isn't around the the um the ability to see the scale free part of it my question is is if i'm a 10 year old and there's something before the plan well how sophisticated is that something relative to the something before the plan has a 40 year old i just don't think that the experiences that a 10 year old has and the resources that they necessarily draw upon are the same so i do think there's also a specific and a scale friendly part to this would you would you push back on that i mean i think that we all kind of just operate on whatever level that like i do agree like that the knowledge of a 10 year old isn't the same as a knowledge of a 40 year old but um i think you know we just operate it at whatever level that we're at similarly like a baby you know like they're just doing their own kind of truth finding but so the process is the same like even though the priors may differ i think the process is similar but this reminds me of in the linguistics guest stream with elliot murphy how there's the view of language as the construct like subject verb object and then there's the language as learned by each learner and so i'm almost hearing like blue saying it's the same process everyone's on the same trip but of course recognizes that each learner is going to have individuated factors including like what they specifically know maybe has some correlations or um whatever with age or with other features and then that's i feel like maybe partially reflected with a specific of the curiosity um like if yeah is the nine-year-old and the four-year-old learning the same science it's either pan-scientism or uni-solopsism-scientism everybody's in their own scientific niche and so it's just a million different granules and then the other hand the other end of that extreme debate would be it's all one integrated extended cognitive process which i wrote some notes on so dean go for that and then we'll think about it as more of an extended continuous process yeah and again i'm not i'm not trying to i don't want to sort of shoehorn something in here and and so i would always i would always hear from people in my history the expression um well i know that's a problem there i guess you're just gonna have to get creative you're gonna have to think outside the box and being the um insurgent and the non-compliant person that i am i would say i would come back with and i was reporting to people i would come back with so tell me about this box what makes up the box what makes up the process and of course that didn't make me very popular because nobody could get specific about what what's inside or what makes up that three-dimensional space and and i i think it's the same thing with we can say well there's a process well tell me but then why a 10-year-old thinks in the same way as a 40-year-old does what evidence what what what what are you hooking that assumption on and again i'm not i'm not trying to be a pill here i want i'm really curious what would make a 10-year-old think the same way as somebody with an additional 30 years of experience they want uncertainty reduction that's a generalization that i can agree with but what about the specifics i'll give a specific in a general i think that for disciplinary education some of the questions might be very similar like why is this leaving group superior to that one in organic chemistry i think if somebody were very precocious in 10 and learning about the halides as leaving groups or if they were learning about it in adult education they would ask that question and then once they got the answer they would have moved on and they're learning in that disciplinary finding out way like here's the periodic table here's why the trend is this way so that's not the whole of education but in that mode which many of us have come from and exist within that was the education that's educational instruction so that's the instructionism component and then i think it becomes different as you move away from that spotlight so then a few other comments like about the specifics of how the different age would be different there's the the biology of how the individuals would be different like in some of their cognitive features whatever it meant for that person from working memory to the specific long-term memories and other capabilities they have perspectives they've learned about um they're a different complex system but they can be a complementary part of the system like older people providing mentorship for younger people but i think that's what kind of comes to some of these notes on science as an extended cognitive phenomena which is partly what has been got at so science has many different possible goals and sometimes scientists in an ad hoc way will appeal to different components like science is correct because there's technology like it's demonstrable or it's useful not always providing a null hypothesis for what we would have if we had done something different but there's all kinds of different uh ways to look at it but it deals with individual cognitive features and learning and action in their niche but it also refers to broader phenomena like maybe a differentiating question is can one person on an island by themselves do science and if the answer is like yes they could explore a hypothesis test then that's one perspective another perspective would be you're only doing science when you're using this interpretation of the scientific method and you're using science as a body of knowledge like you're using the literature with the dois and some people go as far to say that you need to have a career in a certain specific institution or a specific degree or a specific training or approach so that is clearly not applicable to the one person on the island and not that one of them is better but that whole apparatus that's part of the extended niche and then also there's at least in our case the emphasis on the digital or at least informational stigmergy like there's a recognition in science in a different way than like there's many books you could read on tying knots probably a lot of youtube channels is there a doi for information on knots probably not but that in science is a big emphasis reading and writing the literature at least in the way that we know it today but again that could also perhaps not be as much of a core component anyways just like a few points that this maybe now we'll bring it back to this paper but science is kind of the representational paradigm and i think dean's question is getting at what is before what's the before and the after or the subconscious or the submerged of science and that when we're talking about that representational piece and when we're talking about science we're talking we typically talk in terms of isms what is where is how is when we when we set up our our hypothesis and i think what this paper but we're bringing it back to the paper i think what section five of the paper speaks to especially when we're talking about the the functional dimension is the if the counterfactual and how they talk in there about how i'm not sure if it was clark or how ho here one of them talks about the fact that so a