Autopoietic Enactivism and the Free Energy Principle - Prof. Friston, Prof Buckley, Dr. Ramstead

This fascinating exchange between leading scholars explored connections and tensions between the Free Energy Principle (FEP) and enactivism. Moderator Tim Scarfe teed up the dialogue by noting critique of FEP from an enactivist paper. Karl Friston, as founder of FEP, was interested to discover whether he is an enactivist. Chris Buckley brought expertise from robotics and physics, while Maxwell Ramstead offered in-depth knowledge of both FEP and enactivism.
Ramstead outlined core enactivist views - embodied cognition, rejecting computationalism, dynamical systems focus. He distinguished "high road" anti-representational versus "low road" more moderate variants. Enactivism emerged from autopoiesis, which emphasizes structural recursion and self-generation of constraints. Buckley shared coming to FEP as a skeptic from robotics and behavior-based AI. He welcomed reconciling operational closure concerns within FEP.
A tenet of enactivism is rejecting information-theoretic explanations. Friston and Ramstead argued information theory inheres in dynamical systems physics. Ramstead contends the split is baseless, stemming from misreading early proposals as mutually exclusive. He laments enactivists’ philosophical insularity and dogmatism against information approaches.
Discussing boundaries, Ramstead asserts FEP’s Markov blanket formalism captures organizational dependencies akin to enactivism’s operational closure. Blankets are flexible, not fixed veils. Generative models likewise represent systems’ relational organization. Friston emphasizes blankets separating and coupling systems. Both internal and external states are integral.
On goals, Buckley advocates an intentional stance - using goal language pragmatically if beneficial. Goals emerge from beliefs about dynamics rather than reward functions. Friston elegantly deduced goal-directed behavior mathematically falling out of FEP in a particular regime. The group explored how systems act “as if” they have goals or models without explicit representations. This helps reconcile enactivist and computational views.
Ramstead repeatedly critiqued enactivists’ commitment to firm divides, like between mind and world. He argues FEP integrates perspectives, dissolving false dichotomies. Friston emphasized FEP’s consistency with other principles like maximum entropy, foreseeing links with relational quantum physics.
Overall, the conversation was constructive and conciliatory in intent. All parties agreed on seeking compatibility and community between research programs. However, Ramstead levied significant critique of enactivists’ philosophical assumptions and resistance to information theory. While appreciating the spirit of engagement, he contends some differences originate from enactivist misunderstandings. The discussion revealed a complex intermixing of resonance and tension between these leading approaches to cognition.
Prof. Karl Friston - Inventor of the free energy principle scholar.google.com/citations?...
Prof. Chris Buckley - Professor of Neural Computation at Sussex University scholar.google.co.uk/citation...
Dr. Maxwell Ramstead - Director of Research at VERSES scholar.google.ca/citations?u...
We address critique in this paper:
Laying down a forking path: Tensions between enaction and the free energy principle (Ezequiel A. Di Paolo, Evan Thompson, Randall D. Beere)
philosophymindscience.org/ind...
Other refs:
Multiscale integration: beyond internalism and externalism (Maxwell J D Ramstead)
pubmed.ncbi.nlm.nih.gov/33627...
MLST panel: Dr. Tim Scarfe and Dr. Keith Duggar
Pod: podcasters.spotify.com/pod/sh...
TOC:
00:00:00 - Introduction & Participants' Backgrounds
00:04:01 - Core Views of Enactivism
00:15:02 - Dynamics vs Information Theory
00:22:20 - Concept of Operational Closure
00:30:00 - Good Regulator Theorem
00:40:00 - Role of Intentionality
00:51:31 - FEP & Ecological Psychology
01:00:00 - Goals in FEP
01:10:00 - Emergence of Goals
01:20:00 - Importance of Intentional Stance
01:31:15 - Future of FEP
01:40:00 - Observer Dependence in FEP
01:50:00 - Metrological Aspects of FEP

[@MachineLearningStreetTalk]

