[Anthropic Principle Note] - A New Kind of Science: The NKS Forum

A New Kind of Science: The NKS Forum

Pages:1

Anthropic Principle Note
(Click here to view the original thread with full colors/images)

Posted by: Jason Cawley

I thought some might be interested in going through some of the notes in the NKS book, gradually, and discussing some of the big ideas that are touched on there. So I thought I'd pick a note each week - on some philosophic idea that interests me - and explain something about it, and invite questions.

I thought I'd start with a note on page 1026R entitled "the Anthropic Principle". It runs -

It is sometimes argued that the reason our universe has the characteristics it does is because otherwise an intelligence such as us could not have arisen to observe it. But to apply such an argument one must among other things assume that we can imagine all the ways in which intelligence could conceivably operate. Yes as we have seen in this book it is possible for highly complex behavior - ultimately not dissimilar to intelligence - to arise from simple programs in ways that we never came even close to imagining. And indeed, as we discuss in Chapter 12, it seems likely that above a fairly low threshold the vast majority of underlying rules can in fact in some way or another support arbitrarily complex computations - potentially allowing something one might call intelligence in a vast range of very different universes.

That is what I call a big idea, though it may present itself as almost a throwaway. It is really explaining what NKS has to say about an existing debate. There are a fair number of notes that have that character.

The anthropic argument is typically advanced as an attempted explanation of a supposed specialness seen in things like universal constants, the apparent age of the universe, conditions in the solar system or on earth - all supposedly necessary to the existence of intelligent life, because apparently necessary for the particular developmental, evolutionary pathway that led to us. People say things like, without Jupiter stabilizing nearly circular orbits, liquid water doesn't dominate the earth ecosystem, or what have you.

There is something a little forced in this reasoning. The actual is always a tiny subset of the possible. Actualization is a progressive narrowing of possibilities. Something definite happens, and afterward that something can be specified with arbitrary precision. If every precise detail is supposedly "necessary", then one can detect arrangement or coincidence in anything. So what the argument really winds up turning on, is whether the characteristics pointed out, are required not for specific details, but for the general phenomena one imagines things are arranged to produce.

Less abstractly, when one imagine, say, water is necessary for intelligence, one implicitly claims that it is essential to the evolution of intelligence that it happen as it did with us. It is perfectly plausible, a priori, that water's role is accidental. Does it occur in our history? Certainly. Is it necessary to the particular pathway our history took? Sure. So are a lot of things. But is it necessary to the existence of anything like life? Or is intelligence restricted even to life, let alone to water based life? So the argument winds up being a claim, not about how something did happen, but about alternate ways, that supposedly can't happen.

Into this existing argument, NKS makes the point that universality is easier than people expected. We might have thought one needed something as elaborately engineered to have the property of universality as, say, arbitrary Turing machines. One might easily have expected that if one didn't explicitly required it of some system as a property, it would be unlikely it would arise anyway. This intuition turns out to be wrong. Universality can crop up in simple places where it was not engineered in, or expected, in ways we did not see beforehand.

So on an immediate level, the point is to be careful about claims that systems (even simple ones) can't do this or that. We don't know half as much about such things as we often assume. Maybe it can, maybe it has some clever trick for "accomplishing" property X that we just haven't been clever enough to think of beforehand, ourselves. When someone says, "water is needed for life", he is saying "you can't do the trick with methane" etc - and this is not so obvious.

The last part of the note suggests that at least an analog of intelligence might be possible in universes that do not satisfy any elaborate constraints. It is relatively easy for formal universality to arise. It would not take much in the way of rules to allow complexity to occur in a universe, whether largely like ours or not. Complexity, at least, should not be expected to be "delicate".

The note hedges on the possible relationship between complexity and intelligence - "something one might call" and "ultimately not dissimilar to". Personally, I think the case for "necessary" is very strong - what is not capable of complexity generally does not strike us as intelligent. The case for "sufficient" is less obvious, but not easily ruled out.