bacteria doesn't typically ask the question what if when it's sampling its environment so that's that's pre science yeah i think that relates a lot to martin boots's uh guest stream i think 12 on the event based cognition and how if there is an event based structure to cognitive events then for any given parameter in that model it can be otherwise you could say like the ball is rolling off a table the ball is not rolling off a table the cup is rolling off the table like you can start to just do single mutations especially the important counterfactuals for linguistic structures and then that might be closer to what we've seen discussed in other sims papers as the adaptive active inference agent going back to earlier papers citing that but just that was our previous discussion with sims this paper has sims like the pendulum does not have a counterfactual about whether it could go on a different axis it's mechanically constrained whereas cognitive structures can engage in counterfactual play in a way that the pinball machine can't even if the pinball machine has chaotic or other types of seemingly agentic behavior and that and in that in that in that space i can be at the top of the of the ski hill prior to a prior to a gradient descent and i can be adaptive meaning i can close my eyes and do a visualization and i can also be variational meaning that i can try to recreate through my proprioceptive sensors a sensation of floating pre pushing off and engaging with that process right so again the when the when of this i think is something that really matters i think we gloss over it or we don't even pay any attention to it typically because we want to get to in the paper it's figure two we want to get to figure two but i think before there was a figure two before that there there was a plan or a representation we might want to ask when plan if i'm engaging in counterfactual inference prior then doesn't everything in the diagram depend on prospecting up the nodes in that representation all right do not do i not have to provision up a format that will be potentially something that others can comprehend as now organized and don't i have to propose alternative paths as links like all of those things are pre-plan all of those things have to be present pre-science so we can we can say well i'm not paying attention to that now because you know i'm busy putting processing something that i can now transfer to other readers and viewers of this diagram but i think to ignore those things is to ignore the difference between a 10 year old doing a science experiment and a 40 year old doing a science experiment let me give one thought on that it actually speaks to this question of whether science is the same for everybody or different for everybody is this the board exactly as laid out that everybody has to play by that's like the game of life the board game not the cellular automata simulator but the board game there's one path or there's limited points of agency and then everybody on the board on that layer one is moving through life on the same or on a constrained set of trajectories so it's like are these the immutable railroads of decision making and thought or might other people have different connections or different decisions or totally different railroad stops or different dimensionality or just represented in a different format is it like well here are the battle lines that are always going to be the case for philosophy and there's going to be some people whose situation and neurodiversity leads them to be on this side or this quadrant or emphasize this component the regime of attention or is even that layer shifting because there isn't any stability to be found in counterfactuals there you go back to the questions it was a great early point about the rates of change being different because also that does at least talking about rates of change kind of evoked numbers gets the scientists excited because could there be like a number could there be a sentiment score or could it be like this sentence is totally consistent with both or it's inconsistent with this could we annotate the literature and measure or do surveys or actually measure those rates of change in individuals on their learning trajectory or in fields like how could it go from just being a really insightful philosophy paper with a lot of scholarship into being something where we can use some numbers or something so I see the first three columns organization structure and what content as the more stable I think you had another you had another diagram where you were showing that going out over longer periods of time right so which one do you think is the most rapidly changing functional okay so let's kind of recap them so organizational seems to be related to the encapsulation of the the organization like the architectural features of the world such that the representation is separated from the what is represented so like you could have representation of the sun but then that's very far physically from the sun yep the structural aspect is related to representational vehicles to toot that are structurally similar to the state of affairs in the world they stand for so that is the abstract structure of the representation having some congruence or isomorphism or like similarly some similarity to what is being represented this one is super dependent on the definition of representation because someone go well you can't represent a car in your head because your head's on a car content related having internal models that either encode environmental contingencies or sensory motor contingencies specification or description of how the world is taken to be turned analyzed in terms of correctness or truth so that seems to be more consistent with the car content you know it's like the youtube content the car content can be represented if it stands for something the content of that message was an invitation to the birthday party so that is related to the information like the memetics especially in a actionable way and then the functional role this is where we're going to highlight and hopefully Dina I want to hear a little more about why this one is different supports vicarious use before or in the absence of external events of internal variables of model that was discussed in the paper a lot in the context of sensory motor detachment and there have been some blurry lines found there for sure like motor replay and pre-play so we know that it is on the spectrum of existing that neural dynamics but cognitive dynamics involved during are involved echoes or anticipations of them are involved in the before and in the after as well as counterfactuals before and after so what is happening with functional representations and then maybe how does it relate to active and FEP yeah so I don't I didn't know I know enough about dreams and subconsciousness just to get me into trouble but the first three when we reorganize through our dreaming and whatever those representations are or non-representations depending on your your wish those are all backwards looking typically and then the functional that's what segregates it out is it's