Free Energy Principle - Karl Friston

Neuroscientist Karl Friston from UCL on the Markov blanket, Bayesian model evidence, and different global brain theories.
Read the text version here: serious-science.org/free-ener...
Get a bigger picture:
1. Dynamic Causal Modelling - Karl Friston: • Dynamic Causal Modelli...
2. Spatial Navigation - Neil Burgess: • Spatial Navigation - N...
3. Cognitive-Enhancing Drugs - Barbara Sahakian: • Cognitive-Enhancing Dr...
Follow us:
LinkedIn: /
Twitter: / scienceserious
FB: / serious.scie. .
Instagram: / serious.sci. .
Support us on Patreon to see more videos: / seriousscience

[@brainboyben]

Great talk. Friston managed to communicate his very complicated ideas (that leave many academics who work on these problems scratching their heads in confusing) in a very simple manner without insulting the viewer's intelligence. On a side note I found his mix up of the dorsal (where) and ventral (what) streams super reassuring. While it was inconsequential for the narrative he was trying to sell, even the greats can make "simple" mistakes and it doesn't affect how we view their intelligence.

[@luckyyuri]

What a brilliant mind. He's a great explainer too, i was blown away from his article on Aeon.

[@BayesianBrain]

I'm in my last year of college and I recently heard about this. It's relieving to see all the classes that I've taken can finally been sum totaled into a singular concept

[@floortjebouwkamp5282]

Always an inspiring speaker. Just to clarify; at 11:30 the ventral ('what') and dorsal ('where') stream are reversed.

[@z-beeblebrox]

@@msbutterflyz well, he knows what they are, just not where :P

[@mmartinezsaito]

@@z-beeblebrox no, it's called "making a mistake"

[@z-beeblebrox]

@@mmartinezsaito whoosh

[@plafar7887]

Yep, noticed it too. Had to go check to make sure it wasn't me confusing things.

[@klammer75]

Was just going to mention that....amazing talk despite the minor error...it’s the dorsal vs the ventral stream of optic radiation....

[@EgorAnanyev]

An excellent explanation! Usually, he doesn't put so much effort into trying to explain these concepts in simple terms, which makes his articles very hard to read and very heavy on the math. But this explanation is quite good and succinct. Caveat: he mixes up the dorsal (top, "where") and ventral (bottom, "what") streams.

[@iranjackheelson]

what math backgrounds would you say are the most important in understanding his stuff from technical level?

[@biancagoncalvesflores9845]

This is amazing, I feel like I need to return to it after I finish my neuroscience degree and build on my current understanding of it. It also explains so much as to how people see, experience and understand everything in such different ways; same statistical process, different input variables. Literally amazing.

[@HeduAI]

If I understood him right, anything that does not dissipate can be considered living? Planets, stars, rocks, stones are also entities on their own that do not dissipate. Can they be considered "alive"?

[@margrietoregan828]

9:00 a corollary or a consequence of any 9:03 system that doesn't dissipate it looks 9:06 as if it has to behave as if it is 9:10 maximizing actively soliciting 9:13 information from the environment and 9:15 modeling that information as a model of 9:18 the environment to maximize the evidence 9:20 for its own existence

[@DarkNinja-24]

Great principle that unifies many seemingly disparate areas of neuroscience research!

[@mihaianton1386]

So sense and sensibility.

[@lukasberglund1707]

Awesome video. I wouldn't say it's very clearly explained though. As someone who hadn't heard of the Free Energy Principle before seeing this, there are a lot of aspects that are still extremely confusing to me. Great stuff.

[@semidemiurge]

The concept of "modeling" needs to be fleshed-out here

[@rijnder333]

Anyone knows a good article by Friston that sums up most of the arguments presented here?

[@Tracks777]

Great stuff :)

[@davideaston6944]

Okay. Who knew?

[@idankoren]

Pure gold.

[@khalidmf131]

it really catches u off guard when your academic background is solely taking loads of drugs and listening to entities explain seemingly random ideas. then finding out it's all based in reality

[@margrietoregan828]

let's now go back to 7:47 the Markov blanket that comprises the 7:49 active and sensory states and then and 7:52 the internal states that are encompassed 7:55 by the Markov blanket the law the rule 7:57 which says that all of the states 8:00 measurement must maximize model evidence 8:03 which is also known as marginal 8:06 likelihood that is also an inverse upper 8:10 bounded by free energy hence the free 8:12 energy principle all of those states 8:13 have to maximize marginal likelihood or 8:16 minimize for energy including action 8:19 that means action and sensations and the 8:22 internal states are all doing the same 8:24 thing which means that we can understand 8:27 the internal states of the brain as 8:30 modeling the world because they are 8:32 maximized in the Bayesian model evidence 8:34 for me or a model of the world at the 8:38 same time my action is also trying to 8:41 maximize the evidence for my model of 8:43 the world so put very simply almost by 8:46 definition I am in the game of garnering 8:52 information that maximizes the evidence 8:55 why my own existence and that's 8:57 basically the free energy principle it's 9:00 a corollary or a consequence of any 9:03 system that doesn't dissipate it looks 9:06 as if it has to behave as if it is 9:10 maximizing actively soliciting 9:13 information from the environment and 9:15 modeling that information as a model of 9:18 the environment to maximize the evidence 9:20 for its own existence and that's where 9:22 we started with the long history of the 9:26 Helmholtz this notion of unconscious 9:28 inference right through to modern day 9:31 machine learning formulatio

