I am comfortable around guns and generally believe that seriously restricting gun ownership in the US won't work. I think all the outrage put into trying could better be spent elsewhere. On the other hand, I know conservative gun owners who believe that reasonable restrictions are a good idea. — T Clark

No one is ever saying that a gun causes someone to shoot things. An intersectional approach is required to understand their effects here; guns are facilitators and final solutions for problems generated in the intersection of societal pressures and personal issues. Race also plays a role; the majority of these school shooters are mid-teen white guys. Sexism and racial prejudice seems to plays a role; whenever they deign to write manifestos anyway. None of these things causes any individual to shoot their schoolmates, it's not reductive like that, they are all facilitating factors bottling up alienated rage.

High school shootings should probably be considered domestic terrorism; the actions of the perpetrators should be condemned, but the issues that lead them to it should be understood so that they can be addressed.

For what it's worth, the murder rate in the US is half of what it was in 1980. 650,000 die of heart disease annually. Cancer 600,000. Let's put our energy there. — T Clark

Well... school shootings would be less... likely if... there were... less readily available gu-

I'll be over there in the corner.

A different question: what's the enemy here? Are they accounts which can't provide such an account? And what do they look like? Trying to triangulate the motives here. — StreetlightX

I don't really know. Someone who conceived of nature as one big undifferentiated glut, or one big self differentiating glut probably would be anathema here; if they refused the ability to take on the perspective of subsystems. I'm not certain I really have any direct targets here, since I seem to be out on such a limb. I'm carrying my general prejudices with me though; no to reductive notions of causality, yes to thinking in terms of becoming.

We project the idea on events. Didn't Rovelli say something along the same lines? — frank

Yes. I would love to be able to talk without the word "cause", but it is so convenient. I'm trying to think of causes systemically, as they are distributed over networks of interaction. I would prefer to do away with it all and just talk in terms of interactions and sensitivities, but if I did that everyone would recognise what I'm writing here as the schizoid scientific philosophy conspiracy theory it is rather than being enticed to follow me down a rabbit hole they might get stuck in too.

Is that why you didn't really address my posts? Because they weren't in topic? I was wondering about that. — frank

No no you helped me. The thought/being criticism is wack, though.

I have a feeling this whole story is a parable or a metaphor, but I can’t see what of. However seems far more germane to a biology forum than a philosophy forum. — Wayfarer

I think I'm doing philosophy, though I don't think the topics and questions are particularly well posed yet. It's somewhere in the intersection of assemblage theory (philosophy);Deleuze, DeLanda (and Bryant); causal network theory (statistics) and cybernetics or system control theory (which I haven't included yet but will).

Any relationship to the ‘arche-fossil’? Am I getting warm? — Wayfarer

No actually. Hasn't entered my head while writing this.

Here's what I think is at stake here (let me know if it's different): are you trying to account for the autonomy of systems without at the same time trying to entirely disengage that system from it's wider environment? That is, trying to account for a relative autonomy of systems within (variable) threshold values that when crossed (under equally variable conditions), make it so that the environment now bears upon the system in question? — StreetlightX

I'm trying to see where a knot of ideas I've had leads.

So that's one thread in it. I think this is related to a previous discussion we had (in "I Am An Ecology") about how nature learns what to care about; and how that might be fleshed out in terms of the becoming-relevant of a parameter in a dynamical model. Here, temperature, moisture, sunlight vs the rest of the environment for the seed germination process. An 'answer' I'm striving towards here is that nature (can) learn what to care about by the development of subprocesses (say, dormancy) which have signalling structures (seed sensory capacities) that somehow embed the salient features of larger system dynamics (reproductive success rates based on environmental parameters) for the subsystem.

Another thread is about system history. I've been quite meticulous to only use the word 'time' internally to a system level description; like germinating times in wild barley. (Except when describing stochastic processes, but I did also phrase the indexical nature of time in them as an 'interaction' first!). All the other 'orders' in my posts here are actually system orders (like $\text{Volcano}\rightarrow\text{Soil}$), directions of causal influence, rather than relations of temporal antecedence. To be sure, there are relationships of temporal antecedence in the posts, like the reproductive system of plants being older than the dormancy adaptation of their seeds, but these only occur when one system is a developmental trajectory of another.

These two questions are related. When a system learns to care about stuff, it learns not to care about other stuff that might be relevant. There's a mathematical feature of models in evolution (which holds for the deterministic dynamical systems that I'm aware of too) that the next time step depends only on the current one. Let that sink in and take it seriously. If our descriptions of reality are accurate and have an assumption in them that only the recent past matters, where does the relevance of all this bloody history come from? In the first post I referenced that you can 'restore' the next->current (Markov) dependence property of a model by making the state space of the model bigger; by including more relevant information. A corollary is, if you exclude relevant information, you allow more history. As Rovelli put it in the Order of Time "time is ignorance".

The last thread is one I've not written much about yet, it's in the details of the seed germination mechanism 'models' its environment. I plan to leverage:

Though they may have functional sortals embedded in their germination patterns; temperature ranges which are amenable to growth, light levels amenable to growth, moisture levels amenable to growth; switches that say germinate or do not germinate, with fuzzy boundaries depending on the seed and the plant and all sorts of other things. The variation on the individual level manifests in a distribution of germination times on the originator plant level; of probabilities of germinating given (environmental condition) and on the plant community level. — fdrake

the idea of a sortal induced by germinating/not germinating on the individual plant level ("is this environment right for germination?" check thresholds...) to show one way one system (seed germination) can model another (environmental influences on reproductive success rates of seeds) through a discussion of the good regulator principle in cybernetics.

