Daylen's Instinctual Dump
Re: Daylen's Instinctual Dump
Just made a correction in a few spots as I made a blunder initially in typing Travis as ESFP instead of ESFJ. Which is interesting because at the time SF-NT looked all the same to me. Yet, now with a better model of my ENFP mom (along with many others) and a trail of particulars surfacing again, it is clear that he used FeSi. Thus, we shared the same quadra (SiNeTiFe) yet were growing into completely different paradigms (NT-SF vs SF-NT), albeit with similar shadows (TeFiSeNi).
Last edited by daylen on Sat May 07, 2022 1:10 pm, edited 1 time in total.
Re: Daylen's Instinctual Dump
At one of the intersection points of the computational and mystical paradigms we get topological synchronization: https://phys.org/news/2022-04-topologic ... aotic.html
Which basically just means that at some point when simulating reality, in order for our algorithms to halt, we need to stop considering every shape the data takes hold of and start focusing on the shape transitions/boundaries, in doing so two systems that are chaotic can cohere together as one in a fractal manner (i.e. similarly across scale). Hence, the ol' peom I like to write and visualize about a tree being inverted with tangling branches is like a synchronization or "zipping up" of the degrees of freedom at various scales between the gross and subtle edges allowing for a simulatainous zooming in and zooming out of "reality" or simulations of reality with various flavors of zippers. The zippers being slightly different each time the brain gamma syncs due to a path dependence that cannot cleanly unzip yet can know through passive Si nostalgia and Ni prophecy that zipping has and will continue to occur.
Which basically just means that at some point when simulating reality, in order for our algorithms to halt, we need to stop considering every shape the data takes hold of and start focusing on the shape transitions/boundaries, in doing so two systems that are chaotic can cohere together as one in a fractal manner (i.e. similarly across scale). Hence, the ol' peom I like to write and visualize about a tree being inverted with tangling branches is like a synchronization or "zipping up" of the degrees of freedom at various scales between the gross and subtle edges allowing for a simulatainous zooming in and zooming out of "reality" or simulations of reality with various flavors of zippers. The zippers being slightly different each time the brain gamma syncs due to a path dependence that cannot cleanly unzip yet can know through passive Si nostalgia and Ni prophecy that zipping has and will continue to occur.
Re: Daylen's Instinctual Dump
In statistical mechanics, you start with a state and reach a final state in which an asymmetric determination exist. The past determines possible futures, of which the determined future cannot reconstruct a single past (i.e. non-reversibility, information loss, or entropy). In assembly theory, a spectrum analyzer can be constructed to either superficially or deeply(*) determine the complexity of an object allowing for an assembly procedure to be tracked back in time towards the deconstruction and reconstruction of such an object.
Or something like that. Still working this out obviously.
Assembly being accompanied by chemputers that chempute chemical pathways towards complexity horizons in which something like biological selection can take place to confusate algorithmic determination encoded on the chemical equivalent of a digital transistor. An interesting question I saw on twitter being what is the moore's law equivalent for chemputers?
(*) Superficially through bouncing light or sound waves off the object and analyzing the echos. Deeply with further analysis that deconstructs the object into parts to bounce light/sound wave off again.. and so forth recursively which may destroy the object in the process and/or require multiple instantiations.
Or something like that. Still working this out obviously.
Assembly being accompanied by chemputers that chempute chemical pathways towards complexity horizons in which something like biological selection can take place to confusate algorithmic determination encoded on the chemical equivalent of a digital transistor. An interesting question I saw on twitter being what is the moore's law equivalent for chemputers?
(*) Superficially through bouncing light or sound waves off the object and analyzing the echos. Deeply with further analysis that deconstructs the object into parts to bounce light/sound wave off again.. and so forth recursively which may destroy the object in the process and/or require multiple instantiations.
Re: Daylen's Instinctual Dump
Released a golf game on Github here: https://github.com/daylenjbonner/golf
Planning a year-long, self-directed crash course in game development with the Godot engine (using GDscript which is like python). This is practice to help me grasp another project I have in mind called "Volutor", which is a hypothetical simulator with three primary layers:
1. Cellular automata that aligns with macro level geological changes (e.g. erosion, vulcanization, flooding, etc.). Perhaps on a hexagonal grid thus giving each cell six neighbors from which to calculate its next state.
2. Autonomous agents with genetic byte-code. Agents can be asexual or sexual to reproduce offspring (sexuality introducing genetic recombination). A mixture of hard coding with soft coded activation. Speciation could be implemented though a predetermined ancestry tree or through a tree generated from genetic "modules" or genes.
3. Brains as neural nets that have a chance to mutate/emerge and learn during the course of a single agent lifetime. Frames operate sub-agent and allow protocols for learning and teaching that pass some knowledge down from generation to generation.
Here is a great introduction to such topics: https://natureofcode.com/book
I like the idea of playing with perspective, so I am considering some kind of mechanism whereby the "view" of the simulation can alter how it evolves. The big picture being that sectors combine into a ring that can asexually or sexually reproduce offspring rings. Rings combining into tubes or worms.. and so forth.. [space mega-structures] in a way that varies the topological structure of the ground upon which imaginary agents multiply and adapt with seeming intelligence across an array of scales.