[@thewiseturtle]

So, when he uses the term "Markov blanket" he's just talking about categories/sets, right? As in the collection of things inside a given boundary vs. the collection of things outside that boundary.

[@manup8426]

The office is so old school and charming

[@humanbeing6933]

Mark Solms references free energy extensively in his book ‘The Hidden Spring, A journey to the source of consciousness’.. thought I’d understand it better by watching the man himself Karl Friston. Not sure I’m any further forward after that lol

[@danyalakarca7945]

This is excellent

[@kennethgarcia25]

The free energy principle relates to other ideas involving spontaneous organization like self-organizing criticality. These are ways that complex material systems of any scale achieve higher levels of organization by giving up some level of their kinetic energy in the face of intrinsic influences embedded within material at different scales. Thus, the higher level achieves these emergent properties capable of stabilizing a higher degree of organization. Unfortunately, simple Bayesian models are inherently excessively linear/uni dimensional to sufficiently capture the essential multidimensional, non-linear dynamics that provides for emergence of properties that become the organizing influence of the subsequent level of organization.

[@fabiomarson176]

In some points, this basically is "Karl Friston explaining Neon Genesis Evangelion"

[@paddyboylan2523]

@fabiomarson That was *exactly* my first though :) "Ohhhh, so the AT fields where Markov Blankets." I'm imagining Karl, several inches from the tv, frantically taking down notes at 2am...

[@103SideProjects]

Thank you.

[@margrietoregan828]

a world that has some deep hierarchical 10:09 structure in which there is action at a 10:12 distance for example you know so that 10:15 the color of objects around me is 10:18 determined by the incident light as it 10:21 comes almost instantaneously to my eye 10:23 or a falling body is caused by gravity 10:26 then my brain must recapitulate that 10:30 causal structure and of course it does 10:32 the very fact we have nerve cells with

[@keturison]

Excellent! I can see connections to both David Bohm, C.S. Peirce and Gilles Deleuze. Even Edwina Taborsky. Friston's theory of free-energy can be applied into semiotics as a theory of information as well.

[@jasonattar6330]

could you give me an example?

[@denistremblay8029]

the reason is simple ...The basis of Friston perspective is the polarity concept...A polarity can be described in various way ,even geometrical, it is a dynamic object...A duality is only a set of 2 objects....A markov blanket is a polarity....

[@danielderome7312]

Could someone tell me what he is saying exactly at 4'15"?

[@LightGreen5013]

I'm absolutely in awe. Question though - Couldn't a living system be physically disjoint? Would the Free Energy principle still be valid?

[@boriswied6134]

Not in the way we currently conceptualize them i think. Going to the boring old defining traits... (motility, metabolism, reproduction, etc...) it's hard to see how you share a process of movement or metabolism as one whole, without being physically connected, right? You could abstractly and theoretically replace notions of connectedness with mass with a non-massive phenomenon (light) constituting the communicaiton between two points - but then we are just challenging what constitutes "physical" connectedness. You could choose to define two animals (perhaps a fetus and a mother) to be part of the same "living system" and move then far apart, but then we are just semantically toying with the definitions of being "a(one) living system". To a large degree however, this idea isn't even connected directly to old biology definitons of life as i mentioned. Really, much of this theory should be seen as erecting definitions at the outset. On an abstract level you could define a living system as anything you want, and at some point have it not be compatible with the free energy principle.

[@LightGreen5013]

@@boriswied6134 Yes that makes sense. I wonder how/if free energy can be applied to ideas around emergence.

[@ibperson7765]

Just one note, the structure of the nervous system is not modeling quantum action at a distance. Quantum entanglement is not remotely the same as physical activation via electrical impulse to an actuator (like a muscle).

[@paulwolf3302]

A lot of this made me feel the same way. What exactly does he mean by referring to the nervous system having a discrete set of "states?" In what sense is a living cell in a mathematical state?

[@ibperson7765]

@@paulwolf3302 Yes thats another good point. And ad importantly.. claiming it’s state is “not alive” if it doesnt then minimize free energy (over whatever the choice variables are). It’s usually hard to be certain about criticism without knowing the actual model but Im pretty sure there are unresolvable comments made here.

