It's muddled thinking that paves the way to the biggest breakthrough. Together with fuzzy logic. It's an explosive cocktail. Fuzzy mud — Cartuna

Best leave it on the backburner for the moment. Good day.

I didn't read the article thoroughly but I'm struggling to see the utility of the "dark room" model being discussed. — the affirmation of strife

It is a sign of a strength of the free energy principle that the bogus “dark room problem” is the best opposition that might be mustered.

If the only line of attack on an idea relies on a fundamental misrepresentation of the idea, then it’s critics are doing a mighty poor job.

The free energy principle describes the minimisation of surprise in the context of being active. That the organism is busy and striving in a challenging environment is what is taken for granted as the situation being optimised by the cognitive process.

Take away the challenging environment and you don’t even have some gap between the certainty of your actions, and the uncertainty of the world, to be minimised.

So here's a bit more for the pot:

The Dark Room Problem

Again the point is made that an explanation for everything is an explanation for nothing.

And again, the key discussion in relation to this is Watkins. Here's a link to his article: Confirmable and influential metaphysics. There's a thread on it elsewhere in the forums. I commend section seven, in which Watkins shows how to rationally assess what he calls "haunted universe" statements, such as "every action can be explained in terms of avoiding surprise".

I do not think that any single criterion, such as conformity with existing science, can be laid down for assessing a haunted-universe doctrine. This taskis more like assessing the worth of a man's character than the legality of his acts. Conformity with existing science is a favourable factor, but it may be outweighed by a pragmatic estimate of the doctrine's possible influence on the future of science. There re also its bearings on psychology, history, social science, morals and politics to be systematically explored and collectively weighed

SO there is no conclusion to be reached yet, only an interesting area to perhaps follow.

@TheMadFool, there's your metaphysical evaluation. The rhetoric in this thread is also well worth attending to.

Edit: fixed URL for Watkins article. The topical thread is at https://thephilosophyforum.com/discussion/9909/confirmable-and-influential-metaphysics/p1

If the only line of attack on an idea relies on a fundamental misrepresentation of the idea, then it’s critics are doing a mighty poor job. — apokrisis

The thing is that this idea is based on the most significant fundamental error. It tries to oversimplify human behavior, which is wayyyy more complex,with a naive way. Connecting it with one only aspect.
Well sorry but the idea itself is misrepresented on its own from the very beginning.

Not that of course this theory can't attribute useful thoughts, but that's all.

Again the point is made that an explanation for everything is an explanation for nothing. — Banno

Yes. Darwin’s theory of evolution really hammered home that point. Nothing useful can ever result from talk about absolutely general constraints.

I mean what’s with physics and its obsession with the least action principle? Do these guys understand anything about how explanations work?

It tries to oversimplify human behavior, which is wayyyy more complex,with a naive way — dimosthenis9

That is another misrepresentation.

Although I agree that as a formalism, it doesn’t tackle the code side of the semiotic modelling relation. Friston’s Markov blanket is a general physical description of the epistemic cut. But it doesn’t talk about the “how” of the machinery that enables such a cut to actually be made in nature.

There are four levels of such code I would identify. Genes, neurons, words and numbers. Each produce their “worlds” or Umwelt. Only humans have verbal and mathematically constructed Umwelts or world-models.

So yes, there is one general story to be had - a semiotic theory of everything. That is implied in Friston’s approach, but not mathematically expressed in direct fashion. The fact that there needs to be a machinery of semiosis - some system of encoding - is implied by the Markov blanket formalism, but not to be found in that formalism.

And as you protest, humans are more complex. We have words and numbers that lift us beyond the semiotics of neurons and genes. We have social semiosis and techno semiosis. Friston’s free energy principle was directed primarily at the problem of neurosemiosis, and has been expanded to - sort of - include biosemiosis.

So I have plenty of “criticisms” of Bayesian mechanics. But I think it helps to have actually understood what Bayesian mechanics might claim.

presented a particularly neat summation of the Watkins' article. In particular:

The more interesting and hard problem is that all and some statements don't just inspire research hypotheses, they are conceptually related to research hypotheses and thus inspire research hypotheses. "All and some" statements purport to tell us how stuff works, so we go out looking as if it works that way. — fdrake

Hence the theory of evolution inspired wider research that was itself confirmable or falsifiable; and so on. Conservation laws are each examples of scientific statements that are neither falsifiable nor verifiable, but which have the singular advantage of underpinning expansive and efficacious explanations. They work.