Planning a year-long, self-directed crash course in game development with the Godot engine (using GDscript which is like python). This is practice to help me grasp another project I have in mind called "Volutor", which is a hypothetical simulator with three primary layers:
1. Cellular automata that aligns with macro level geological changes (e.g. erosion, vulcanization, flooding, etc.). Perhaps on a hexagonal grid thus giving each cell six neighbors from which to calculate its next state.
2. Autonomous agents with genetic byte-code. Agents can be asexual or sexual to reproduce offspring (sexuality introducing genetic recombination). A mixture of hard coding with soft coded activation. Speciation could be implemented though a predetermined ancestry tree or through a tree generated from genetic "modules" or genes.
3. Brains as neural nets that have a chance to mutate/emerge and learn during the course of a single agent lifetime. Frames operate sub-agent and allow protocols for learning and teaching that pass some knowledge down from generation to generation.
Here is a great introduction to such topics: https://natureofcode.com/book
I like the idea of playing with perspective, so I am considering some kind of mechanism whereby the "view" of the simulation can alter how it evolves. The big picture being that sectors combine into a ring that can asexually or sexually reproduce offspring rings. Rings combining into tubes or worms.. and so forth.. [space mega-structures] in a way that varies the topological structure of the ground upon which imaginary agents multiply and adapt with seeming intelligence across an array of scales.
Re: Daylen's Instinctual Dump
Making progress towards cellular automata on hexagonal grid with the following tiles for various biomes and/or weather effects..
The goal for now is to make the world feel alive before agents are introduced at the individual tile/biome level. Inevitably this involves careful calibration of several geological and climatic trends operating on a wide range of frequencies. In the long run, gravitation and erosion flatten while plate tectonics shift into novel continental configurations. In the medium run, hot and cold fronts bring green houses and ice houses. In the short run, weather effects such as floods, droughts, and fires occur regularly (as well as seasonal variation in temperature, humidity, radiation, etc.).
A wide range of event frequencies requires a sparsely updated cellular automata that uses weighted coins or die to draw out slow processes relative to quick processes. Rare conditions are checked first resulting in something like..
if rare outcome and condition A:
change cell to A
else if less rare outcome and condition B:
change cell to B
else if less rare outcome and condition C:
change cell to C
...
Some processes are not so simple and require memory (i.e. as a script-wide variable or data output to a file). Clouds temporarily cover other cells and thus require a dictionary called clouded that tracks prior cell states. Once clouds uncover a cell, the cell is switched back or hydrated into soil if dirt before. Volcanic eruptions have a chance to occur on static or dynamic fault edges and thus this information needs to be remembered/tracked. Other climatic effects will likely also require some data storage to achieve desirable realism.
All of this is relatively easy compared to the next part of the project which is to automate agents with interesting behavior! In order to scale up arbitrarily, the singular "biome" scene needs to be reasonably independent from the "valley" scene which instantiates a tile map. Later the "mind" scene will be reasonably independent of the "biome" scene. Yet, there should be some room for the mind to determine agent behavior and for agent behavior to influence collective agent behavior (leading possibly to some influence over the biomes/tiles).
The biome scene will need to render agents into sprites on the fly so that 1000+ sprites are not constantly being updated in the background. Should be simple enough given agents will just be black circles of various sizes. Activities can be highlighted through particle animations. The minds and genetic codes of agents serve to model what they are doing even when not being viewed in their current biome. This may lead to some bizarre updates when zooming out and in if the model is not robust, though this may be a necessary trade off in order to simulate more agents with more interesting minds than would otherwise be possible.
Lineage and language trees will also be remembered. Users will be able to analyze why an agent engaged in some behavior by looking at their mind history, their genetic ancestry, and/or their emergent language translated back into the universal language underlying the mind logic (partially developed here as a model of agency).
The goal for now is to make the world feel alive before agents are introduced at the individual tile/biome level. Inevitably this involves careful calibration of several geological and climatic trends operating on a wide range of frequencies. In the long run, gravitation and erosion flatten while plate tectonics shift into novel continental configurations. In the medium run, hot and cold fronts bring green houses and ice houses. In the short run, weather effects such as floods, droughts, and fires occur regularly (as well as seasonal variation in temperature, humidity, radiation, etc.).
A wide range of event frequencies requires a sparsely updated cellular automata that uses weighted coins or die to draw out slow processes relative to quick processes. Rare conditions are checked first resulting in something like..
if rare outcome and condition A:
change cell to A
else if less rare outcome and condition B:
change cell to B
else if less rare outcome and condition C:
change cell to C
...
Some processes are not so simple and require memory (i.e. as a script-wide variable or data output to a file). Clouds temporarily cover other cells and thus require a dictionary called clouded that tracks prior cell states. Once clouds uncover a cell, the cell is switched back or hydrated into soil if dirt before. Volcanic eruptions have a chance to occur on static or dynamic fault edges and thus this information needs to be remembered/tracked. Other climatic effects will likely also require some data storage to achieve desirable realism.