[@payam-bagheri]

He also sounded to me to be claiming that even gravity is action at a distance, which is not.

[@ibperson7765]

@@payam-bagheri i believe he’s been good in other research. Maybe physics is the issue

[@THash-qs5qg]

He's completely wrong about the "What and Where" streams in the brain. At 11:20 he talks about this. The "What" stream is in our temporal lobes and the "Where" stream is in the dorsal area. Surprised a smart person like him got this totally reversed.

[@jimmybolton8473]

More amazing than the first time i seen this thats for sure..😟

[@anthonybrett]

So in other words, from a physics perspective, this is the second law of thermal dynamics (entropy) in action. The free energy principal is the containment of information?

[@lemonsys]

Maybe this is a weird question but would an atom be modelling its world?? Any system that endures/doesn’t dissipate? Or perhaps molecules? I’m especially thinking noble gases because these are relatively isolated/non-interacting.

[@paulwolf3302]

Reductio ad absurdum. That's a technique of taking an argument to its limit to see if it still holds up. If you end up with something absurd, like atoms having brains, you've disproven the argument.

[@lemonsys]

@@paulwolf3302 i mean, the mathematics shows that a droplet of water “represents its world” in terms of the dynamics of staying a droplet - is that an absurd conclusion?

[@michalchik]

He swapped the what and where streams anatomical locations. Two-streams hypothesis - Wikipedia

[@malcolmvargas45]

Is it safe to say that maximizing the model of the world, minimizing free-energy is driven by the anatomy of the systems tendency to increased efficiency?

[@MartinLichtblau]

Living beings have the aspiration to increase their efficiency with regards to their aims.

[@MartinLichtblau]

Markov blanket as seperation sounds just like Luhmans systems theory.

[@ewaldneumann8665]

Why in the world did he get the what and where systems exactly reversed with regard to the dorsal and ventral streams?

[@Juka161]

There is a branch of science is called statistical physics, some correlation here?

[@MrCatalincorj]

Your creapy program on vshared not working... is getting blocked at question: Which Best Describes You?

[@timkbirchico8542]

So, you're in a room, the room has objects within, maybe a sofa, a table, a computer and an organic observer of itself and the external. What is the fundamental difference between these objects? Is the observer a complex combination of other molecules present outside the observer? As in matter observing itself.

[@___Truth___]

Autonomous Movement, Delineation, & Structure: the organic observer unlike the table, sofa, etc., can move autonomously in and of itself within an environment while delineating between itself and the other objects in the environment. Think about the difference between a drop of oil in a clear bowl of water vs. a tadpole in a clear bowl of water, even though the drop of oil is delineated from the water in the environment , the drop of oil cannot move autonomously in and of itself within that environment because it lacks the structures to do so, but a tadpole can move autonomously within that environment and also delineate itself with what's in the environment because it has the necessary structure to do so. The more structures an object has that allows autonomous movement within an environment and the more structures an object has that allows itself to delineate itself with what's external from it, becomes the difference between a drop of oil and a tadpole, or a sofa and a Human Being, or a Personal Computer and an Organic Observer.

[@johnphil2006]

Flow of molecules simply moving uphill on the probability distribution where the molecules should be ; consciousness would be an emergent phenomenon happening at the time of this distribution.

[@jondodkins]

I like the idea that this guy sometimes just has chips and ketchup for lunch.. then back to it

[@jacobjorgenson9285]

He has a lot of alcohol in his office

[@JinKee]

gravitation opposes dissipation, does star nucleation mathematically look like the "living drop of ink" from the free energy principle perspective?

[@danzigvssartre]

Jin Kee Good point. I was thinking the same thing. Furthermore, does the Sun make a Bayesian Inference about the rest of the universe?

[@ibperson7765]

I thought same for a minute, but.. no. The principle is even beyond maintaining internal order, which the star doesnt even do. The first step in maintaining life is to note that total entropy is increasing but the living system is not increasing consistently in entropy, so it is casting off entropy to its environment to maintain its internal order. Gravity doesnt even help with that. But I saw that the free energy principle goes beyond that idea though.

[@danzigvssartre]

Art Boman The Sun is mostly plasma. David Bohm observed that electrons in a plasma behaved in an ordered unison as if they were alive. There is nothing really new about the so-called “Free Energy Principle.” It’s just Erwin Shrodinger’s idea that life is negentropic coupled with Helmholtz’s model of perception as unconscious (Bayesian) inference.

[@danzigvssartre]

LAUCH3D Your comments are a bit confusing. You mention the FIP has nothing to do with consciousness, but then later talk about hallucinations. Only conscious beings hallucinate. As for your comments about coupled systems, two synchronised clocks are a coupled system. Are synchronised clocks living systems? Saying that meaning is derived from a probabilistic model developed by evolutionary design is non-sensical. Evolutionary accounts do not explain meaning, it presupposes living systems already have meaning and works backwards from the presupposition.