It tries to oversimplify human behavior, — dimosthenis9

Indeed, it does. So the question becomes one of how successful it is at doing so. It might be - indeed it seems likely - that this approach will lead to a better understanding of the function of various neural bits and pieces. It seems less likely that it will be able to explain why one person likes vanilla yet the other chocolate. And it will not demonstrate why you should vote Green. The temptation to look to unifying explanations is to be avoided if it leads to oversimplification.

Cheers. I'm more interested in just causing mischief, but will take a look.

The temptation to look to unifying explanations is to be avoided if it leads to oversimplification. — Banno

You do realise that that point is something specifically contained in the Bayesian maths of Friston’s approach?

The trade off between model complexity and model accuracy is the algorithm driving the whole optimisation process.

Perhaps you haven’t read that far and discovered your generalised truism now has been given a formal mathematical description?

Heh, I'll even save you the trouble of looking. Skip to minute 53 of this presentation to see the equations.

Interesting. But that reinforces, rather than helps dispel, an instinctive distrust of theories that explain everything.

You do understand that I am agreeing with you that this is an interesting area of research? But you seem to think that this little exercise has explained consciousness. That strikes me as overreach, and it seems I am not alone.

I'm too old to watch videos. If you have a point you wish me to understand, write it or link to an article.

It is for the contrived definition of "living" that seems to be used here, almost entirely by definition. If life is nothing but avoiding non-anticipated stimuli, then minimising non-anticipated stimuli means living longer? — the affirmation of strife

I'm not sure that's it. The argument is apparently along the lines that avoiding non-anticipated stimuli explains not only individual behaviour but environmental fit with respect to evolution. That's an astonishing claim, if it can be maintained.

But that reinforces, rather than helps dispel, an instinctive distrust of theories that explain everything. — Banno

It might do that in your mind. You do tend to argue a hell of a lot from "instinct" rather than reasoned fact.

I just think it funny that your instinctive generalisation - the one you just tried to use against Bayesian mechanics - is in fact the principle found at the core of Baysesian mechanics ... it being a generalised theory of reality modelling itself.

You do understand that I am agreeing with you that this is an interesting area of research? — Banno

Interesting. You and your weasel words. If it is so "interesting", why are you so unwilling to invest the time needed to have a valid criticism of it rather than your usual "instinctive doubts".

But you seem to think that this little exercise has explained consciousness. That strikes me as overreach, and it seems I am not alone. — Banno

I love it. Classic Banno argument - tailored to pander to the Dunning-Krugerism of an internet forum. "You confident, me doubt, and look at all the others who also don't understand what is being said but are just as eager to knock down tall poppies."

I'm too old to watch videos. — Banno

I only asked you to scroll to a single slide at minute 53. The fact that you don't want to disturb your cosy solipsism says it all.

That has to be the lamest excuse ever made.

(No wait. That would be "lunch is calling". Or "The spiders under my flower pots and the rotting garlic in my garden are suddenly more important than this internet discussion I started but which seems now to be biting me on the butt." I forgot that you are the master of lame reasons for fading from sight as soon as anyone calls you out on your BS.)

So the question becomes one of how successful it is at doing so. It might be - indeed it seems likely - that this approach will lead to a better understanding of the function of various neural bits and pieces. — Banno

Well as I wrote before, sure it has to offer some useful thoughts. And maybe it also become a step for a better idea. Who knows.
But it's not the idea that troubles me mostly, it's the supposed purpose that this idea tries to serve. To explain human behavior like that. I find it naive and that's where my biggest objection is.

Again the point is made that an explanation for everything is an explanation for nothing. — Banno

This - not so much this article, but your complaint that PP/PEM seems to have an explanation for everything - reminds me of a common creationists' complaint about the theory of evolution, which often follows a series of failed challenges to its ability to explain evidence. The fact that a theory can explain all evidence doesn't distinguish between a good theory and a vacuous theory. And there is no way to establish which is the case other than scrutinizing the theory and how it purports to explain evidence. There are no easy shortcuts here to dismissive judgements.