All of this is relatively easy compared to the next part of the project which is to automate agents with interesting behavior! In order to scale up arbitrarily, the singular "biome" scene needs to be reasonably independent from the "valley" scene which instantiates a tile map. Later the "mind" scene will be reasonably independent of the "biome" scene. Yet, there should be some room for the mind to determine agent behavior and for agent behavior to influence collective agent behavior (leading possibly to some influence over the biomes/tiles).
The biome scene will need to render agents into sprites on the fly so that 1000+ sprites are not constantly being updated in the background. Should be simple enough given agents will just be black circles of various sizes. Activities can be highlighted through particle animations. The minds and genetic codes of agents serve to model what they are doing even when not being viewed in their current biome. This may lead to some bizarre updates when zooming out and in if the model is not robust, though this may be a necessary trade off in order to simulate more agents with more interesting minds than would otherwise be possible.
Lineage and language trees will also be remembered. Users will be able to analyze why an agent engaged in some behavior by looking at their mind history, their genetic ancestry, and/or their emergent language translated back into the universal language underlying the mind logic (partially developed here as a model of agency).
Re: Daylen's Instinctual Dump
After playing around with cellular automata for a while I decided to simplify the tiles a bit and focus on a procedural generation algorithm instead. It was difficult to tune the automata to create something like simplex noise, so I may as well just implement it directly. Plus, for the weather effects I was going to require stored info outside the tilemap which sorta defeats the purpose of using cellular automata (although it still is pretty cool and will have its place in a future project).
To gain perspective on where to go I stepped back and started integrating more of the meta/macro work on ring inheritance. I read a paper on engineering ring structures (like in Halo, Ringworld, A Space Odyssey, Elysium, etc) here https://cdm16363.contentdm.oclc.org/dig ... cdi/id/15/
And that got me thinkin' more about origin myths in this universe.. and whether this universe is this universe.. which I decided yeah, but in billions of years after the front of habitation (rim worlds) reaches the edge of the galaxy and bounces back in towards the black hole at the center (while some activity is sent off to neighboring galaxies in the local cluster). Instead of being "rim worlds" they are "once rimmed worlds" or a ring super colony consisting of long line(s) of ring descent(s). Rings being made of carbon blankets (nano weaved) that wrap around to spin holding rock, water, and air.
Some rings are of first descent in the sense that the agents that initially inhabited the rings emerged as if from a petri dish. Many other rings are built from mythical "great ring forges" at the edge the of galaxy to be populated by the ancient ones [more like manys] (although from the perspective of the characters generated in volutor such myths are just that.. mythical). As the integrity of words fades over many generations of storytelling.
I was going to divide rings into valleys but then decided to go back to the original plan of having a potentially continuous space to scroll around. Though, rings tend to have a singular divider that serves as a control center for the ring itself (and thruster and lazor). Such areas are initially sealed off from future inhabitants in mountainous terrain (running on autopilot or by pirates that hijacked the ring due to overpopulation on their own rings). The ring scene then in godot is responsible for the main simulator loop that updates tiles and their weather, as well as volving agents and activities over time. Agents are bounded entities that have births and eventual deaths, and activities are processes that start and eventually end. Agents have a genetic code "oxxooxoxooox..." that is expressible in the environment, thus evolving upon reproduction. Some agents have a mind that involves a centering amongst tension. Extentions seem to come from external circumstances (tile changes, weather, communication, sensational activities, thoughtful activities), and intentions seem to come from internal circumstances (genetic code expressions, dreams, contemplation, feelings, intuitions).
Agents with minds could potentially be updated all at once every so often or every main loop so that tensions negotiate to update activity engage and disengage. All agents across their lifetimes are either engaged or disengaged in an activity at any given time. Mindless agents are constrained by strictly extensive responses and lack intention. Intentional agents prioritize between activities (F) and project circumstances onto more general cases (N), allowing for emergent, active creation in which an agent leads a new activity that then becomes a staple of culture.
Still debating whether to use collision logic between agents in the tile scene. Perhaps having a realistic movement pattern is not as interesting as having a clear sense of active locations and social relations. So, each tile could have recurrent (potentially seasonal) resource points around which agents can gather (leading to layered orbital patterns). These orbital configurations could encode information about the activities utilizing the centered point, the agents involved, and graphs that help infer evolutionary adaptations and agent relations. Also, instead of points being the center of attention, agents can make communication the center of attention (e.g. campfire conversation) and also other agents (e.g. initiation rituals, tree chopping, hunting, witch trials).
Below the tile scene being the mind scene which is a nest of circles with a history of impression and expression. Tensions press the mind in at least four ways: out -> in, out -> in -> out, in -> out, and in -> out -> in .. visually appearing as triangular reflections off of a centering point. This leads into various agent states like awake, dreaming, sleeping, etc. having different influences/purposes in the development of the mind. Many reflections will skip over thoughts and feelings near the boundary of agency (remaining perceptual rather than judgemental).