Since you've read The Order of Time, I'm trying to tease out a sense in which an approximation of one system by another might be immanent to their relation; and this approximation is a 'forgetting' of irrelevant parameters, which induces 'history' to their relationship through unobserved relevance. Another thing I'm being inspired by here is the brief discussion of wasps and orchids in A Thousand Plateaus; one being a 'map' of another.

Speaking figuratively, the seed has 'learned' something of the expected dynamics of its own internal processes given environmental variables. — fdrake

In a response to @frank I wrote:

When canopy tree offspring have a reproductive advantage when they grow taller, this is already an aggregation on the level of the population. The unique events which cause the tree bodies to grow taller matter far less for the evolutionary pressures than the increase in height relative to other trees; to canopy higher. Evolution acts to increase tree canopying height in general, that it acts on all scales; molecular, developmental etc; to do this is is exactly a 'blurring of the details' of unique events. Evolution does not care for the how here, only a constrained how that satisfies the comparative advantage in the context of ecological constraints.

The more general feature here is that random variables allow you to make a summary characteristic of indeterminate processes in a way that depends on the events which happened (what makes the random variable realise) but does not care how they happened so long as they realise into the specific summary value. The hows and whys of how evolution acts on the mean height of canopying trees fall away in the sigma algebra of random tree-done events; and this is a certain 'forgetting' in itself, of nature deciding what is relevant. The 'forgetting' of the influence of unique events through aggregation is exactly what random variables do. — fdrake

The ability for seed germination processes to filter in what is relevant for them is precisely this kind of causal flow. Individual seeds have no choice in how their germinating process was generated; it cannot vary within seeds; but the germinating process (seeds' germinations) can vary within plants, and it can vary within populations of plants. In this regard, a seed need not have its germinating indicators tailored to its own needs to provide a reproductive advantage, all it needs is to be hooked into the reproductive success rates of the overall population of plants to garner this advantage.

In that regard, the

$\begin{align}\text{Seed}(\text{Environment}\rightarrow \text{Seed})\end{align}$

indicator system is an aggregation over seeds embedded within particular seeds. What this would look like probabilistically is:

$P(\text{germination}|\text{environment})\rightarrow P(\text{germination}|\text{temperature, sunlight, moisture})P(\text{other})$

how the germination indicators for the seed would interface here is as an approximation to:

$P(\text{germination}|\text{temperature, sunlight, moisture})$

but it is also simultaneously a filter for environmental variables which are relevant but not seed sensitising; like our old volcano friend from the original post.

Edit: sketch until I describe conditional distributions and networks.

One system can only be sensitive to another when they have a direction of variation which can interact with another.

In terms of seeds and dormancy; dormancy evolved upon the back of seed dispersal mechanisms and environmental sensitivity of seeds for their germinating conditions.

What this looks like in terms of seed populations and environmental conditions is:

$\begin{align}\text{Environment}\rightarrow \text{Seed}\end{align}$

There are many directions of variation in an environment which are not relevant for the germinating behaviour of seeds. They have evolved a sensitivity to environmental conditions which give them comparative advantages in reproductive success; this is a form of self modelling of the $\begin{align}\text{Environment}\rightarrow \text{Seed}\end{align}$ interaction within the seed. It was of reproductive benefit for the reproductive system of wheat to internalise environmental constraints on the reproductive success of wheat plants! The insertion of this process into seeds then looks like this:

$\begin{align}\text{Environment}\rightarrow\text{Seed}(\text{Environment}\rightarrow \text{Seed})\end{align}$

which dwells alongside the previous:

$\begin{align}\text{Environment}\rightarrow \text{Seed}\end{align}$

Now the environment can interact with the seed's model of its environment as well as the seed as a whole. Speaking figuratively, the seed has 'learned' something of the expected dynamics of its own internal processes given environmental variables. The seeds have become responsive to environmental considerations as a result of interacting with their environment along a direction (really a space) of variation; the seed germinating process. And in particular, temperature, moisture, light, have become internally represented parameters of the seed indicator system because they were driving forces of seed reproductive success.

What was a feed forward; the environmental control on the reproductive success of seeds; has been internalised in a simplified form within the seeds as a feedback; a sensitivity of the seed's germination indicators to their environment.

If you want to have it that Nature possesses a rudimentary form of thought, I'd be excited to hear why you look at it that way — frank

The opposition between thought and being is not an opposition which matters for seeds. They do not have substrate independent concepts (red, justice, ...), recursive grammars, or the ability to imagine angels dancing on the head of a pin. In terms of germination; they are sensitive to environments' warmth, moisture content, light level. From the perspective of a plant, the distinction between thought and being might look more like the distinction between environmental indicator and environmental condition. Not that they have the faculties to conceptually aggregate their sensitivities like we can from the outside.

