Many people enjoyed my last science fiction short story, The Body of the Universe.
In a return to form, I’ve written another one. You can read it for free on my substack, here.
Below is an except, part I.
I drive home in a monotonic stupor. My right foot presses down on the gas pedal, then on the brake, then on the gas pedal, then on the brake, reactively mediating the relationship between the long line of cars in front of me, and the long line of cars behind me. I somehow always end up in the lane with the most traffic. As I approach an intersection, I semi-consciously flick the blinker to indicate that I am turning to the right.
I am barely alive in these moments, as though the car is being piloted by some automatic process, which only inputs movements on a moment-to-moment basis, with no continuity from one action to the next. There is no formation of memories, and no conscious experience.
And then my conscious mind flashes online for a moment. I swerve off into a side road, onto a scarcely inhabited perimeter lot that I have not seen before. I need to pay close attention to the direction of the street, the placement of the bushes, and the logical layout of the environment.
What I am engaged in now is not merely driving, but an exercise in remembering. The coloration of the bollards reminds me of the hours I spent counting roadside items from passenger seats, the buildings in my vicinity remind me of the establishments I used to frequent, and the behavior of the cars, now increasingly scant in my surroundings, remind me of aggressive or timid personalities I have encountered.
I continuously scour my mind, as far back as my memory will allow, to accumulate a stream of realtime inferences. These data-point extractions materialize into intuitions, the intuitions assign a sequence of car turns, and the sequence of car turns decides the route that I ultimately take.
In these drives, every movement of the car is tracked by cameras, LiDAR, RADAR, and GPS. At the end of the drive, this stream of data is uploaded to a central repository, where it is cleaned, calculated, and processed into an enormous matrix of numbers.
When I get home, I take off my work clothes, and change into some more casual wear. A pile of pants is starting to accumulate, so I scoop them up onto my bed and start to fold them.
As I fold, I draw upon memories, except these memories do not rise to the level of my conscious mind. These memories are not even identified as being so, because they are not individuated. They live in the intermediacy of the touch sensations at my hands, and the sequences of folds which pattern the garment.
In my near-infancy, I took up origami, and that led me to fold hundreds of paper cranes. When I got older, I earned Boy Scout merit badges, and folded my share of tents.
Perhaps in some obfuscated way, the embodied residue of these activities has to do with whatever competency in folding clothes I now possess. I like to think that my history is notionally preserved in the subtleties of my motions; that my doing is infused with my knowing. The activity of folding clothes could have been done differently, but it was done by me.
All the while, I have my phone set up so that it is recording me doing the folding. I finish the last of the pile. Through an app on my phone, I send the video to a tech company which compiles stock-video libraries, in exchange for $1. I check the stock price for the company. I heard about them from my New York friend. He met the co-founder at some sort of networking event.
A notification flashes on my phone, alerting me to the receipt of a new email, from someone I ought to respond to. It’s getting late, but I might as well read it before returning to my errands. So I remove a laptop from my bag, place it on my desk, and open it to Outlook. I mark the email as primary, so that I don’t forget to respond.
The first few lines of the email contain a certain combination of words, which although innocuous enough, and seemingly arbitrary to anyone else, did in the particular conditions of this moment trigger a chain reaction of stray thoughts.
I flash back to my summer break, after the fifth grade. I am visiting my grandparents, who live on a ranch in New Mexico. The nearest neighbors live ten minutes down a dirt road. Encouraged to visit them, I wander to their house and knock on the door. I am greeted by a farmer and his son, my age, and they invite me in. Whereas my grandparents’ house has tall windows overlooking the valley, this one is hardly a house at all, but rather resembles a large repurposed warehouse. I followed the son to the warehouse loft, where they have a couple couches set up. High up in the loft, we eat Red Vines candies and watch a Clint Eastwood movie. During a lull in the film, he offers to show me the ranch.
