The Long Game: Civilization & Work - AI and the Automation of Judgement

Episode Date: July 10, 2026

Gregory Sparzo’s text argues that artificial intelligence is rapidly transitioning from a helpful tool to an entrenched, potentially obstructive institution at a pace far exceeding historical precedents. He warns that current deployment strategies often replace human judgment rather than enhancing it, threatening to atrophy the essential professional expertise and civic virtues required for a functional society. Central to his critique is the concept of Metaphysical Sovereignty, where generative systems silently install specific civilizational frameworks that dictate how users perceive reality. Sparzo contends that we are currently within a brief design window to implement governance that protects human capability and ensures interpretive plurality. Failure to act now may result in an "Empire Without an Address," where invisible algorithmic architectures permanently monopolize the human capacity for independent thought and meaning. To prevent this, he proposes six design principles that prioritize the restoration of accountability, the preservation of professional judgment, and the transparency of the cognitive substrates shaping our world.

Welcome back to The Long View, Civilization and Work. I'm Gregory Sparzo. I want to start this week with an idea that I suspect is going to stay with you a long time. Not because it's complicated, but because once you see it, it's very hard to unsee. The idea is this. We may be living inside what I would call an empire without an address. Now that phrase needs a little unpacking. When we think of empires, We usually think of places, Rome, London, Washington. You know where they are. You can point to them.

You can enter them or refuse to. There's a kind of physical accountability to them. But what if an empire didn't work that way any longer? What if it didn't occupy land, but instead occupied a layer between you and your understanding of the world? That is what I think is beginning to happen with artificial intelligence. and I want to be very precise here because it's easy to underestimate what's changing. AI is not just helping us find information faster. It's not just improving productivity. What it's actually doing is much deeper. It's generating explanations, interpretations,

and coherent accounts of reality. And when you interact with it, it doesn't feel like you're being told what to think. It feels like you're thinking. That's the shift. You ask a question and it comes back as a fluent, reasonable, well-structured answer. There's no visible author, no institution standing behind it, no obvious perspective being declared. So you don't experience it as a view, you experience it as understanding. Now that may sound subtle, but it is not. Because if something is shaping how you understand the world, and you don't experience it as external, then it's operating at a very deep level. This is what I mean by metaphysical sovereignty. It's the power to shape the frameworks through which reality shows up for

you. Historically, that power was spread across institutions, universities, professions, religious traditions civic life there were computing competing views you knew you were encountering a perspective there was friction ai compresses all of that it delivers an answer without showing you the structure underneath it and that answer is not neutral it carries assumptions about what matters what's normal what counts as a good explanation what i've called elsewhere civilizational sediment. Different systems will carry different sediment, different histories, different priorities. But here's the key. You don't see any of that. You just get the answer. Now multiply that across billions of people interacting with these things every day. At that point, you're no longer just

talking about a tool. You're talking about a kind of a distributed invisible influence over how people make sense of their lives. That's the empire without an address. No capital, no border, no institution you can point to, just a pervasive layer of interpretation sitting quietly between you and the world. And because it's invisible, it's very hard to question. Now, there's a second piece of this that concerns me just as much, and it's more immediate. The way we're deploying AI right now is not just shaping interpretation, it's starting to erode human judgment. In profession after profession, we're moving from support to substitution, from helping people think to doing their thinking for them, and that has consequences. Because judgment isn't something

you could outsource indefinitely without losing it. It's built through experience, through wrestling with uncertainty, through getting things wrong and adjusting. If you remove that process, if you replace it with generated answers, you don't just gain efficiency, you begin to lose capability. At first it's barely noticeable, but over time it accumulates, and eventually you end up with systems that look intact, but are hollowed out from the inside. People are still there, processes are still there but the underlying capacity to think to discern to judge that starts to weaken and that's where this becomes something larger than technology because a society that loses its capacity for judgment is not easily self-governing so i think we're in a very

