Daylen's Journey

Where are you and where are you going?
daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Daylen's Journey

Post by daylen »

Welcome to my new journal. Old journal here: viewtopic.php?t=11160

Figured I am entering a new phase that is less instinctual and more deliberate. Not much has changed as far as lifestyle goes. Still live with mom, slowly saving, and planning my journey through life. Still working on an agent-based simulation and thinking a lot about all kinds of stuff. This journal serves as a sub-activity for the activity of "research". One of the core models orienting my research being behavior trees: https://arxiv.org/pdf/1709.00084.pdf

Behavior trees can be thought of as generalizing both decision trees and webs of goals. Behavior trees can be used to control an artificial agent or a robot. They are a natural progression from state machines that improve modularity and reactiveness. Individual behaviors or leaves of the tree can be reused in multiple contexts across multiple agents. Agents with behavior trees appear reactive and can even appear to plan purposefully.

Artificial agents modeled by behavior trees are strictly using deductive reasoning in the sense that depth-first search is iterated from the root of the tree to run behaviors that eventually succeed or fail. Actual agents that can make active inferences from the "bottom-up" generalize behavior trees. We are such agents and thus while behavior trees can be used to model us, they fall short of us. This will not actually stop me from attempting to model my own life trajectory with them.

It is possible to approach or approximate the behavior of actual agents by developing an ever finer behavior-space and optimization protocol. I suspect this can go a long way towards the creation of convincing artificial agents, but perhaps it is much easier as far as engineering goes to introduce some kind of wetware like electrochemical RAM or the like. I don't know, but I suspect we may find out soon enough.

In the meantime, I will be using these approximate agents to explore social and economic outcomes in a semi-controlled, evolutionary simulation I am calling "Volution". Still a long ways to go on this front, but I will try to keep my journey updated here as I hit all kinds of computational bottle necks. Feel free to save me time by pointing out flawed protocols or potential improvements.

I may also post various ideas and ideals here and there relating everything from mitochondria to space mega-structures to black holes.

Henry
Posts: 499
Joined: Sat Dec 10, 2022 1:32 pm

Re: Daylen's Journey

Post by Henry »

daylen wrote:
Wed May 17, 2023 10:06 am
Still live with mom.

Feel free to save me time by pointing out flawed protocols or potential improvements.
I'd start here.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

Behavior Trees

A more detailed description can be found in the document linked above but here is an overview. Behavior trees are directed trees where the inner nodes control the flow and the outer nodes (i.e. the leaves) execute actions or behaviors. Ticks or processing units are routed down the tree and the processing statuses of nodes are routed back up. The status of a node is either "running", "succeeded", or "failed".

There are three basic control flow nodes: sequences, selectors(*), and parallelizers. Sequences execute the next sub-node upon success of the current sub-node and returns success upon completion of all sub-nodes. Selectors execute the next sub-node upon failure of the current sub-node and returns success upon completion of any sub-node. Parallelizers run all sub-nodes despite their statuses and return success upon success of some proportion of sub-nodes.

(*) Selectors are also referred to as fallbacks.

When cooking a meal, a selector can be used to select what meal to cook from a list of possibilities, a sequencer can be used as a recipe to figure out to do next, and a parallizer can be used to manage the flow of tools and ingredients in the kitchen while cleaning up. Perhaps the selector is stochastically sorted each day to send recent meals to the back of the list. Another possibility is to have a sequencer that does meal planning throughout the week. With behavior trees, there are often multiple ways to solve a problem.

When exploring an unknown territory you could use a sequencer to do a grid-search while using a parallelizer to map the territory. If some features of the territory are all that matters then a selector could be used to pursue leads to those features. Perhaps you see a stream of water and follow it back to the source or maybe you are tracking an animal you just struck with an arrow to recover the prize.