To gain perspective on where to go I stepped back and started integrating more of the meta/macro work on ring inheritance. I read a paper on engineering ring structures (like in Halo, Ringworld, A Space Odyssey, Elysium, etc) here https://cdm16363.contentdm.oclc.org/dig ... cdi/id/15/
And that got me thinkin' more about origin myths in this universe.. and whether this universe is this universe.. which I decided yeah, but in billions of years after the front of habitation (rim worlds) reaches the edge of the galaxy and bounces back in towards the black hole at the center (while some activity is sent off to neighboring galaxies in the local cluster). Instead of being "rim worlds" they are "once rimmed worlds" or a ring super colony consisting of long line(s) of ring descent(s). Rings being made of carbon blankets (nano weaved) that wrap around to spin holding rock, water, and air.
Some rings are of first descent in the sense that the agents that initially inhabited the rings emerged as if from a petri dish. Many other rings are built from mythical "great ring forges" at the edge the of galaxy to be populated by the ancient ones [more like manys] (although from the perspective of the characters generated in volutor such myths are just that.. mythical). As the integrity of words fades over many generations of storytelling.
I was going to divide rings into valleys but then decided to go back to the original plan of having a potentially continuous space to scroll around. Though, rings tend to have a singular divider that serves as a control center for the ring itself (and thruster and lazor). Such areas are initially sealed off from future inhabitants in mountainous terrain (running on autopilot or by pirates that hijacked the ring due to overpopulation on their own rings). The ring scene then in godot is responsible for the main simulator loop that updates tiles and their weather, as well as volving agents and activities over time. Agents are bounded entities that have births and eventual deaths, and activities are processes that start and eventually end. Agents have a genetic code "oxxooxoxooox..." that is expressible in the environment, thus evolving upon reproduction. Some agents have a mind that involves a centering amongst tension. Extentions seem to come from external circumstances (tile changes, weather, communication, sensational activities, thoughtful activities), and intentions seem to come from internal circumstances (genetic code expressions, dreams, contemplation, feelings, intuitions).
Agents with minds could potentially be updated all at once every so often or every main loop so that tensions negotiate to update activity engage and disengage. All agents across their lifetimes are either engaged or disengaged in an activity at any given time. Mindless agents are constrained by strictly extensive responses and lack intention. Intentional agents prioritize between activities (F) and project circumstances onto more general cases (N), allowing for emergent, active creation in which an agent leads a new activity that then becomes a staple of culture.
Still debating whether to use collision logic between agents in the tile scene. Perhaps having a realistic movement pattern is not as interesting as having a clear sense of active locations and social relations. So, each tile could have recurrent (potentially seasonal) resource points around which agents can gather (leading to layered orbital patterns). These orbital configurations could encode information about the activities utilizing the centered point, the agents involved, and graphs that help infer evolutionary adaptations and agent relations. Also, instead of points being the center of attention, agents can make communication the center of attention (e.g. campfire conversation) and also other agents (e.g. initiation rituals, tree chopping, hunting, witch trials).
Below the tile scene being the mind scene which is a nest of circles with a history of impression and expression. Tensions press the mind in at least four ways: out -> in, out -> in -> out, in -> out, and in -> out -> in .. visually appearing as triangular reflections off of a centering point. This leads into various agent states like awake, dreaming, sleeping, etc. having different influences/purposes in the development of the mind. Many reflections will skip over thoughts and feelings near the boundary of agency (remaining perceptual rather than judgemental).
Re: Daylen's Instinctual Dump
A month later and I am still working on the declarative structure, ha. With godot it is quite easy to build up structure without implementing functionality (i.e. the imperative part). This is part of my process considering that I would like to have a blueprint that I am satisfied with before proceeding to connect everything up. There is a lot to think through in terms of what scenes are responsible for what functionality and how to arrange and manage nodes within scenes. This project is forcing me into learning a lot of low-level details of the godot engine as well as programming patterns that will make things easier down the line. Here are some current thoughts rumbling around in my current washing machine of thought... which will likely be spinning for another few months at least.
Civil Topology & Topography
Something that has been on my mind for a while now is how the structure and stability of a civilization could be influenced by topological changes such as from a sphere to a ring. Poke out two holes where the poles are and stretch/compress what remains into a ring. In the case where a ring has a river connected around the ring the incentive landscape could be drastically different in that water returns back to you so pollution could be suicide. This seems a bit unstable for large populations, though perhaps a population below a million would thrive on the shared fate of pollution coming back around to..
I am generally aiming for like 10,000 - 100,000 agent simulations, so like a small city. Though, I am working with a large range cause I am not sure what the cost will be to run these simulations yet.
A prototype for the ring scenario could inspire other simulator topologies that then help towards establishing a [mostly computational] science that looks at the connectedness of civilizations in a very general sense and how these connections or lack there of constrain the volutionary paths from the past leading into the future.
Interface
The first prototype will likely be overkill with the degree of control in the interface scene to alter the ring simulator. This goes hand and hand with testing. The ring scene will be instantiated from the main scene along side the interface as a sibling and is meant to be completely automated (no interaction necessary to see life emerge). Controls like adding sources, replicators, and agents, and like brushes that change properties over an area, will allow the user to alter the destiny of dots on a screen.