forward and backwards so that's to me a huge differentiator right there in terms of how much how much or how long we expect something to stabilize versus how much we want to act on something and and generate that change or outcome that we're seeking that's why that's why I get back to that search field being different than a framework okay it's a lot to think about but it's interesting the organizational um maybe there's some dissenting view from this but this is related to how the system is arranged so this one is can be thought to be slowly changing because we're talking about things which means we're talking to something that has persistence and so the organizational aspects of systems tend to be the slowest changing perhaps like the organization of the servers changes on a very slow timescale relative to the processor then the structural aspect also it's very important that it has persistent like storing a variable in memory like you want things to stand for each other for a long time you want the doi to be a representation of that paper for a long time those are also related to like it's sort of the functional side of organization like form and function a little bit but i know for functional is used later so not to confuse the term function but this is related to some of the performative aspects of organization and infrastructure and architecture so that's also like that's you don't want the screwdriver to be a different thing in your hands that's too slow but maybe be cool if you can modify or tweak the tool or like you know unscrew it and bolt something in a little slower to make it different but usually you don't want to change too fast then um but what about the content related this seems to be also changing somewhat faster because environmental or sensory motor contingencies that seems like that could change on the fly and then i agree that the functional rule as defined here like vicarious detachment including counterfactual and self-modeling counterfactuals this is like some of the most rapidly mutating but what do you think about content also changing fast well i think whenever i make a mistake and i accidentally hit the the delete button on my notes how upset i get because the stability of the content is lost so i actually see it is still relatively stable as opposed to something that's changing rapidly so these four uh columns here there is a different temporal aspect which relates to what you asked earlier about when representation so instead of the possessive do organisms have representations or the identitarian are organisms representations or the definitional what are representations this is the temporal or the dynamical when our representations well that's not complete you have to say what it is you can't just say when it is well couldn't you just flip the words wouldn't it be just the other person's perspective you can't talk about is without when so why is that one okay why are there papers about what is a representation what is it like to be a bat rather than when is it like i don't think it has to be or i think it's both the other thing that i would i think we need to maybe think about that is the first three um tend to be on a continuum that's orthogonal to the fourth i can organize i can structure and i can content build functional now is orthogonal to that i will because it's going it's used before so again back to that temporal aspect it's both i wouldn't drop any of the four columns i'm just saying that i think that the fourth column is maybe on a y axis relative to the first three that are on on the x plane okay a few notes here so before or in the absence of right now isn't it the case that usually anticipation entails something being both before and in the absence of i wonder what it will be like when my friend comes over that's in the absence of the event and it's a counterfactual before but it may happen it's a prediction it's an expectation and maybe even an active inference it's a preference so i agree like even the way that the authors have written it these are related to architectural and performative features of like basically functional or structural definitions of representations what they are what they do and then this is even uses temporal words as well as counterfactual words though and those are probably like used elsewhere in the paper but that's a very interesting distinction and so it's like there's our map but you know how easy is it to just lay out the map and go all right here's northern california here's a map instead of the temporal aspect and i think that also shines a light actually on another feature of what people would call science which is like science the metric system and then like okay you've gone too far with the metric time but then pretty much on a lot of the other fronts except in the united states the metric system has hold as how people are sense making in terms of like distance and time and other units and that's chronos like like like chronology and time in the decimal time and then kairos is like the timeliness and there's other ways to think about time so we kind of open it up and it just makes me wonder okay what else is handed with the map is there a time object how do we share time along with the map is that attention is um focus times time times all these other things so how do we convey the temporal aspect and instead of just this gene regulates running network or this is that the map here or the map and the math and the territory how do we bring the temporal along so i wonder like about the temporal also and especially like now that we're kind of delving into um like quantum aspects because you know like the the simultaneous existence of more than one fate right is possible and so is the the is it possible to be like building and like executing your model at the same time like can is it input then output is it always sequential or is there some kind of like simultaneous happening right like that's that's going on there i don't know oh blue that's perfect for the next question i wanted to ask so is one of the assumptions here that there's a continuum like if then between i can and i will because this was a debate that i got into with people forever when i was in education that there's some sort of an assumption that because i can therefore i will well i wouldn't jump off a cliff even with a parachute on there's no will for me to do that because that's just not my it's not my jam but but orthogonally speaking if if if organization is on the on the x and structural is on the x and content is on the x but functional is on the y now i can separate the two things out and and and as you said if i go back to the quantum level now i'm i'm pre decision branch in that diagram do you notice that the diagram is basically unidirectional it's always heading to a particular outcome it's always i can make this decision branch moment therefore i will arrive at representational or non-representational and i'm with you i'm not so sure that previous to this that it isn't bidirectional it isn't functional it isn't orthogonal i think it could be all three of those things blue would you like to say