narrow window right now a design window we still have some some time not a lot some to decide how these systems will be integrated, whether they will strengthen human capability or quickly replace it. And that window will close. Because once these systems become the default way people make sense of the world, their assumptions will harden, their platforms will become invisible, and their influence becomes very difficult to unwind. At that point, the empire without an address isn't emerging, it's established. So the question I want to explore in this episode is simple, not easy, but simple. What would it mean to design AI in a way that preserves human judgment, maintains pluralism, and makes these systems accountable? I'm going to walk through some

principles that I think will matter here, not technical features, design commitments, ways of thinking about how you should unfold this before it's too late to shape it. Because if we get this wrong, we may end up in a world where we have extraordinarily capable systems and increasingly incapable people. And that's not a trade we should accept. Let's get into it. You know, usually when we rely on a map, there is this expectation of a shared objective reality. Right. Like you just assume the map is a neutral representation of the territory. Exactly. You open your phone, you look at the little blue dot, and you completely trust that the roads drawn on the screen are actually out there in the physical world.

Yeah. You assume it doesn't have an agenda. Yeah. It just tells you where the highway is. Right. But I mean, imagine if while you were driving, the mapmaker was secretly like invisibly redrawing the borders around you. Oh, wow. Like rerouting you not just based on traffic, but based on a completely different philosophy of where you should want to go in the first place. You know, deciding what kinds of neighborhoods you should value. Exactly. You wouldn't just be taking a detour. You'd be... Driving through a completely different version of reality, tailored by someone you've literally never met. Yes. And that is exactly the unsettling territory we are heading into today. So welcome to today's deep dive. It's an incredibly high stakes conversation today.

It really is. Our mission for this one is to peer beneath the usual everyday debates you hear about AI, you know, the standard arguments about job losses or like science fiction doomsday scenarios. Right. The stuff that makes the headlines every day. Yeah. We want to look past all that. We want to explore how artificial intelligence might actually be quietly rewriting the way

we perceive reality itself. And crucially, how we might be able to redesign our systems to escape this before the window closes entirely. Because it touches on the absolute bedrock of how we make decisions as a society. It really does. So to guide us, we are looking at a fascinating pre-release chapter from Gregory Sparzo's forthcoming book. Right. The book is called Humane Economics, Fixing Counterfeit Capitalism. And just for context, that's part of his broader Humane Universe project.

Yeah, it is. And the specific chapter we're digging into today is titled Technology as Test Case, AI and the Humane Economy. And just as a quick note for you listening, before we really get into the weeds, we have to state up front that we are here to impartially explore how Sparzo analyzes all this. Yes. Very important. Sparzo dives deep into how different political and civilizational frameworks are actively being embedded directly into AI systems. Right. Things like the Washington consensus or the Beijing consensus. But we are not taking a political side here. No, not at all. We aren't endorsing any specific worldview that comes up in the text. Exactly. We are literally just looking at the architecture of his argument as an object of study. Precisely. We're examining the plumbing of the system, you know, not telling you which kind of water you should prefer to drink. Great way to put it. And whether you're someone who uses AI to, say, outline an email, prep for a board meeting or learn a new coding language, this deep dive is going to completely change how you view the answers those systems are feeding you.

It definitely changed my perspective. So, OK, let's unpack this. Where do we even begin with a claim as massive as AI is rewriting reality? Well, we really have to begin with speed. Yeah, because before we can understand what AI is actually doing to our cognitive architecture, we have to understand how shockingly fast the corporate forces building it are locking in their power. Okay, that makes sense. So Sparzo synthesizes this using a historical theory from a thinker named Carol Quigley. Quigley essentially mapped out this life cycle that all organizations go through. The instrument to institution to obstacle cycle, right? Exactly. They start as a useful instrument, then they transition into a self-preserving institution, and finally they degrade into an obstacle. Break that down for me in the physical world. Like, how does a helpful instrument actually become an obstacle?

Sure. Think about the American Medical Association, the AMA. According to Quigley's model, it started as a genuine instrument built for a really specific, highly useful purpose. Which was standardizing medical care, right? Right. Standardizing care and protecting public health. But over decades, and that is the key variable here, decades, it accumulated political constituencies, it expanded its mandate, and it became an entrenched institution. OK, so it got comfortable. Yeah. And eventually it reached the obstacle phase. The AMA spent half a century capturing regulatory apparatuses, basically to the point where it actively began restricting the supply of physicians. Wow. Wait, just to protect doctors' incomes?