[@connis2299]

@@danzigvssartre I wouldn't say conciousness is a prerequisit for "hallucinations". If processed an audio signal, trying to find a certain sequence, and set it so very little evidence convinces you to have found the sequence (because you strongly believe the sequence to be contained - a too strong prior), then even random noise will ocassionally convince you you've found the sequence. This is a very similar situation to what @LAUCH3D described, b ut I wouldn't argue that makes half a page of MATLAB code conscious

[@margrietoregan828]

start just by acknowledging that if you 2:54 want to talk about something there has 2:56 to be a separation between the thing 2:58 you're talking about and everything else 3:00 and in fact there was there were no 3:02 boundaries there would be nothing 3:04 because there'd be no distinction being 3:06 a thing and not that thing so 3:09 statistically speaking that distinction 3:11 or that boundary is called a Markov 3:13 blanket it's just a mathematical way of 3:18 separating states of some abstract world 3:22 system organism culture life cell brain 3:27 into things that are internal to the 3:30 boundary that are owned by that system 3:32 and things that are outside the boundary

[@harpothehealer]

'Excuse me Professor could you tell me where the free car parking area is?' 'Free energy?' 'Err yes I guess that will be it. ''Take a seat' 'Will it take long my cars outside blocking all the road.' 'Not really, it's so simple and seen in a variety of context. There are 2 roads to understanding it. You can apply the principle to any living system..'................................ Show less

[@jaimaa1512]

harpothehealer hillarious!! this is definetely a double collateral flow

[@CoolCat6131]

harpothehealer I LOVE THIS SO MUCH

[@galaxyxie9748]

Love this

[@electric7309]

🎯 Key Takeaways for quick navigation: 00:10 🧠 Understanding the Free Energy Principle - The Free Energy Principle is a fundamental concept originally from systems neuroscience, providing a principled way to understand how the brain functions. - It has broad applications, extending to various living systems, making it an organizing principle for life's characteristics. 01:04 🌐 Two Approaches to the Free Energy Principle - The Free Energy Principle can be explained from two perspectives: one tracing its historical development in unconscious inference and psychology, and another, a high-level approach based on the fundamental characteristics of life. - Both approaches lead to the same conclusions about the brain's role in constructing explanations. 02:25 🤖 The Brain as an Inference Engine - The brain actively constructs explanations by sampling information from the world, a concept known as the Bayesian brain hypothesis. - It gathers sensory input and selects which information to sample, shaping its model of the world. 03:23 🔄 The Markov Blanket and Boundary - To discuss something, there must be a distinction between the subject and the rest of the world, known as the Markov blanket. - The presence of a Markov blanket implies that a system is modeling the world, which has implications for its behavior. 04:45 🧩 Components of the Markov Blanket - The Markov blanket consists of sensory states and active states. Sensory states mediate sensory input from the outside world, while active states influence external states and depend on the internal states. - These components play a crucial role in modeling the environment. 06:04 🌊 Resisting Dispersion and Self-Organization - Systems that exist over time resist dispersion by random fluctuations. A living system retains its structure, appearing to self-organize. - This self-organization involves states moving uphill on the probability distribution, linked to the Bayesian model of evidence. 08:17 🧠 Maximizing Model Evidence - Systems behave as if they are actively soliciting information to maximize the evidence for their own existence. - Actions, sensations, and internal states all contribute to this process by maximizing Bayesian model evidence. 11:22 🧐 Brain Structure and Universe Characteristics - The structure of the brain reflects the characteristics of the universe it inhabits. It can reveal whether the universe exhibits action at a distance or causal relationships. - The brain's hierarchical structure mirrors the hierarchical nature of the universe. 14:37 🤯 Free Energy Principle's Implications - The Free Energy Principle helps connect various global brain theories and provides a unifying framework for understanding brain function. - It leads to testable hypotheses about the brain's processes and how it achieves the maximization of model evidence. Made with HARPA AI

[@jaimaa1512]

how many times do we maximise??????

[@malcolmvargas45]

infinite process, as long as there's new input there's new cycles, new updates of the model.

[@BruceThomson]

Lovely stuff, 'enjoyed it - partly because it seems to confirm my own discovery (which comfortably encompasses his model) that the most useful general model of the driver of all 'life', is a blind, mechanical curiosity. E.g. the jumbling of genes that we call sex, is the curiosity of 'see what happens' towards accidentally creating a better new individual. Feel welcome to provide any contrary example that defeats my model. If I can, I'll counter it. =)

[@user-ic7ii8fs2j]

does this remind anyone of the yoneda lemma?