Seed to an Immersive World
Speaking of dots on a screen, I am thinking perhaps AI art is headed in a direction that will allow third-person accounts to be colored and shaped into first-person experiences that humans can immerse themselves into. The art would initially come from humans as many disconnected models of various styles to be averaged and extrapolated over an entire world. The future could allow for the design and experience of new worlds with minimal effort distributed widely.
Ring and Biome Scenes
One challenge I have been working through is how to allocate functionality between the ring and biome scenes (I called the biome scene the tile scene in prior post). Though, as I learn more about the design and use philosophy of godot, things have become clearer. Agents that move across biomes will be instantiated and managed by the ring scene and agents that are stationary will be handled by the biome scene. Simple. Perhaps not so simple is how to start building an agent! Here is where replicators may be a step in the right direction.
Replicators & their Genomes
Replicators superclass agents to implement the volutionary essentials like movement, hydration, and consumption. They reproduce asexually and have a small chance to spawn near sources. Perhaps there are five categories of replicators that can bond to become agents of various species: eastwise, westwise, frontwise, backwise, and stationary. Stationary replicators don't move and are handled by the biome scene. Eastwise replicators migrate east indefinitely, westwise replicators migrate west, frontwise replicators migrate clockwise within and between biomes, and backwise replicators migrate counter-clockwise within and between biomes. The ring is an object in space that spins to imitate gravity on the inner surface. East and west are relative to which side of the ring is front and which is back. The front of a ring is considered the side that spins clockwise.
Replicators have room in their genomes to mutate. Currently the genome is just a dictionary in the replicator script that can be accessed by the ring or biome scenes upon the replicator signalling a replication event. Bonding is a mechanism whereby replicators of opposing kinds (e.g. eastwise and westwise) can bond to produce a sexual agent. The details of how these bonds lead to gendering and perhaps speciation is unclear at the moment.
Sources
Sources are simple diamond objects that spawn quasi-randomly over the landscape, decay due to entropy, and are accessible by replicators/agents. Replicators can feed on them and agents can build with them given the right adaptations.
Agents & their Connectomes
Agents will be subject to an adaptation tree with various lines having different affordances towards the persistence of genetically encoded phenotypes implementing a survival strategy. Some adaptations are dependent on others (e.g. swimming ability is required for diving ability). The volutionary algorithm will explore various expressions of the adaptation tree allowing for the emergence of a technology tree in which agents utilize an object with turing complete capabilities. In the process of volution, agents learn within their lifespan to connect their genome to the biomes making up the ring via a connectome. The connectome being modulated by a hierarchical state machine (awake: active or inactive, or asleep: dream or deep) to conjure up possible mutations in how a survival strategy in executed.
This is where the STFN model comes into play. I am thinking that the active state will bridge SN (sensed objects to points of origin), the inactive and dream states will tend to suggest branches of a light weighting stemming from completed bridges (adding noise), and the deep sleep state will tend to dampen unnecessary noise and sharpen focus in the next activity. Still lots to work out as far as how all this works analytically to favor robust minds.
Utilities
This square kind of object will subclass into perhaps trans, builds, tools, and bots. Agents will be able to create various technologies from sources. Trans to move agents and other utilities, builds as structures agents can enter/exit, tools that agents can carry on them, and bots to allow for programmable combinations of all the above. Utilities do not yet need to be implemented. For now I am focusing primarily on rings, biomes, sources, and replicators.
Civil Topology & Topography
Something that has been on my mind for a while now is how the structure and stability of a civilization could be influenced by topological changes such as from a sphere to a ring. Poke out two holes where the poles are and stretch/compress what remains into a ring. In the case where a ring has a river connected around the ring the incentive landscape could be drastically different in that water returns back to you so pollution could be suicide. This seems a bit unstable for large populations, though perhaps a population below a million would thrive on the shared fate of pollution coming back around to..
I am generally aiming for like 10,000 - 100,000 agent simulations, so like a small city. Though, I am working with a large range cause I am not sure what the cost will be to run these simulations yet.
A prototype for the ring scenario could inspire other simulator topologies that then help towards establishing a [mostly computational] science that looks at the connectedness of civilizations in a very general sense and how these connections or lack there of constrain the volutionary paths from the past leading into the future.
Interface
The first prototype will likely be overkill with the degree of control in the interface scene to alter the ring simulator. This goes hand and hand with testing. The ring scene will be instantiated from the main scene along side the interface as a sibling and is meant to be completely automated (no interaction necessary to see life emerge). Controls like adding sources, replicators, and agents, and like brushes that change properties over an area, will allow the user to alter the destiny of dots on a screen.
Seed to an Immersive World
Speaking of dots on a screen, I am thinking perhaps AI art is headed in a direction that will allow third-person accounts to be colored and shaped into first-person experiences that humans can immerse themselves into. The art would initially come from humans as many disconnected models of various styles to be averaged and extrapolated over an entire world. The future could allow for the design and experience of new worlds with minimal effort distributed widely.
Ring and Biome Scenes
One challenge I have been working through is how to allocate functionality between the ring and biome scenes (I called the biome scene the tile scene in prior post). Though, as I learn more about the design and use philosophy of godot, things have become clearer. Agents that move across biomes will be instantiated and managed by the ring scene and agents that are stationary will be handled by the biome scene. Simple. Perhaps not so simple is how to start building an agent! Here is where replicators may be a step in the right direction.