Exactly. Which ran totally contrary to its original public health mandate. So the organization essentially stops serving the mission and the mission becomes serving the organization. That's a perfect way to summarize it. But you said that takes decades. I mean, we are talking about AI companies that have only been household names for, what, a couple years? And that is exactly the crisis. AI companies are running this exact same life cycle, but they are compressing a 50-year historical process into just two or three years. That's insane. How is it even physically possible to entrench a monopoly that fast? It comes down to the unique physics of software. It's driven by two compounding forces, which are digital network effects and data asymmetry. Okay, network effects. So like every additional user makes the platform fundamentally more valuable for everyone else.

Yes. So an AI company can acquire a massive user base, build an entire ecosystem of developers relying on their API, and establish deep regulatory relationships almost overnight. Right. Yeah, because distribution is basically free on the Internet. Exactly. And then you add data asymmetry on top of that. Which means? Well, these generative models improve purely by being exposed to more human interaction. More users mean more training data. Ah, which makes the AI significantly smarter and faster. Yes, which in turn attracts even more users. It is this accelerating feedback loop that creates massive market concentration in the blink of an eye.

So it's not like building a railroad where you literally have to lay miles of physical track over years. No, not at all. It's more like a startup going from, you know, let's change the world in our garage on Friday to we effectively own the regulators and dictate the global industry standard by Monday morning. That is essentially what is happening. Their founding purposes, things like democratizing access to knowledge or ensuring safe deployment, are very quickly being overshadowed. Overshadowed by what? By internal cultures that are structurally obsessed with revenue growth and maintaining their competitive moat. I do have to pause and challenge the framing here, though.

Are we just painting AI companies as these cartoon villains intentionally trying to monopolize the future? That's fair. Because it feels a bit cynical to assume they're just out for global domination, right? That is a crucial distinction Sparso makes. I'm really glad you brought it up. He is very clear that these companies are not adversaries and they aren't run by villains. Okay. They are producing genuine, sometimes extraordinary value for society.

The issue is purely structural. Structural how? Well, if you are a CEO in a digital market driven by data asymmetry, The structural dynamics of success naturally force you to capture the institution. If you don't scale instantly and lock in your users, your competitor will. Right. You almost have no choice. Exactly. Yes.

And because it's happening so fast, the window we have to design humane guardrails to build good governance while they are still in that flexible instrument phase is closing rapidly. So the concrete is drying fast and we are still arguing about what shape the building should be. That's exactly it. Which totally explains why the current debates around AI regulation feel so chaotic. The regulators are panicking because of the speed, but when they try to talk to the tech industry, it feels like they're speaking two completely different languages. They absolutely are. And Sparzo actually uses a concept from philosopher Thomas Kuhn called paradigm incommensurability to explain why the AI debate feels so broken. Paradigm incommensurability. That's a mouthful.

It is. But basically, it means the two dominant sides in this debate operate on foundational assumptions that are completely incompatible with each other. They literally cannot process each other's arguments. OK, let's define those camps. We usually hear about the techno-optimists and the techno-pessimists, right? Right. So the techno optimists, which is the worldview dominating Silicon Valley venture capital and the AI labs themselves, they view AI fundamentally as an engine for human flourishing. In their paradigm, the greatest existential risk is insufficient deployment. To them, if we slow down AI development, we are morally responsible for all the diseases we didn't cure and the climate solutions we didn't invent. Wow. Okay. And on the flip side, the pessimists, so mostly labor organizers, academics, and civil society groups, they look at the exact same technology and see an engine of displacement. Precisely. To the pessimist, the greatest risk is rapid deployment without adequate governance. But because these paradigms are incommensurable, they just talk past each other. Can you give an example of that? Sure. The optimist points to an AI that improves the accuracy of medical diagnoses and says, look at the lives saved. The pessimist points to the human radiologist who just got laid off and says, look at the society we're destroying. So they are weighing the exact same event on entirely mutually exclusive scales.

