On Fine-Tuning and the Functionality of Physics

[ConradDJ]

Are you saying here that life 'does the reproduction dance' rather than act as an agent for reproducing DNA? You then disagree with that polemic biologist, Richard Dawkins? Not that this is a bad thing, of course --- he is very strident. Your invention, "functionality", is I think too unspecific to separate such possibilities.

Dawkins may be pretty dumb about some things, judging on hearsay about a recent book of his that takes on religion. But the "selfish gene" thing is good, and I especially like his little book River out of Eden as a reminder how evolution works and how powerful this self-replication business is.

"Functionality" is purposely un-specific, because I'm more interested in raising the question about what the basic functionality is, than coming up with a definitive answer. Even in biology where we know the thing quite well, conceptually, evolution is complicated and gets more and more so over time. So while it's accurate to say it's essentially all about things making copies that make copies... really what "functionality" points to here is whatever it is that evolves, so that it can keep on evolving.

In case it's not obvious, I'm using the term in the sense of the functionality of a button on your computer screen, or of a piece of software or hardware --i.e. a description of what it does, what it's good for.

In the case of physics, I think that to describe the "basic functionality" as measurement, or observation, or the communication of information, comes close. But none of these terms are really well-defined yet, in physics, though we know what they mean well enough in daily life. But I'm not trying to give a precise definition, at this point. First we need to get a feel for what's going on with this business of determining information through interaction that then gets passed on as part of the context for determining other information, and so on.

I'm thinking that eventually we may be able to picture this process as the kind of thing that can evolve, just as we can picture the evolution of self-reproduction in biology. Then we'll be in a better position to describe just what's needed to make this work.

 

[Fra]

I have feeling that I'm unable to convey what I mean. Which is a bit weird because i think your idea of the measurement process is not far from it.

What I describe is both a deeper idea of things, that's somehow abstract which is why I think it's hard to convey, but it's also very simple.

The problem is describing the underlying “functionality” in terms of which the laws are useful.

My suggestions is inference. The functionality is inference. What we need to describe is the physics of inference.

An physical interaction process = measurement process, is nothing but an inference process.

Expectation -induces-> action; backreaction -induces-> revision of expectation, which takes place at two levels; information preserving revision and non-information preserving revision.

It's selfpreservation of the inference systems that causes equilbrium points to evolve. The "logos" of the interactions are then nothing but equilibrium strategies. They are stable because no observer/subsystem benefits from changing theirs. This is how I envision that we will explain the SM for example.

What I'm suggesting that each physical system is loosely speaking one-2-one with a particular inference system. This inference systems has predictable interaction properties as we can predict how it ratioanally would respond to input. But the predictions are not deductive, their and inductions, again only inferences, that further determines the action of the physical systems where the "logos" lives.

It's a picture that may seems circular and chasing itself, but that's why evolution is part of the expectations. There is no objective or global equilibrium. This IS a game, and all "predictions" and testing of predictions come in the form of "play the game". This is the deeper insight I advocate and is why I think we should study evolving and interacting inference systems.

But this is a generalization of the inference we have in QM. I'd say the inference we have in QM, is a subset of the more general case. Although still a generalization as compared to classical inference (thermodynamics).

The exact mathematics to use for this is still unsolved, but I insist that the key idea is very simple. The problem is somehow that we seek OTOH a "background" independent inference model (because most inference models do have such backgrounds) but then again, the way to do it is NOT to REMOVE the background, it's instead of understand how the background evolves, and with it the new inference system.

/Fredrik

 

[Fra]

The obvious question one can have is: Ok, I have my "pet-model", that is that everything is inference. Why is this better than all other "pet-ideas" like the idea that "everything is geometry" or everything is just abstract algebras etc.

I think the difference is that inference models is the perfect match with science. Science is essentially an inference process itself. You know the old debates of the problem of induction and Poppers resistance against this description. Not to mention that any learning process is nicely abstracted as evolving inference systems.

I think aside with this, the ideas like "manifolds" "geometry" really seems almost pre-historic remnants of the realist history of human science.

/Fredrik

 

[ConradDJ]

The functionality is inference. What we need to describe is the physics of inference... A physical interaction process = measurement process, is nothing but an inference process.

Expectation -induces-> action; backreaction -induces-> revision of expectation, which takes place at two levels; information preserving revision and non-information preserving revision.

It's selfpreservation of the inference systems that causes equilbrium points to evolve. The "logos" of the interactions are then nothing but equilibrium strategies. They are stable because no observer/subsystem benefits from changing theirs...

The exact mathematics to use for this is still unsolved, but I insist that the key idea is very simple.

Hi Fredrik –

My sense is that you may well be on the track of something important. As in the “Inductive Inference” thread currently running – there seem to be several lines of research that identify the basic structure of QM with the logic of inference. And I very much agree with your idea that we need an evolutionary understanding of basic structures, instead of just assuming them in the background of the theory.

But, I think your idea is probably too simple... I think you may be falling into the trap of hoping for a single, simple mathematical principle that explains the whole show.

Specifically, I’ve been trying to show that measurement is not “nothing but an inference process.” As you’re well aware, inference needs to work on incoming information, and there is no information without a context of other information – in fact, a context of other kinds of information -- that let it mean something. My emphasis is on the way the different elements in the structure of physics work to make each other meaningful and communicable.

So the logic of inference may well play an important role here, but it’s not the whole picture. Unfortunately “the whole picture” is very difficult to assemble, in physics. The methodology of formal / mathematical analysis strongly tends to isolate one element – there is still the Platonic goal that we will be able to derive everything from One idea.