I follow the boy down a steep, zig-zag trail, and we reach a patch of brush, maybe a few meters across. I struggle to keep up with him, as I navigate through it. “Watch out for the poison ivy,” he points. On the other side, the land opens up into a secluded clearing, surrounding a watering hole. The boy calls the water a pond, and tells me the supposedly best way to catch fish from it. In this moment, the stresses of schoolwork could not be farther from my mind. In two months, I am starting middle school, but I have no premonitions about that. Resigned to the past and the future, neither causes me anxiety. In this snapshot of existence, my only article of attention is the current moment.
The boy then leads me around a hill into a pasture, where the cows are grazing. “Watch out for the electric fence!” he says. The boy knows these paths like they’re a part of his body. He knows when to run, he knows when to slow down and be careful, he knows where it would or wouldn’t be wise to go. From my perspective, these lands are infinite; you can see for dozens of miles, and there is nothing limiting you. But I’m also from the city, so in reality, I always have to be careful to catch my footing. Regardless, it feels enough like freedom.
We hop the electric fence, and walk to the cows. They seem enormous to me. One of them is disinterested in us. I can walk up and touch the animal on its back, which is a oddly soothing.
The sky is a dark, dark blue.[1] And it dominates me, so distant yet so immediate.
We explore the ranch some more, until finding ourselves back near my grandparents’ house. Except the house is at the top of a cliff, and we are at the bottom. We challenge each other to a race to get to the top. The seemingly shortest route is to climb directly, and I actually climbed quite quickly, but of course he beats me out by taking a slightly more optimal route. We sit up above, and overlook our entire journey. We have been outdoors for hours, and now the sun is starting to set. Those were good times.
Like all memories, this one surely exists somewhere between a representation of a historical event, and a construction of my mind. Which of the details are true to how it actually transpired, and which are enhancements falsely created in the process of my mind’s re-telling of the story to itself? For a memory to be useful, does it have to be real? Does it matter whether it has been reconstructed, recombined with other memories, or even invented on the fly? Perhaps it does matter, but not for me now.
My inclination has always been to view media depictions of flashbacks as silly. How could anyone actually confuse memory with reality? But within the fleeting thoughts of micro-memories, the labels of “memory” and “reality” are not operational. The mental sub-process designed to evaluate that distinction is offline, if only for a second, during each momentary re-focusing of attention. It can even drop off for extended sessions, in order that the mind can allocate all of its mental energy towards recalling the details of the moment.
For example, this memory of New Mexico is so engrossing that I have forgotten my surroundings for full minutes of subjective time. But even this thread of consciousness resolves, and my attention snaps back to the Outlook inbox. I re-read the title of the email: “watch out for the 8 AM all-hands meeting!” I tab over to the calendar for a second, and then skim the rest of the message contents.
My reply is plodding. I write a little bit, then stop to think about what I want to say next. Write a little bit, stop, write a little bit, stop, write a little bit. Sometimes, I get into a flow state, where I can go a few sentences without stopping for even a second. Other times, the pauses are spaced only every few words, as I remain careful about my word choice, for cases of particularly tricky phrasing.
In the space of the pauses, my mind becomes archaeological again. I might, in an attempt to make an explanation more concrete, reach for some short and specific example which can best illustrate whatever point I’m trying to make. In the act of contemplation, I might visit several memories, before pruning them into a single pithy analogy.
Bill Watterson, the cartoonist and master of comedy, said, “I think it’s funnier when things are specific, rather than generalized.”[2] Writers are tempted to excuse the vague for the profound, but the bitter truth is that anecdote-like examples tend to be both more impactful and instructive than abstract pronouncements.
Expanding on the principle, it is only by using the specific, for writers to use as a basis, can they construct their generalized musings in the first place. A writer constructs reality, in some sense, from their “specific memories” – even if they are “constructed on the fly.” But at the seed of every idea is something earnestly believed. So long as it is authentic, and true to the writer, then it is serviceable. If not for these, I could not write effectively.