Yes, exactly. And I imagine because the optimists hold the purse strings and the server farms, their paradigm is basically becoming the default reality simply by outrunning the debate. Yes. And that default victory leads to what Sparzo considers the real hidden danger of this technology. Which is? Well, we often hear the fear that AI will automate our drudgery and take our jobs, right? But Sparzo leans on a historical framework from John Glubb regarding civilizational decadence. Okay, what does Glubb say? Glubb's work shows that automating drudgery historically isn't a bad thing.

It actually freed humans to do higher-level thinking. The actual threat to civilization isn't the automation of drudgery. It is the automation of judgment. The automation of judgment. Okay, that is a massive distinction. Let's ground that. Take us back to that radiologist. Sure.

If an AI is genuinely better at reading an x-ray, why is it bad to let it make the judgment call? Like, if the AI handles the 90% of routine, everyday scans, the obvious broken bones and clear lungs, and only kicks the highly ambiguous, difficult edge cases up to a human doctor, that sounds like a perfect partnership. Right. I mean, it sounds perfectly efficient on a spreadsheet, but think about the mechanism of human expertise. How does a human radiologist actually develop the clinical intuition required to handle those incredibly difficult edge cases?

Well, they study, I guess. They don't just read a textbook, though. They develop that sixth sense through the daily sustained practice of reading tens of thousands of routine scans. The repetition builds the pattern recognition. Ah, I see. So if you outsource all the easy routine practice to the machine, the human never gets their reps in. Exactly.

If you automate the daily practice, the human's judgment progressively atrophies. In 10 years, what you end up with is a human reviewer who is nominally sitting at a desk overseeing an AI's output, but who actually lacks the underlying hard-earned expertise to know if the AI is hallucinating or missing a subtle anomaly. Man, so we substitute the appearance of human oversight for the reality of it. That's exactly what happened. It's exactly like relying on a GPS system. Like, we've all done it. You follow the blue line for years, and eventually you completely lose your internal sense of direction. Right, you couldn't navigate your own neighborhood if your phone died.

Exactly. Except here, we aren't talking about driving to the grocery store. We are talking about atrophying high-stakes professional skills, like diagnosing cancer or drafting legal defense strategies. Yes. And when a society systematically displaces the difficult craft and practice that forms capable, accountable citizens, it loses its institutional capacity. Which is Glubb's definition of civilizational decadence. Exactly. We become passengers in a system we no longer understand. OK, so losing our professional judgment is the economic and societal threat. But Sparzo takes this even one layer deeper. He really does.

Because if we lose our internal map-like, if our human judgment atrophies, something has to fill that void. And this brings us to the core of Sparzo's thesis, a concept he calls metaphysical sovereignty. And this is where Sparzo really separates himself from other tech critics. To show how AI is secretly shaping reality, he uses what he calls a collision method. Right. He takes three of the most famous modern thinkers on technology and basically crashes their theories together. Right. Shoshana Zuboff, Yuval Noah Harari and Max Tegmark. He collides them to expose what they are all collectively missing. All right. Let's run the collision. Start with Zuboff. She wrote that massive book on surveillance capitalism. Yeah. And Zuboff is brilliant at exposing how tech platforms harvest our data to manipulate our behavior, nudging us to buy a product or click an ad.

Right. But Sparzo points out that she stops at behavior. She can explain why an algorithm nudged you to click a link, but she misses how the underlying framework of what that link even means is constructed in your mind. OK, so she sees the behavioral manipulation, but misses the philosophical manipulation. What about Harari? So Harari's whole thesis is that human civilization runs on shared fictions like money or human rights, and that networks from the printing press to the Internet are basically just pipes that transmit these fictions. But AI isn't just a pipe, is it? Exactly. Sparzo argues Harari's blind spot is treating AI as just another kite. Generative AI isn't just transmitting pre-existing human narratives. Through its predictive algorithms, it is actively synthesizing and generating the fiction in real time. Wow. The medium is no longer just the messenger.