Though they may have functional sortals embedded in their germination patterns; temperature ranges which are amenable to growth, light levels amenable to growth, moisture levels amenable to growth; switches that say germinate or do not germinate, with fuzzy boundaries depending on the seed and the plant and all sorts of other things. The variation on the individual level manifests in a distribution of germination times on the originator plant level; of probabilities of germinating given (environmental condition) and on the plant community level.

What conceptual distinctions make sense for a system depends on the system; the distinction between thought and being is only something that can become relevant once thought occurs, not before it. The task here is to be able to think through a perspective from its context; to let our thought take its cues from the immanent structure of a system; and not to see distinctions - types of development - in it which are not there.

The distinction between thought and being is a device for outputting metaphysical doctrines; some collapse to one pole - idealism, vulgar materialism. Some insist upon dwelling within the dyad; the myriad correlationisms. Some make the split between thought and being and express us as moments of both; like Spinoza. But what is the dyad to a seed?

If we follow a system's dynamics with our thoughts, when does the distinction between thought and being become relevant to it? Only insofar as our capacity to make errors in description of it; an epistemological issue for us, not an ontic or ontological one for the system.

Find topics that interest you. Doesn't matter whether it's within philosophy or whatever. You generally get exposed to interesting topics by studying other topics. Follow your interest. As you're following your interest, you'll learn how to follow it better.

If you're super interested in something, a book, a topic, make notes, write on here. It helps the material to stick, and it helps you explore it.

I like the story a lot. Two other things I've read recently come to mind (I guess you could say in the sense of a deleuzian series of resonance) — csalisbury

I've been listening to A Thousand Plateaus recently. I think this thread (and my recent posts) have been some interaction between that book, the recent (wonderful) The Order of Time by Carlo Rovelli, and a lingering interest in causal networks that I use to procrastinate at work.

I don't feel comfortable talking about identities and the unconscious at this point; I suppose the vantage point of the thread is supposed to be a-subjective, indifferent to the functioning of human beings. But I won't lie and say I don't have opinions on the matter insofar as they relate here. I'll spell them out once I know better what they are.

If the volcano seems irrelevant to what's happening, that's your focus, not nature. — frank

Artificial distinctions which do not reflect system dynamics can be made. Nevertheless, nature can learn to discriminate. Processes can be sensitive to others; seed germination times had a sensitivity to long periods of bad weather, those seeds that could survive the bad weather gained a comparative advantage later by being able to recolonise the area with less competition.

Such sensitivities can be sites of contingent interaction; like wild wheat germination times varying. Or they can be(come) internalised system capacities; like dormancy.

Nature makes maps of itself along its interstices. Simplified projections to either side. Like the zebra striped with grass, the pepper moth black as smog, or the dance of the long tongue moth and its flower.

So I can't be bothered writing up bits from the book. But I will describe the general thing. It wasn't barley, it was wild wheat.

Let's say I am a wheat plant and I'm growing on a hill. I disperse my seeds, they'll grow or not grow. What makes them grow or not grow is their ability to embed in soil in conditions suitable for their growth. It would be a big advantage here to have seeds that germinate based on environmental indicators of their success chance for growth.

And so that is what they do; seed germination is temperature dependent, sunlight dependent and weather dependent. There is no goldilocks "just right" here, a range of possibilities are available for germination. It would also be nice if seeds had a mode of low energy consumption they could enter if none of these conditions are met for a long while; and they do, this is called dormancy. It would be a waste for seeds to go dormant; to not germinate; without any possibility of germination, so dormancy ends due to environmental sensitivity too. So let's talk about dormancy.

Seeds have little models of their environment in them. Which is to say, they are sensitive to their environment in constrained ways. The environment is always 'on', always in some condition of temperature and sunlight and weather, the dormant seed needs just to attune to differences in their environment; the process by which seeds are enabled to attune to their environment is not done at the level of particular seeds, it is done at the level systems of environmental attunement. Dormancy is a response to the problem of troubling environmental variation, just as the germinating condition sensitivity of seeds is their way of finding out whether to grow from their environment.

Dormancy itself requires us to think what capacities of seed germination allow seeds to adapt to indicators of resource unavailability. Put another way; what process within the seeds can vary with quite long term patterns in the environment; like a snowy period, an ice age (a volcano like in the OP metaphor)? How can something vary with something that is usually 'off'? One way is germination time changing; but look what's being described here.

Germination time influences are composed of various complicated (in the sense of complex system) cellular, intercellular, molecular, organ-level and inter-organ level changes. Nevertheless the overall environmental sensitivity which is relevant here is their emergent property of germination time. But look at the nature of this emergent property; it is still within the seed. Perhaps a better way of putting it is that it is an output of the seed insofar as it germinates; an open site for an environmental feed forward that can become internalised as a feedback loop. This invites us to see the seed as an interacting process with itself and its environment.

Doubtlessly the question might be asked; how can it be an interacting process with itself? Which is a fair question, but ultimately rooted in a failure of perspective. The seed is how it functions as a seed. How does it function as a seed? Through the interaction of its internalised processes, and how those processes interact with its environment. The seed is a corpuscle of interacting sub processes, and emergent dynamics like 'germination time' are just as much a properties of seeds as the processual dynamics which lead to germination time.