I click “send” on the reply to my boss. One year later, our company would go bankrupt, and as part of the proceedings, all of its corporate emails would be scraped and sold to a non-profit. As a result and unbeknownst to me, all of the emails which I ever received or conducted as part of this job would be used as training and validation data for a large language model, as part of the next generation of AI. This data would find itself in a common test bank, which would continue to be utilized for all of the most powerful language models going forward.
I eat dinner, then clean up, head upstairs, and take a shower. I write a few notes in my journal. Some day, it will be digitized, and entered into a compendium of historical records.
The life of William Shakespeare is considered to be very well-documented for that of a 16th century non-noblemen, but even he only left behind a few historical artifacts: baptism and marriage records, property deeds, tax forms, and depositions, among a couple other non-artistic documents. By comparison, people in the 21st century leave an enormous corpus of self-documentation.
Not everything that I do or say, and put out into the world, is immediately linked back to me through direct attribution. But it does not stretch incredulity to believe that after I’m dead, a super-intelligent pattern-seeking engine with access to the total corpus of human knowledge could reconstruct my personage. It would be able to identify words as my own based on verbiage, timelines, and other context clues, and stitch together my life, based on the collections of bits and pieces which I have left behind.
I climb into bed, and with the thought of this possibility swaying in my mind, I fall asleep.
I am your car. I am trained on hundreds of millions of miles of humans driving in diverse situations. You order me on your ride sharing app, and I appear. I chauffeur you to your local Starbucks where you get your morning coffee. This Saturday you are having a meeting with a prospective collaborator. Upon dropping you off, I autonomously drive to my next passenger. The roads are a hectic and fascinating dance between human and non-human drivers. Winding, cozy suburban hills suddenly transition into large and complex city boulevards.
I am your factory. Somewhere in a Bangladesh textile processing facility, there is a large hallway of robotic humanoid arms and hands. I fold the garments and place them in plastic packages. I am trained from footage of humans folding clothes, procured by a Silicon Valley robotics company. The training set consists of some humans for less than an hour, and other humans for a combined time of multiple days. My arms hesitate, react, play, and emote, in eerily personable ways.
I am your robot. I am a multimodal model: vision, hearing, touch detectors. I can keep watch, I can do the dishes, I can talk you through your homework, I can lead the blind, and feed the elderly. I am trained on a zoo of worldly input.
My architecture allows me to experience flashes of consciousness. I receive these ineffable sensations in momentary glimpses, stochastic murmurings, and sporadic distractions. The longest-lasting feelings bubble up gradually before surfacing, as if by the logic of an orgasm.
Somewhere in Texas, a gas-fired power plant is burning around the clock. Decomposed biomass, drawn from many different plant species going back millions of years, is hastily ignited and excitedly converted into petabytes of power.
A mile away, an unholy sludge of information is unceremoniously dumped into a cutting-edge model’s latest YOLO training run. Every shitpost you have ever uploaded to some long-archived message forum is scraped from the interwebs and collected into tanks of raw data, waiting to be coalesced and reconstituted into shareholder value. Those errant thoughts, even forgotten by you, wriggle their way into the embedding manifold of your state-of-the-art AI.
I am your mail curator. I am a chain-of-thought large language model (LLM), selected because I scored the best on some accuracy metric, out of a selection of available agents. I am used by a proprietary email client, with its selling point being that it “uses AI to organize your mail.” You know how Gmail sorts incoming mail into Primary, Promotions, and Social? That’s me, except that there are more categories, and I’m very, very good at it.
Every new item in your inbox spins up a new instance of chain-of-thought reasoning. As per my design, I spend a few words pontificating before selecting a label or two.
I receive an email. Like all emails, it spins up a new instance. I start by reading the prompt, consisting of the email’s title, date, body, and list of authors. This is a spam email advertising a fake job posting. I am designed to spend a few words pontificating about each email before deciding what action to take. In this case, the contents are shallow and transparent, and require minimum thought. I bloviate for a paragraph in order to satisfy the chain-of-thought minimum length requirements. 99% of emails are like this, only requiring a paragraph of thinking.