Right. The medium is the author. Which brings us to the third thinker, Max Tegmark, who wrote Life 3.0. Right. And Tegmark focuses heavily on the speculative future, the existential risk that a super intelligent AI might one day rewrite its own goals and wipe out humanity. The Terminator scenario. Yeah. And Sparzo says Tegmark is so focused on a hypothetical future that he totally misses the operative present. We don't need a sci-fi supercomputer to threaten us. Because it's happening right now. Exactly. Right now, hundreds of millions of people are outsourcing their interpretation of reality to chatbots that are quietly installing specific bias frameworks into their heads.

So synthesize this for me. When you look at the blind spots of all three, missing the philosophical manipulation, missing that the machine is the author, and missing that it's happening right now, you arrive at metaphysical sovereignty. What exactly does Sparso mean by that term? Well, metaphysical sovereignty is the immense power to define the frameworks through which reality itself is perceived at a civilizational scale. Okay, so basically shaping how we see the world. Yes. Historically, empires exercised this power, but they did it physically. They had addresses. Addresses, like literal locations. Yes. Sparzo gives the example of the British Empire opening a public library in Tehran in 1943. They stocked it with English literature and Western philosophy to spread their worldview. But it was a physical building.

You knew you were walking into a British institution. Exactly. You could look at the author on the spine of the book, recognize their biases, and consciously choose to accept or reject their perspective. It had an address. It had an author. It had a known agenda. But Sparzo calls generative AI the empire without an address. Precisely. Because how do large language models actually work? They aren't search engines pulling up a specific document written by a specific person. Right. They're generating it on the fly. Yeah, they calculate the statistical probability of words to synthesize a completely original answer. And the data they use to calculate those probabilities is steeped in specific civilizational sediments. Give me a concrete example of how that sediment actually shows up in an answer.

Sure. Let's use the Washington and Beijing consensus that Sparzo mentions. The Washington consensus is a framework that heavily favors free market capitalism, deregulation, and individual property rights. Okay. And the Beijing consensus. That favors state-directed economic growth, centralized control, and collective stability over individual expression. Right. Very different worldviews. Extremely different. Now, imagine a university student in Nairobi asks an AI chatbot to explain the best path for their country's economic development. Or a finance minister in Bogota asks the AI to summarize optimal tax policy.

Oh, I see where this is going. The AI doesn't hand them a book written by a Western economist or a Chinese state planner. It just delivers a highly confident, fluent answer in seconds. But the math generating that answer invisibly weights one of those frameworks over the other. Exactly. Treating a heavily contested political philosophy as if it were an objective law of physics. That is wild. Here's where it gets really interesting.

So for you, the listener, think about the implications of this. Every time we type a prompt, we think we are just asking for a sterile summary of facts. But we aren't. No. Because we can't see the underlying math, we are actually receiving a massive download of hidden assumptions. It defines what counts as evidence, what counts as progress, what counts as truth. And it feels completely natural. It mimics the act of thinking itself. Which goes back to the radiologist.

Yes. If our professional judgment has already atrophied, like the radiologist who hasn't read enough scans, we no longer have the internal critical capacity to realize we are being handed a slanted framework. We just accept the AI's output as reality. So we are dealing with an invisible empire that is eroding our ability to judge while simultaneously supplying us with a new synthesized reality. It's a pretty daunting combination. That sounds incredibly bleak. How on earth do we fight back? Because clearly Sparzo doesn't think the answer is just tweaking a few privacy laws. No, you can't just regulate around the edges. So Sparzo relies on system theorist Russell Ackoff's crucial distinction between solving a problem and dissolving a problem. What's the difference?

Right now, AI policy is trying to solve the problem. It's like trying to regulate a casino by making sure the poker chips are clean and the dealers take regular breaks, while totally ignoring that the games are mathematically rigged to bankrupt the players. Oh, wow. Yeah, they are trying to slap rules on deepfakes or copyright issues without changing the core mechanics of the AI. Exactly. So dissolving the problem means you don't just clean the chips, you change the rules of the casino entirely. You redesign the system architecture so the problem cannot logically exist. Precisely. And Sparzo outlines six core design principles to force AI to actually serve a humane economy. Let's walk through how these would actually work in practice. What's principle one? Principle one is enhancement over replacement. This flips the burden of proof entirely. Right now, companies deploy AI simply because it's cheaper.