Intuitively one might think there is still a great gap between the two, and this intuition may eventually prove correct. (There may be some additional trick involved, e.g.). But it is not easy to put limits on the possible behavior of a system complicated enough to support universality. Granted in simple enough conditions, some rule may yield behavior that just looks like overall mixing or randomness. Besides the possibility our own analytical limits may be partially responsible for that impression, there is also the question what it might do with special conditions.

When you try to formulate a denial as a theorem, you get an idea why this might be hard to prove. If you can specify any finite formal output you'd be satisfied by, you can see the problem. You wind up saying something like, "universal system X, from any initial conditions, will not produce finite state Y". When in place of a known universal system, you have instead a class that includes many complex behaviors, amongst which there are probably members that support universality, the problem is pushed back a step or two, but fundamentally remains.

So, overall, what does NKS have to say about the anthropic principle? Mainly that we should not assume intelligence is as special as we have typically taking it to be in the past. We should not assume the particular pathway that brought us here is the only one that might produce intelligent observers. These were observations available before. The results in the NKS book have simply added evidence for these points, making it more plausible than before that our particular developmental pathway, in all its detail, is not uniquely necessary to something like intelligence arising.

I hope this is interesting.

Posted by: Tony Smith

A comprehensive response to this thread could take a lifetime, so I will only focus on a few points.

The actual is always a tiny subset of the possible.

This is a very important but grossly understated point. Failure to appreciate just how "tiny" leads directly to common but absurd claims that we would have twins elsewhere in an infinite cosmos. It is the difference between countable and uncountable infinities.

NKS makes the point that universality is easier than people expected.

By the same argument, universality alone is not sufficient. The initial conditions need to be credibly small for the implied claim of "natural" to stand scrutiny.

we should not assume intelligence is as special as we have typically taking it to be in the past.

If we are talking about the intelligence needed to hold this discussion and recall that there has been at most marginal increase in the average complexity of life in the 250 million years since the Permian extinction, that claim needs to be considered against the fact that even within terrestrial vertebrates, that kind of intelligence is an historic aberration.

Personally, I think "intelligence" has become way too loaded a term to be useful in this kind of discussion. "Intentionality" is a much clearer criterion and I would be last to claim that intentionality might not arise in quite different mediums that the boundary of solid, liquid and gas which has been colonised by CHO-based biology.

And as much as it might be the opposite of what any of us would wish, placing too much weight on "intelligence" at the same time as raising important questions about the directedness of biological evolution just plays into the hands of those who promote the argument for design.

The weak anthropic principle tells us things that are just as important as other (and equally poorly understood) principles that have been revealed through the study of complex systems. The fact that we are here and can observe our world provides useful information about that world.

Despite arguments elsewhere that recursiveness may be overrated, to my way of thinking it is the power of shared recursive language, of conversation, rather than misplaced faith in "intelligence", which leaves open the possibility that our era may be the seed of a change in the universe of possibility which ranks with the Cambrian explosion. Or our misplaced faith in our own intelligence may lead as the way of the Ediacaran fauna.

Posted by: Fiona Maclachlan

Re:

The note hedges on the possible relationship between complexity and intelligence - "something one might call" and "ultimately not dissimilar to". Personally, I think the case for "necessary" is very strong - what is not capable of complexity generally does not strike us as intelligent. The case for "sufficient" is less obvious, but not easily ruled out.

Intuitively one might think there is still a great gap between the two, and this intuition may eventually prove correct. (There may be some additional trick involved, e.g.). But it is not easy to put limits on the possible behavior of a system complicated enough to support universality. Granted in simple enough conditions, some rule may yield behavior that just looks like overall mixing or randomness. Besides the possibility our own analytical limits may be partially responsible for that impression, there is also the question what it might do with special conditions.

I don't know if the great gap is intuition or just the way we've been brought up. I find the idea of intelligence arising as you suggest foreign but can see no obvious objection.

The note on non-Western thinking (1196L) addresses in part the question of whether NKS leads into domains that traditionally we would dub unscientific. But if we allow for the possibility of non-human intelligences out there, I don't see the grounds for a denial.

Posted by: Phillip Craig

The interesting post on the Anthropic Principle got me to re-read chapter 9 of the NKS book, and parts of it a few times more.