That goal was in a sense reached, by Darwin’s insight in biology. But even the simplest self-replicating systems may well have involved several types of molecules interacting in many different ways. What’s simple in biology was never the specific structures or processes, but the overall “thing” that they all accomplish, working together, i.e. self-replication.

You describe a system of interacting “observers” reaching an equilibrium, which can then be disturbed by new data, which might then require a new equilibrium, and so on. The system would evolve by requiring the schema shared by the observers to become more complex, so as to be able to make sense of more of the data. But again, this presumes a mechanism for defining each observer’s data-schema and comparing it with others – a kind of “language”, in effect. That not only shares data but (separately?) shares the schemata.

A parallel in biology might be a species reaching a state in which it’s well-adapted to its environment... but then the environment changes (maybe due to another species adapting to it), so a new “equilibrium” is needed. This is surely an important process in evolution, but it depends on the self-reproduction of each species, which is not basically an equilibration-process. (Though it does involve multiple processes of equilibration both internally and with the environment.)

So again, I imagine your inference-model may be an important part of the “functionality” we need to understand – but it doesn’t look to me like a whole picture, yet.

Thanks for your notes -- Conrad

 

[Fra]

Conrad, I see your objections but I still think that this fits into the inference picture. But as I've tried to stress several times, I'm not talking about just probabilistic induction - this is a special case of inference. I'm considering en EVOLVING general inference.

I'll try to be brief for clarity:

I do not have an illusion to find a simple mathematical principle from which all follows. This should be clear from the type of reasoning I constantly insist on. There are not 100% confident premises and no deductions. This of course applies even to me - I do not have this illusion is a simple one-line TOE from which everything follows in an instant. What's simple is more the evolutionary mechanism, which is pretty much darwins BUT I have a slighly more SPECIFIC suggestion, in terms of general inference.

It seems you think that natural evolution and selection in biology is a good example of whose analogy we seek - right? But as I think you also agree, evolution is NOT just variation and selection on DNA, because obviously evolution started way before the first stable DNA or RNA was on Earth right? To just consider evolution of DNA sequences in the space of all possible DNA is to think that there is a fundamental level on which to apply optimations, but this is a simplificaiton.

So here's my analogy.

The "DNA" of the inference systems is coded in each physical system (each observer) and gives rise to the rules that yield expectations of the future knowledge, based on the present knowledge. This expectation determines the action of this physical system, as a form of generalized "diffusion" (but over non-commutative discrete structures, which si exactly why it ISN'T normal diffusion, and the reason why we need a NEW mathematical model for this inference, the probability based does not work. But borrowing the words helps since the original intent is the same.).

Each inference systems, thus acts and gets reacted on, with it's environment. This puts selective pressure on evolving the FIT inference systems, that are self-preserving.

To described the process of how the inference systems for example handles an inconsistent feedback is at two levels, sometimes a simple state revision helps, sometimes there is no consistent correction and then this implies breakdown of the inference system, because in my view the inference systems is like a steady state structure, unless it's continously supported it decays.

This picture thus contains small variations, and selection.

"Reproduction" can be pictures in a more sophisticated way - by induction of a environment hospitable for "similar" inference systems. So the reproduction mechanism is then simply to provide the right breeding grounds for "similar thinkers".

Then I think you say that this is still too simple, because ALL of what I say still requires a context. Yes, but this is why even THIS description lives within another inference systems! In this case it happens to be my brain, but obviously I am not an expcetion. I'm constrained to the same principles as is an atom.

What I'm saying is the equivalent to that each physical system has a "model of reality" encoded in their internal state and structure. IT's just that for simple things like atoms, this is of course not near the complexity of my brain, but that's why the action of an atom also is MUCH simpler than the action of a human.

I'm not saying that the mathematics for this is worked out, but I don't see how your concerns are not taken into account. The prediction of this, is not a simple thing form which all follwos. However, if we look for the smallest DNA of physical law that we humans can distinguish, then looking at the subatomic systems or Planck regimes seems like the way to go.

I do not think it's unreasonable to think that maybe we can make inferences about this, that may yield us first principles undertandings for example why we have 3 dimensions, why we have some values of the masses and parameters of the SM.

I think the reason why ST just gives as a gigantic landscape is because they fail to connect the excitaitons to the evolving background. The analogy of that in my view, is to understand how the inference system evolves in response to the state. LEt's just say I have more specific but very immature ideas on this, but the overall idea I think seesm to be in your direction.

/Fredrik

 

[apeiron]

That goal was in a sense reached, by Darwin’s insight in biology. But even the simplest self-replicating systems may well have involved several types of molecules interacting in many different ways. What’s simple in biology was never the specific structures or processes, but the overall “thing” that they all accomplish, working together, i.e. self-replication.

Again, evolution is only half the story. You need development as well. There is replication, but also metabolism.

If we are talking analogies, this does fit a decoherence approach to QM. A QM wavefunction freely develops some indeterminate potential, and the environment imposes constraints that collapses the wavefunction, selects some actual outcome. There is an evo-devo story there.

 

[dpackard]

The function of the universe is to make actions stationary just like it is a falling body's telos to be at the center of the earth. It doesn't seem to me you're adding any new information by describing things in terms of functionality, but returning to a teleological view that is unnecessary. With evolution, the functionality is a product of reality - only the fittest replicators for survival will survive. It's tautological. We only see functionality because we are viewing the whole history from one end. Maybe a functional perspective could be a good one to look for from a psychological point of view, that could lead to progress by changing our viewpoint, but it doesn't seem to me like such a thing could ever be fundamental. Or that we would ever be justified in calling it fundamental.