anything or can i all right two points i would be remiss not to make first the tetrahedron and fuller synergetics gives us a mnemonic to have four spatial dimensions so this four column scheme all of them can be four intersecting spatial dimensions they're coordinated at 60 degrees not 90 degrees and then the second point is like this um one directional maze reminded me a lot of the way that people study decision-making and path finding or figuring whichever one may be in ants which is there's some laboratory maze like a y maze or a t maze or maybe some branch on a tree in nature and then for the purposes of the science so-called the paper it's going to be like we counted them as going down a certain branch making a decision at the node when they went five centimeters past the branch point and didn't return so it's like this is the the line that you trip when it goes from zero to one on that number and that's what gets fed into the scientific process and the method observations of the literature and the statistics and all of that but then if you're watching the ant or if you look at the video maybe they dwell at the node or they turn around or they walk down one and then they come back or like even they go far down one and they come back so it is very interesting and it's a subtle graphical point but one that would be almost seen as like what do you mean how could these be undirected edges well what if we say hey let's start at representational FEP it's it's just where we're going to start today and then let's pull back to the vicarious use of variables now where can we go from there to sensory motor detachment or we can take no and all of a sudden we're a non-representational FEP because they're only one step away because they're both outcomes of the decision about the vicarious use of variables so I hope those are some interesting points and I think that relates a lot to the discussions that we've been having in edu about cyclic curriculum and you touched on it I'm going to just bring it down I'm going to reduce it not over reduce it I'm going to reduce it to the to your point Daniel there can be a yes there can be a no and there can be a maybe and the maybe part is what's being left out of this cool I'll read a comment from survival in the chat uh who I was also thinking of when we talked about like science is a human phenomena and all of that I think survival is a lot to say on that survival wrote considering many world hey Dean I'm going to read survival's comment considering many world type theories is it possible that can entails will it would be a distributional property though not a trajectory property okay I have to think of that in terms of what I how I set up entailment um maybe I'm not trying to I'm not trying to duck I think there will be times when my confidence through my ability to say that I've done something before leads me to a lower threshold around will but then sometimes it doesn't so I'm not I'm not trying to be evasive but I can think of times when it can entail and I can also think of times when it very much does not so will it's a short English word it's could refer to something that will happen like the ball will go to the bottom of the bowl unless anything else stops it from doing that or it can refer to sort of that motivational psychological energy which is related to agency and so it's really it's kind of an interesting congealed word because agency and will are very related and I'm just trying to think of they're positively related like to feel that you have the choice is to have the agency and that can be seen as willpower whereas willpower doesn't come into play in situations where there's no agency if you just need to sit in the seat on the airplane it doesn't require any willpower agency for that action to occur so I'm just thinking about Surveille's comment and the distributional property like counterfactuals do we engage with them as a possible cosmological reality like in the many world type theories or just cognitive possibilities so they can be engaged with without taking on like a cosmological meaning and then what does that mean about our n equals one trajectory the specifics of our kind of entity in motion and then the the distribution of perhaps different types of counterfactuals or different kinds of possible stochastic rollouts of the system given really similar starting conditions and that's where the will can entail the can I can start out by asking what if but I still have to select the can the which if so that's what again I at some point we can have we can see a relationship of entailment but I think as a as a quantum question they're orthogonal can and will are not one leads necessarily to the other so just to kind of like play on words and loop back around to what I was talking about earlier um like can is an affordance right like so if you have an affordance like like that doesn't mean that that you'll use it right and will like means I will choose or implies that you have some agency right I choose to um you know not start a fight with a flight attendant while I'm sitting in my seat on the airplane or whatever because you can do all kinds of things any any you get many affordances um and I just wonder how this ties into these representational and non-representational aspects of of the FEP like really when I was talking about free will earlier if internal states are a direct representation of external states it seems to me like I kind of feel like there's less options there um whereas internal and external states have coupled dynamics that kind of leaves some like mystical fluff outside like that I can potentially leverage with my own choice or something like that I mean I I don't know what to call it mystical fluff but but um it leaves something unset right like there's a coupling it's like a hidden there's a hidden states there that like are only that that's where I can impose my will well I'm going to put that fluff on a pedestal I'm going to honor the fluff because I do think that there are actual times when the the ability not to be captured by our observations is actually an advantage so mystical may may be uncertain yes and and is and there are there are there advantages to that at times yes so we can call it fluff but I think sometimes we don't pay it enough respect I think that relates to the cloud of unknowing which is perhaps a text if somebody wants to go into a more esoteric avenue but I'd like to return to the figure and ask what either of you thought about figure one how does that play into this conversation we've been having like what does this entity and action partitioning have to do with science is this partitioning unique to active inference or other partitioning is possible does it could it be another way I think this is a pretty stable representation and it and I think it and I think it does honor to the the scale free aspects of what the free energy principle is pointing to in terms of a in terms of a statistical pathway yeah I mean how many debates and uncertainties could be prevented by saying this is a representation of an entity and having people know what was meant