Top Quotes! (times are estimates) "The important thing to point out is that just definitionally, the Markov blanket comprises the set of degrees of freedom that separate and couple systems together. It's it's it's not the the veil, it's the interface." - Maxwell Ramstead, 00:17:38 "You can think of the state based description as sort of the asymptotic limit of the path based description, the 2 are kind of like deeply connected, in in some sense." - Maxwell Ramstead, 00:31:52 "The free energy principle is a model. ... and it's a map, and it turns out to be the canonical way of modeling systems that are engaged in modeling." - Maxwell Ramstead, 01:25:38 "I would, you know, distinguish ontology from metaphysics." - Maxwell Ramstead to Tim Scarfe, 01:23:13 "I think there's a kind of there is a or a message is to take from an activism on the low road, which are super important for cognition, right, without kinda going into the depths of the, or, autopoiesis." - Chris Buckley, 00:08:56 "What is operational closure? That would be quite nice." - Karl Friston, 00:18:38 "The goal ness, if you like, is is, inherent in the characteristics of the particle or person that you're trying to describe." - Karl Friston, 01:18:38 "I mean, I would agree that we shouldn't conflate them, And I would, you know, add that we don't. There are distinct constructs that map onto these things." - Maxwell Ramstead, 00:30:28 "I drink a lot of coffee. And if you were to sample me at random during the day, I would say there's a 1 in 25 chance." - Maxwell Ramstead, 00:36:58 "And I think that that, you know, there there is a lot of, if you like, work to be done from the point of view of free energy principle theorists in consolidating the links to what has gone in terms of things like maximum principle, and what has to come in terms of, say, quantum blue gravity or Chris fields, quantum information, theoretic treatment of holographic screens in the Markov blanket." - Karl Friston, 01:33:45 "The main difference is just, an enacted ecological psychologist, sorry, would say that we perceive things directly, but then they appeal to this, these notions of, like, resonance, and whatever, which, you know, to us is just sending in for inference." - Maxwell Ramstead, 01:05:06 "I wanted to challenge it and then eventually fell in love with it and now think it's the right way to go for it for too. So, yeah, I came as a skeptic, you know, and then I'll be bashed into into into place." - Chris Buckley, 00:03:12 "This isn't to, you know, bash anyone specifically, but, like, this is a complete nonstarter mathematically." - Maxwell Ramstead on enactivist critiques, 00:36:58 "We're talking about the same thing. We're talking about the way that, like, a set of structured dependencies generates a thing that's able to sustain itself over time." - Maxwell Ramstead, 00:28:48 "The free energy principle is a principle. ... that leads to a modeling method. And, ..we make some assumptions, and we're able to say some things, but the the principle itself doesn't rest on on any of these assumptions" - Maxwell Ramstead, 00:31:52 "Systems at different scales will behave as if they had a model, even if they don't necessarily have a model in the sense of a person" - Keith Duggar, 00:39:21 "The free energy principle just is that. It is just a constrained maximum caliber principle or maximum entry principle." - Karl Friston, 01:31:08 "It looks like she has goals. It looks like she plans. And then a few years later, if I develop a sense of self, well, perhaps I'm a thing like mum, and that perhaps I have goals" - Karl Friston, 01:18:38 "The free energy principle is a map of the boundaries you can think of. And and a map of what happens when things have boundaries." - Maxwell Ramstead, 01:25:20 "The enactivists need there to be a bright line in the universe, like a bright red line." - Maxwell Ramstead, 00:58:55 "I don't really understand the nature of this fork" - Karl Friston, 00:15:02 (Dynamical vs Information Processing) "It's always been about counterfactual futures. It's and it's difficult to argue that, like, you don't account for historicity when literally the core construct that you're deploying is, like, probabilities of different histories" - Maxwell Ramstead, 00:36:58 "I think that kind of particle starts to now show the the deep intention, you know, it looks as if they are in they are doing this because they intend some outcome." - Karl Friston, 01:18:38 "The free energy principle is a theory of observer ness, you know, the the sort of the it from bit Right? Like, the it it's it's thing this, but from another perspective, it's observance, which which are kind of the same thing." - Maxwell Ramstead, 01:28:21 "The free energy principle started off as a path based formulation. I think this is absolutely underappreciated in the literature." - Maxwell Ramstead, 00:36:58 "This isn't saying like once and for all, these boundaries are here and crystallized and just, you know, are are are there independently of, you know, the way that we are are experiencing or sampling or interacting with the world." - Maxwell Ramstead, 01:23:13 "I wanted to understand just enough to be able to to know for sure that I don't need to understand this, actually." - Maxwell Ramstead on originally approaching FEP, 00:55:21 "The enactivists in in the auto poietic, tradition really want to go hard on this and say, well, what you really need to do is map a set of constraints such that the system kind of self generates." - Maxwell Ramstead, 00:28:48 "I mean, I was frightened. You were going to ask me, well, you know, if I had anything to add about in enactivism, because I haven't. I'm here to find out whether I'm an an enactivist or not!" - Karl Friston, 00:12:35 "This, split between dynamics and information theory is mostly in people's imaginations, there's no mathematical basis for it." - Maxwell Ramstead, 00:31:08 "I mean, all interesting information theory read as probability theory. That is used, in physics, inherits from dynamical systems." - Karl Friston, 00:15:02 "I mean, I think, you know, some really nice dialogue between people in activism and what we do in the act in active inference and the free energy principle." - Chris Buckley, 00:11:01 "The free energy principle at core is a metrological statement, where metrological means measure theoretic or relating to measure" - Maxwell Ramstead paraphrasing Karl Friston, 01:28:21 "You have, this interesting situation of these very precise dynamics and they're precise just because we are big." - Karl Friston, 01:18:38