Another thing to note is that there are different ways to respond to a challenge. One is to make a positive argument that the challenge misses its target, e.g. by conducting a decisive test, or by showing that what is alleged to be the case necessarily is not the case. Another is to argue that the challenge may not hit the target. For example, when creationists rhetorically ask "what good is half a wing?" one response is to argue that the equivalent of "half a wing" can be adaptive (maybe not for flying). This second mode of response doesn't provide an additional argument for the theory, but it does forestall the challenge and tasks the challenger (or alternatively the defender) with going deeper and doing more work. This is how we can view Friston's response to the "dark room" challenge in the comment article that you shared.

That's the long-expected post. Your habit of attacking the messenger has been noted by others.

Cheers.

So yes, there is one general story to be had - a semiotic theory of everything. That is implied in Friston’s approach, but not mathematically expressed in direct fashion — apokrisis

But that's the exact base that he builds his argument on.Even if not expressed in mathematics,he uses mathematics to serve that initial premise he makes.

I don't dismiss the idea of free energy principles totally. It sure has some food for thought, maybe in different fields also. But, imo, the way it is presented make people not to take it seriously and dismiss it all together.

Maybe if it was presented differently would be much more helpful as people to take notice on it . And be serious about it. Somehow, in that way, I think it underestimates its own self.

...naive... — dimosthenis9

Not sure this is the right word. It seems to me that the very sophistication of the approach leads some to over-applying it.

And I don't mean the authors of the articles referred to here, who seem almost surprised at the potential of their theorising, but those who might take their work as already having solved the problems of consciousness.

Not sure this is the right word. It seems to me that the very sophistication of the approach leads some to over-applying it. — Banno

Yeah probably isn't the right word indeed.But the meaning it's the same. Just couldn't find an appropriate word to summarize what you wrote above.

The fact that a theory can explain all evidence doesn't distinguish between a good theory and a vacuous theory. — SophistiCat

Well, not quite. We want a theory that rules out things that are contradicted by the evidence.

...there is no way to establish which is the case other than scrutinizing the theory and how it purports to explain evidence. There are no easy shortcuts here to dismissive judgements. — SophistiCat

Yes! I quite agree. Small steps.

Edit:

...seems to have an explanation for everything — SophistiCat

Just to be sure, I'm not saying that PP/PEM seems to have an explanation for everything; but that it would be wrong to read it as having an explanation for everything. Does anyone here do that? Perhaps.

For a lively state to energy, free energy simply has to be increased. If there is a huge number of interacting particles, the state of the least free energy or the least chemical potential, is a dead one. Structures evolving in a constantly non-equilibrium dynamical dissipative state (between starlight on the day and the dark in the night, to and fro, every time a bit further away from equilibrium)) always away from, and not towards that equilibrium, are bound to evolve in life. I saw a seemingly dead trunk of tree. But a newly forming bark structure was evolving on it! Truly a beautiful miracle to see. Hence the need for the least of free energy is nonsense. Maintaining free energy means maintaining the status quo. Some organisms, like viruses, have chosen for this. Others like to evolve. People have reached the level to freely play with free energy.

That's the long-expected post. Your habit of attacking the messenger has been noted by others. — Banno

You are a curious chap. If you are indeed the messenger, why do you keep disappearing at the very moment you are asked to deliver the message you claim to bear?

You said, aha!, the problem with all totalising generalisations is that they risk leaving out particulars essential to telling the larger story. I said, check this single slide. You can see that the totalising generalisation makes that same totallising generalisation - but now as a mathematically specified and biologically relevant fact.

Now you want to seek cover as the victimised. You want to answer reasoned argument with social game playing.

Well I guess when you are hoist by your own petard trying to play one game, you must scramble to make it a different game instead.

Maybe if it was presented differently would be much more helpful as people to take notice on it . — dimosthenis9

I dunno. The more common complaint might be that Friston seemed to have changed names for his story so many times down the years. When I first met him 25 years ago, it was all about coupled neural transients and generative neural networks. Then it became Bayesian brains, Free-energy principles, Markov blankets, a Unified theory of unifying theories.

So there have been a ton of different presentations. Much ink spilt. And now he has come up with equations - a general model condensed to a single slide or two. He has proposed a Bayesian Mechanics.

If nothing else, this is huge because it allows neuroscience to finally kick computationalism and Cartesian representationalism out the door.

Cognitive neuroscience started off with the idea that the brain is a prediction machine. The formalisms were already there in the 1800s with Helmholtz and his mathematical definition of free energy, Bayes and his theory of probability.