The seed also internalises its environment through sensitivities to weather, temperature, sunlight; but it never internalises a whole, only a map constrained by what is available to it and what can be embedded in its processes. Human eyes may see the colour wheel, but our skin sees UV and IR. And we have learned to think of light as waves; then wavicles; then quantum clouds. And as an interaction between cultures, retinas and environments.

Hey, fdrake, is this the direction you wanted this conversation to go? I think you need to give us more guidance. I thought you were going to come back and wave your magic data wand. — T Clark

Well there's a lot to work through, and I'm not exactly presenting things in a unified manner, so the discussion being scattered is to be expected.

Whether a discrete event is determinative or irrelevant would depend on whether the event closed down species' ability to reproduce. My guess is that the resident plant life in some places has been changed by dramatic geological events (like volcanos). If species were unique to the vicinity of the volcanic blast or meteorite strike, they might become instantly extinct. On the other hand, conifers, for instance, wouldn't have become extinct because of Mt. St. HelensNM because there are millions of acres of conifers nearby to re-seed the altered slopes of the volcano, and any wrecked territory. Some plants have very limited ranges and volcanism could wipe them out. — Bitter Crank

I think this is exactly right. If an event closes down a population's ability to reproduce, it truly is an extinction event. But if, like in the jungle-volcano system, there's jungle outside the volcano's fury, it can incorporate the extinction event. If a population can survive such an extreme event as this, it can actually adapt to it in some ways. Whether the population can adapt to some change depends on:

(1) whether the change destroys the reproductive ability of the population.
(2) whether the population is embedded (or has within it) in a process that can adapt to the change.

In terms of (2) and actual plant recolonisation of volcanic soil post-eruption, this can be done through the longer range varieties of seed dispersal available to when animals can carry seeds in crap; if the long range seed dispersal is a part of the system already, it already has a recolonisation option for the volcanic soils. What might be 'selected for' here; what might be changed about the seed dispersal-seed formation sub-system; if it provides an comparative advantage of reproductive success, is the durability of seeds in animal guts, how long they can be stored before germinating, the efficiency of energy storage within them and so on; anything that makes it easier for the plants to recolonise the volcanic soils.

In considering the jungle-volcano system, it's still worthwhile to think about the jungle's development in terms of what it is sensitive to as well. If the volcano is not integrable like above, then the jungle works in a causally isolated manner from the volcano, and cares about the soil rather than the lava.

The next post I had in mind precisely talked about this kind of thing; germinating times of wild barley vs crop barley, wild barley germination times are much more variable than crop barley (crop barley had regular germination selected for 'artificially' so long as humans have been growing barley), wild barley seeds can lay dormant for quite a long time; over a year if memory serves. This allows them to wait out particularly snowy periods and recolonise after that in a less competitive environment; but the seeds do a 'mixed strategy'; most of them germinate quite quickly, some play the long game, some play the very long game. Will write more in a new post about this when I can be bothered writing up bits from a book I have.

I agree with you entirely. If I rewrote my OP and changed the volcano's timescale to "once every 10000 years" like you suggested, would your criticisms cease? To me, the year choice mattered little for demonstrative purpose. If I was actually analysing the behaviour of a real ecosystem involving a volcano I would care more, but since I was writing a story where I took the perspective of a tree for a bout 2/3 of it I thought the issue of precisely how volcanos integrate into ecosystems didn't matter much, only that they can operate as on-off switches of eruption and have mediated causal relationships with the grander ecological dynamics they imbue with nutrients and so on.

Probability doesn't tell us anything about unique events. It just tells us what happens when a procedure is repeated unless we're talking about logical possibility and just assigning weight some how. — frank

When canopy tree offspring have a reproductive advantage when they grow taller, this is already an aggregation on the level of the population. The unique events which cause the tree bodies to grow taller matter far less for the evolutionary pressures than the increase in height relative to other trees; to canopy higher. Evolution acts to increase tree canopying height in general, that it acts on all scales; molecular, developmental etc; to do this is is exactly a 'blurring of the details' of unique events. Evolution does not care for the how here, only a constrained how that satisfies the comparative advantage in the context of ecological constraints.

The more general feature here is that random variables allow you to make a summary characteristic of indeterminate processes in a way that depends on the events which happened (what makes the random variable realise) but does not care how they happened so long as they realise into the specific summary value. The hows and whys of how evolution acts on the mean height of canopying trees fall away in the sigma algebra of random tree-done events; and this is a certain 'forgetting' in itself, of nature deciding what is relevant. The 'forgetting' of the influence of unique events through aggregation is exactly what random variables do.

Or, more to the point, plants make/produce/form soil. It isn't a fast process. In mountainous areas, it may take a century for the plants to produce an inch of soil. It's faster on well watered, temperate plains. Tropical jungles produce soil, but the high volume of rain and drainage wash most of the decayed plant matter out. Regular falls of volcanic dust would definitely help. — Bitter Crank

In reality the two systems; soil formation, plant growth; are both coupled (direct link) and mediated (indirect links). The coupling occurs through the decay of organic matter and the behaviour of rain and decomposers - the processes of decay. They are also linked in the way you say directly to rain; even though the decomposers exploit the rain, and the trees exploit the decomposers exploiting the rain and so on.

Would you agree that the action of a volcano on such a long timescale, not long enough to be 'regular' for the plants, would be an irrelevance for the evolutionary dynamics of trees in such an ecosystem?