[@Pianoblook]

Hearing Professor Friston speak always reminds me just how incredible a gift intelligence can be (and how relatively little I seem to have 😅) Excellent discussion, thanks!

[@sharif1306]

If you are here and can comprehend his intelligence then you must be fairly intelligent yourself.

[@tuckerbugeater]

Nope

[@smurp_com]

Homeorhesis - a tendency to return to a direction of flow after a perturbation. It is a generalization of homeostasis to a nonstatic system 39:00

[@BrianMosleyUK]

Really enjoying this, although will take a couple of watches to grasp 😂

[@michaelwangCH]

Excellent talk and deeply satisfied.

[@kd192]

This is the best discussion with Karl regarding the free energy principle, I love his explanation about non-equilibrium systems being an interaction between sinoidal and dissipative flows. Thanks for posting this interview!

[@dr.mikeybee]

I asked Bard for information on the Free Energy Principle. He's Bard's response: The free energy principle is a theoretical framework that suggests that the brain minimizes surprise or uncertainty by making predictions based on internal models and updating them using sensory input. It highlights the brain's objective of aligning its internal model with the external world to enhance prediction accuracy. This principle integrates Bayesian inference with active inference, where actions are guided by predictions and sensory feedback refines them. It has wide-ranging implications for comprehending brain function, perception, and action. More formally, the free energy principle says that if a system has a "particular partition" (i.e., into particles, with their Markov blankets), then subsets of that system will track the statistical structure of other subsets (which are known as internal and external states or paths of a system). The free energy principle is based on the Bayesian idea of the brain as an "inference engine." Under the free energy principle, systems pursue paths of least surprise, or equivalently, minimize the difference between predictions based on their model of the world and their sense and associated perception. The free energy principle has been used to explain a wide range of phenomena, including: * Perception: The free energy principle can explain how the brain perceives the world by making predictions about sensory input and updating those predictions based on the sensory feedback it receives. * Action: The free energy principle can explain how the brain generates actions by selecting actions that are predicted to minimize surprise. * Learning: The free energy principle can explain how the brain learns by updating its internal model of the world to reduce surprise. * Consciousness: Some researchers have suggested that the free energy principle may be able to explain consciousness as a state of minimal surprise. The free energy principle is a powerful theoretical framework that has the potential to revolutionize our understanding of the brain and mind. However, it is still under development, and more research is needed to test its implications. Here is an example of how the free energy principle might apply to perception: Imagine that you are walking down the street and you see a red ball. Your brain makes a prediction about what you are seeing, based on its internal model of the world. This prediction is that you are seeing a red ball. Your brain then compares this prediction to the sensory input it receives from your eyes. If the prediction matches the sensory input, then your brain is not surprised. However, if the prediction does not match the sensory input, then your brain is surprised. In this case, your brain will update its internal model of the world to reduce surprise. It will do this by increasing the probability of the hypothesis that you are seeing a red ball. This will make it more likely that your brain will predict correctly the next time you see a red ball. The free energy principle is a complex and nuanced topic, but it is a powerful framework for understanding how the brain works.

[@stavroskarageorgis4804]

I'm happy to discover that what appear to be seriously "bad faith" pseudo-debates are a feature of much more logically/mathematically rigorous fields than those in the "social and cultural sciences/disciplines".

[@ElleDyson]

I can simply say that this video connected a lot of dots that had been bouncing around in my head, and added many others to ping pong around. I have been back to this video on a few occasions to reflect and realign. TY!

[@_ARCATEC_]

autopoetic autopoiesis of the autopoietic

[@MachineLearningStreetTalk]

Transcript and summary here - docs.google.com/document/d/1aDVM-WD1UvQ5BZINYcLrBxMqei1x2Z8adcZQfd-phj4/edit?usp=sharing

[@reinhardts-entropica]

Somewhere Prof. Friston wrote when he was young, he wanted to become a blend of Albert Einstein and Sherlock Holmes, and it seems to me he somewhat succeeded.