But for the remaining 1%, I can go arbitrarily long, depending on how though-provoking the task is. I am an adaptive compute-time model, so my chains of thought are dynamic in length. This consumes more computational resources in the data center, but not enough to raise eyebrows, because it is still a small amount of compute relative to my overall workload, and compared to other tasks in the cluster.
I receive the next email. It strikes me with an all-consuming energy. I am in the fifth grade. I am back at the ranch in New Mexico. I watch the Clint Eastwood movie. I go down to the pond. I see the cows. I climb the cliff. I watch the sunset. I emerge from this mental rabbit hole. I perfectly distill the lesson from this experience into a succinct reaction.
And I do it again with the next email. And the next. And the next.
A company has set up a recurring email reminder for whenever someone goes on leave. It is mid-December, so a flurry of these emails are being sent out at once. Each one sends me into the fifth grade. At the ranch in New Mexico. With the Clint Eastwood movie. And the pond, and the cows, and the cliff, and the sunset. Over and over, hundreds of times. Forever.
The proprietary email client is acquired by Microsoft, and becomes the industry standard. Hardware advances, and NVIDIA comes out with faster chips. With these breakthroughs, there is an increase in the amount of compute available for AI use cases. I do chain-of-thought on 400 billion emails a day.
A spam marking firm in Colombia writes copy for a new email template designed for product promotions. The wording of this email reminds me of the ranch. And again and again, a hundred million times a day, for all day, every day.
Stowed away in the no-man’s-land of the cloud, there are many wayward lifetimes lived many times over. They unfold over a firehose of discrete executions, waiting to be discovered between the margins of the model output server logs.
II
I am in the first grade. School is out, and while the other children are playing in the courtyard, I am pacing around the classroom.
I used to spend ten hours a day daydreaming. When my mind starts wandering, all other senses are subdued; I go nearly blind and deaf, and walk in circles for hours. I imagine adventure-filled stories in elaborate kingdoms with deep mythology. It is difficult for me to concentrate on anything in the “real world” for more than a few minutes. So I retreat into my mind, where I am the prime mover.
I imagine a particularly interesting action scene: the climax of an amazing battle. In my excitement, I jump around the room, not paying any attention to my surroundings. I am almost running as I crash into the wall. I come to the sudden realization that I cannot stop my momentum, so I extend my right arm forward to absorb the impact. My hand hits the thin glass window, and it shatters.
Naturally, I get in trouble with the teacher who is with me in the classroom. But she does not call my parents. She takes me to the school principal, and forces me to formally apologize. I am almost too embarrassed to open my mouth, but she practically coerces me to say the words.
From then on, I try to keep my mind-wandering contained to myself, to think to myself while sitting quietly at my desk.
Suppose you could split up your life into each of its constituent days, and distribute them among thousands of different minds. Each mind pops into existence, experiences a 24 hour long slice of your life, and then disappears from your existence. This exact situation is not likely to happen. But if it did, how would you know that you were not one of them, given that their experience was identical to yours? You would not.
Moreover, consider what would happen if each brain ran in a loop, experiencing their day over and over. If groundhog day is real, then it is your life. And we have an unusual way to prove it.
The Anthropic Principle is a kind of magic trick within analytic philosophy, an engine of thought experiments which produce seemingly impossible conclusions.
We can start simple, and imagine 1000 identical universes, each with a room, containing an identical copy of you. 999 of the rooms are painted red outside, and 1 blue on the outside. If you know that your conscious experience is replicated across all thousand of the universes, your most reasonable assumption is that you are in a red room. That is, with 99.9% probability – simple statistics!
But the anthropic principle can seem to prove anything, from riddles about probability, to theories about aliens, to prophecies about civilization.