Replicators & their Genomes
Replicators superclass agents to implement the volutionary essentials like movement, hydration, and consumption. They reproduce asexually and have a small chance to spawn near sources. Perhaps there are five categories of replicators that can bond to become agents of various species: eastwise, westwise, frontwise, backwise, and stationary. Stationary replicators don't move and are handled by the biome scene. Eastwise replicators migrate east indefinitely, westwise replicators migrate west, frontwise replicators migrate clockwise within and between biomes, and backwise replicators migrate counter-clockwise within and between biomes. The ring is an object in space that spins to imitate gravity on the inner surface. East and west are relative to which side of the ring is front and which is back. The front of a ring is considered the side that spins clockwise.
Replicators have room in their genomes to mutate. Currently the genome is just a dictionary in the replicator script that can be accessed by the ring or biome scenes upon the replicator signalling a replication event. Bonding is a mechanism whereby replicators of opposing kinds (e.g. eastwise and westwise) can bond to produce a sexual agent. The details of how these bonds lead to gendering and perhaps speciation is unclear at the moment.
Sources
Sources are simple diamond objects that spawn quasi-randomly over the landscape, decay due to entropy, and are accessible by replicators/agents. Replicators can feed on them and agents can build with them given the right adaptations.
Agents & their Connectomes
Agents will be subject to an adaptation tree with various lines having different affordances towards the persistence of genetically encoded phenotypes implementing a survival strategy. Some adaptations are dependent on others (e.g. swimming ability is required for diving ability). The volutionary algorithm will explore various expressions of the adaptation tree allowing for the emergence of a technology tree in which agents utilize an object with turing complete capabilities. In the process of volution, agents learn within their lifespan to connect their genome to the biomes making up the ring via a connectome. The connectome being modulated by a hierarchical state machine (awake: active or inactive, or asleep: dream or deep) to conjure up possible mutations in how a survival strategy in executed.
This is where the STFN model comes into play. I am thinking that the active state will bridge SN (sensed objects to points of origin), the inactive and dream states will tend to suggest branches of a light weighting stemming from completed bridges (adding noise), and the deep sleep state will tend to dampen unnecessary noise and sharpen focus in the next activity. Still lots to work out as far as how all this works analytically to favor robust minds.
Utilities
This square kind of object will subclass into perhaps trans, builds, tools, and bots. Agents will be able to create various technologies from sources. Trans to move agents and other utilities, builds as structures agents can enter/exit, tools that agents can carry on them, and bots to allow for programmable combinations of all the above. Utilities do not yet need to be implemented. For now I am focusing primarily on rings, biomes, sources, and replicators.
Re: Daylen's Instinctual Dump
I find myself thinking a lot about patterns of behavior. Using behavior trees, the design of artificial intelligence is made more accessible but still quite complicated. Behavior trees have composite, decorator, and leaf nodes. Composite nodes control the flow of ticks through the tree (parallel, sequence, and selector). Decorator nodes condition the ticks of leaf nodes. Leaf nodes execute behaviors. So, the root of a behavior tree is built from composites that support decorations and terminate in leafs.
The classes of organisms I am currently considering are plants, replicators, reproducers, and agents. Plants may use a sequence composite as a root, replicators may use a selector as a root, reproducers may parallel into both a sequencer and a selector. An agent may then build on the reproducer to parallel into a sequencer, a selector, and other parallelizers. Sequences constraining/freeing the selectors to act towards reproduction.
Parallelization across multiple levels is essentially what game engines allow for and is the most difficult of the three composites to manage. Sequences tick from left to right until a node fails(*) and only succeeds if all nodes succeed. Selectors tick from left to right until a node succeeds and only fails when all nodes fail. Sequences put goals last and selectors put goals first. Though, this may be a weird use of the word goal since agents will likely have the goal of death be all the way over on the left of their selectors. At first reproduction will likely be right along side death implying the purpose of life to be sex, though this can be adjusted so that if an agent is too successful at finding mates then boredom might open up more doors in the parallel part of the behavior tree besides their core. Relates to sex being a solved problem in the White Trash Book Smart thread.
Anyway, I am still in the thick of all this behavior tree stuff... slowly going insane... luckily I don't really know what it is that is going insane and what it is that is going sane.
(*) thus sending a failure signal down the tree towards the root, beyond which the simulation loop might loop back to try that sequence again.
The classes of organisms I am currently considering are plants, replicators, reproducers, and agents. Plants may use a sequence composite as a root, replicators may use a selector as a root, reproducers may parallel into both a sequencer and a selector. An agent may then build on the reproducer to parallel into a sequencer, a selector, and other parallelizers. Sequences constraining/freeing the selectors to act towards reproduction.