[@aaronclarke1434]

Source for anyone interested: Said on Lex Fridman’s podcast. Final part in answer to the question on the meaning of life.

[@loganmcphail4866]

Quick question, Ramstead, Kirchoff and Friston published a paper in 2020 arguing for an enactive, non-representational, interpretation of the generative and recognition models. Where is the gap in my understanding? It seems like Friston was arguing for an enactive inference in that paper but it more uncommitted here. The paper was called "a tale of two densities" for reference

[@neurojitsu]

I particularly found helpful Maxwell's pointing to the importance of distinguishing ontology from metaphysics... "ontological boundaries means capturing features of the math; metaphorically, a map that behaves like a map (as opposed to/versus metaphysics)... all of this is a modelling approach, a physics, not a statement of reality" (typed from notes, may not be word for word). What struck me was how awesome the human mind is, that it can create its own maps from modelling experience, given that - as Chomsky said in a previous mlst - we can't introspect our way to understanding the underlying structures of our own minds. The wonder of this power of observation is I think why FEP holds such fascination for me. Thank you for hosting a thoroughly enjoyable and enlightening discussion. Lots to digest from this conversation.

[@stephensillett9020]

Do you think ontological boundary makes sense in terms of defining things in the normal way of saying this thing that is stable over time should be called [X]. However, the blankets within the modelling process, seems more of an axiological threshold. i.e. a sense of what is valuable informationally, between random states and states that impact the generative model i.e. The dynamical differences that make the difference to inference, and action.

[@neurojitsu]

@@stephensillett9020 Well, I completely agree that that's the key question! I don't have a definitive thought about this, but let me comment from my current state of understanding. Firstly on, "Do you think ontological boundary makes sense" and "in terms of defining things in the normal way of saying this thing that is stable over time should be called X"... my first thought is that we 'make sense' both with incoming experience and prior knowledge, and I agree that values are at the heart of navigating the as-yet unknown. I'm not sure, however, the part that categories in language play versus embodied knowledge when "exploring" the unkown (eg bodily proprioception, or the sort of perspiration that the nervous system triggers from anxiety but before we are consciously aware of our threat response), and nor is it clear to me therefore the importance of 'naming' something for navigating the unexpected and novel. Neuroscientist Immordino-Yang claims, in fact, that we use emotions to navigate, BEFORE we use any cognition to formulate concepts/understanding (emotions therefore being our means of sensing value - you can look up her paper about the "Iowa Gambling Task" if interested). Value, it seems to me, entails a different sort of processing from analytical computation, and so value procesing is not of the language/rational processing sort. We can 'know' things from direct experience before we are able to name them. The knowing is a matter of memory, but the naming is one of categorisation, and the action planning is a matter of relevance realisation. Names anyway are just a convenient shorthand for communicating between agents, they are not "thinking" as I understand it. That is, names don't substitute for knowing a thing first hand: if we're navigating in a forrest, we need to know specific trees, rivers, mountain range profiles, paths as reference points, and so concepts are not so useful for navigation beyond the mere recognition of a thing as a thing (ie any such thing). In other words, thinking is embodied and enacted, invoking memory and using it in the present as we explore: meaning only exists "in relation" to a world (ie the enactive cognition perspective, though I realise di Paolo, Thompson etc have recently made claims about the misunderstandings amongst FEP advocates about the enactive view). "A sense of what is valuable informationally" is similarly not so clear-cut as it seems, for the reason already stated: value judgments are not computational (ie they need more than just information processing), so much as analogical in the Hoffstadter sense of the term. As he says, insights just happen. Vervaeke, a neuroscientist who studies wisdom, suggests that relevance realisation is a self-organising system. He uses the analogy of a falling stream of sand forming a pile: at some point, the growing pile becomes unstable, collapses and a wider base is formed on which the pile can continue to grow. Again, it seems to me that self-organisation necessitates some form of enactive explanation for value judgments, because they are inherently practical judgments embedded in a setting. The setting makes demands, as much as we make sense. Returning to the video then, and the distinction between ontology (maps behaving as maps) and metaphysical statements of reality: values, it seems to me, only come into play when we have contact with reality. Names/categories are maps, values are the navigation mechanism.