by representation right because it's like I made this representation of an entity I think it does this useful thing those points cannot be false they're assertions by a person and what would be disagreed with that it's not a representation someone could say there's a more useful representation or more good true beautiful simple but then let's see it and then we can use science to compare representations because that would be okay right and then you said it does honor to the scale free so the only thing that kind of hints otherwise is the person and the world kind of gives a scale indicator for this model but without that iconography it would be and also adding in perhaps a legend or description of what the edges mean because it does say schematic of reciprocal exchanges and it uses statistical vocabulary but it also uses actual like ontological vocabulary in the terms of actuators so are these the physical things that are doing the actuating or are they the statistical or informational exchanges or is this the representation the form of the scientist model so we're only ever talking about scientists using these models so we're only ever talking about that kind of nexus where we're referring to both loosely and so it shouldn't be interpreted as either of them specifically it's just dialogue there's a lot of ways to interpret what the edges are and they might have similar different labelings and different schemes but it is statistical it is quantifiable it's amenable it's a qualitative topological partitioning that's amenable and beyond to certain quantitative analyses like if you can draw it like that you might be able to do a Bayesian graph and then there's software toolkits and there's implementable algorithms and then there might be some quite interesting guarantees or tendencies with certain architectures for example to perform predictive processing or to have learning anticipation of memory or to engage in control or anticipatory control but that's definitely some of the most core iconography that we've seen I mean 50% of papers or what we'll find out blue have the action loop because what do you what the little side hustle of you blue got going here we're trying we're trying to um read more papers together and learn about them okay but it's interesting how often this has come up and sometimes authors draw on the same image other times they do different and sometimes the arrows are just the clock like just a unidirectional flow other times there are some bi-directional arrows and um the there are several guest streams on the more technical aspects there but um let's return to the general questions anyways anything else to stay on figure one though like just on this like what is this and how does it I think this represents better the the I I fixating on the on the potfly than it does on the physics model and so I think it's until somebody as you said until somebody comes on with a better representation one that that seems to give us the maximum amount of information that we need versus the maximum that we can handle I think this is a this is a good rep great and very satisfying conclusion it yeah go ahead sorry I I really think that there um what's not shown in this representation and what I feel like is always left out is the fact that the there are some states are hidden right and so like where is the hiding here what is hidden from what um in this and so like the sensory states we don't even know if like we perceive I mean we know that we all perceive temperature differently like some people I'm cold in this room but many people would be comfortable and some people like my favorite colors purple my daughter's favorite colors yellow so like we don't perceive things in the same way and so there's there's always something hidden from everyone else and I think that this this schematic kind of leaves that out I've seen the hidden states represented well and other like action perception loops but I think that it's important to recognize like what's obvious and what's not if I could give a one thought there then Dean so I think there's there's two levels of like what are the hidden states from the perspective of the internal states the system of interest the external states are hidden states implicitly or explicitly modeled but they're the ones that are never directly observable they're only observed through their proxy sensory states and enacted or the transition frequencies are modified through policy selection through action so within the model within the tetrahedra the hidden states are relative to the internal states on the other side of the markup blanket and then there's like the implicit tetrahedra which is like us modeling this fourfold partitioning and then everything that's not included as a variable in that statistical model is like implicitly a hidden state but then what is that but then we'd have a different model if it wasn't a hidden state and the other thing I think is that the vertical line between sensory states and action states if you tip that 90 degrees it would look like a blank even though it has two points on the end of it and I think that's that blank is by definition hidden state it's to be filled in so I think I think again you can you can fill in the blank or you can leave it blank and say because it's blank that is by definition hidden so another related point thanks for bringing that up Dean is um there's there's one edge that's not shown here but has featured very importantly in recent discussions on actinth and that's the sigma or the mapping function from number 26 and number 32 a mapping function of internal states to external states now look at axel constant and yellow burners work to think about how the niche is also doing like anticipatory modeling of the individual so the other way but let's just focus on the entity and it's mapping to external states it's predicting external states or at least acting as if or can be modeled as if so there is an edge there that we've seen in modern work it's not drawn here which actually might suggest either the inheritance of a legacy graphical pattern or some preferences or conclusions about what these edges do represent again even though they're unlabeled so that's one edge that we've been hearing more about and then I think another edge that Dean just highlighted is this one directly between sensory and action states and again the clock flow doesn't have this the one that's just internal action external sense that doesn't have the sense action bridge and it depends a lot on what these things mean but it'd be cool to understand what is meant by the backwards connections as well as by the vertical connections and why are there not backwards connections between action and internal and sensory and external it's like it's really a four by four matrix with some connections and that's how we've seen it represented as the sparse connectivity matrix in number 32 and thinking about the way that we model complex systems and generate approximations that sometimes can also retain some other features like the highly operative exponent so some thoughts there blue so we've definitely seen