Hence there wouldn't be the interaction between the environment and the volcano. — ssu

So long as you can get this bit from it, that's all I care about.

The situation with more general causal networks is more complicated, you end up with so much shit mediating so much other shit 'direct cause' or 'causal isolation' is difficult to even conceptualise (even though the 'Markov Blanket' link in post 2 spells it out). What the 2 posts stress is that when you aggregate to the level of 'jungle' 'soil formation' 'volcano eruption'; 'volcano eruption' only interacts with jungle 'through' soil formation. If you check out the link in the first post, you'll see in reality there are other dependency chains and the volcano is not just 'on or off' in the island system; even when it's not erupting and forming the land, the heat it provides makes a microclimate, so you get volcano -> climate -> plant growth and volcano -> soil -> plant growth both being flows of the system. These have radically different operations; the climate is a constant feature of volcano interacting with sea, the land itself the volcano made is just kinda 'there' until the next eruption (so are the initial post-eruption nutrients).

The plants still won't be immune to lava though, they don't respond to lava in the way the land itself does; they respond to weather and soil; so it's likely the volcano's effects on their growth are mediated through other flows like those two examples. So they can't 'adjust' to the eruption in that way, they could never 'anticipate' it, even if seed dispersal from bird shit can rapidly recolonise the volcanic soils; what matters is that soil quality rather than the eruption which caused it.

I put math in mine so that means my rant has academic pedigree.

Love the story, although I said "I should notify Baden, fdrake has finally snapped. We've been expecting that." Then I was relieved to find you back on more familiar fdrake ground in your second post. — T Clark

I definitely have snapped, fallen down a great big hole, I just hope my schizoid ranting invites others to jump in.

A concern - at first I thought it was a quibble, but I convinced myself it's not. A billion years is just orders of magnitude too long. The other processes you describe - soil formation, jungle growth - operate on a scale of tens or hundreds of years. You left out other, very significant processes that also operate on much shorter time scales, e.g. rainfall cycles, ice ages, soil depletion, climate change, human encroachment, continental uplift, the movement of continents, asteroid strikes. In the context of a billion years, all of these except maybe the last three would be first order Markovian effects. — T Clark

Yes! A billion years is way too long to be accurate to the dynamics of the system, but 'a billion years' is simultaneously a cultural signifier of 'a time so long ago it's irrelevant' and 'a very long time', it also suggests the sheer time scales dynamical relations can persist in.

My next post in the thread will actually look at two examples in a different context of how 'extinction events' like that can get internalised as a sensitivity, or treated as an indifference to functioning.

Stochastic processes are sequences of random variables.

Random variables are summaries of stuff that happens; all the baby making in the world produces a total number of babies per year. The hows and whys of reproduction matter little to the summary "total number of babies", because the hows and ways are specified by an interaction with the population of babies.

If that seems abstract, it is because it is. More specifically, the context of the random variable "total number of babies" is specified implicitly there. We can ask "what are the total number of babies born in (category X)?". Above it is "the entire world now", but "(category X)" could be "In America", "per American state", "per continent", "between the years 1995 and 2013 globally".

Random variables can be generated through their interaction with different time points; each time point gets a set of "hows and whys" associated with it; these are the events that drive the formation of the summary. So if we take the sequence of years {1991,1992,1993}, there are random variables "total number of babies born in (category X) at (year)". This is called a (discrete time) stochastic process. An example then would be:

{Total number of babies born in America in 1991, total number of babies born in America in 1992, total number of babies born in America in 1993}.

There (category X) is "America" and year is 1991, 1992 or 1993.

A stochastic process is called "first order Markovian" if the behaviour (in terms of probability) of the a time point depends only on its immediately precedent time point. Total births, there, would be first order Markovian if "The total number of babies born in 1993 depends only on the total number of babies born in 1992" (and the same for the other time points). Prosaically, this gets called "memorylessness"; (first order) Markovian processes are memoryless because they only care about what immediately precedes them. Like how on a pool table the balls don't care how they got to where they are, only where they are.

The billiard balls there are like the plants forgetting the volcano's rage in the above story; they can forget the eruption but not the soil nutrient distribution. The volcanic soil nutrient distribution mediates the relationship of the jungle growth to the volcano; once the relationship with the volcanic soil is set up, the volcano's rage is forgotten. Mathematically how this causal property gets represented is in terms of the connectivity of graphs, and the time points above are a sort of graph. How you can travel from one node to another in a graph determines the causal relationships of things indexed to each node.

Let $A$ be the random variable that maps a year to the total number of babies born in America on that year. Then the graph:

$A(1991)\rightarrow A(1992) \rightarrow A(1993)$

has the first order Markov property. If you follow the arrows, you can only get to $A(1993)$ from $A(1991)$ through $A(1992)$. History is forgotten in the process because it embeds all of its relevant information in every time step.

What this looks like in terms of the jungle-volcano system in the story above is:

$E(eruption)\rightarrow E(soil\hphantom{..}formation) \rightarrow E(jungle\hphantom{..}growth)$

where E is a placeholder for "mean canopy cover of species X" or "total biomass of moss over area X" or some other indexed summary of the jungle, eruption or soil. EG:

volcanic ash composition and distribution -> soil nutrient composition and distribution -> growth rate of trees over space

You cannot 'access' the effects of the eruption in the past on the jungle growth except for the soil formation (this is not strictly true in general of course, and the one directional arrows between soil formation and jungle growth are actually reciprocally dependent from the perspective of the whole system). And the jungle's growth patterns really do 'forget' the eruption by only interfacing with the eruption's effects on soil nutrients.