The material around pages 504-505 dealing with multiway networks, and a lot of the ideas in this chapter in general, raise significant questions about human perception of our universe. For example, if it is possible that there are multiple histories of the universe, each one offset from the other by a small but critical step, then our perception of the universe is limited in the extreme.

The idea that the universe was designed for us, or for life in general to be more precise, would in that case be based on one particular history. Other histories might look very different even to the point of having no intelligent life. Could such multiple histories coexist and be invisible to each other?

This could be interpreted as the many-worlds theory, or the theory that, at the quantum level, all possible states exist. But, multiple histories might also mean that there are very different streams of existence. We just happen to be in one that produced patterns that behave like intelligent life as we know it.

The idea of multiple histories also raises some interesting questions about the Big Bang theory.

Posted by: Interwoven

An element of NKS I have trouble accepting is the notion that there is nothing 'special' about our universe. It is apparent that every property of our universe favors the apperance and evolution of life. If any property of particle physics, the 4 forces or even the basic subtle elements of our universe is even slightly tweaked, then life could not exist. One has to imagine there would be an infinite number of ways (with different rules) that existence could form in ... logically nearlly every form of existence would be void of life (atleast life as we know it). So, out of an infinite number of forms existence could take, it has taken this curious form ... allowing the universe to form itself into beings that can experience itself through consciousness.
Why didn't the universe form only a tiny amount of mass, or with completley different laws of physics, with different basics then time and space? It would also seem absolute nothingness should 'exist'. The existence of anything, even a single particle seems to imply some form of intention ... now apply that to a vast universe that seems to contain a sense of logic in the rules that makes it tick. Thease traits of our universe could suggest some form of a 'creator' though that human term would likely be inaccurate.

Posted by: Jason Cawley

The philosophic problem, not really related to NKS incidentally, is that of anything at all you could say the same. Yes the actual is always special looking compared to the space of possibilities. That is really just another way of saying "actual". Something definite happens or occurs. You can imagine arbitrarily many other things that might have happened instead - though of course it is imagining. Of how many of them would you say, gee that is so special because of x, or because of y? Virtually all.

The anthropic principle is addressed to the idea that conditions for intelligence might be rare. But typically it refers not to any imaginable intelligence, but to the particular processes that gave rise to it in the one case we know about. And what of all the other special phenomena the universe doesn't give rise to but might, if it were different in this way or that? Every definite thing that happens excludes a whole bunch of other things, any one of which might seem just as "special" if it occurred. To me this just comes down to saying "the actual is a tiny subset of the possible" - it will always look "special" and in some sense it is.

Posted by: Karl Smith

I think the Anthropic Principle often taken too far as an "explanation"

In my mind it is a subset of the principle "We observe that which tends to be observable"

This is the fundamental princple behind evolution, economc equilibrium, the exact calibration of physical constants, etc.

However, it is only to remind us that the set of "possible" states of the world and the set of "observable" states of the world are not at all the same.

In science we can ultimately only deal with observable states of the world though we often fancy ourselves dealing with possible states of the world.

It is "possible" that a universe could have grown up in which the electromagnetic force was slightly different and therefore the stars would have never formed.

Such a universe is "possible" but not "observable." Hence, we should not be surprised to find that all of observations are made of universes for which that is not the case.

The same thing with evolution. It is certainly possible that, and even highly likely that, mutations decrease the fitness of an organism. However, such organisms by definition do not thrive and reproduce. Thus they are not highly "observable".

Hence, we should not be surprised that all of the organism we observe are created from a hugely long stream of fit mutations.

The probability of picking such an organism out of all "possible" organisms is absurdly low. However, the probability of picking such an organism out of the set of all "observable" organisms is essentially one.

Posted by: Adam Lloyd

I'm just a fly on the wall here, but could the limiting factor for intelligence not lay in the fact that different forms are possible to produce it -- but rather that only certain 'ingredients' that make up the form have the properties that are in fact necessary to build that ultimate form which is called 'intelligent'?

I realize the initial post is about NKS saying we no longer need to assume that any particular ingredient is necessary. That's fine. So we throw away the assumption. But let's come back to the issue again right away and instead of assuming, let's ask. Does intelligence need particular ingredients?