internal states and sensory states like that reciprocality happen but I do I always wonder why like especially since we did the mental action paper like how action states don't influence internal states and so for me like when you take action like when you practice playing the piano that affects your internal states totally like not just through external states or even mental action like where you're doing nothing with external states it's entirely internal and so I always wonder why that action and internal is there but I less so wonder why there's not a reciprocal loop between external states and sensory states like what like I never want the sensory states to go back to the external states for me that doesn't make sense and maybe you guys have an example where it does make sense but I always want the action and internal states to loop together to have it bidirectional here's one possibility on the sense back to external but it depends of course on a specific interpretation of those a photon let's just assume it's out there it hits the retina if we talk about the retina as a sense state using it very loosely not the statistical variable correlating to the sense state but the actual retina it's the absorption of the photon that changes the niche to be sensed like if you're going to be sensing molecules from the ant's cuticle you have to take some you could sense from the ground the ones that have already been deposited but if you want some from the thing you have to take a sample and so that may and whether that would be better understood as being an action in the niche is another question but like you do have to change the external state to be half skin in the game no and I just thought of it as you're saying that like the observer phenomenon right like so there's that too right like just the just the act of sensing it changes it or is the observer phenomena does that come from sense does that come from internal states does it come from action states are the observers just acting in a way that might fit in within an action state but those are really interesting questions Dean no I just think you're right I think both of you we're getting we're getting closer I don't think well it will ever fly directly into the sun but we're getting warmer yeah this uh we'll have another few minutes to talk this has been a really great discussion and I look forward to like re-listening to it and discussing more then of course feel so appreciative that we'll have a .2 as well so we can return to some of our early ones free will maybe save it for the .2 or if we want to talk about it now I think we've blurred and fuzzed some of these lines a little bit so we did kind of get there let's see what else we had down and of course if anyone in the last 30 minutes has any questions they want to write so Stephen wrote I like your method to work backwards in the flow diagram start at representative or non-representative assumptions or perhaps conclusions like starting at those end points as assumptions then looking for prediction errors in our action perception this maintains uncertainty in ways that does not happen when following the causal effective correlations that's sort of like 20 questions is framed as a game of um oh I'm sorry I didn't know that mentioning common children's games was funny um that's framed as a game of going from the most uncertain towards more certainty like if you were asking uninformative questions what are you doing if you're not if you're just getting like and that's information theory kind of exemplified you're reducing the set of what you can be um guessing with the preference to like win within 20 questions and then this is as Stephen is pointing out kind of the other direction but then what does that look like and that is almost inflating uncertainty at least in the information theory sense so here's Stephen's question and then either of you if you have a thought on this do we model the temporal lags in dynamical change between sensory state in relation to changes in action state so maybe a related question would be how do timescales and temporal lags play into our modeling of action perception and cognition this is why I kept saying to young people when in doubt zoom in zoom out and it wasn't just zoom in zoom out in a sort of a um in a visual sense you have to zoom in and zoom out on the temporal sense as well you have to think about the you have to think about the unit of analysis both as a as a as an object but also the as a time of the length of time something can remain stable or not I don't think we I don't think we're naturally inclined to include the time frame as much as we're inclined to include the did it change did it float did it burn whatever whatever it is that we're waiting to see what happens next we don't tend to give as much attention over to how long did it take I'll give a complimentary thought which is more related to the kinds of models that we've specifically seen in act inf so the deep temporal models or sophisticated active inference in for example a three time step model at time step one it's doing inference on one two and three that's planning as inference and anticipation at time step three it can still be calculating the model in light of incoming evidence about time step one two and three which is related to memory and learning and all these other sorts of so-called backwards looking types of inference so there's like anticipation there's now casting and then there's reconstruction of your recent and long term past and how that's related to identity so I think the the question do we or how do we model the temporal or just the temporality of sense and action states the answer from model based science is what's the time horizon parameter on the model that we ran oh it was 15 days okay well then we modeled the temporal lag up to 15 days and that is what we modeled and that is sort of where the uh scale free becomes the scale friendly becomes the scale specific and then that's the the most that it contracts down to is that one specific model as deployed by those researchers here's another question from Stephen when do we sense and when do we act when do our sense states update when does our generative model update and when do our action states update I'll start with the model based science one which is again it would just be the specifics of how the pseudocode and the actual computer code were written which line of code updates first that's the narrow answer but what about a more broader answer does it depend on how open we are how how how how confident confident we are in staying in an uncertain non-specific space and can we can we necessarily know going in how long that will be I don't I don't know I don't know how long and what about precognition right like like if we you know expand to that possibility like like I knew that was gonna happen I mean that happens to all of us right we all experienced that like oh I knew that was gonna happen like if I you know step on this ice cube it's gonna like this ice pile it's gonna crack through to water or whatever like I knew that was gonna happen so so we're able to