 

[ConradDJ]

It doesn't seem to me you're adding any new information by describing things in terms of functionality, but returning to a teleological view that is unnecessary. With evolution, the functionality is a product of reality - only the fittest replicators for survival will survive. It's tautological.

It's true that in general, when we talk about the function of something, we're thinking in terms of teleology, i.e. something that has a purpose. But evolution is an exception; it doesn't have any extrinsic purpose, it's not happening for any reason. It's just random accident. And you're right that "survival of the fittest" only means the survival of those that happen to survive.

But the point is that random accident almost never gives rise to structures anywhere near as complex and finely-tuned as an organism. This happens only when a very unique type of "functionality" happens to get started -- as self-replication happened to get started Earth. Once you have entities that can make copies of themselves and more or less successfully replicate variations, then mere random accident can eventually generate all the complexity of life.

The suggestion I'm exploring is that there is another such unique functionality at the basis of physics. I'm thinking that the things we tend to take for granted about physical reality -- e.g. that things have determinate characteristics, and "obey laws", and that the lawful interaction between things communicates observable information -- constitute this other special functionality. We don't have a word for it, just because we've always taken it for granted, but we might call it a kind of "self-determination" -- i.e. something like the ability of an system of interactions to define its own characteristics to itself.

The idea is that, like self-replication, this functionality of "self-communication" is such that once it gets started, then merely through random accidents of succeeding and failing it can evolve more and more complex ways to do this thing... eventually producing a very finely-tuned system of many types of entities and interactions, in which everything both "measures" and "is measured by" everything else.

So you're right that just to take whatever happens in physics and call it "functional" would add no new information. On the other hand, my argument is that the pervasive fine-tuning of our universe is at least prima facie evidence that there may be something special going on here, similar to the special functionality at the basis of life.

 

[ConradDJ]

The "DNA" of the inference systems is coded in each physical system (each observer) and gives rise to the rules that yield expectations of the future knowledge, based on the present knowledge. This expectation determines the action of this physical system...

Each inference system thus acts and gets reacted on, with its environment. This puts selective pressure on evolving the FIT inference systems, that are self-preserving...

This picture thus contains small variations, and selection...

"Reproduction" can be pictured in a more sophisticated way - by induction of a environment hospitable for "similar" inference systems. So the reproduction mechanism is then simply to provide the right breeding grounds for "similar thinkers".

What I'm saying is the equivalent to that each physical system has a "model of reality" encoded in their internal state and structure.

Fredrik – Apologies, I take it back about your scheme being “too simple”!

Maybe it’s just because you’re speaking my language here, but your last post gives me a much fuller picture of what you have in mind. It does seem as though you use the concept of “inference” more or less the way I use “communication” or “measurement”. That is, it refers not just to the specific act of “making a guess” based on certain data, but to the whole physical context that contains the data and gives feedback on the success or failure of the guess, and also relates the guesses of different inference systems to each other.

So I can see that if we could find a way to translate this into physics – how “expectations” are encoded in the properties of systems, how they get expressed in a system’s “actions”, what the feedback is and how it gets incorporated into the inference-system (or else kills it off), and how the ”model realities” of different systems affect each other – then it could be the sort of special functionality I discussed in the previous post, that supports a genuine evolutionary process.

To put this into mathematics – is there a way to specify logically what the elements of such a system would need to be? I imagine something like a flow-chart that pictures both what happens internally within each inference-system to test the model against data and to generate actions, and also what the external linkages would need to be. Does the “data” for each system consist simply of the “actions” of the other systems?

I know this is work-in-progress and I don’t mean to press you for answers... I’m just wondering, seeing if I can get the picture you're imagining.

 

[Fra]

Conrad, now we are more in tune!

it refers not just to the specific act of “making a guess” based on certain data, but to the whole physical context that contains the data and gives feedback on the success or failure of the guess, and also relates the guesses of different inference systems to each other.

Yes. The logic of "making an isolated guess" is one component only of the total picture. It's the problem of placing bets optimally.

So I can see that if we could find a way to translate this into physics – how “expectations” are encoded in the properties of systems, how they get expressed in a system’s “actions”, what the feedback is and how it gets incorporated into the inference-system (or else kills it off), and how the ”model realities” of different systems affect each other – then it could be the sort of special functionality I discussed in the previous post, that supports a genuine evolutionary process.

Yes, this is exactly what I mean.

To put this into mathematics – is there a way to specify logically what the elements of such a system would need to be? I imagine something like a flow-chart that pictures both what happens internally within each inference-system to test the model against data and to generate actions, and also what the external linkages would need to be. Does the “data” for each system consist simply of the “actions” of the other systems?

As it's clear that this is an open issue and I don't have any final or complete answers, I can go ahead and give some ideas just for the discussion w/o misunderstandings.

is there a way to specify logically what the elements of such a system would need to be? I imagine something like a flow-chart that pictures both what happens internally within each inference-system to test the model against data and to generate actions, and also what the external linkages would need to be.

I have some of my own thinking here but it would be wrong to call them "logically unavoidable" since by consistency of reasoning here we see that even the description of the evolving inference systems itself is subject to the same constraints and is itself evolving too. So all *I* have is an expectation of this "logic".

My abstraction is composed of some components

* A finite set of distinguishable events that we can think of as indexing the expected types of events that the observer can distinguish. I picture this as a finite index, which we can consider one-2-one with the set of bounded integer. The bound represents the bandwidth of the horizon.