Parallelization across multiple levels is essentially what game engines allow for and is the most difficult of the three composites to manage. Sequences tick from left to right until a node fails(*) and only succeeds if all nodes succeed. Selectors tick from left to right until a node succeeds and only fails when all nodes fail. Sequences put goals last and selectors put goals first. Though, this may be a weird use of the word goal since agents will likely have the goal of death be all the way over on the left of their selectors. At first reproduction will likely be right along side death implying the purpose of life to be sex, though this can be adjusted so that if an agent is too successful at finding mates then boredom might open up more doors in the parallel part of the behavior tree besides their core. Relates to sex being a solved problem in the White Trash Book Smart thread.
Anyway, I am still in the thick of all this behavior tree stuff... slowly going insane... luckily I don't really know what it is that is going insane and what it is that is going sane.
(*) thus sending a failure signal down the tree towards the root, beyond which the simulation loop might loop back to try that sequence again.
Re: Daylen's Instinctual Dump
Zoom out a bit and each class of organism is a character body(*) that has a body, a boundary, a blackboard, and a behavior tree. The body is a sprite/image, the boundary a collision detector, the blackboard contains stateful data, and the behavior tree contains stateless behaviors. When interacting organisms signal up to either the biome or ring level where certain key inter-organism behaviors are to be implemented. As much as possible should be done at the organism level, though the ring and biome levels will inevitably need to act as intermediate players between the organisms and perhaps as local and global limiters for the population.
There will likely be many more replicators and plants than reproducers and many more reproducers than agents. Though, not overwhelmingly so as to not feel crowded. At least 100 agents is a good target at first with perhaps 200-300 reproducers and 400-900 replicators and plants.
(*) Characters can move_and_slide, move_and_collide, and so forth. From which algorithms for fleeing, wandering, arriving, pursuing, evading, etc. can be coded.
There will likely be many more replicators and plants than reproducers and many more reproducers than agents. Though, not overwhelmingly so as to not feel crowded. At least 100 agents is a good target at first with perhaps 200-300 reproducers and 400-900 replicators and plants.
(*) Characters can move_and_slide, move_and_collide, and so forth. From which algorithms for fleeing, wandering, arriving, pursuing, evading, etc. can be coded.
Re: Daylen's Instinctual Dump
In lifestyle design guides, especially those by Barbar Sher, a common exercise is to think about what you were passionate about before you reached the age/stage of puberty.Though, this may be a weird use of the word goal since agents will likely have the goal of death be all the way over on the left of their selectors. At first reproduction will likely be right along side death implying the purpose of life to be sex, though this can be adjusted so that if an agent is too successful at finding mates then boredom might open up more doors in the parallel part of the behavior tree besides their core. Relates to sex being a solved problem in the White Trash Book Smart thread.
Re: Daylen's Instinctual Dump
For sure, one way to think of it is that the sequencer and selector are performing a temporal pincer where early sequencing and late selection blend into late sequencing and early selection. Not sure how all this will be connected yet though I imagine that the "natural" course of development will tend to be deviated from with random coin flips or die rolls and that such coincidences will later help anneal organisms towards a web of goals or activities given a somewhat stable or predictable environment. Plus each organism will have their own unique set of weights for their coins and dice.
Re: Daylen's Instinctual Dump
Another angle being that sources appear and disappear with a static rgba value (color + transparency) corresponding to a factorization of chemical substances bundled/pooled/crystalized together. Alpha or transparency being a density measure. Red being animal-like structure that is protein and fat dense (due to animal decomposition and excretion). Green being plant-like structure that is carb dense (due to plant cleavage and decay). Blue captures other substances leaning more towards the inorganic side and enables utility construction (metals, minerals, ores, rocks). Each source being spawned with a value from either the ring, the biome, or from another source (that signals to the ring/biome).
Both plants and replicators(*) having some kind of absorb or siphon behavior that "recharges" rgba values from existing sources. Reproducers and agents being larger animals that grow and consume with movement/steering acceleration being inversely correlated with body size/mass. Reproducers and agents can be carnivores, omnivores, or herbivores, each having different trade offs. One universal being that animals have a tightly coupled homeostatic process that targets a particular rgb state with little slack. In contrast, plants have a loosely coupled homeostatic process that is free to vary rgb state by seasons and beyond.
(*) replicators being simple animals that do not grow, consume, or interact much asides from decomposing dead plants and animals.
Both plants and replicators(*) having some kind of absorb or siphon behavior that "recharges" rgba values from existing sources. Reproducers and agents being larger animals that grow and consume with movement/steering acceleration being inversely correlated with body size/mass. Reproducers and agents can be carnivores, omnivores, or herbivores, each having different trade offs. One universal being that animals have a tightly coupled homeostatic process that targets a particular rgb state with little slack. In contrast, plants have a loosely coupled homeostatic process that is free to vary rgb state by seasons and beyond.
(*) replicators being simple animals that do not grow, consume, or interact much asides from decomposing dead plants and animals.
Re: Daylen's Instinctual Dump
The activities of sources, plants, and animals could also enter into feedback with local biome weathering and global ring climate. Each biome itself having an rgb value that varies based upon prolonged plant and animal habitation (carbon-nitrogen balancing for the soil). Day and night being simulated by modulating biome colors with greyscale (white for light, black for dark).