kind of perceive things before they happen and so I don't know like is there not necessarily the the temporal separation this makes me like think back to like Shana Dobson and like the flattening of the time diamond right like so it's a really like is that is it necessary that there's sequence here or can it just all be happening simultaneously that kind of relates to what I was asking earlier to you like at is there some kind of simultaneous update sensing action possibility and I think that there is I mean I don't think that not not in a computational model because you know code just runs like it is but I think in our model it's possible okay one more general and then more specific comment the 4d spatial model tetrahedral synergetic geometry you have space and time four dimensions spatially represented rather than the xyz you have three spatial dimensions and then there's like the special question of how does t come into play so some is known there and some is not known there but when and where and how are of course very linked and if we take model-based science seriously then they're also very enabled and constrained by the kinds of models we have so maybe it's related to how we think about geometry and then the sense and action are they continuous or are they discrete we've seen recently Ducosta at all like the synthesis of active inference on discrete state spaces and then we've also seen some of Alex Chance's work on the continuous control settings with like the mountain car which is not just the discrete case so it's just interesting to see how some of the theory goes out in front and then there's parallel lines of mathematical development with continuous and discrete modeling like digital and analog modeling and sometimes those types of modeling scientifically are very well reconciled and other times they result in pretty fundamentally different patterns like there are certain equations where the discretized form or the version that's computed without really high precision numbers on a computer has really different behavior than the other kinds of models which gets at the importance of thinking about how the model is is framed and then also applied because like if the precision of the processor matters for what the paper discovers then shouldn't that be a part of the consideration okay here's a few of the other comments and for 3612 so by the way we'll write down any other questions so if anyone in the last few minutes has some questions they can ask and we'll be just writing things down getting excited for .2 and if anybody wants to join who wasn't here today they're also very welcome to how is representation related to auto poesis so refresher auto poetic system is capable of producing and maintaining itself by creating its own parts so how does representation differ or does it differ for a system that is just engineered and secreted into the niche like a computer versus potentially something that is reassembling its parts continually why does that matter if it does is that a thing to look at in .2 daniel that sounds good yeah yeah i'm trying to push it on but i mean the reality is is that we can take each of the of the different segments the organizational the structural the content related and then put them through the auto poetic grinder and see what pops out the other side okay great suggestion apply this to the four columns what other things would be fun to discuss in .2 or what else can go through the four column grinder or what would be any other question that any of you have in that in the in the point to also we can we can spend a little bit of time on some of the some of the comments that were in the section five of the paper specifically around how we face state from potentially they didn't say that this is true but they said if this claim is true we could be faced stating from non-representational to representational maybe maybe could we spend a little bit of time not just looking at the auto poesis aspect of that but how does how does that density form in the in relationship to this idea of modeling ourselves great question here's a question that steven wrote so feel free to give a thought on it or we can push it it was will robotics or neuro phenomenology help most in testing these complex temporal dynamics in a realist way i'm just going to say that's a great question and i think we need the person who asked the question to come on and chat with us about that a little bit we're just delegating so hard here blue go ahead so something that i've brought up i think in the dot zero is like i would really like to tease apart the similarities and differences between like structural aspects and content related aspects i'm not entirely certain like that's a fuzzy blurred line for me there so maybe we could get into that a little bit more i agree it's fuzzy for me too these are awesome questions too thanks for writing it's all blue we'll we have multiple pages we have auto poesis from non rep to non rep to rep and back and forth or whatever else how does that form in the context of self representation i think there's a few hamlet quotes that'll come into play there then steven's question about technology and realism structural and content aspects detached representations in the neurobehavioral systems in the sensory motor loop niche modification stigma g and the digital case extended cognitive systems that include digital and uh mere adaptive systems steven's singing 80 songs in the chat i believe thinking through other minds in multi-scale integration one of sims's other papers how are they related does the fp exist independently or only in the mirror a lot of ways to answer that one um one part we could look into now or maybe in the dot two would be the free expected free energy and variational free energy and how they play different roles in the fp these are also some things like if people are watching live or in the next few days then they can write a comment on it or um get in touch with us or like if they want to submit a question or they can come on and discuss it in the dot two variational and expected free energy there was a lot of parts to the paper oh and action-oriented representation we didn't get to that slide in the dot zero but just what is what is action-oriented representation we've talked about representations today but then did we talk about were we implicitly talking about action-oriented representations are those a subclass how are they different any other yeah blue blue it in our in our like system and in like human cognition there are some um like hard-coded representations that we have um like action-oriented maybe one of them like kinesthetic memory and also like visual recognition systems and it's really interesting that um and perhaps easy to apply the fpp to specific cognitive systems like subsystems but it's really much more difficult when we have like multiple competing processes like we're not just playing the piano and we're not just looking at the piano keys like we're doing both of those things at the same time and lots of other things like we might be you know having some memory of one time that we