* An internal structure (in general a composite structure of sets of sets, which defines flows between the sets, that can be interpreted also as a non-commutative information space). This represents the observers "memory of the history of events", but it's not optimized to "store time histories", rather the information that initially arrivs as streams, are processed and optimized for the benefit of the expected future. Ie. we "remember" what we think is useful for the future, the rest is discarded. which is pretty much like we believe the human brain works. The human brain is not optimized to remeber correct historical sequences. Our memories instead have the purpose of helping us navigate into the future. Only some disorders of the brain cause humans to have extremelty good detalied memory, such as amazing photohtaphic memory etc. But then of course these disorders have other side effects, that has to do with forseeing the future, social interactions etc. I picture this as a finite set of sets of event histories. Where information can flow from one set to the other, and this is defined by various datacompression methods. I imagine a total bound of the set of sets, so that there is a complexity bound on the total memory system. This represents the information capacity.

* Implicit in this internal struture, is internal flows that are entropic in nature and are defined by the relations between the sets. This defines rataional expectation and natural actions, that are simply entropic to it's nature. But the different between simple thermodynamics is that we have here non-commutative structures - this means that more sophistication and in particular CYCLIC (an not just dissipative) flows appear! I'm working on mapping out these expected flows (in this language the DETERMNISTIC evolution of QM) will be replaced by the observing systems EXPECTED evolution of the image of it's environment. Now, these two things conincide in situations where the expectations have reached and equilibrium and the expectations of the future stay consistent with the actual future so that a revision of just the state of the structure and not the entire inference backbone is needed.

There are many compnents in this that are still troublesome. The above is still just a "snapshot" of the entire story. In particular are the index sets, the internal structre and the transformations, as well as the SIZE(bandwith) of the index, and the SIZE (capacity) of the internal memory structure also evolving! To determine their values we get even more self-references. But the size and evolution of these measures I associate to the problem of the origin of mass and inertia. The SIZE of the microstructure (information capacity) is a measure of the intertia or "mass" of the memory strcture. And to understand how an inference system can gain or loose this inertia is one of the KEYS to understand the origin of gravity in this picture. It's hard to explain shorlty but this entire picture actually predicts that TWO inference systems, does "attract" in a way that depends on their inertia and in the sense that their "information divergence" decreases the more they interact. Similary the resistance against change, when exposed to conflicting information is also directly related to thsi complexity. Anyway, I'm working slowly on some of these things, and the expressions for the expectations and calculating say "probabilites" for different future gets very involved, but they are essentialyl combinatorical expressions. But in the continuum limit or commutative cass they have interesting similarites to transition amplitues that goes like e^-M*SKL there M is the "size" of the conflicting informationm and SKL information divergence. But when the structures are non-commutative (which they of course are in all but trivial cases) things get more hairy and the problem of counting the set of possible transformations is still something that's bugging me.

The problem is that all this ideas, remain completely incoherent to anyone not tuned in on the thinking. That's why I have to take this to the next level and make contact to some of the current physcs and at least come up with some predictions in order for anyone to even care.

But this is of course just like any othre programs, say ST. Except that are big programs, with ALOT of people working on it. As far as I know there aren't much "communities" working in this direction.

Does the “data” for each system consist simply of the “actions” of the other systems?

Yes, this a decent way ot putting it. Also "the other systems" from the inside point of view, simply form "the environment". So the systems data is the actions (including the environments backreations from the systems own actions) are originating the "boundary" to the unkonwn. Which I mahtematically picture as an indexed horizon the defines the boundary between the obsevers knowledge and new input. Where each index represents and distinguishable event. From the inside the environment does not have a definite structure. All we have is a communication channel, and the already present expectation that indirectly contiains "an image" of the environment. But still this is nothing but an expectation of the environment. Which for example contains strange things such as superpositions etc.

/Fredrik

 

[ConradDJ]

Fredrik – there are many interesting details in your post... I wish I were better equipped to get into the specifics. But I see that you’ve made a lot of progress with a difficult question – how to define what’s minimally necessary for the evolutionary process you have in mind.

I think what you take as fundamental about the physical world is its lawfulness. Not that things are lawful a priori – but it’s only insofar as they do act in accordance with certain de facto laws that there can be any predictability or coherent communication between them.

So you envision a lawful universe gradually coming into being as a result of many different interacting participants, each trying to interpret its own environment as lawful, and acting in accordance with that interpretation... which contributes to the lawfulness of the environment for the other participants.

To make this work, you need to assume certain minimal structure is just given – i.e. the sets of distinguishable events, and information flows between them, etc. Of course this is the same procedure as other theories – you set up a mathematical model based on a certain elementary structure, and see if something like actual physics can get generated out of that. In your case the basic structure is fairly complicated, because – like me – you expect that the basic structures actually have to do something – i.e. to generate and test hypotheses about what the “laws” are, and to express those laws in “actions” that can be interpreted by others.

My own guess is that the “lawfulness” of physics evolves in response to something more basic. If we can assume there exists some kind of definite information in the world – e.g. “distinguishable events” that can be indexed and ordered in some way – then I can see that the key issue would be how to generate non-random orderings.

But to me the primary lesson of QM is that the existence of definite information can’t be taken for granted. Nothing is “determinate” in physics that isn’t actually determined in a measurement-interaction – which always involves other kinds of interactions determining other types of information. So physics is not only about evolving toward an “equilibrium” in which everything is predictable according to some set of laws. More basically, the laws have to support the different kinds of interaction-contexts in which physical information can be defined and communicated.

I think you might say – measurement is also an inference-process. It’s not just a matter of inferring the “laws” that make data-streams predictable... it’s also a matter of inferring the contingent “reality” that the data-streams are telling us about. But this inference still needs some sort of definite observable data to work with... and that for me is the key issue. What we observe may be partly lawful, partly random, but in either case it’s determinate, in some degree. So where does this observable definiteness come from? How is it physically accomplished?