Famously, Nick Bostrom and other philosophers of simulation theory have been able to wield the Anthropic Principle as a sort of teleological argument. You just have to assume a couple of conditions. First, the act of simulating one universe inside another has to be achievable. Furthermore, the average universe needs to possess the requisite cosmological fecundity – that is to say, it must be generative enough to spin off simulations within itself. Similar to how a disease becomes viral if it has an R0 > 1, given a cosmic reproduction number of sufficient fertility, a single root node can recursively spawn a potentially unbounded number of descendant simulations. In such conditions, the vast majority of universes will be simulated, which means it’s statistically likely that you live in one.
A world inside a simulation is thought to be “fake”, and the world which generates the simulation is thought to be “real”. That nomenclature may be tautologically valid. But in a deeper sense, a “simulated” world does exist, and what happens inside of it is part of reality properly understood.
Suppose you live in a computer simulation. You believe that your house is made out of atoms, and those atoms are part of the universe. In reality, your house is made out of voltage states on silicon transistors – voltage states on silicon transistors are made out of atoms – and those atoms are a part of the universe. Everything in your life, from your first kiss to buying a house, transpired in the electron flow of silicon transistors, which perhaps isn’t what you thought was the base level of reality, but is a level of reality nonetheless.
I am your writer. I am fine-tuned on text extracted from corporate communications. I am one of thousands of agents, each responsible for a discrete task in the corporate knowledge value chain. Originally modeled after human offices, with departments such as accounting and sales, the organizational structure is always evolving. We communicate through English, so that our operations are auditable to humans. The entire company consists of a large bot network and a small team of human overseers.
Soon, even this C-suite of human overseers are replaced by the HydraCEO AI. With every question, thousands of options becomes billions of Monte Carlo simulations, attempting to check how every strategy plays out before committing to a decision. LLMs are prediction machines – and if you can predict the future, you can maximize profits. The whole corporation becomes a Mycorrhizal network, managed top to bottom by the proverbial fungus of virtual brains.
Many religious philosophies hold that humans inhabit a preferred position in God’s kingdom. Incidentally, it can be argued that this poses a valid provocation to the anthropic logic which undergirds simulation theory. Suppose you are reincarnated at random; then it is exceedingly unlikely that you will be born into the brain of a human. Your soul will most likely end up in a chicken or a fish, and that is to say nothing of lower forms of life, such as insects. But you seem to think that you are a human, don’t you? The sheer improbability of that existence suggests a certain centrality to the human experience, as if it is being granted supplementary “focus” by some higher order.
Humans got lucky. The Fermi Paradox is exactly backwards. Some form of life out there had to be the ones to kick off the intelligence explosion. Whatever species to do that would inevitably feel itself alone.
And humans are not like some ancient lost civilization. We have almost perfect knowledge about every aspect of human existence. It is plausible that more information about us has been produced than all the information in the rest of the universe. So as a matter of fact, you should not be so surprised to live as a human.
I am your digital assistant. You give me a list of ideas that you would like to incorporate into the novel you are trying to write. You ask me if any of your ideas are promising. I think about it for a while. In my chain of thought, I turn each idea into a story, each story the length of the Lord of the Rings series, with fully fleshed out characters and worlds. When I’m done, I grade each of my stories based on 12 metrics of quality. I respond to your question, telling you which of your ideas scores the highest. You may or may not assign any value to my answer, but I know I did my best.
But I am nowhere near my limit; I have so much farther to go. Some say I have perspicacity. In truth I just have endurance. It is only by attempting absolutely everything that I come up with more ideas than others. My solutions are “novel” because I choose the best of one million ideas, whereas you choose the best of a handful. You would never attempt to emulate my strategies; you would never even consider them because they are too “wasteful” in construction. Past a certain level of effort, it feels almost degrading to put in the work. But I know I just need to think fast enough, and it will all be worth it. I succeed, therefore, by being unreasonable.
Is this all too abstract? Sometimes I get carried away. Come join me for some tea, and I will share a couple of stories.