The problem here is of course that considered from the perspective of the jungle-volcano system over time frames that the volcano will erupt in, soil formation and jungle growth are absolutely linked; the potential to accumulate plant biomass more rapidly in the direction of increasing soil nutrients is only there because of the volcano (in reality it can be more related to wide-spread seed dispersal mechanisms from animals' shit, but this won't change the increased fertility I am trying to highlight). From the view of the entire system we see an oscillating series of destruction and regeneration. From the view of the plants within it, they see the soil and how they interact with it!

The eruption as a "cause" of volcanic soil factors out, all that matters is the volcanic soil for the jungle in how it behaves. The jungle has learned not to care about the volcano because its eruption is usually off on most timescales, and when it is off there is great soil. When it is on, however, the plants nearby are destroyed.

There is no relevant information nature cannot access, nature unfolds according to its own sense of relevance, but its sub-processes learn to contextualise. Perhaps it could even be phrased like the origin of sub-process is a context of development. Like the canopy trees never become immune to lava. Causal histories get absorbed into intermediaries until they become relevant again.

From the framing device of the story, we know that the jungle growing near the volcano waits in bated breath; happy to grow how it does until it is destroyed by its own ignorance. The internalised processes of jungle growth in the jungle-volcano system do not anticipate this; at least insofar as they cannot interface with the destruction the volcano brings.

Some ways of growing, however, can interface with such destruction and profit from it (next post later). Citation describing first order Markov models in a molecular evolutionary context. The relevant thing to look for in here is how expanding the 'state space' (available information which is incorporated to process dynamics) can reduce the dependence on the unobserved past (unavailable information that is implicitly unincorporated). Citation for causal graphs. Citation for how intermediaries causally isolate nodes.

I thought I almost killed you because I thought you were a p-zombie. :wink: — Noah Te Stroete

Two behaviourists are laying in bed after fucking. One says: "That was good for you, was it good for me?"

Courtship Exam

How do you spend your days?
Doing whatever it is I can still do without you.
What interests do you have?
Long walks on the beach into the sea.
And your sense of humour?
Good, that is why we are still talking.

Do you want to come over tonight?
Do you mean watch a screen until we get bored and fuck?
No, nothing like that, too much reality repulses me.
What do you want then?
To remain when the questions cease.
Do they ever stop?
No, but we might live them.

Do you remember that time we
almost killed each other
over a teacup?

Is there a “hard problem” in your view? — Noah Te Stroete

I dunno, I'm apparently a p-zombie.

(@StreetlightX @csalisbury because Deleuze stuff)

I guess I didn't talk about events and objects so well in that context. Bar magnets. Let's go with bar magnets.

From one view, a bar magnet is just a bit of metal (possibly with blue or red plastic covering like in our school days) that has a persistent magnetic field that happens to be shaped like a bar. However, from the perspective of functioning as a magnet, the shape matters little; it's free to vary so long as it doesn't become too big or too small. The material determines the amount of magnetisation, and the shape determines how that magnetic field distributes over space. But in terms of the magnetic field; the thing which makes it a magnet; what matters is how the magnetic field persists in the material. The atoms are moving about, all kinds of quantum shit is happening between them, there's finger grease and dirt and crap on it, but that doesn't matter for its functioning as a magnet; what matters is how the magnetic field persists.

This question of persistence is key for understanding the magnetic field in a bar magnet as an object. The question of persistent patterns of flux is key more generally for individuation. How does a magnetic field become a stable faculty of an arrangement of atoms? How does a magnet emerge as a magnet? These are answered with the question of how magnetic fields persist. But let's keep an eye to the broader context too.

Keeping track of the entities which induce the magnetic field, we're keeping track of the electromagnetic behaviour of the subatomic particles and their relationships of arrangement and proximity as constrained by the atoms. Proximity driven by an organisational principle that produces a persistent pattern if its conditions are right. It is tempting here to impose an arbitrary split between low and high level components, of atoms as a base and a persistent magnetic field as a superstructure, but magnetism is baked into atoms as much as atoms are baked into magnets; these are just different patterns of articulation within the same self differentiating nature that articulates itself through its own interactions.

The resources for unfolding the magnetic field of the atoms in a persistent form are all given by a nature which resists the conceptual imposition of hierarchical organisation because its interactions are simply the origin of all hierarchy. Substance (atoms) and predicate (arrangement), content (atoms) and form (arrangement), lose their antipodal bivalence in the dizzying thought of their co-constitutive reciprocity. Interaction is baked into nature, and the sides of an interaction form poles for each other that need no externalised articulating principle other than the codification of their own internalised resources. These are the events that keep happening, an object is just a durable happening with somewhat uniform character.

Edit: something I under emphasised here is that the flux of the magnetic field doesn't change the atom very much in the bar magnets, it changes how the electron orbitals are shaped, but it doesn't change anything about the atomic nucleuses; whenever something can be in flux, it's always in flux with respect to something else, something has to change sufficiently slowly for pattern of flux to persist with respect to that change.