[@ibperson7765]

So Robert Craig has said this video is basically wrong. My tentative understanding is that he’s saying if you really understand the mathematics and formulation of a markov blanket, it does NOT actually create a boundary between the system and environment so the fep results derived do not really obtain. Robert if youre seeing this is that correct? I have other thoughts. One is that a system modeling its environment seems, at least intuitively, correct. How else can it interact in a complex way that accounts for many steps ahead? It must model its environment. There must be some at least rudimentary system within it that encodes some things about the current environment and is capable of changing when the environment is different. Finally, how can it possibly be true that the system is maximizing bayesian inference or minimizing free energy, or maximizing or minimizing anything? Because if so there is presumably a unique solution to that optimization implying only one way this system can exist or act or be alive. However it is that we characterize the system (short of a complete and total description of every last atom by location and by momentum vector and by energy state), there will then undoubtedly be multiple ways for a system to exist and act given that characterization... and still be called living. In other words, claiming a specific maximization must be done to count as being alive is the same as saying only this one exact series of actions must be done in order to count as being alive. But that cant be. One alternative is that the maximization solution is not unique. That seems unlikely to me though. Im being vague because I dont know the primitives and the model and what is free to maximize over etc. Mire generally, Im very skeptical that all life can be said to be doing this maximization though. What constitutes being alive is so complex and varied even for the tiniest and simplest living thing. Is a suicidal ant choosing to jump into a volcano maximizing as she jumps? Because she is certainly alive as she jumps. And the same ant except that it runs away at the last moment is also alive, as is the one that sits there. This is just one second in the life of an ant. We’ve declared the ant is minimizing free energy. That required minimization cant predict which she will do?

[@ibperson7765]

When we say minimize free energy, I assume it means the actions minimize it over some future time or times. Still knowing little about the model, I would venture a guess that the ant is not minimizing this when it decides to jump into the lava two inches below it. So it is not alive when it jumps? Well, it is alive then.

[@manuelsantosbermejo3732]

Problem here is (if I'm not wrong) that you misunderstood the "maximizing free energy" expression. Karl doesn't refer to the thermodynamic concept of energy. His free energy is a proxy to "surprise" when comparing the model of the world with the sensory signals coming from the real world. If we treat them in a bayesian way as probability distributions, the role of the living being is to minimize surprise by generating new models (posterior) or by looking for sensory signals that reduce the gap between the model and the previous sensory feedback (likelihood). In a way, reducing surprise can be achieved by doing suicide practice even if it goes against the homeostasis of the organism. This ant is indeed reducing free energy despite commiting suicide.

[@ibperson7765]

@@manuelsantosbermejo3732 makes sense

[@starxcrossed]

Sounds like something I heard Stephan Wolfram say about the brain and the universe. Also Kant and Don Hoffman.

[@arseniklas]

What comes first? The model of the world or sensory information from the world forcing the human mind to create a coping model?

[@ibperson7765]

Niklas H It also applies to any living system like an ant. He should have talked about that first went straight for the human brain sheesh. Anyway the answer is the model comes first. It is built into the anatomy and physiology of any living thing. Like even with the brain he said the anatomy of brain subsystems for what and where thinking exist as model of a what/where world. The structure of those brain subsytems didnt come from sensory information. But it updates from there

[@kevankwok01]

So if something with a Markov blanket system exists over time, it must resist entropy by gathering evidence of its own existence, minimising free energy or surprise, via a change in its modelling or actions. The first line of thinking makes sense, if something exists over time and is consistently able to resist entropy, it must have a strategy to do so whether consciously or not. Whether that strategy is gathering evidence of its own existence, I'm not sure. For example, I'm not trying to confirm my own existence by any means, I'm looking to understand it, even if understanding it moves me towards an increasing accumulation of evidence that disproves my existence. I'm not actively seeking the blue pill if you will, even though it is more comfortable to do so, I seek the red as it moves me closer to the truth. This may be my strategy for resisting entropy, by better modelling the world, others and my self, but the aim isn't reduction of uncertainty, it's growth. Growth in consciousness. Am I missing something here?

[@johnthescott2409]

seems like science is gradually asking the questions that advatia vedanta addressed eons ago, namely the role of our conciousness in the construction of both our private and shared notion of reality. i recommend videos by james swartz, for a western perspective on vedanta.

[@golagaz]

This is brilliant.

[@margrietoregan828]

think what 5:09 it means for a system to exist over 5:13 periods of time what that means is that 5:16 it is effectively resisting a dispersion 5:21 by random fluctuations so perhaps the 5:24 simplest example here would be if I 5:26 dropped or placed a drop of ink in a cup 5:33 of water then almost immediately would 5:36 start to disperse as random fluctuations 5:38 disperse all the molecules around and I 5:42 would not call that drop of ink a living 5:45 drop of ink because it has dispersed if 5:48 however I placed a drop of ink in some 5:52 water and then to your amazement you saw 5:55 it gather itself up

[@robertlawrence2318]

Intriguingly, the Gibbs Free Energy principle is at the root of Physical Chemistry. Throw a bunch of chemicals together and by minimizing Gibbs Free energy you can predict the resulting chemical compounds. Perhaps Friston's Free Energy principle is the logical extension of Free Energy to organic chemistry.