This idea is far less developed than yours, so far as mathematical modeling goes. But then my personal goal isn’t to create a new theory. What I would love to be able to do is to show how each of the various structures described by existing theory contributes to making a world of observable information.

We know that whatever the physical world may be, it does in fact succeed in making a huge amount of information observable. QM even suggests there is no information in the world that is not actually “observed” in some way by something. But it’s very difficult not to take the existence of observable information for granted... to see it as something remarkable that the specific, finely-tuned physics of our universe makes possible.

 

[Fra]

Hello Conrad, I agree there are so many things to discuss here and it makes it hard. I'll comment more later but just a reaction on your first paragraphs.

I think what you take as fundamental about the physical world is its lawfulness. Not that things are lawful a priori – but it’s only insofar as they do act in accordance with certain de facto laws that there can be any predictability or coherent communication between them.

We may enter the grammar here but I'm not native english speaking and I am not sure I clearly understand your distinction between "lawfulness" and "beeing lawful".

But let me say this: I certainly do not assume that there are forcing laws on the observer/inference systems. Neither do I assume that there are forcing logic in the inference system FORCING them to behave "rationally". There is still an uncertainty here.

It's howver my conjecture that it's the best possible guess to assume rationality. This does NOT mean that I think that rationality is forcing. This is a big difference for me. Perfect rationality does not even exist as there is no certain way to decide what is rational and what's not. The entire construction containts natural uncertainty, this also causes a small, but important "variation", that's important in the big evolutionary picture.

I call it a conjecture but in a certain perspective it almost follows unavoidable, since by construction a "rational action" is the only self-preserving action. Thus, in certain context I think one can even almost "prove" that irrationally acting inference system are not stable, and thus aren't observed in nature - except possibly really OFF equilibrium (think big bang or something similar).

So in my view, the fact that we EXPECT rationality BUT it's NOT PERFECT is even a key feature.

Ok.. this was just my reaction on your "lawfulness" vs "beeing lawful" but maybe I got you wrong. If so please explain. I like to think I've got a "decent" english but sometimes it's obvious that it's not my native language.

(I haven't raedh the rest of your reply yet.. mor lateR)

Edit: One can even summarize my conjecture in the very plausible conjecture or axiom that the only rational conjecture, is to assume rationality in unknown inference systems. However rational action does NOT mean deductive certainty. It's not what it means. It just means rationally placing your bets and play the game. Some people objected to this by referring to rational theory in economics and arguing that not all market players are rational, but that objection is based on a misunderstanding of what I suggest. Rationality they speak of is not of intrinsic kind.

/Fredrik

 

[Fra]

God morning,

So you envision a lawful universe gradually coming into being as a result of many different interacting participants, each trying to interpret its own environment as lawful, and acting in accordance with that interpretation... which contributes to the lawfulness of the environment for the other participants.

When I compare this phrase with what I mean I think I see what you mean. I think the word lawful could be replace by "rational" and it is like how I would gave phrased it. Lawful makes me associate to deductive logic, but that's now what I mean. I mean rational and "lawful" in the general sort of inductive sense.

The key point in view is that the REASON for this "rationality" is not it's logical necessity in the deductive sense but rather it's necessity in the sense that it's the only constructive or self-preserving way. Since this is hard to prove formally, although I see it as aslmost obvious, I like to call this a conjecture or axiom.

If we step back and ask what our or any observers basic task is?

Is the task to predict the future given the present? NO, not quite.

The basic task is, howdo we act in a situation where we in fact to not know the future, but only have incomplete guesses? Here my conjecture of rational action comes in. This is idea, is what I am trying to formalize and translate to mathematical inference models and then ultimately to connect to physical interactions.

To make this work, you need to assume certain minimal structure is just given – i.e. the sets of distinguishable events, and information flows between them, etc. Of course this is the same procedure as other theories – you set up a mathematical model based on a certain elementary structure, and see if something like actual physics can get generated out of that. In your case the basic structure is fairly complicated, because – like me – you expect that the basic structures actually have to do something – i.e. to generate and test hypotheses about what the “laws” are, and to express those laws in “actions” that can be interpreted by others.

Yes, that sounds in tune! I need certain structure, BUT this structure is ALSO evolving, which is why I called the previous description just a snapshot.

Your also right that the importany point is not just to "predict the future", the more basic point is "what actions to take" given a certain expectation of the future. Also the important question is not just to falsify or cooroborate a theory - the really important point is howto EVOLVE the theory.

This is a basic trait of a learning model. To just fire a statement and evaluate it as true or false is a trivial matter. The deep part is how the statement was generated, and how the feedback of evalutations evolve the next guess. Anything that doesn't care about that detail in the "scientific model" is just sterile an unable to intelligent development. Ultimately I picture this in connection to "survuval" and fitness of the inference systems. A system that fails to rationally revise it's opinon as new information arrives, simply gets ripped apart by it's environment.

Again, this idea is what I try to formalize, and turn eventually into a new framework for interactions and phenomenology in physics. But one has to be fair and say it's a massive task unfortunately, but I find it so plausible that it's nevertheless - irresistable to and even irresponsible no to - try.

/Fredrik

 

[ConradDJ]

The basic task is, how do we act in a situation where we in fact do not know the future, but only have incomplete guesses? Here my conjecture of rational action comes in...

...the important point is not just to predict the future , the more basic point is what actions to take given a certain expectation of the future. Also the important question is not just to falsify or corroborate a theory - the really important point is how to EVOLVE the theory.