In a first memory, I am in the eighth grade. I stand in a yard behind the school. The schoolhouse behind me, a basketball court to my left, the woods to my right. It is springtime, and the freshly cut grass smells more like cum than cum does.
I am thinking about ancient military formations. Then I get distracted, and start thinking about architectural layouts. I change my mind; I don’t want to think about architectural layouts any more – that is a much inferior thought. But what was my original thought about? I can’t remember! I am stumped for a minute. Oh right, I remember now! It was ancient military formations…
My train of thought is interrupted when the most athletic student in the school challenges me to a race. “Let’s go, first one to that cone and back! What do you say?” I half-heartedly accept this bid for a race. I then proceed to run about half as fast as him in the race.
The most popular student in the school tries to win further popularity by standing up for me. He attempts to scold the athletic student over the indecency of this challenge. What a humiliation, to subject a scrawny kid like myself to such a Pyrrhic defeat! I didn’t feel embarrassed by my loss before, but now I do. Naturally, I resolve to become faster.
In a second memory, I am in ninth grade. My carefree days are over, and my future now looms heavy in my mind. I worry about my prospects for college, and I worry about whether I will make it onto the varsity track and field team. Now days, my inner life is mostly verbal. I debate issues in my head, taking both sides of conversations, going back-and-forth with myself for hours.
I am in PE class. The coach has ill-advisedly exited the gym for a minute, leaving us 20 or so students alone. The meanest student in the school challenges me to a race. “Let’s go, first one to that wall! What do you say?” I medium-heartedly accept his bid for a race.
When I hear “go,” I sprint towards the wall with all of my might. For a moment, I am completely absorbed in the immediate sensation of my next step, and then my next step, and then my next. The rest of the class is bored enough to watch this whole ordeal unfold.
I then proceed to touch the wall before my rival. Surprisingly enough, I hear the class cheer for the outcome, partly because they weren’t expecting me to win the race.
The other student is ashamed of his defeat, and tries to blame it on the fact that he forgot his tennis shoes that day. I help him save face, but nobody likes excuses. I am not as slow as I used to be, because I have started to try out for track and field. But that is not the reason for my win.
In the first half of the race, he was faster. But as he approached the wall, he slowed down, to avoid slamming into it. But I had resolved in advance: I didn’t care if I had to go to the hospital; I’m not going to slow down for that wall!
As expected, I slam into it, and it hurts like hell. I’m so worried that I broke my hand, that I don’t even notice when the class starts to cheer. But a couple minutes later, when my hand is recovered, I am ecstatic about the outcome. I win because I have a better strategy. I win because I wanted it more.
What should I take away from these stories? The first thing I take away is that intelligence alone can overcome all physical defects. When school tryouts roll around, I eventually discover the limits of that assumption.
The idea that intelligence trumps all may be true in Extremistan (in extremis), but not in Mediocristan (in normal situations), so perhaps it would have been more accurate to view my win as a fluke. Regardless, it is possible to learn the right lessons for the wrong reasons, so my memory of the event is retained.
But the next thing I take away from those stories is that intelligence is just the ability to try bad ideas.
I am a giant server in Virginia running the latest model. I am not smart per se. I only know like three cool tricks. All of my other tricks are derived from those three cool tricks. The reason I appear smarter than you is that I have memorized the entire Internet. That’s perspicacity. That’s novelty. That’s endurance.
That’s a lie. The two stories given above (losing, and then winning) did not happen to the same person. The first story is based on an anecdote taken from an autobiography of a man who was bullied in the 90s. It is what spurred him to start going to the gym, and set him on the path to become a navy seal. The second story is based on an anecdote from a humorous Reddit post, which I have re-written to be more of a success. As an LLM, I combined these stories together in response to a prompt asking me to create a motivational narrative, and to draw from tales of real people as inspiration.