[@Ucedo95]

In Physics too: all bodies, particles, and radiation, follow the trajectory of minimal energy.

[@mickwilson99]

It's analogous. It can be applied to economics, ecosystem dynamics and in fact any dynamic system in which an abstract element is conserved despited being distributed unevenly. "Energy" is not a "thing"; it is a descriptor for the repeatedly observed propensity for dynamic systems to evolve or change state over time. The abstraction becomes useful only when it can be used to successfully predict the evolution of novel systems. Chemical, mechanical, economic, ecosystemic, whatever.

[@paulwolf3302]

The laws of thermodynamics including Gibbs free energy apply in organic chem also. But there is no such thing as a thermodynamics of information. If that's the bottom line, it suggests a kind of religous belief in mathematics, as Pythagoras' cult had.

[@marcopivetta7796]

that which is below is as that which is above, and that which is above is as that which lies below

[@hn6187]

But assuming not all the energy we expend to defy gravity / the universal average rate of entropy is spent on an efficient survival brain, what else is the mind up to? If an alien took our brain would they conclude it would trash the planet? ( ... Yes probably, since it evolved to gobble more than it needs, then learnt to burn oil )

[@jccusell]

7:57 - 8:18? Why? I've been looking for this all over internet can't find a layman's explanation. Why does a system need to minimize free energy and what exactly is free energy?

[@connis2299]

From my understanding, free energy is the portion of the world unexplained by the model? Which is therefor a measure of how good your model is, minimizing it leads to a better model (better predictions)

[@KalebPeters99]

Read "The Hidden Spring" by Mark Solms

[@abrangemavimbele9421]

It is very good question, I am also interested in the answer, Have you got it yet? If yes, Kindly share with me [e-mail]

[@valueengines2184]

Maximisation of understanding not minimalisation free energy surprise

[@sciencebeartimberwoods7610]

18 tin foil hats came to this video just to find it to be real science.

[@BugRib]

Good one! 😆 I don't think anyone else got the joke.

[@dfearo]

Karl, where have you been all my life? A path from George Herbert Mead, Ilya Prigogine, David Edelman, Ralph Stacey, and David Fearon lead right here. Check out his collaboration with Mark Solms too.

[@DistortedV12]

Finally some advanced psychology

[@khwajawisal1220]

@The Iguana psychology is a subset of neurology I guess.

[@siarez]

If I drop a stone in a cup of water, it too resists dispersing. Does it have a markov blanket? Also when a compress a rubber ball and let go, it restores itself back to its original shape. Does that have a markov blanket? Can I call them alive?

[@jigonro]

Siarez yes and yes

[@siarez]

@@jigonro Friston sells the free energy principle as something that living things abide by. But if it also applies to stones and rubber balls, then does it really say anything?

[@simoncrase5360]

No. The stone doesn't have a natural tendency to disperse on a human timescale. If someone were to devote their whole life to looking a a stone they would not see it disperse.

[@margrietoregan828]

it but is this a 7:11 beautiful observation that the the 7:13 defining dynamics of any system that 7:16 does not dissipate over time is that 7:19 they on average will move of their 7:23 states will flow so as to maximize model 7:28 evidence bayesian model evidence so that 7:31 means that if a system exists then it 7:35 will appear to maximize Bayesian model 7:38 Evans it will appear to be a little 7:39 Bayesian engine it'll appear as if it 7:42 has a model of its world why well

[@victoralmeida4955]

11:28 Pretty sure it's the other way around. "What" corresponds to the ventral stream, the dorsal stream is concerned with "where". Lapse of attention, maybe

[@mauriceebner5555]

11:14 10:31 02:12

[@Peter-rw1wt]

If I meditate, I am no longer maximising the evidence for my own existence, am I ?

[@atomicV1sion]

Wow.

[@oddmentings]

How does the free energy principle apply to a rock, as a thing that exists and is resistant to dissipation?

[@priced80]

It is still dissipating, just very slowly. He is really talking about entropy, the idea that matter tends to go from an ordered to a disordered state. Rock would eventually decay, it just takes a long time. In the meantime, it is not moving or doing anything to counteract that decay so it is not alive. If one rock that normally decayed in 1 million years actually took 20 million years to decay then you might say that something is going on there. It's the same argument as the ink drop. The counteracting of entropy is judged relative to the typical behaviour of the matter in question.

[@sharpenedge]

Note that referred to the cyclical nature of the dissipation resistance. The 'thing' has to have a cyclical property... i.e. reproduction.