This is a basic trait of a learning model. To just fire a statement and evaluate it as true or false is a trivial matter. The deep part is how the statement was generated, and how the feedback of evaluations evolve the next guess... Ultimately I picture this in connection to survival and fitness of the inference systems. A system that fails to rationally revise its opinion as new information arrives, simply gets ripped apart by its environment.

One can even summarize my conjecture in the very plausible conjecture or axiom that the only rational conjecture, is to assume rationality in unknown inference systems.

Thanks, Fredrik. Here’s the picture I’m getting from you so far –

A population of inference-systems evolves toward collective “rationality” because to the extent that each system acts “rationally”, they create an environment for each other in which each system can better succeed in behaving “rationally”. Given that what it means to be “rational” is not given a priori but discovered or invented over time, as the population evolves.

This works because in order for any system to compute its own rational behavior, it has to assume rational action on the part of the others. Understanding that “rationality” involves guesswork, inductive approximation rather than deduction, and that the action of other systems too will only approximate “rationality”.

So there will always be a certain degree of randomness in the environment, because systems are always making partly-wrong guesses about the rational principles other systems are following. Also the ability of individual systems to compute actions based on the given data may be very weak, to begin with. But over time, as individual systems learn to make better predictions and use them more effectively to generate actions, the environment will become more predictable.

That’s because systems that can’t guess successfully eventually cease to function, and aspects of the environment that don’t tend to become more predictable become destructive, undermining the possibility of rational inference.


Now it’s not clear at this point whether there’s anything in fundamental physics that can operate like this, as an evolving inference-system. But whether or not this picture of interactive learning-systems turns out to be relevant to physics, it’s very interesting in itself... and if a mathematical model is feasible, I would think it would be relevant in quite a few other areas where learning-systems certainly exist.

My guess is that this model assumes too much to work as a basis for physics. Not that it assumes more than many other theories – Smolin’s CNS as one example. But as I suggested in the OP, I think we need to question some of the key things we take for granted in physics. So I would not want just to assume that physical systems can store indexed data-sets over time, and perform mathematical operations on them.

Even more basically, I don’t want just to assume that the “actions” of one system can produce “distinguishable events” for other systems. I think the fact that there are distinct systems that communicate information to each other is the main thing we need to understand about the physical world and how it works – given that information is only “distinguishable” (or “determinate”) in a context of other information, communicated through other channels by other kinds of systems.

But apart from my own peculiar notions, I get where you’re coming from. I’m glad someone takes seriously that there’s a functional dimension to physics, and that we can try to understand it in terms of the requirements of an evolutionary process. And the method of your madness seems to be similar to mine – i.e. to trust a certain basic intuition about how things work in the world, and try to find language – in your case mathematics – to make it more concrete.

 

[Fra]

Now I think your summary of the main idea is right in tune with what I mean!

But whether or not this picture of interactive learning-systems turns out to be relevant to physics, it’s very interesting in itself... and if a mathematical model is feasible, I would think it would be relevant in quite a few other areas where learning-systems certainly exist.

Excellent point. I'm glad you see this. This is why part of what I'm doing is not just physics, it's more than that. It's rather a new framework for theory building and inference in science, but my main focus lies on physics. Although there are very clear parallells between other fields. This does not only allow utilising this idea in other fields, it also means that intuition and understanding of other fields can HELP develop our understanding of physics.

The deeper aspect of this, which underlies also my choice of analysis is that in order to understand a scientific theory in the deepest way, one has to understand the scientific process, and function a theory has.

Here many physicists seem to have a very superficial view. This is much more complicated than just falsification and corroboration. Unfortunately a popular opinon here, where Popper joins in, is that these "problems" are not scientific problems, they are largley problems of psychology of the human brain. I am convinced that that is a mostly confused position. When I read poppers book some years ago it struck me as the work of someone what seeks a way to deny the confusing but true nature of inference.

Somehow who does seems to understand this extended application to social theory is Smolins "side-kick" in his evolving law contexts - Roberto Unger :) I think Unger understands his better than Smolin, at least that is my impression from listening to some of the Perimeter talks of them both. So from my perspective, I think R.Unger has had good influence on Smolin. Unger may not be a physicist, but he seems clever.

Now it’s not clear at this point whether there’s anything in fundamental physics that can operate like this, as an evolving inference-system.

You have a point here. If it was really CLEAR, then everybody would work on this, but they don't.

I can just speak for myself and all my understanding an intuition tells me that this makes perfect sense even for physics. It's how I understand QM for example, except of course that "my understanding" implies that QM as it currently stands can not be fundamental.

Without this, I have to admit that QM would be very confusing for me. I would even say it was my attempts at understanding physics, including QM and the issues With GR and infinities that has lead my to this position.

So for ME, this interacting and evolving inference-system stuff has EVERYTHING to do with physics.

But you are still right that from the somewhat "mainstream" or objective scientific perspective of today this is NOT clear.

It's about as unclear as ST is to me. ie. what does these wriggling strings has to do with physics?

But this will remain unclear until somehow shows how clear it is, and what this can do. Also the connection between physics and science, and scientific process which is generally agreed to be an inference process, is I think undoubtful. Therefor my suggestion seems logical and rational, and in this sense I think it's LESS unclear than say ST philosophy, where it's a total ad hoc trick.

What I'm suggesting here is not ad hoc tricks. It's more a general appeal to analyser the process whereby theories and laws come to be, in a more proper way than say Popper did.

So I would not want just to assume that physical systems can store indexed data-sets over time, and perform mathematical operations on them.