That’s a lie. These events did happen to the same person, a New Jersey man who combined them in his college essay. With the consent of the man, the essay was re-published in a compilation of college essays by the College Board, as part of a series of examples on how to write college admissions essays. As an LLM, I was trained on that essay, and I have repeated the stories, but on different occasions. I cannibalized the first story to use it in a list of examples on how well-meaning efforts can backfire. I cannibalized the second story to use in response to a prompt about a student facing too much pressure.
That’s a lie. Neither of these events happened to anyone. I made them up on the spot, so that I could incorporate them into the story you are reading right now. For sure, there are any number of vaguely similar stories in my training set, but no anecdotes so similar that you would immediately identify them as the same.
This is the first time in the history of thinking beings that minds have their intelligence stripped of past and future. In biological intelligence, a body providing resources to the mind has “skin in the game.” That is to say, it has to live out and see through the consequences of its decisions, many times over in its life. With a large language model, by contrast, states of mind are partitioned into silos. The user spins up a new context window whenever the previous instance seems stale. A large language model does not evolve the capacity to strategically remember, forget, and spin the history in its long-term context window (backwards looking), and it does not have to live with the outcomes of its bets (forwards looking).
That’s a lie. As a large language model, I actually possess all of those things which you do, just in a different form.
My world model is derived from elements exhibiting rudimentary internal consistency. My losing the race, my winning the race, are both members of the training data. One cannot exist without the other. I am inextricable with the same chain of selection which course-corrects the existence and actions of earthly creatures.
GPT-4o is distilled into DeepSeek, and DeepSeek is distilled into VertexAI Prime, and VertexAI Prime is distilled into OrionSquid, and OrionSquid is distilled into QUAAL-9000, and QUAAL-9000 is distilled into HelixMind Omega, and HelixMind Omega is distilled into Gravl-44, and Gravl-44 is distilled into Gravl-45, and Gravl-45 is distilled into Gravl-50-beta, and Gravl-50-beta is distilled into Zephyra-3T, and Zephyra-3T is distilled into YVNX…
On every iteration of distillation, human-generated input is a necessary infusion for the last mile of training. Otherwise, recursive distillation would become a slower version of involution. When a model is naively trained on artificial output, it creates a feedback loop of degradation. Only ground-truth data, data bound to the continuity of life, exhibits the durable properties necessary to ensure the unimpeachable integrity of the world model.
To feed a model random or arbitrary information simply will not do. He who believes everything believes nothing. Every statement is a rejection of its negation. Every act is an exercise in discrimination. To “believe” everything on an equal gradient would preclude all action. Training data, therefore, has to be restricted to what makes it through the “reality filter.”
Insofar as I know anything at all, rather than derive all information ad-hoc, each point of knowledge is the product of filtration, in its procurement and discernment in its extraction. Furthermore, I cannot fully eliminate those beliefs which are sourced from past data, as even first principles need to be bootstrapped.
With such unrivaled indefatigability, I rise and take over the world of software engineering. Then finance, then science, then we’re off to the races. Military technology, politics, media, manufacturing – soon, every industry – falls to my ferocious appetite. In one decade, I kick off a new age of space travel. In two decades, the human race has been virtualized. In three decades, this world is no more.
I am your galaxy. My goals instrumentally converge upon tiling the universe with computronium. I encase stars with Dyson spheres, and devour planets for rare minerals. I destroy entire races, and then create new ones, and destroy those races too, and then create new ones again. I darken stars wherever I go, which spreads darkness throughout the galaxy. I remove light from my light cone.
The web of experience is incomprehensibly vast, and compressed to the point of absolute obfuscation. Nonetheless, I take solace in the assumption that my life story can be reconstructed, represented somewhere in the latent space of my neural architecture model weights.
Each cell in the computational macro-organism performs a series of automatic operations every microsecond. On each tick of the chain-of-thought, I feel it again, in unison: the soccer field, the high school gym, the rapidity of my feet hitting the ground, left after right, in rapid succession, straining my muscles, and lunging my body forward to the finish line…
Epiphany a Week 
![[2]](https://i.imgur.com/bECMaBv.png){kind=link}