[@LMalintzi]

the free energy principle refers to living things / living forms.

[@priced80]

LMalintzi the question is valid and I have already answered it, so your response is both unhelpful and superfluous.

[@LMalintzi]

@@priced80 thank you for contributing to the discussion by providing your opinion. I did the same.

[@JackSarfatti]

External pumping creates the Markov blanket - essentially shown by Herbert Frohlich in his biological membrane coherent phase transition analogy with a laser.

[@jasonattar6330]

Jack, we all know that Herbert F. Was a rotter and a scoundrel. And you above all people; you should know better. Has he paid you back for burning your garage down?? Me thinks I know the answer to that one. He still owes me for shooting that Bio membrane laser at my dog Sizzles. The poor mutt still sicks-up small macaroni type yellow tubing when nervous! Grrrr

[@aquaqglass]

I love you Karl. I hope ive told you that. ')

[@michalchik]

Perhaps he addresses these problems in other areas, but the idea is explained here is fatally flawed in at least a few ways. A basically is saying that living things maintain homeostasis and in order to do this they have to be information processors and all information processors are ultimately reducible to Bayesian systems the, and all problems that we face in maintaining homeostasis are ultimately problems of minimizing Gibbs free energy. Which is fine but it's an oversimplification. There is no neat Markov blanket, clean division between self and everything else in biology. First you have evolutionary principles were each and every sequence of DNA is competing for its own reproduction, which is semi collectivized in the process of reproduction of the organism, which in turn creates things like kinship relationships which are also being optimized. So you kind of wind up with this Kluge that have this continuous and multipolar Dynamic tension and no neat boundary to wrap a Markov blanket around. Secondly it's all great think about these sensory inputs and motor outputs, but we get inputs into our system that are not sensed. Many chemicals and forms of energy and radiation that are sensors at no levels respond to, but actually add random noise into the structure of our internal systems. Not noisy input but the scrambling of the system itself. 3rd and again perhaps he addresses this elsewhere but when you're talking about a living system you're talking about systems who's structure is not solely determined by its ability to fulfill some abstract function, but by the Dynamics of its components which in and of themselves have their own functional quirkiness, from two major sources. The constraints of what a finite number of molecules can do given the thermodynamic environment they are part of, Plus a whole bunch of series of evolutionary Frozen accidents. That is random events that just happened to be the first good enough one that life could work with, whether or not it is Optimum or has strong intrinsic biases. The classic example being the peculiar distribution of codons in the genetic code oh, but also things like, the use of DNA RNA and protein, lipid bilayers, or the vertebrate body plan and the other survivors of the Cambrian explosion. another important principle to take into account here is that even the best adapted organisms are adapted to their ancestral environments, not their current environment. their structure and ability to maintain homeostasis is not a model of their current environment, but a partial model ancestral environment. They're only modeling their current environment to the extent that it resembles the past and random mutations gave it the ability to model the past. Don't get me wrong, I absolutely love with Gibbs free energy. It certainly does apply to living systems as it does everything else in the universe. Nevertheless there is an irreducibility of biology to just thermodynamics because you have to take into account the Dynamics of darwinian evolution and the history of life, as well as the specific physically constrained limitations of physiology, as well as the fact that there is no clean biological separation between inside and outside. We have structured maintenance and competitive antagonism all the way from partials stretches of DNA up to ecological systems.

[@khwajawisal1220]

you got it all wrong sir, Gibbs free energy is similar to free energy but both the things are completely different in the terms of how they work and govern the laws of nature. Free energy minimization is simply minimizing prediction error which you just did with your explanation of his thoughts In your above comment the only thing that Dr. friston wants to convey is that the basic principle guiding every living organism is to interpret the environment in which it exists in order to make predictions that co-incide with what is actually out there and that is what I am doing while writing this comment and that is what you did when you wrote the comment because according to your generative model the explanation you have is more closer to what is actually out there and that is all that this principle stands for apart from the mathematical notions that it involves, this is the gist of it

[@artandculture5262]

I’m reading his paper “Knowing One’s Place...” which explains what he is speaking of here. Does anyone know what they mean by “a cell has a belief”? I ordinarily find that scientists think in a goal oriented tangibles way, as that is what it took to create the effort to become one, and that most people’s work either defends their life choices, or is guided by their capsule of how their life choices work. I don’t think the ordinarily person is goal oriented in a similar way. Sometimes goal oriented behavior is dealing with unfortunate things, like trauma and dissociation, or parents who are addicted, or a plethora of things. So goal oriented behavior can be “hidden”, as he mentions in the paper and in this interview. It looks like how AI people try to program humans, as I read the paper. There are a lot of assumptions about what free energy is, and what that means in different lives. I worked where Karl’s collaborator Hinton also has someone who collaborates with Hinton - who is a principle Facebook AI person - so I have background that suggests in my “hidden way”, that this is social reconstructive theory. It used to be humans studied as animals, and now it is humans studied as machines. Artists do something else with their goal oriented behavior. But, if anyone can explain to me what his paper means by “a cell has a belief”, I would be thankful.