heard the song that we're playing or something like that so it's really interesting to think about um representation and all the ways that it exists also even in memory right like the memory representation is um is it's like a it's not that's not a hard-coded um thing it's really squishy and it's it's um subject to change through time and you know we might remember a situation like you can tell me your phone number right now and i'll remember it you know in five minutes but in five days no like or something like that so um i don't know it's it's really interesting to think about the fpp and how it can apply in this overarching way um where there are multiple competing goals and priorities and um yeah that one thought i hope i'm not misrepresenting in a empirical result was in fruit fly the drosophila they found that uh short-term memories didn't require protein synthesis like just second to second some sound or some symbol could be associated with a preferred or repulsive outcome but preventing protein synthesis blocked the consolidation of memory and perhaps that's similar to other systems and so there's kind of like the resonant neural structures that are just using the systems connections as they are that's sort of one perhaps faster mode of cognition that's like running around figure two and then there's a restructuring that requires changing the model like structure learning but that also connects to realism because we're talking about the synapses and the neurogenesis like of specific cells that could be maybe measured and then the neurophysiology and the behavior can be observed as well if we have integrative frameworks like actinth so like it also relates to dean's comment about the scales the time scales of change like neural circuits cannot be changing at a time scale faster than neural firing so how does the spiking component or the other sensory transduction or other chemical components how do those overlapping time scales coordinate this reminds me too daniel of something that we talked about like the very first time we met because it was a recent experiment at that time about the memory transfer between snails via injection of RNA so that was like an interesting possible memory storage and I'm not sure what else has you know developed from that but also we've talked about you know ants and what how they leave like they like told me like that they're just throwing up RNA like they're throwing up all the time all over the place and like maybe leaving chemical like RNA traces for each other we talked about that too so maybe I don't know if you want to expand on that or does that tie into this protein synthesis you know thing one thought on that it was a recent ant paper where basically the act of mating which we won't describe on this family friendly discussion directly infuses neurotransmitters into the recipient and that changes their behavior and induces all these other hormonal changes and so the realism of cognition like the embodiment and the physicality of these systems it is perturbable by infusions of neurotransmitters in other neural systems and in us so that's sort of where realism can't be eradicated amidst all of these representations because things do hit our head and molecules do diffuse it's kind of like grounding it and that's I think why it's such an interesting and integrative area with a lot of contention because the 4e side is holding down that real side and then the science side so called is holding down the abstractions and the representations and then cognitive science especially is applied to humans is kind of where it meets in the middle it's like well yeah you're talking about embedded cognition and active cognition but that's your cognition that's modeling and so then you get all kinds of fun viewpoints and ideas because of how mixed and complex that area is thanks blue and dean and those who participated in chat for this really great discussion I hope you can digest it and learn in the next week and reflect on some of these questions and that anyone can ask more questions or start to prepare some thoughts so that they can also be involved because this was really fun all right one question yes statement every one of these live streams I don't know ultimately how it's decided which order that we do them in but the fact that we've gone from Connor's paper to Magie's paper to this one whether it was done with some sort of intent and looking back I'll I'll back to the future or it was just okay well I think we'll we'll do this one next and see what happens I think the order of these last three papers in particular has been pretty just the order in which we've looked at them has generated a whole bunch of interesting ideas including sort of the the idea of in transitivity in the last paper now potentially influencing how we've discussed this paper so again I don't know how much pre-thought goes into that but if there was some pre-thought that went into putting these papers in this particular order it's been very helpful and one other thing if if I find something happy because it's true I'm going to laugh and if I find something happy because it's playing a game like 20 questions or any game I'm going to laugh so if I can't be if I can't enjoy and really appreciate the sophistication of the conversation here I probably wouldn't show up so thanks to both of you you can make me read active papers but you can't make me have fun but just to give it one closing answer information for anybody who wants to like get involved with the discussions here's our coda for the .com's unit and so here's the livestream table the one that gets represented on the viewable public one so we have all the papers planned like in this case up till 37 38 39 40 up through the end of march so we try to plan the papers in advance so that people can have time to read them and commit to collaborating on the .0 background video and then we have papers to potentially discuss and so here people add papers and then they click a heart and dean we just sorted by hearts and then we picked the papers that were on the four top papers and we said okay we're going to read these four for 36 789 um yeah those and then we emailed the authors and so we have different papers and we said well what dates might work because we always will prefer to schedule a paper that we want to read when the authors want to join and so that's kind of our first um like the way that we pull papers out like we said the dates for um the quantum free energy discussions in march and then we work backwards from there and then just try to not do two um not even overlapping papers just try to provide rhythm and novelty and philosophy and technical details so it's sort of a local decision making based upon the votes that people provide who are participating in the .coms unit and pretty random yeah and just what people surface that's good to say that's a pretty serendipitous order then that's fantastic it's fun um and and if anybody watching or listening wants to suggest a paper for the live stream uh please do so yeah all right thank you dean and blue so see you all next week for the .2 bye