When gardening you could check and harvest multiple plants concurrently using a parallelizer. You could use a selector to harvest some plants for a meal or to weed out some unwanted plants. A sequencer could divide the labor and maintenance out across the day or week.

When investing, a selector within a selector within a selector within ..etc.. could filter possible securities until only a handful are left for deeper analysis. An index being like a parallelizer for all kinds of securities. Sequences could be used in rebalancing back towards the appropriate allocation after a given period of time.

As you can see, behavior trees are quite universal. In fact, with a blackboard or data-context they are turing-complete. So, it is not so much a question of what can you do? ..but rather how can you do it at scale while retaining readability? This is where designing the appropriate behavior-space for your own purposes comes into play. Hierarchical differentiation of various sub-trees can also become a handy exercise. Combined it is possible to search a space of behavior trees using an evolutionary algorithm that directly or indirectly imposes a fitness function on the space and eliminates low-fitness trees from the population.

User avatar
mountainFrugal
Posts: 1125
Joined: Fri May 07, 2021 2:26 pm

Re: Daylen's Journey

Post by mountainFrugal »

What aspects of mitochondria?

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

@Henry

You must be lonely. I hope for your own sake your mindless jokes are entertaining enough.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

@mountainFrugal

I am sure you know quite a bit more than me on this topic! From what I understand, they are the power-houses of cells and have quite an interesting role to play in the very early evolutionary history of multicellular life. Which seems to be one of the major, if not the major, filters of life in the universe.

User avatar
mountainFrugal
Posts: 1125
Joined: Fri May 07, 2021 2:26 pm

Re: Daylen's Journey

Post by mountainFrugal »

Yes. They exchange bits of genetic information (they have their own genomes) with the nucleus of the cell. Over evolutionary time, mitochondrial genomes are much smaller because they have transferred that information to the nucleus of the cell where gene expression is much more fine tuned for the entire cell. The same is true for chloroplasts. In plants, the information is exchanged between the three organelles and certain subsets of proteins that are coded for by the genes in the nucleus are then targeted to either the mitochondria or chloroplast. :). Information layering.

7Wannabe5
Posts: 9369
Joined: Fri Oct 18, 2013 9:03 am

Re: Daylen's Journey

Post by 7Wannabe5 »

Interesting. Since a human agent can't really consciously parallel process, would it really be more like leaving multiple tabs open in a browser? Maybe the open tabs might represent current contracts or commitments? I see some easy (likely too easy :lol: ) analogies in terms of the relative frequency of use of sequencers, selectors, and parallizers when considering monogamy vs polyamory, etc.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

@mF Cool as hell!

@7W5

It has more to do with proportional success of the sub-nodes. So, say you are going around your garden checking if planets are ready to harvest and roughly 17% of them are, if your intended success rate is 15% then the parallizer itself succeeds.

I think it actually may be debatable whether humans can parallel process. Attention as a high-level mechanism may not be capable but if you include the many latent or subconscious layers of processing then we certainty do. Aww, just saw your "conscious" qualifier.

avalok
Posts: 277
Joined: Fri Mar 05, 2021 4:42 am
Location: West Midlands, UK; Walkscore 73

Re: Daylen's Journey

Post by avalok »

@mF please correct me here but isn't there an idea that mitochondria were once distinct organisms (perhaps bacteria?), that then formed a symbiosis with other single/simple celled organisms? It's been a while since I did biology, but I recall this being in vogue at one point.

User avatar
mountainFrugal
Posts: 1125
Joined: Fri May 07, 2021 2:26 pm

Re: Daylen's Journey

Post by mountainFrugal »

@avalok yes for mitochondria and also for chloroplasts (cyanobacteria-like).