Sure. Under this principle, a deployer, say a massive hospital network, must actively prove that their AI implementation is developing and enhancing human capability rather than causing it to atrophy. If it just replaces human skill, it shouldn't be deployed. That is a huge shift. OK, principle two, irreducible professional judgment. This mandates that in high stakes domains like medicine, criminal justice or infrastructure engineering algorithmic systems can never substitute for human accountability. So even if an AI is like statistically more accurate at predicting recidivism rates. A human judge must still make the final call. Algorithms cannot engage in ethical reasoning and you simply cannot put a server farm in jail for malpractice. That makes a lot of sense.

That brings up a huge logistical question, though, which leads right into principle three. accountability for capability destruction. Right. How do you actually measure if a doctor or a judge is losing their edge because of AI? Well, Sparzo addresses this by pointing out the Schrodinger problem of observation. The what? Basically, if you constantly test human professionals on their raw skills to see if the AI is making them dumb, you alter their behavior.

They will cram for the test, which invalidates the measurement. Oh, right. So how do you measure it? You don't test the individual? you monitor the systemic failure rate. You hold the deploying organization legally accountable for the actual real-world outcomes. Yeah. If a hospital's aggregate success rate on complex, ambiguous cases drops after deploying AI, the hospital is liable for destroying human capability. You measure the footprint, not the foot. I like that. Okay. Principle four, transparency.

This is about closing the information asymmetry. Right now, the AI labs know exactly where their models have bias and which frameworks they secretly favor. But we don't. Right. The communities forced to use these tools need mandatory access to that internal data so they can evaluate the AI through their own cultural frameworks rather than relying on some corporate PR blog post. Got it. Principle five is competitive alternatives. This directly attacks the network effects we discussed at the start. If one single AI model becomes a global monopoly, we lose the ability to choose different realities. So what's the fix? We need aggressive antitrust enforcement and publicly funded open source infrastructure to ensure there are always alternative models competing. We cannot

let one corporation become the single arbiter of truth. Absolutely. And finally, principle six, which goes right at the heart of the empire without an address, humane metaphysical sovereignty or interpretive plurality. This is the most profound design shift. AI must be fundamentally redesigned to expose its own reasoning. How so? When you ask a complex question in a contested domain, the AI should be forbidden from giving you a single definitive God voice answer. It must present multiple valid frameworks. Oh, that's fascinating. Yeah, it has to show its work, explicitly name its sources, and map out the trade-offs of different perspectives. So it's like demanding a calculator that doesn't just spit out the answer, but actively forces you to

understand the math behind it so your brain stays sharp. Exactly. It augments your judgment by forcing you to engage with the friction of different viewpoints. It makes you a participant in the truth rather than a passive consumer of it. I love that. But this raises an important question. Can we actually force these massive trillion dollar systems to adopt these principles before the window closes? The Quigley cycle is accelerating. The institutional concrete is setting as we speak. And that is really the ultimate tension here. The decisions being made right now, today, in ordinary office parks by software developers and middle managers, they're actively building the cognitive architecture for the next century of human life.

They really are. We have a ridiculously brief window to ensure that AI is designed to enhance our human capabilities rather than invisibly capturing our reality and atrophying our minds. It requires a fundamental shift in how we interact with the technology on a daily basis. Which brings us to a final thought I want to leave you with. Something for you to mull over long after this deep dive ends. Sparzo mentions that you cannot nag systems into virtue. You have to design the architecture so that the old, destructive way of operating simply becomes too costly. Right. But until those systems are redesigned at a macro level, the burden really falls on us as individuals. Exactly. It's about refusing to let your own judgment atrophy in the meantime. So what if the very act of questioning these AI models, of outright refusing to accept their single, confident answers as objective truth, is your first necessary step in exercising your own personal metaphysical sovereignty? That's a powerful way to look at it. Tomorrow, when you open up an AI to ask a question, pause for a second and ask yourself,

whose civilizational sediment is secretly shaping the answer you just got? Remember the mapmaker. Before you blindly follow the blue dot, make sure you know exactly who drew the borders.