At first the biomes can be drawn procedurally with a bag of blues, a bag of browns, and a bag of golds. Blues being water biomes, browns being dirt/soil biomes, and golds being sandy biomes. Initial climatic values related to heat and humidity maps generated by noise being factored into biome placement around the ring. Weathering erodes rock to spawn the first sources that then go on to mutate into the first plants and replicators. As plants and animals coexist in a biome, greens will be overlapped onto the map demonstrating ecological succession (grassland -> forest -> dense forest).
Height being encoded into slopes that gradually scale and overlay greys onto the biome until a tipping point where white snow is covering parts of the biome (or the whole hexagonal biome). This is one of the variations in the environment that incentivize alternative movement capabilities (climbing). Water promotes swimming. Both water and height promote flying and leaping.
At first the biomes can be drawn procedurally with a bag of blues, a bag of browns, and a bag of golds. Blues being water biomes, browns being dirt/soil biomes, and golds being sandy biomes. Initial climatic values related to heat and humidity maps generated by noise being factored into biome placement around the ring. Weathering erodes rock to spawn the first sources that then go on to mutate into the first plants and replicators. As plants and animals coexist in a biome, greens will be overlapped onto the map demonstrating ecological succession (grassland -> forest -> dense forest).
Height being encoded into slopes that gradually scale and overlay greys onto the biome until a tipping point where white snow is covering parts of the biome (or the whole hexagonal biome). This is one of the variations in the environment that incentivize alternative movement capabilities (climbing). Water promotes swimming. Both water and height promote flying and leaping.
Re: Daylen's Instinctual Dump
Depth can be encoded through collision layers that entities can switch between and monitor. Allowing one organism to jump, fly, or swim over/under another without collision.
Many of the uncertainties I have with the overall design have to do with how organisms update what they know about their environment as it changes without bothering the main loops/nodes too much. This matches up well to the paradigm of incomplete information and self-world modeling being done within the organism scenes with dynamic memory allocation.
The sequencer might lend hand to behavior that does not necessarily move the body (freeing the selector to move aggressively). Reminds me a lot of the right/left brain distinction in that the sequencer is the master and the selector is the emissary.
Many of the uncertainties I have with the overall design have to do with how organisms update what they know about their environment as it changes without bothering the main loops/nodes too much. This matches up well to the paradigm of incomplete information and self-world modeling being done within the organism scenes with dynamic memory allocation.
The sequencer might lend hand to behavior that does not necessarily move the body (freeing the selector to move aggressively). Reminds me a lot of the right/left brain distinction in that the sequencer is the master and the selector is the emissary.
Re: Daylen's Instinctual Dump
Now open to hosting free tutoring/talk sessions through video chat to members of the forum on MBTI, spiral dynamics, or other esoteric topics discussed here. I am easy to talk to, patient, and can be enthusiastic!
Such could even be recorded and shared here or on YouTube. Definitely not required.
Such could even be recorded and shared here or on YouTube. Definitely not required.
Re: Daylen's Instinctual Dump
Not really, I grew up having lots and lots of time to reflect on what I thought was the absurdity of American consumptionism. If there were obvious paths in this society towards becoming a monk.. I might have. For a while now (or sense birth?) I have had a sociopathic and asexual leaning which turned advertisements of all varieties into comedy and fascination to me. I have learned empathy slowly, allowing for a steady integration of my thoughts with my feelings.. hopefully in a way that is conducive to the functioning of wholes I am currently embedded in.
Re: Daylen's Instinctual Dump
I thought your answer might have been something like only seeing beauty in the life of the mind...
Nonetheless there has to be a sense of values in play here, otherwise you could have turned your clear intelligence and what you are labeling sociopathology into a lot of money, at least as multiples of your burn rate. I imagine you saw the game not only as absurd, but basically it had to be evil or repellent.
Going sub-JAFI without during the 5 years to get the stash is the harder way, but is some ways that only makes it more admirable.
Nonetheless there has to be a sense of values in play here, otherwise you could have turned your clear intelligence and what you are labeling sociopathology into a lot of money, at least as multiples of your burn rate. I imagine you saw the game not only as absurd, but basically it had to be evil or repellent.
Going sub-JAFI without during the 5 years to get the stash is the harder way, but is some ways that only makes it more admirable.
Re: Daylen's Instinctual Dump
Putting volutor on a side burner.. hard to juggle all the components right now. In the meantime, thinking I might carve out a niche in developing puzzle games. Started thinking yesterday about "Maze" which is a puzzle game where you try to find the exit of a huge maze with a bunch of smaller mazes and perhaps puzzles within it. Two different NPC's being guards and guides. Guards block your way and guides show you the way, but it might not be clear who is who (you may get lead astray).
This will be good practice with behavior trees as well as with some other godot nodes I haven't touched yet like 2D joints (springs, grooves, and pins).
This will be good practice with behavior trees as well as with some other godot nodes I haven't touched yet like 2D joints (springs, grooves, and pins).
Re: Daylen's Instinctual Dump
Wrestling with "reality", whether through dialog, manipulating equations, or doing science experiments is simultaneously the most personal and the most impersonal activity. Otherwise, you are either just pushing against the universe or the universe is just pushing against you.