I think there is a lot for @daylen to dig into with genome evolution for packaging and layering information. There might be some good empirical examples for his other projects and models. Happy to set up a call some time.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

@mF

That would be fantastic! Recently, I have been looking into Nick lane's and Michael Levin's work. I am sure you have heard of them. There work seems to be pushing some of that information processing up in scale. For instance, cells appear to search morphogenetic space to find the correct configuration (i.e. as an organ) and this process can be interrupted with electromagnetic fields to alter the shape of simple organisms like planaria. Also cells in some sense solve the prisoners dilemma with gap junctions that partially dissolve the individual identities of cells, allowing a collection of cells to feel the stress of a minority.

User avatar
mountainFrugal
Posts: 1125
Joined: Fri May 07, 2021 2:26 pm

Re: Daylen's Journey

Post by mountainFrugal »

I am familiar with both of them as names within this field but I have not delved into their work specifically (only tangentially related to what I worked on in the past). Do you have a reading recommendation? The Vital Question by Nick Lane is available right now as an audiobook. I will PM you to set up a call.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

Here is one way to hierarchically classify sub-trees of an overall behavior-tree: phases -> schedules -> activities -> projects -> goals -> behaviors (leaves)

Phases

Phase branches correspond to bifurcations in the state space of either the agent or the arena.

For the arena, this bifurcation could be the high and low energy futures of the planet for instance. The low-energy future lacking oil and being limited to mostly wood, while the high-energy future manages to preserve the energy density of oil without relying on it. Possibly with some combination of fusion, distributed renewables with improved battery tech, artificial fuels, space-based solar, and so forth.

For the agent, bifurcations could be the result of a major injury, a disease, or a paradigm-shift. Such branches can limit or otherwise enable behaviors stemming from the sub-tree below.

Schedules

Schedule branches correspond to bifurcations in the space of cyclic couplings between an agent and the arena. Cyclic couplings exist at several scales of time which tend to align to multiples or divisions of astronomical/geological cycles. The earth rotation relative to the sun being a day which can be roughly multiplied by 365 to get a year which is also a rotation of the earth around the sun.

Practically speaking, these branches account for different ways to organize behavior within a day or a year, or every period in-between and beyond.

Activities

Activity branches correspond to bifurcations in the space of games agents can participate in given a shared arena. These games can be of the finite or infinite type. Basically, anything you can name and "do" to fill up a period of time or schedule. Including football, investing, gardening, research, golf, skydiving, and on and on.

One distinguishing criteria from projects below being that activities tend to stick around in a civilization or culture once they have emerged.

Projects

Project branches correspond to bifurcations in the space of sub-games agents can participate in given a shared arena. These sub-games are always finite and apart of some larger game. Unlike activities, projects do not tend to stick around but usually have a rather high turnover rate. Exceptions that can stand the test of time might include large construction projects (e.g. the pyramids, space mega-structures).

Goals

Goal branches correspond to bifurcations in the space of optimizations. A few possible optimizations being metabolic-energy minimization, run-time minimization, and free-energy minimization. The latter being based on the "free-energy principle" and can be thought of as minimizing surprise through updating model of self and world.

Behaviors

Lastly, behaviors are the leaves of the tree. They can be thought of as conditional actions in the form: if this, do that. This is where the function space of the agent is expressed. Some possible functions being turn(), walk(), run(), look(), locate(), throw(), point(), frame(), and so forth.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

There are three high-level activities that seem to be attractors in my own life and in many others as well. These activities being homesteading, research, and investing.

Homesteading

Homesteading is typically associated to country living and a high-degree of self-sufficiency. In the future, it is likely that I will orient more towards this kind of life. This will be especially true if the low-energy phase of the future arrives. Homesteading as an activity however can be scaled down to all kinds of living situations. Any tie to land or space across periods of time longer than say a year might qualify. Currently, my mom and I live in an apartment where the sub-activities of cleaning, cooking, potted-plant care, and neighbor interaction can be thought to scale into a future where we live near family in the country. One possibility enabling this future being the inheritance of land outside Lawrence. Before then it is important to save sufficiently to build a portfolio capable of paying the taxes, as well as health care and other ongoing expenses.