[@jovanyagathe2299]

The truth about my identity... Is that I'm a Markov decision process you are all contributing to through sensory and active states. You might call me the weave of our mutual blanket.

[@ubaldinarosati8906]

Traduzione per favore

[@Well_Meaning]

I knew it was only a matter of time until science discovered AT Fields

[@sodiumsalt]

AT fields??

[@MrTheLollooo]

Absolute chad.

[@brunomartel4639]

A genius

[@kevankwok01]

So if something with a Markov blanket system exists over time, it must resist entropy by gathering evidence of its own existence, minimising free energy or surprise, via a change in its modelling or actions. The first line of thinking makes sense, if something exists over time and is consistently able to resist entropy, it must have a strategy to do so whether consciously or not. Whether that strategy is gathering evidence of its own existence, I'm not sure. For example, I'm not trying to confirm my own existence by any means, I'm looking to understand it, even if understanding it moves me towards an increasing accumulation of evidence that disproves my existence. I'm not actively seeking the blue pill if you will, I seek the red as it moves me closer to the truth. This may be my strategy for resisting entropy, by better modelling the world, others and my self, but the aim isn't reduction of uncertainty, it's growth. Growth in consciousness. Growth in level of awareness of being, in relation to understanding of self, other (people/species) and whole (nature/universe/God). Perhaps this reduces down to a systems ability to process information at the various Markov blankets in which it exists. Or awareness of that within its blanket, outside its blanket & the interface between the two. The quantity and quality of its data set, the ability to process, model & synthesise that information, then translate those learnings in to revisions of its modelling or actions, in a real time continuous learning loop. Minimising uncertainty might be a key survival strategy, but is mere survival the highest motivating force in the universe?

[@vegesenaprudhviraju]

excllent

[@maartenvercruysse358]

"If you can't explain it simply then you don't understand it well enough..."

[@dodosama6502]

what he said

[@galaxyxie9748]

听不懂

[@troelsvejenchristensen8850]

So the Sun is a living creature!

[@jcchew4086]

who is here from west world?

[@mmmjjjlll]

Sounds like Schrödinger negentropy. And also Haken's Synergetics.

[@ibperson7765]

Michael Lyons Yeah I kept thinking entropy would come in. Because total entropy is increasing but the system is not increasing consistently in entropy it is casting off entropy to its environment to maintain its internal order. But I saw that the free energy principle goes beyond that idea though.

[@anshul.infinity]

So it means of brain is trying to understand the true nature of world including brain. Is that it?

[@tdreamgmail]

Egomaniac. A lot of hot air. Oh everything relates back to my theory.

[@jacobjorgenson9285]

You didn't understand a thing did you?

[@Wulffie86]

Westworld brought me here

[@dflatminor]

drops oil inside the glass of water *gasps* the oil is alive!

[@havenbastion]

...in other words?

[@paulwolf3302]

I couldn't understand a word of this.

[@michaelparrish777]

interesting that even someone so intelligent and accomplished as Karl can still make minor (but important) mistakes, such as the function of the dorsal and ventral streams of visual processing

[@pimplydickballs]

At his age its a miracle there is any brain function at all. Enjoy your aging brain while you can

[@digitalsamurai42]

when a camera is pointed on you and the whole world is watching, might feel a tiny nervous which we know that excess cortisol shuts down proper functioning of brain. For him to only make a single mistake is amazing, me thinks.

[@DataWaveTaGo]

His left cuff is unbuttoned. Can we trust him ...

[@ibperson7765]

He was also wrong about action at a distance being modeled in the structure of the nervous system

[@roberth7921]

that 'mark of blanket' explains why state and church had to be separated haha...

[@yurirodrigues2216]

argmax (?)

[@EddoRats]

So Are we all like Maxwell's demon?

[@davideaston6944]

Stop, start, stop, start, stop, start - it's all one could do. There's no time given to even process the thoughts this guy's suggestions create. I guess you're packing the most information into the small time capsule as possible, but ... leave some time to breathe, no? I think I would be very frustrated if I bothered to try listening to this 25 times until I actually caught everything he was saying. Obviously not stuff to listen to without a pretty thorough grasp of the subject at hand, and all the theorems referenced, books cited , etc. Tough sledding... I mean, not even challenging; simply not relevant, unless you're already almost an expert in the field. Cheers.

[@antoniofdez]

He made a mistake about the what and where, it's the opposite of what he said.

[@CONNELL19511216]

Wow. Difficult for an aging brain to absorb fully, but fascinating nevertheless.

[@agt5jx87]

Things are strange.