Even more basically, I don’t want just to assume that the “actions” of one system can produce “distinguishable events” for other systems. I think the fact that there are distinct systems that communicate information to each other is the main thing we need to understand about the physical world and how it works – given that information is only “distinguishable” (or “determinate”) in a context of other information, communicated through other channels by other kinds of systems.

I've actually give my assumptions quite some thought. They aren'y ad hoc starting points. After all my vision is not just mental understanding, my vision is a mathematical model for physics that allows predictions and computation of statistics etc. I have really tried to find the simplest possible constructive starting points. I could give some further elaborations and details beyond the main ideas we've alreadyt discussed, but instead of that maybe I can firsst ask:

Given the basic idea here - what starting points would you chose? After all, I assume that the result we need is something that makes a difference? And without a quantiative predictions (=mathematics) what difference do we make?

The human brain, can understand this without mathematics. Because even a humanist frame does all this magic. But to translate this "insight" into a mathematical model, is the difficult thing, but this is what we need to do. So in some way, we need to start putting mathematical clothes on these ideas...I've been thinking about this a lot and what I'm now trying to do is simply the simplest way I could come up with. But if there is a better way I'm open for that.

/Fredrik

 

[Fra]

But apart from my own peculiar notions, I get where you’re coming from. I’m glad someone takes seriously that there’s a functional dimension to physics, and that we can try to understand it in terms of the requirements of an evolutionary process. And the method of your madness seems to be similar to mine – i.e. to trust a certain basic intuition about how things work in the world, and try to find language – in your case mathematics – to make it more concrete.

Yes, I think although we may have slightly different emphasis I think there is a similarity in the our thinking. I think that's why I find your points highly readable.

If we go back and consider how mathematics was developed, it's basically first founded by notions of logic, true and false, and various ways of COUNTING. This is exactly where my notions of distinguishability (boolean) states come in. Either you can make a distinction, or you can't. Then I add a notin of counting. To basically count how many times this occurs.

Eventually this boils down to a way to "count evidence" and thus form a rational opinion.

Next, comes the issues of howto keep counting records, when resources are limited. Then distinguishable patterns in the data, can be used to instead count patterns, so that this works as a compression of data.

I think that Quantum mechanics is the result of optimal inferences made during the constraints of bounded resources. Then the non-commutative information spaces are simply more fit than the commutative cases. The competition requires quantum or non-commutative logic for survival. What I'm offering here is a way to understand also what "non-commutative logic" is and how it naturally extends the commutative logic and normal probability, into quantum logic. But this no just inventing explanations from something we want to undertstand (QM), this framwork also comes with suggestions how to extend QM to incorporate gravity.

/Fredrik

 

[ConradDJ]

I've actually given my assumptions quite some thought. They aren't ad hoc starting points. After all my vision is not just mental understanding, my vision is a mathematical model for physics that allows predictions and computation of statistics etc. I have really tried to find the simplest possible constructive starting points. I could give some further elaborations and details beyond the main ideas we've already discussed, but instead of that maybe I can first ask:

Given the basic idea here - what starting points would you chose? After all, I assume that the result we need is something that makes a difference? And without a quantitative predictions (=mathematics) what difference do we make?

Fredrik – I agree that your assumptions seem well thought-out. And I agree with you that this business of “inference” is a remarkable evolutionary mechanism. The history of science is fascinating to me, because it shows all kinds of strange pathways that led to the major discoveries.

And I agree that Popper’s attempt to identify a logical core to this process is off-base. I’m not a mathematician, so I’m not in any position to judge whether your construction is viable from that point of view... but I’m happy to trust you that this approach is a good bet.

As to my starting-point – The basic thought I’m working from is that there are separate systems connected to each other through momentary interactions, where the interactions in some way "make a difference." Like you, I want to avoid assuming any background reference-frame in which these “differences” can be defined. The basic fact that impresses me about our universe is that everything observable in physics is in fact definable in terms of other observable aspects of physics.

Of course it’s possible that the physical world is built on structures that are unobservable, such as the intrinsic geometry of spacetime, or “hidden variables” built into things as intrinsic properties. Traditionally, physics has always assumed an underlying, unobservable reality of “things-in-themselves” with intrinsically determinate characteristics. The goal of theory has always been to identify the simplest set of hypothetical unobservables that can account for all the phenomena we observe.

My basic idea is that no unobservable reality can in principle account for the fact that things are observable. In order to be observable, something has to be defined in terms of other observables, and those have to be defined in terms of still others. So even apart from QM, the fact that anything is observable, in our world, shows that physics has a remarkable kind of self-referential structure – since it clearly provides the physical means of measuring each of its observable parameters.

QM adds to this strong evidence that there is in fact no determinate information in the world that isn’t part of this self-referring measurement system – no definite “reality in itself” that underlies the web of phenomenal interaction. But even if there were such an unobservable reality that “causes” everything to appear the way it does, it still would not explain how or why all these appearances work together to provide measurement contexts for each other.

So I take this as the primary thing we need to analyze and explain. That leads me to a starting-point involving the topology of networks, per the suggestion I made in another thread.

https://www.physicsforums.com/showthread.php?t=331008"

But this is, as you say, just a “mental understanding” – and one that seems to be quite difficult to get across, at that! So my sense is that you and I share some basic assumptions, but are working on different levels in trying to clarify what those assumptions imply.

 

[Fra]

As to my starting-point
...
the suggestion I made in another thread.

https://www.physicsforums.com/showthread.php?t=331008"

But this is, as you say, just a “mental understanding” – and one that seems to be quite difficult to get across, at that! So my sense is that you and I share some basic assumptions, but are working on different levels in trying to clarify what those assumptions imply.