This activity also melds nicely into permaculture or gardening which has been an interest of mine going back to college when I took a permaculture course. During the course I was able to tour many farms nearby and observe many WL6+ agents in their natural habitat, so to speak. After the course I was able to work as a farm hand and learn some of the ropes. This extended my skill-set from the animal-husbandry focused projects I would help my grandpa out with in youth. Combined, the transition back to farm life or homesteading makes sense for my own arc. The tranquility of such an environment lends itself towards a deep focus on other activities like research and investing.

Research

Research is a fairly broad term that I am applying to include talking, writing, coding, experimentation, and a general seeking of knowledge. It would be impossible to narrow down this activity in my own mind to a particular field like math, physics, biology, psychology, or so forth. Rather it is systemic at heart and can be considered to revolve around what I call agent-arena relationships. I extend the concept of agent to include any system that is assumed to have agency. Including all life and potentially silicon life in the near future. Arenas being the planets or structures on which agents can live. In particular, I am curious about how changing the topology of the surfaces of these structures can influence a society of agents. Agents living on the inside of a torus or ring, for instance, may organize differently than agents living on a spherical planet. A ring being topologically equivalent to an O'Neil cylinder.

The relationship between agents and arenas existing in a nexus of game theory, biology, astrophysics, engineering, economics, ecology, and more. All very much intertwined when building civilizations in a box so to speak. Simplifying assumptions are necessary for this building process. Volution being a project under the research branch that simplifies agents to 2D circles that can interact and move in a 2D space that wraps back around on itself (i.e. a ring or inner torus). Anatomical details of agents being abstracted away into simple scaling laws and metabolic processes. Objects or items being abstracted away into interactions or processes agents engage in. Emergent institutions being abstracted away into instances agents can enter corresponding to regular polygons from triangles to hexagons.

Investing

While I am investing 6k in index funds, I do not own a sufficient mass of funds to devote much time to this activity. However it has been my assumption for a while now that this mass will eventually increase enough to start treating portfolio management as a part-time activity. Jacob's posts throughout the years have been a major source of inspiration and have helped bootstrap theoretical foundations that will serve me well for years to come. Not quite sure how active I will become on this front as this largely depends on the phases of the future and how my portfolio grows into those futures. Time will tell.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

Ring Estimates

Let a ring be a torus that is half-way filled up with solid and liquid matter and half-full of gases. The ring spins inducing a centrifugal force on the contents inside. What is the range of possible ring constructions that are capable of providing habitat for life as we know it? A few problems arise. Such as how to protect agents from high-energy radiation? How fast to spin for a given ring size to induce earth-like gravity? How can light be supplied? How can these structures be built? What population could these structures support given the resources available in the solar system? I am in the process of writing a paper on these questions, but I will briefly give some estimates.

As Earth has a magnetosphere, lithosphere, hydrosphere, atmosphere, and biosphere, rings have a magnetotorus, lithotorus, hydrotorus, atmotorus, and biotorus.

Volume is given by cross-area x circumference which is 2 pi^2 r^2 R, where r is the cross-area radius and R is the ring radius. We can simplify these calculations at first by presuming either a 10:1 or 100:1 ratio of R:r. So say we focus on structures ranging from 100m:10m to 100km:1km. Much smaller and the ring must spin very fast and provides little surface area. Larger is possible but may require expensive composite creation. This roughly gives a volume range between 10^5 meters cubed to 10^12 meters cubed.

Mass is volume x density, which can vary significantly depending on the material composition. Air can be ignored for now. Water is roughly 1000 kg per meter cubed, and rock can range from 1500 kg per meter cubed to 3500 kg per meter cubed. Though, given that habitable surface area per a given mass is generally being optimized for and a soap-bubble formation could be used, we can bring that range down to 1000-1500 kg per meter cubed. Giving us a range of about 10^8 kg up to 10^15 kg.