But then the science of mind got infecting by computer-mania in the 1950s. The maths of the Universal Turing Machine tool over at the putative new foundations. This resonated nicely with the Cartesian representationalism of popular culture where consciousness is treated as the witnessing of some kind of data display.

So now we have Friston emerging as at the top rated cognitive scientist of his age because - as well as also sorting out the maths that underpins the credibility of functional neuroimaging - he returns neuroscience to the wisdom of the 1900s. He unifies the maths needed to ground formal cognitive theory in the 21st Century.

Anybody who studies neuroscience, or even biology in general, always knew life and mind aren't merely machines - neither Newtonian physical machines nor Turing computation machines. But it was difficult to resist the mathematics of those paradigms for as long as neuroscience hadn't established a maths of cognition it could call its own.

That in a nutshell is what is at stake.

Now of course Friston's claims to arrived at a final maths of everything cognitive needs challenging. I've already detailed the challenges I find warranted.

But this "darkened room problem" is a tedious misrepresentation of the maths. And as I say, if this is the best you have got, you ain't got nothing.

It is a sign of a strength of the free energy principle that the bogus “dark room problem” is the best opposition that might be mustered. — apokrisis

Not sure about sign of strength, and as you quoted I hardly see my comment as the best opposition. I was responding mostly to the OP and what I quickly read in some of the other comments, as well as the linked article itself, which I find difficult to parse. I've re-read your comments and thankfully you seem to be arguing for something different, and this "Markov blanket" is a much more interesting idea. Thanks for the video link, it looks interesting. I wil need more time to digest it.

The trick is then to act in ways that only increase your certainty about the sensations you will experience. If the certainty of your actions effectively reduces the uncertainty of your sensations, then the two sides of the equation are tightly coupled in a way that optimises your ability to exist in the world. — apokrisis

You are winning to the degree your plans for your future don’t encounter the unexpected. But an organism lives in the world. It exists because it can tame environmental uncertainty through its actions. It can feed itself, protect itself, reproduce itself, etc. It can act in ways that reduce the world’s uncertainty. So it doesn’t need to retreat to the refuge of a darkened room to escape the environment’s capacity to surprise. — apokrisis

This makes more sense.

But the theory actually states that life expresses the drive to avoid becoming randomised by its environment. — apokrisis

I think I get it now, and discussion in terms of this framework would be more interesting. Unfortunately, the language and form of the article and the "dark room" idea are confusing, it seems not only for me.

I am also unsatisfied by the computationalism in popular representations of neuroscience.

Sorry to add to the "tedious misrepresentations". Other than the video, is there maybe a more up-to-date article on this topic that you can recommend?

Maintaining free energy means maintaining the status quo. — Cartuna

I think you have some misunderstanding here.

Life and mind exist by constructing paths for entropy dissipation. The sun shining on bare rock winds up returning the radiation to space at the much cooler average temperature of about 60 degrees C. The same planet surface covered by a mature biofilm radiates at about 20 degrees C.

A small improvement so far as the second law is concerned. But even a small difference is one that counts and so is the path the physics will select for.

The problem for life (and mind) is then to separate its materiality from the entropy it transacts. To do better than bare rock, it has to invest in complex material structure. It must recycle its atoms, so to speak - keep them recirculating and so creating the path down which actual free energy becomes actual physical entropy.

This leads to a delicate balancing act in biology - a matter of considerable theoretical debate. Is life caused by the principle of maximum entropy production, or instead the contrary, the need to evolve towards a zone of minimum heat loss/maximum efficiency?

The solution is that existing as a degrader is indeed a delicate trade off. Plasticity and stability both count. An organism that grows too fast is like a cancer. It burns bright, but burns short.

Once you get this, the whole of ecology and evolutionary history follows. You need weeds, but then it is also natural that forest takes over. Complexity is good, but also the simplicity of bacteria is nature's other enduring pole or attractor.

So there is an optimisation function to be had. Yet - as a dynamical balance - it also affords a wide spectrum of solutions. And it is this scalefree variety that we must then expect to find actually expressed in nature.

And so it goes. Life has organised the earth from bacteria that can eat rock to the Gaian cycles that have stabilised the Earth's climate and retained its benign watery condition, its fine-tuned balance of CO2, its unusual reservoir of free oxygen, for several billion years longer than would be the case for the bare rock thermodynamic counterfactual.