Conrad, I skimmed the first two of your posts in that thread, and you are right that these ideas are hard to convey. Probably partly due to somewhat difficulties of finding a unambigous language for it, and second because it suggests a new way of thinking that's simply alien to many. It may take a while to digest and see the new "logic of reasoning" in what first seems as madness in the light of the old thinking. So it's not hard to understand why it's hard to get across.

I agree that there are clear similarities in our thinking.

In particular things like this

Instead I wanted to focus on an aspect of the world’s structure that I think is underappreciated -- that is, to the extent information in the world is observable, it must be able to be defined entirely from a point of view inside the interaction-web, in terms of the “experienced” structure of interaction.

This is exactly what I mean with intrinsic(=inside perspective) inference. To me, what you say here is pretty clear.

What I suggest is that all we have are "interacting" inside views, but unlike Rovelli I do not assume that interactions are constrained to OBEY the mathematical structure of QM, with fixed hilbert spaces etc. Instead the question I then ask - that Rovelli does not (*) - is, how can this insight also lead to predictions of the interactions? Ie. can our further analysis of this "idea" put constraints on the interactions themselves? OF course I've convinced the answer is yes.

So although as you say, you seek a more philosophical stance and I try to find some kind of formalisation of this, I symphatise with a lot of what you say.

(*) In addition to this, I just have on more comment with regards to your reference to QM. It seems to me that you sometimes use QM as an argument supporting your argument? And that I THINK you enjoyed rovellis RQM? Here I just want to say that, as far as I am concerned, what I am talking about here, and what I think you are talking about as well although it's hard to convey details is definiteily MORE radical than QM indeterminism and Rovelli's RQM!

This is a very important point for me. As you know, QM is still deterministic. The evolution of the information state in QM, is deterministic. Also Rovelli's nice talk about relations and observers needing to commmunicate to compare results are broken when he still simply assumes that all communication OBEYS the mathematical rules of QM.

So Rovelli's RQM is an attempt of something say "in this direction" but not sufficiently radical as he does not want to CHANGE or GENERALIZE QM, all Rovelli does is to find a different interpretation which keeps the same mathematics.

My conclusion is that sticking to the very reasoning I think we both have praised, implies by consistency of reasoning that we are forced to conclude that the rigid structure of QM, simply CAN not be the proper framework we seek, it is rather a special case where the observer is NOT interacting with the system, and qualifies as an infinite information sink. Two conditions that are completely off-sense if we require that the pictures we seek is a proper inside-view.

/Fredrik

 

[ConradDJ]

What I suggest is that all we have are "interacting" inside views, but unlike Rovelli I do not assume that interactions are constrained to OBEY the mathematical structure of QM, with fixed hilbert spaces etc...

As you know, QM is still deterministic. The evolution of the information state in QM, is deterministic. Also Rovelli's nice talk about relations and observers needing to communicate to compare results are broken when he still simply assumes that all communication OBEYS the mathematical rules of QM.

Thank you for your kind words!

I interpret Rovelli's RQM paper a little differently, which is why I like it. But in any case, Relational QM appears to have been a curious side-step for him -- I don't see any solid connection with his work on Loop Quantum Gravity.

For me, it's not that QM gives us deterministic laws that things must obey. It's just that empirically, these are the laws we've discovered (so far) -- including the fact that measurements turn out to be random, "unlawful" events, without which nothing would be "determinate". So the "determining" process is always only partly lawful.

Then the question is -- what kind of universe would operate this way? Rovelli gives only a partial answer, but it's in the right direction... He says -- it must be a universe in which systems don't have real, intrinsic states "built in". So he gives a dynamic picture of systems communicating with other systems -- "asking questions" of each other and giving answers -- that somehow eventually constitutes a shared "objective reality".

He doesn't try to tell us how this happens. His goal is just to derive the formalism of QM from axioms representing this questioning process -- so in other words, he tries to make it plausible that the Hilbert-space structure of QM is what such a universe would look like.

As you know, others have worked at deriving the formalism from the logic of inference, also with some success. And any successful theory would have to show why QM works so well. But it’s still an open question – what’s ultimately going on here that gives rise to this kind of observed structure? So the QM structure is what we want to explain, not the ultimate (and incomprehensible!) explanation.

The way I use QM is at a more "primitive" and philosophical level, but in parallel with Rovelli's argument. I use it to suggest that a real-time process of “determining” reality is basically what the universe is doing. Like Rovelli, I believe that the basic obstacle is our traditional belief that some kind of reality just has to be given to begin with, in the nature of things. (But in his LQG work, it seems he may slip back into that belief?)

You focus systems that use data from past interaction to direct future interaction, according to rules that evolve by trial-and-error. This seems to address an aspect of this “determining process”. You’ve identified what seem to be the basic structures needed to make this work. But it’s possible that the kind of system you’re dealing with evolved from something more primitive.

As an analogy – human language is an extremely effective system of communication based on words, syntax, grammar. But it presumably evolved from a more primitive, wordless connection between people that had none of these kinds of structure. Or likewise, the biological process based on replication of RNA/DNA must have evolved from more primitive mechanisms that did not have available any way to store huge quantities of information in long-lived complex molecules.

So it's possible that my efforts to understand how communication channels work, in physics, are complementary to your evolutionary process, which seems to assume the existence of such connections between systems (or between data-sets within a system).

Given that "communication" in physics is not basically a simple and reliable data-transfer mechanism, but more like the process involved in human interaction. This is a process of mutual "guessing" about what the other person might mean, where we can gradually come to a common view of the world, but it's also a process of trying to make and maintain connection with the other person at a level more primitive than words or ideas.