Based on the fact that different ingredients have distinctly different behaviours, I think that even with the assumption thrown away we can reasonably come right back to it as a "logical probability."

Let's assume that everything in the universe is ultimately made out of a single kind of thing, 'thingness' itself. Energy. (And we know that the energy of a thing is what gives it existence at all -- mass.)

I use carbon to explain but it's just chosen for the example because it fits with our own brain structure -- we know it works in other words.

Now, it is true that there are any number of ways that initial energy can be built up, but only one of them creates the thing called 'carbon', and only carbon has the properties neccessary to create more elaborate forms that will result in emergent intelligence.

What it means is that there is nothing theoretically wrong with saying intelligence could be formed with anything. It means that there is only one "thing" to form it with, and that the carbon element is a part of the form necessary to complete the statue. A statue may theoretically be just an organization of things to create an image, but it turns out that clay does the job and dry sand doesn't.

The issue is that the ingredients just past 'thingness' in general are a part of the form. The formation of intelligence starts with the formation of it's constituent parts. In other words, it is possible that not all carbon becomes intelligence but all intelligence is made with carbon -- why? Because carbon behaves in a certain way, that barium for example does not.

Is there any merit in this?

Adam

Posted by: Jason Cawley

It is possible, but it is not proven. On your analogy of clay and dry sand, maybe there is also wet sand that can do the job and we've just never seen it done. So we might think sand can't, clay can, that clay is necessary to statues. Which could be false even if every statue we'd ever seen had been made out of clay. (I know, you will tell me "clay is wet sand, at least sort of", but bear with me. That is just your analogy).

The point is to be careful about specifying certain items in a catalogue of "actual" as being "necessary" to a phenomenon seen. Of the particular instance of it in front of you, it may be necessary (carbon as underlying brick for DNA based life). Of some other instance it may not be necessary (imaginary silicon based life or "wet sand" - which of course is entirely imaginary, to be sure).

What NKS has suggested that is relatively new in all this, is that very simple arrangements of vanilla elements can track behaviors of arbitrary complexity. Arrangements so simple one can imagine some instances of them actually arising in natural systems, as opposed to in artificial ones deliberately engineered to have it as a property. And without a millioneth of the specified complexity of e.g. DNA. Bricks that are much less special than hitherto supposed might suffice.

The anthropic argument was an attempt to explain how special the bricks were. (That they were special, it took as given - it wanted to explain how something so special looking could arise). The argument was that they have to be or we wouldn't be here bothering our heads about them. In other words, it is supposed to be a sort of "survivorship bias". It is not transparently a silly argument, but it is not the strongest argument either. Because it frankly applies to too much. But if the bricks don't have to be so special, the need for it declines.

Posted by: John Newman

The fact that many different rule systems and initial conditions could possibly give rise to intelligence does put a major hole in the ID argument.

However, does it disprove the anthropic principle? Actually, I think it doesn't. If anything, NKS pushes the anthropic principle forward--only a different kind. No, our particular universe may not be all that "special," to the extent that it has life and intelligence in it, but rather many universes can have life and intelligence in them and thus the whole realm of possibility is �special� in that regard. In fact, those universes exhibiting 'intrinsic randomness generation' seem that they would give rise to intelligence inevitably. That, to me, shows anthropic principle in the very construct of reality and existence, independent of our subset 'actual universe.'

Under this paradigm, life and intelligence seems much less like a fluke or accident, and much more like an inexorable effect, or conclusion (depending on perspective).

Posted by: Jason Cawley

I mostly agree with John Newman's comments. If NKS undermines aspects of the Anthropic Principle (AP), it is only by providing a better argument against ID than AP is to begin with. I would slightly qualify it, though, as evidence in favor of this rather than having shown it. Because there is still a gap between universal complexity (UC) on the one hand, and intelligence on the other. It is not obvious whether the latter has additional requirements besides a universe that supports universal complexity easily, because we don't really have an adequate theory of intelligence.

Also, while the principle of computational equivalence (PCE) suggests universality, and perhaps with it intelligence, could arise in numerous ways, PCE is a hypothesis not a theorem. What NKS shows rather than just suggests, is that just about any universe will support UC, that it isn't special that it arises in ours. Because UC arises in formal systems beyond a very low level of underlying variety; essentially only incredibly simple "universes" would lack UC.