To induce an artificial gravity similar to Earth's, we need an acceleration of 10 meters per second squared which requires an angular velocity of (10/R)^1/2. Giving a range of 10^-1/2 to 10^-2 rad/sec. In frequency of rotation this is about once ever 20 seconds up to once every 10 min.

The next step of stress-strain analysis is more involved, but thus far I seem to find that towards the larger end of this range reinforcing metal with graphene is required. Though, steel or titanium can work up to a point, especially if smart design in used (e.g. lattice with bubbles).

Probably the biggest challenge is radiation shielding (i.e. the magnetosphere). The two major options being heavy electromagnetic shells or plasma. It is tempting to just make the whole torus an electromagnet but this doesn't work because life cannot stand the magnetic fields required and Faraday cages do not block magnetism. So, this calculation is a whole can of worms but for now I will assume that plasma works. :lol:

Light could be reflected using large mirrors or supplied from a power source. This shouldn't be that hard.

As for energy, rings are supported by swarms of fusion-drive spaceships. Allowing for flexible grid setup and transfer of energy across rings. Rings could also link together into a worm or cylinder to simplify long-distance travel and stratify a manufacturing-heavy economy. The economy being centered around asteroid mining, gas giant harvesting, and potentially solar lifting. Recently, I heard about a process for converting metals in asteroids into gases that can be transferred and molded into shapes with few imperfections and without forging. As you can imagine, mining asteroids in space is vastly different from mining on a planet. Applying force will scatter all the pieces bounded together loosely by gravity! So, instead you can create an atmosphere around the asteroid that breaks up substances and suck up all the gases to be redistributed accordingly where they can be converted back to the appropriate solids.

Given the asteroid belt is estimated to hold roughly 10^21 kg of matter distributed nicely between Carbon-type, Silicon-type, Metallic-type, and so forth it may be possible to build upwards of 10^21/10^8 = 10^13 rings of min size or 10^21/10^15 = 10^6 rings of reasonable size. That is something like 10^6*4pi*10^5*10^3 = 10^15 square meters or a billion square km which is double Earth's surface area. Though, this is a low-estimate since the larger structures could use mass more efficiently by becoming more cylinder-like than torus-like.

Anyway, lots of cool stuff to explore here.

basuragomi
Posts: 419
Joined: Tue Oct 15, 2019 3:13 pm

Re: Daylen's Journey

Post by basuragomi »

The Stanford Torus is 10^7 kg and is 95% rubble for radiation shielding, it seems like your mass estimates are rather generous given you're assuming no mass devoted to shielding. Maybe because you assumed 50% solid infill?

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

Aww, I figured people had already done these calculations, but I haven't heard of The Stanford Torus. Thanks, I'll look into it.

50% fill would account for less than an order of magnitude difference. How the radiation shielding is done makes up most the difference. With plasma shielding, you would still need extra mass for magnetic confinement and the plasma itself but potentially much less than just rock. I have no idea how much less though.

daylen
Posts: 2528
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Daylen's Journey

Post by daylen »

Oh, wait I do remember that Wikipedia article. It says it weighs 10 million tons which is 9x10^9 kg, rounded-up is about 10^10 kg. That is fairly consistent with the 10^8 to 10^15 kg range above for 100m:10m up to 100km:1km, given The Stanford Torus is about 1km:100m (ratio of R:r). Presuming plasma shielding doesn't account for more than an order of magnitude (which it wouldn't even if 90% of mass).

7Wannabe5
Posts: 9369
Joined: Fri Oct 18, 2013 9:03 am

Re: Daylen's Journey

Post by 7Wannabe5 »

Are you applying the principle of indifference to low energy future vs. high energy future in your model? Could there also be a moderate energy future given steady, rapid, but not too rapid, population decline?

Also, book I recently picked up that you might enjoy "Probability Theory: The Logic of Science" by E.T. Jaynes.

Post Reply