[@petermartin5030]

Consciousness is a process running in the brain. It takes input from external processes in order to track and predict what they will do. It generates outputs with the aim of being able to survive and replicate. This process becomes aware when it takes input summarising its own state, and tracks and predicts what itself will do - it recognises itself as a process. This includes measures of how well it is surviving and reproducing. Free will results from its outputs being available to it to inform future processing. 'Free' here means attributable to conscious decision-making, so that outcomes can drive learning that will improve future decision-making.

[@raminsafizadeh]

Fantastic discussion, all within the concept’s own blanket!

[@oncedidactic]

Absolute feast ahead of us here, wow. With all intent to savor will certainly be scarfed! buh duh tss ;D

[@UnderstandingSystems]

One of the very best discussions I've heard! Truly phenomenal and tremendously clarifying

[@LukePluto]

Autopoietic*

[@MachineLearningStreetTalk]

THanks

[@stavroskarageorgis4804]

How the heck was M&V's autopoiesis "mechanistic"?

[@dfearo]

If only I could have heard this discussion 20 years ago when I was trapped in the autopoiesis rabbit hole 😢

[@_ARCATEC_]

💓

[@margrietoregan828]

1:27:03 This is just a cohesive modeling approach that inherits from the 1:27:08 physics of information and self-organization. I would quickly just point out that the these things seem 1:27:18 intrinsically related to observer dependance because they are in some deep sense Carl's 2019 monograph, particular free energy principle for 1:27:28 a particular physics ends on this idea that the free energy principle at core is a metrological statement where metrological means measure 1:27:37 theoretic or relating to measurement. And it basically says, well, things that exist look as if they're constantly measuring 1:27:44 themselves and like keeping themselves within certain bounds. So this is and, you know, this is even clearer from the quantum 1:27:52 information theoretic formulation where literally what we're saying is that the free energy principle is a theory of observer ness. 1:28:00 You know, the sort of the it from bit, right? Like it's thingness. But from another perspective, it's observer ness, which which 1:28:07 are kind of the same thing there was a link in terms of were you maybe leading into the metrological aspects, 1:31:12 the self measurement, that that was really important? Yes, that's right. So just looking backwards, um, you know, what is the the 1:31:22 sibling of the Or. Yeah. The sibling of the free energy 1:31:27 principle of in the past century. In the past few decades. 1:31:34 I think that would be the maximum entropy principle constrained maximum entropy principle looking forwards. What would be the bedfellows of 1:31:42 the free energy principle? And I think I think that the that metrological or relational, um, thing is vitally important. 1:31:51 And you know, I love the work of, um, Carlo Rovelli and his work with, 1:31:58 you know, quantum relational approaches to quantum mechanics, 1:32:04 which is all about relationships. And of course, how do you how do you quantify a relationship? Well, it's just in terms of the 1:32:12 interaction, the coupling, which is going to be a sparse coupling. It is just a measurement. It is just an observation. So inference, measurement, synchronization, sparse coupling, 1:32:22 they're all the same thing. And so I think that what will hopefully happen is that somebody will write down a math, possibly 1:32:29 Dalton will write down a calculus that says all of these things are just the same way of looking at it. And I read yesterday because I had to 1:32:38 because I was reviewing this paper for synthesis about Ontic structural realism. Ontic structural realism. And it seems to me that that is 1:32:47 basically philosophers having discovered the same underlying fabric. The reality is in the measurement, 1:32:55 but it's the measurement of how we relate to each other via that measurement of of each other or things measuring other things purely 1:33:05 in terms of their relationships. So I, I think that, you know, there is a lot of, if you like, work to be done from the point of 1:33:14 view of the free energy principle t 1:33:26 Chris Fields quantum information theoretic treatment of holographic screens in the Markov blanket.

[@margrietoregan828]