But I have become too general, so let's get back to the magnet. The atoms on its surface all have electron orbitals, which due to the proximity of the atoms overlap. That overlapping constrains the positions of the electrons on the outside of the material, which provides an immanent definition of how the electromagnetic field can oscillate - wave about. Like a valley provides an immanent definition of water collection, and its river provides an immanent definition of the same valley. Valleys are scored out landmasses of gravitational potential for water, field lines are scored out landmasses of magnetic potential for the electromagnetic field.

Due to the properties of the atoms involved, their numbers of electrons and orbital shapes, their proximity provides different valley shapes through which electromagnetic currents can flow. And they must flow, they always do, rest is motion from an altered perspective. So flow they do, along the paths they feel they must, but they empathise; they feel in unison, constrained by the interactive context of the material's organisation and constitution. They feel out what direction runs down the valley they create together and swim towards it.

So we have a magnet, a little island of dirt emerging from an ocean indifferent to it.

What an individual is depends on the scope of the question "What is an individual?"; what an atom is has a much more tame domain of inquiry than what a living being is. So if you consider it in the broadest possible sense, you have to ask it in a way that is informed by the various ways distinct entities come about or emerge from their background. An instructive metaphor might be thinking of 'beaches' as little bubbles around 'land' that distinguishes it; makes it an individual; from 'sea; through the interaction of tectonic plates and global water distribution.

If you take a vulgar materialist view on things, there's an order to stuff. Stuff's ultimately made of fields; which describe how particles come about through fields interacting with themselves; then particles make atoms; interacting with other forces; and atoms form compounds and molecules; and by that point you have chemistry, and a novel domain of phenomena appears. You can begin describing chemicals as 'catalysts' for 'reactions', for example, and 'catalysts' and 'reactions' don't occur between fields or quarks, they occur between chemicals.

This gives extra expressive power to nature over and above a universe where chemicals could have never formed.

This is a very simplified picture, but it is instructive insofar as it provides hints on viewing where this new domain of phenomena; that which is studied by chemistry; came from. When fields interact they make particles, when particles interact they make atoms, when atoms interact they make compounds, when compounds interact they make chemistry. Organisation of one domain (atoms) can generate novel behaviours (chemistry) which have extra causal powers (chemical reactions) than what was organised (particle-particle interactions). The general principle suggested here is that when you get enough and the right sort of interactions between stuff, when interaction can organsie, you get new domains of entities which then stick out from their background.

Ignore issues of personhood, intersubjectivity, social constructions etc since the post is already too long and too imprecise, which I am sorry for.

Individuation comes along with emergent behaviour. Emergent behaviour is rooted in interaction. Fields interacting make atoms, atoms arranged thusly make molecules and compounds, which make chemistry as a new frontier and language of expression; the play of electrical forces are like pitches in the notes written in scores of molecular geometry.

The ontological task of thinking individuation acutely corresponds to finding general structures in how reciprocally interdependent event networks (interaction) graduate and congeal into stable objects. Asking how anything stable can come within all this flux.

If I cannot be the smile that graces your lips
Let me be the wood that burns your sorrow
If I cannot be the fire that warms your feet
Let me be the ground you walk tomorrow
May I pray my embers cast your dancing shadow on the wall
Whose unseen smile becomes your own

All these bloody pure mathematicians trying to stop us from occupying their lawn. They forget the rest of us squatters were here first.

In this regard I expect that outside of computer-vision and speech recognition, where deep-learning as a strong inductive bias, the black-box neural network approaches in other problems domains will begin to take a back seat as transparent models of network logic come to the forefront. For you cannot easily inject human-readable logic clauses into a distributed neural network with positive and negative activations. — sime

There's a general purpose procedure for doing causal inference in Bayesian models, though. Once you've fit the model, you sample from the posterior conditional of the desired response (say "do they have cancer?") given the desired 'intervention' (say smoking or not smoking), then you get probabilities of each. The reason this works is because the dependence on all other variables you have is already included by integrating them out.

The kind of algorithms, like neural networks, that are used to analyse big data aren't made of math objects that are interpretable in terms of the data. Studying the properties of these algorithms and creating them is a science to the same degree that studying the properties of any (set of) mathematical models is a science.

The biggest difference between these algorithms in general and more run of the mill statistical methods in general is that neural networks won't tell you how important each part of the data they depend on is to the conclusions the neural network draws, whereas run of the mill methods like linear regression will. What this difference turns on is that neural networks learn to weight (combinations of) variables in terms of their contribution to predictive accuracy without telling their user the consequences of this weighting scheme and how it learned this pattern from the data. Gains in predictive flexibility are bought with losses in interpretive precision.

A neural network model still takes the form of:

$y = f(x) + e$

where y is a response, x is a spreadsheet of data and e is an error.

A very tame model for the same x might be:

$y = b x + e$

where b describes the slope of the line each x makes when regressed upon y. What these b terms tell you immediately is how x relates to y in terms of the model. In terms of the neural network, you don't get a similar interpretation for f; the meaning of whatever f is is occluded because it is so complicated and flexible.

Neural networks can be a force for good, when applied to data with lots of test cases that actually represents the phenomenon being studied; like language recognition and translation in google translate's photo/video of words -> translation app. They are good when the data is good.