The other bit of qualification concerns the possibility of finding an ultimate rule for the universe. Suppose we do, will we be able to infer anything about AP from how complicated or simple the rule turns out to be? It seems to me it could suggest different answers. If it is the simplest possible rule or close to it, then it will suggest no need for AP, or for anything much beyond our universe. If there are millions of other rules before it that give rise to fully consistent universes but not to ours, on the other hand, that would suggest - at least to me - that our universe is a speck in a lot of other stuff. In between, there might be a role for AP, or arguments it competes with. Since we don't know such things yet, it seems to me reasonable to leave this much open, undecided.

Posted by: John Newman

When I hear "Anthropic Principle," I think, 'the universe is such that sentient life should come about.' When I hear "Intelligent Design," I think, 'the universe is _tuned_ such that sentient life should come about.'

A view of nature based on universality and intrinsic randomness generation is a damning arguement against ID theory, IMHO. But, concerning AP as I define it above, this view of nature at the very least makes sentient life more probable than we may have previously assumed.

I guess what it comes down to is what is the distribution 'intrinsic randomness' generating functions across all possible functions?

If the threshold for universality is relatively low in terms of complexity, meaning more complex functions are more likely to be universal, then it follows that more complex functions are likely to potentially exhibit intrinsic randomness generation. Plus, even without regard to universality, I thought I might remember it being mentioned in NKS that the rate of randomness generating functions increases with the complexity of the functions (considering only simple initial conditions).

Mind you, I'm no expert on any of this, so you'll have to correct me whenever.

I guess my point is that the strength of the AP arguement depends on what the ratio of 'intrinsic randomness generation' causal networks is relative to all possible causal networks. Would that be a correct statement?

Another question is, is one intrinsic randomness function more likely to develop intelligence sooner than another? Well, the output of some functions grow faster than others, right? So if the portion of irregularity in the output of two different functions is equal, then they should be traversing the search space at roughly equal rates, right? Unless there is some actual regularity to the irregulity, then two such functions should generate intelligence at generally the same times.

I know all this depends on the ability of intrinsic randomness functions to manifest intelligence (though I'm already sold on the idea, being that we are dealing with the fundementals of causal networks here), but if they do, then one shouldn't differ from the other by much if they both output irregularity at a similar rate, right?

I guess another good question would be, among all functions below a given threshold of complexity, how much potential irregularity exists?

That's probably way to many questions, and some of them might not make a lot of sense, but I'm just trying to get an idea of what the parameters are here.

Posted by: Jason Cawley

Start from the least possible complexity in a formal rule, and slowly increase its number of states, connections, etc. At first you will see only simple behavior, then that and a little nesting. If the rule is particularly bad at it, you may have to add more elements or allowed connections before you get to real complexity.

These changes are always discrete and, at the low end of things, are each relatively large. (By that I mean, you add an allowed color, or an extra state of a head, or a connection a step farther away, etc.) So effectively you are moving the overall complexity of the algorithm upward in a set of jumps, not a slow continuous dial. The first time you see real complexity you will see it relatively rarely. But go even a little farther, and it gets common enough you see it in a few percent of rules of that type, up to something on the order of 10-15%.

Some rules are still going to give simple behavior. The portion can change with additional changes to the components, but marginally - you more or less plateau. The overall effect is a threshold - nothing, nothing, nothing, barely there, several, about as many as you will ever see. This is an empirical generalization. It can be different for a specific scheme, but the pattern above does repeat in numerous cases with large differences in the rule scheme. That is the formal evidence.

Next one has to get clear what it means for a universe's rule. And here there are intuitive pitfalls to avoid, because in general one underlying rule generates behaviors that mimic many higher level ones, by course graining, by arrangement of initial conditions, etc. If this possibility isn't clear, consider the universal CA construction starting on page 644 -

http://www.wolframscience.com/nksonline/page-644

The section on emulation and universality continues through page 675.