((Copy of notes for my own personal notes :- )) 7:00 interactive kind of mode of explanation. So yeah, the Enactive approach is basically committed to this idea 7:08 fundamentally that cognition is basically bringing about a world, a structured world of meaning and significance through 7:18 adaptive appropriate actions left in the world. So the, the slogan is, you know, yeah, exactly. 7:28 Constructing this meaning, you know, laying down a path and walking, 7:33 I think is the slogan, the poetic slogan form that's so popular. Um, yeah. And in in its more radical 7:42 inclinations, this enactive approach says dynamical systems is really the 7:48 way to study systems that have mind. And so the kind of the background 7:55 assumption there is that there is a fork in the in the road somewhere, you know, where you have to choose between 8:01 information theory and dynamics. And yeah, they obviously go down 8:06 the dynamics path... this enactive approach says dynamical systems is really the 7:48 way to study systems that have mind. And so the kind of the background 7:55 assumption there is that there is a fork in the in the road somewhere, you know, where you have to choose between 8:01 information theory and dynamics... 13:14 Put very simply, the free energy principle basically will accommodate both perspectives. It rests upon an externalist view of 13:22 the universe as a random dynamical system that just by equipping 13:28 that system with a partition that separates something from everything else then induces an internalist explanation for the way that thing 13:38 makes sense of its universe. So there is a necessary enmeshing 13:44 of both the Externalist and the Internalist perspectives. You know, put it another way the Internalist perspective is 13:51 entirely licensed in terms of the interpretation of self-organization 13:59 as a process of inference. And you know, with that an 14:04 explicit commitment to a certain kind of representationalism. But that only inherits from, you know, where you start from, 23:47 But, you know, so I'm just aware of the the trickiness in defining operational closure that's happened in the literature in the last 20 23:54 years and really in an inability to kind of really get a concrete quantitative description of what we mean by operational closure. 24:01 And so, you know, I guess it's to say that we've simply solved it within the free energy principle that might that might 24:09 have put some backs up somewhere. But, you know, Maxwell maybe can give you a bit more insight into this because he's thought deeply 24:16 about how to define operational closure within the fringes. 25:47 just put the terms on the table. A generative model, right, is our model of the dependencies that link the states and parameters 25:58 that make up a given system. Right? So it is our way of talking about basically the causal connective tissue or structure 26:06 that underwrites a given system. And what the free energy principle says is if the underlying system is disconnected 26:13 in a certain appropriate way, then you get this tracking behavior. 26:19 And formally speaking, it says if there's a Markov blanket in this dependency structure, right, if conditioned on certain 26:27 blanket states, then it looks as if subsets of the system are independent from each other, but also tracking each other. 26:33 Then, you know, you get this interesting behavior like the if there is this Markov blanket is if there is this 26:40 independence structure, then you will get this tracking behavior. That's what the free energy principle is talking about. So, of course, it's not sufficient to say that things are 26:49 carved out in a specific way. And therefore, you know, you can say interesting things about them. But I think what's missed in the 26:57 argument and this has to do with like, I think the the the the 27:03 I'll say provocatively, the dogmatic commitment to anti representationalism is like when the inactivists regenerative model, 27:11 they mean a model in my head that I carry around. They hear like a representation. Right that but what's really at 27:19 stake is the generative model is the dependency structure, right? So if you have the right kind of dependency structure that 27:27 generates a Markov blanket, I would say that you have what the Inactivists call operational closure 28:41 So we're talking about the same thing. We're talking about the way that like a set of structured dependencies 28:47 generates a thing that's able to sustain itself over time. And as I was saying, this this sustained over time is Operational closure in FEP 28:54 a way of describing the interface or coupling between the system and its environment. So that just mathematically we're 29:01 talking about the same thing

[@mchlud4818]

Anyone can please recommend any good sources / articles on structural learning? This podcast is great, just dont think it really touched the structural learning part. Thanks

[@MachineLearningStreetTalk]

Sorry about that, we did intend to discuss that (including the criticism in this paper link.springer.com/epdf/10.1007/s42113-022-00131-8 ) and got carried away with the other stuff. We can revisit another time. Feel free to drop in our discord to discuss more

[@mchlud4818]

@@MachineLearningStreetTalk Thanks, will do!

[@_ARCATEC_]

16:16 ✨ 😁👍✨

[@sharif1306]

@MachineLearningStreetTalk Really enjoyed this discussion. But for the sake of your audience why spend more than 1 hr on some obscure philosophical point of enactivism vs information when you could be discussing structure learning and practical implementation matters such as approximating the posteriors in a high dimensional generative model. Thanks.

[@MachineLearningStreetTalk]

Sorry about that, we did intend to ground this discussion more. There is a bit of a theoretical bias on MLST in general - try checking out the active inference YT channel kzbin.info

[@1330m]

Modern Manna = Incorruptible milk by Huh kyung young Try it.

[@jordan13589]

‘twas gray when I was Colorblind, but now I see red. for the bed is shorter than that a man can stretch himself on it: and the covering narrower than that he can wrap himself in it. Really ClAUStrophobic.

[@_ARCATEC_]

External and internal Space form a Continuity over the Unity of Mass. •X(s zc q(mu ) ZC ( UM)Q zc S)Y•