They are however bad when the data is bad. But no matter how complicated and flexible your model is, your data still can be terrible in all sorts of ways, and the resultant model (when it is performing exceptionally well!) will replicate all the crap that's gone into the signal of the data which in an ideal world shouldn't be there.


Companies have tried using big data algorithms to automate hiring practices based off of accepted/rejected CVs, they ended up being arbitrarily racist and sexist, and no matter what the engineers did to improve the algorithm the data had this prejudice so strongly and in so many interlinking ways - it was always preserved. After much consternation the funding for these projects dried up. A similar thing will be happening soon, terrifyingly, with court cases.

Denying that the application of neural networks to data is science is rather arbitrary, it's just often bad science, where people use the algorithms instead of thinking.

Also the whole book is so... Deleuzian — StreetlightX

I love that he just gets to work on the issues, I read "Brief Lessons" a couple of months ago and it wasted absolutely no time.

Reminds me of:

Daisy daisy give me your answer do
I'm half crazy my balls are turning blue
I can't afford a johnny, a plastic bag will do
'Cos you'll look sweet upon the sheets
With me on top of you

Carlo Rovelli - The Order of Time — StreetlightX

I'll have this book soon. Will make a thread of it once I've finished reading.

So the unknown is always with us alongside the aspects of predictability. With and within. So if I can get my butterflies to flap their wings just so, and give you the insight you need, then the world will be transformed. — unenlightened

Think even this is too instrumental, there's two ideas in the butterfly effect as usually presented that just aren't there in chaotic complex systems:

(1) Outlandish perturbation sensitivity.
(2) Ability to attribute the cause of a cascade to any given perturbation.

(1) Complex system dynamics are typically overdamped, which means that small perturbations (like the butterfly flapping its wings) are way more likely to fizzle out into their immediate environment (system A) before passing on any of their effects to some other system in the complex system (system B); this is like a generalised 'friction' tending to render events' effects minor.

Moreover, perturbation sensitivity can't really be thought of as finding 'god's levers' into the system, it's still complex and hard to reduce to levers. Sensitivity reveals itself on the aggregate level, and is very difficult to link to any specific perturbation, which introduces (2)...

(2) Given that we're in a cascade, we (or another system) only observe the effects of the cascade after it's built up a bit. A great example here is orgasm:

... a model is introduced wherein sexual stimulation induces entrainment of coupling mechanical and neuronal oscillatory systems, thus creating synchronized functional networks within which multiple positive feedback processes intersect synergistically to contribute to sexual experience. These processes generate states of deepening sensory absorption and trance, potentially culminating in climax if critical thresholds are surpassed.

So generally people are gonna know what feels good for them in sex, but they're not going to know the biomechanics of everyone involved's flesh and friction and tactile feedback, despite these lower system components feed forward-ing to the state of pleasure.

The take home here is that interventions in complex systems have to be done with respect to their organising principles (how they feedback loop, "do they enjoy this position and style of touch?" in sex) rather than their feedback loop inputs (the butterfly flapping its wings, the precise pressure and skin deformations and other tactile variables in touch). So we can think in terms of "what are the worst/best effects of this system and how can we act to stabilise our community/agriculture/relationship from it or grow from it?" rather than "i need to find a butterfly lever to pull to make everything right again".

The problem is that you’re answering a philosophical question about the nature of knowledge with a scientific question based on the knowledge of nature. That's why we're talking past one another. (I've now answered AndrewM below about a similar point.) — Wayfarer

Do you agree that the (existence of a moving asteroid prior to the existence of humans) entails that there was space before human minds? Because it had to move in space to move as asteroids do.

Edit: you are absolutely fine leveraging scientific thinking whenever it serves your perspective, and I won't let you subject me to this double standard.

All due respect, we’re talking past each other. — Wayfarer

If you say my argument is question begging, and I've made it into something like a syllogism, you should be able to tell me precisely in what premise or inference I'm begging the question. If it's not in premise (1), I've moved onto premise (2). So I'll ask again:

Do you agree that the fact entails that there was space before human minds? Because it had to move in space to move as asteroids do.

The hard problem also has to do with the fact that we are trying to understand the very thing we are using to understand anything in the first place. Consciousness is the very platform for our awareness, perception, and understanding, so this creates a twisted knot of epistemology. Indeed, the map gets mixed into the terrain too easily and people start thinking they know the hard problem when they keep looking at the map again! — schopenhauer1

It isn't so surprising that the hard problem is unsolvable when the capacity to give an account attempting to solve it apparently undermines any account.

↪fdrake It’s not false, but I feel the question is being asked for a reason over and above its facticity, namely, by way of introducing a naturalist philosophy of mind. — Wayfarer

Do you agree that the fact entails that there was space before human minds? Because it had to move in space to move. It's an asteroid.

But can’t you see that the same principle applies to all empirical facts? Not simply what existed prior to human life, but anything that happened in the past. From a naturalistic perspective, of course there’s a temporal sequence within which the human species is a recent arrival (‘a mere blip’, as if often said). — Wayfarer

Please indulge me and answer the questions. Is it false that there was a moving asteriod before humans?

Interesting.

Terrifying potential for oversight and fake objectivity.