Our universe clearly has real instances of subsystems that support universal complexity. Its underlying rule must support that - though perhaps only up at some aggregation level, making use of arranged initial conditions of a subsystem e.g. A single underlying rule (a hypothesis, but assume there is one for the sake of the argument) that only gives simple behavior won't work. (You could imagine all of it coming from initial conditions I suppose. But it is not a practical modeling scheme). Presumably, then, the underlying rule has enough elements and allowed behaviors etc, to be up on the plateau described above.

What role is there at this level for AP? One can imagine all the other possible rules of the same general type, with about the same number of elements, states, or allowed connections - or with exactly the same number - or with the same number or less. Out of that set, how many might support anything we'd recognize as a universe? We don't know because we don't have such a rule. But it may be anything from - it happens naturally with the first few that support complexity, to - making a universe at all requires not only support for universality but also meeting this and that constraint, and making one with physical laws matching ours means also matching this or that other set of constraints. Someone might appeal to AP to explain this exact rule out of that possibility space. Universe candidate rules have to work as universes to have observors, say.

But beyond that point, will there again be a giant hurdle between possible universes and intelligent observors? That is where NKS seems to me to greatly weaken AP. Because practically any rule that makes a possible universe will support universality. And it is very hard to say intelligence can't happen where you've already got universality. So most of the "alternate possibilities" under consideration - in the subset, "makes a viable universe" - either would have intelligent observors, or could, or we won't be able to rule it out easily.

If a theory S has to invoke something like AP to explain why the actual universe has this and that characteristic, out of 10 to the 500th power possibilities the theory also allows, it is a sign of weakness in explanatory power of that theory. The theory S does not rule out much of anything. All the particularity of our observed universe's "content" is referred to AP.

If, on the other hand, an alternate theory T suggests rules of type ABC would result in viable universes, and typically so (even from simple initial conditions in the rule's terms) once beyond some level of complexity; practically all of which might have intelligent observors; and this particular rule alpha out of ABC also recreates the known physical laws; and there are only a few billion or trillion rules of type ABC or simpler; and practically all the significantly simpler ones fail to produce anything like a universe - then that is a sign of great explanatory power in the theory T. The theory rules out simpler rules, as not capable of making universes. It appeals to the existence of observors hardly at all. Most content is explained by the theory itself.

An intermediary case would be a theory like T, with non-trivial initial conditions. If there are billions or trillions of possible initial conditions for T that all describe viable universes - simpler ones forming a smaller subset failing to do so - then T explains less about the particular laws seen in our universe. If out of those billions or trillions, many, say one in ten, still support universality and (stipulate) therefore intelligence arising under those initials can't be ruled out, then AP might still be invoked in a minor way, to explain that one out of ten. But this is not what AP typically means. It is not the kind of factor one needs something like AP to "span".

Understand, a theory can be insulated from possible contradiction simply by withdrawing its claims more and more from possible contact with real particularlity, by e.g. including more and more possibilities in the same overarching formula. It is a sign that this is going on if the possibility space compatible with the theory explodes - by which I don't mean even trillions, I mean much larger numbers than the number of particles in the universe - and specifying which version or "rule" within the theory is actually responsible for our universe is left to initial conditions, or AP, or "it just happens to be this one".

In the limit, I can rule out nothing, never face contradiction, and point to everything that happens and say it is "just so". Then I have explained nothing. At the other extreme, I can have one obvious intuitive simple possibility that leaves nothing to initials, chance, AP, etc, yet explains every particular detail actually seen. Then I've explained everything. The first is easy but not helpful. The second is probably impossible. But you can get a lot closer to the second than to the first, and that is the idea. In this theory judging sense, the more use a theory finds it has to make of principles like AP, the less explanatory power that theory has.

(Please note that how much a theory explains is a different question than whether or not it is true. In general, the more a theory explains the easier it will be to determine that it isn't true, if it isn't. But it doesn't ensure anything).

I hope this helps.

Forum Sponsored by Wolfram Research

© 2004-2009 Wolfram Research, Inc. | Powered by vBulletin 2.3.0 © 2000-2002 Jelsoft Enterprises, Ltd. | Disclaimer
vB Easy Archive Final - Created by Xenon and modified/released by SkuZZy from the Job Openings