Biology has figured it out. Life is self-maintaining, code-based, dissipative structure. But the status quo is a dynamical balancing act - always an optimisation, a trade off. Every organism has to have a goal that then defines it as a self in its world. There must be stored information - genetic information - that encodes the homeostatic set-point for the kind of species of thing it is. A weed or an oak, for example.

And neuroscience is recapitulating the same intellectual journey that biology got started on back in the 1960s. Mind is a higher level of organisation for constructing dissipative structure.

There are all kinds of mindful organisms. Bacteria included. They can sense environmental gradients. They can swim towards nourishment and away from toxins.

And mind continues the effort to solve the essential optimisation problem. To have stable existence as an organism, that organism must transact some quantity of entropy while extracting some sufficient degree of work from the dissipative flow it thus creates. It must repair, and even reproduce, the material body - the structure of metabolism - that achieves the feat of both serving the second law and also defying it by maintaining the integrity if its material structure.

Mind science thus doesn't talk about consciousness as such. In the lab, they talk about researching cognition as some self-optimising balance of attention and habit. Or voluntary and automatic processes. Or top-down and bottom-up information flows in a hierarchical architecture.

Always it is about detailed balance. The optimisation of two complementary factors.

So the idea of the minimisation of free energy speaks to that. There has to be some baseline of entropy production that supports its "other" of investing in negentropic structure. The structure that is doing the entropy production.

Nothing makes sense until the two sides of the equation are cybernetically coupled. Action and perception have to be coupled - as in Friston's Markov blanket formalism – so that together they will track the goal of being optimised in a way that produces both enough entropy, and also enough negentropy to "live the longest". Or better put, as I have said, "reproduce the most successfully".

Ecology amazed the world back in the 1960s with its maths modelling predator-prey relations and other systems-level facts about the dynamics of evolutionary processes. Biology could claim to have its own maths. And this maths was itself rooted in the right kind of physics - thermodynamics.

Friston continues the story.

Other than the video, is there maybe a more up-to-date article on this topic that you can recommend? — the affirmation of strife

You can easily search for Friston's recent papers - https://scholar.google.com/scholar?as_ylo=2020&q=karl+friston&hl=en&as_sdt=0,5

But you will see how he is engaged in doing exactly what @Banno complains there is so far a lack of - and that is putting theory into practice.

If you want to start with the general theory, this video is the summary which he has been hawking around the past couple of years as the one for a less technical audience.

His next step beyond that is then sketched out in this other workshop presentation from June - https://youtu.be/T711im7ZgmU

you keep disappearing — apokrisis

I'm still here.

But are you keeping up? I hear your usual plea for “small steps”. So probably not.

But this "darkened room problem" is a tedious misrepresentation of the maths. And as I say, if this is the best you have got, you ain't got nothing. — apokrisis

So if I get it straight, your main objection is the way Banno presented Friston’s idea here on this thread? Using the "dark room problem"? You find it misleading?

this is huge because it allows neuroscience to finally kick computationalism and Cartesian representationalism out the door. — apokrisis

So in your opinion, and in a few lines what exactly this theory tell us about consciousness? Or what it implies at least as to rephrase it. Why you find it so huge?

It's not an ironic tone, my knowledge about this is only the article Banno linked. But as I read it I didn't get very enthusiastic about it and I wrote my objections above. I guess I just act lazy and I want you to give me the juice(hmm.. That sounded weird....).

Well, not quite. We want a theory that rules out things that are contradicted by the evidence. — Banno

The thing is that when you reduce a theory to very general and rough slogans, like "minimizing surprise" or "survival of the fittest," you will readily find both apparent examples and apparent counterexamples, which then prompts complaints that the theory is either contradictory or explains too much, or even both (@Kenosha Kid). The devil is in the details, as you acknowledge. Without getting into those details you can't really say anything one way or another.

I am sympathetic to your attitude towards totalizing theories. But there is a difference between a general unifying idea and a detailed treatment of a subject. Evolutionary biology as a whole is a complex and diverse field, appropriate for its complex and diverse subject. And yet Darwin's basic insight pervades it throughout. I think there is room for more such insights in cognitive science and biology.

By the way, looking Friston's publications, you can see a rather typical pattern where the further he gets from his own field, the wider he casts his net, the more diffuse and light on details and empirical support are his (team's) works, shading into pop-science and philosophy-lite. (There is also an inverse correlation with the number of citations...) Then again, if he got something essentially right, then these kinds of big-picture narratives can be valuable as setting directions for future research and providing an insight into large-scale patterns.