A growing number of people truly believe AI systems are already conscious, or will be soon. A growing number of other people consider this view bizarre or dangerous. The trouble, as many in both camps admit, is that their disagreement cannot be resolved with the usual tools of empirical science. Neither the Turing test, nor IIT, nor J-space, nor any similar conceptualization of conscious activity as an observable quantity can verify the reality of any other being’s subjective experience.

Are these two camps therefore doomed to lock horns unproductively forever? No! At least I hope not? Because the question can be sidestepped. There is no need to resolve the question of consciousness empirically, because all the important ethical and legal questions that seem to depend on it — moral patienthood, status as a fellow responsible agent, etc. — turn much more cleanly on natural teleology. In short:

Aristotle to the rescue!

Instead of asking whether machines can feel, we should start asking what their natural end is, or what they exist for the sake of. As it pertains to artificial objects, this inquiry is affected but not determined by the intent of the maker: Aristotle described the action of the user of any artificial thing as “architectonic” over that of its maker. (Physics II.2, 194a33–b8.) He also allowed that the form of the tool itself could affect the hexis or character of the user in a kind of feedback loop, a theme further explored by McLuhan and others.

Happily, to gain insight into the telos of artificial things like AI, one need not resolve any factual question beyond the reach of empiricism. The relevant questions are simple and logical, like: Why do people make this thing? How do people use it? How does it tend to influence people? The respect we owe to artificial things follows organically from the answers to these inquiries.

In a sense, teleological inquiry is more complex than the binary question whether AI is conscious and can suffer. But while the question of AI consciousness does not turn on observable facts, teleological inquiry does.

A Genealogical Critique Of The Idea Of Consciousness

The term consciousness arose during modernity as a substitute for older ideas like soul and telos.

In early modernity, thinkers like Descartes and Bacon argued that Aristotle’s four types of causality (efficient, material, formal, and final) could be simplified to just two: efficient and material. [1] This was a brutally effective reduction of Aristotle’s world-model, and an amputation of his metaphysics. It made it far easier to think about how to model and manipulate material objects, unlocking scientific and technological breakthroughs. But it came at a cost, because it meant that the scientific-technological worldview contained no space for things like timeless ideals, or the essential inner natures of things.

Prior to modernity, Aristotle’s notions of formal and final causation acted as conceptual homes for many ancient intuitions about the human soul, morals, and afterlife. But these intuitions did not simply disappear when modern thought ascended. Thus, many old morally-important notions had to find modern homes. They did so in new concepts, such as human dignity and equality before the law.

The notion of “consciousness” arose as another one of these: a modern home for pre-modern ideas. It smuggled notions of soul and telos across the centuries — although not without considerable mouldering — by designating a place that science and technology could not reach.

Here is the n-gram, showing how the idea of consciousness surged in high modernity.

It might seem puzzling to us today that the notion of consciousness was simply absent prior to modernity. Were people back then unaware of this important, mysterious aspect of being? Of course not! The concept simply wasn’t needed, because formal and final cause, especially when combined with other concepts like “substance” and “person”, covered the same ground with equal or greater subtlety. Here are a few illustrative quotes from Aquinas (all of which require some vocabulary translation which I will not belabor here; but for example, it helps to know that in Aristotelian thought the soul is the “form” of a living body):

“Everything is good insofar as it has being, and the being of every thing follows upon its form; so each thing is good by its form and by the ordering to an end which follows from its form….’Person’ signifies what is most perfect in the whole of nature, namely, a subsistent individual of a rational nature….The universe of creatures does not exist for the sake of man alone, but for the sake of God’s goodness, since each creature exists for the sake of its own act and perfection.” Summa Theologiae I, q. 5, a. 5; q. 29, a. 3; and q. 65, a. 2

To me, all this adds up to a rather withering “genealogical” critique of the notion of consciousness. It unveils consciousness as not at all a normal object of scientific interest, like the interior or the Earth or the backside of the Moon, but a hardcoded limit to modern thought itself. It is a rough blanket that modernity threw over a large group of phenomena that were described in richer terms before modernity. It indicates what modern thinkers decided not to think about in order to begin thinking modern thoughts.

Thinking About Telos Instead Of Consciousness

This doesn’t mean consciousness doesn’t refer to anything. Quite the contrary. It just means that it is a rough and confusing concept. We are better off using the older Aristotelian vocabulary to attune our minds to what it refers to.

And this suggests a new way of thinking about “moral patienthood”. To discern what we do and don’t owe to AI systems, we should not ask whether they are conscious or even whether they suffer. Again, I want say this multiple times so as not to be misunderstood: It’s not that suffering isn’t real or doesn’t matter. But subjective experience matters derivatively, not absolutely, and with regard to the telos of the suffering thing. Subjective experience is like a photograph of a sunset, which might be good or bad, clear or obscured, but telos is the sunset.

In fact even stipulating to strictly modern-scientific terms, it is not hard to notice the strangeness of the Benthamite idea that pleasure is the ultimate good, and suffering the ultimate evil. From an absolutely hard-boiled biological functionalist point of view, experiences and sensations are not goods or evils themselves, but only indicators of external factors helping the organism toward its ends or impeding it from them. Pleasure is only evidence of something good, like, say, healthy food. Conversely, pain is not inherently a bad thing, but rather, in general, a beneficial-but-imperfect indicator of the possible presence of a bad thing. It is an alarm system. Thus, saying that we should not mistreat humans, animals, or AIs because they might suffer is like saying that we should not burn down buildings because the fire alarms might go off.

That is not why.

We should not burn down buildings in order to respect the good inherent in the buildings.

So too with beings.

—

[1] In a canonical illustration, a mug’s efficient cause is the activity of its crafter, its material cause is the clay it is made out of, its formal cause is the idea or concept of a mug, and its final cause is its ultimate socially-embedded role, its “for-the-sake-of-which”, such as drinking hot coffee at breakfast.

This is a modest talk but it’s the fruit of quite a bit of hard grappling.

It is hard to articulate exactly what the problem is in humanity’s relationship to technology. Increasingly, Aristotelian vocabulary feels almost indispensable. This talk explains why.

I conclude that building things that can do many things while simply punting to the user the question of what they’ll do with it — that is, “empowering” users in a nonspecific sense — ought to be regarded as the cardinal sin of ethical tech.

Good builders know exactly what their tools are for.

At first glance, Carissa Véliz’s new book appears to be making a familiar argument, that Silicon Valley manipulates us all through data-driven prediction. So I didn’t expect to learn much from it, but the central conceit of framing prediction as prophecy caught my eye. I’m glad it did. Véliz is not beating a dead horse, but exploring a live mystery: Silicon Valley’s narrative of inevitability. Why do we keep building, funding, and using technology we wish didn’t exist?

To be fair, 15 and 20 years ago, naivete was real. It is not an excuse. But we cannot doubt that many people really thought social media was an excellent thing, or at least had difficulty understanding what their salaries required them not to understand. This is no longer the vibe. The AI labs are staffed and funded in no small part by penitent souls who think someone else would swing the scythe if they didn’t.

This is bizarre. Are we being compelled? Maybe. Maybe our technological trajectory is being dictated to us by the icy hand of capital, or the blind anxieties of the security state, or the chaos of lowest-common-denominator decision-making. Or maybe we are simply sleepwalking toward dopamine, convenience, and beggar-thy-neighbor greed. But even if all these stories have their share of truth, they leave something out. Something else is happening too, something truly strange, which hasn’t quite been articulated yet. This is where Véliz’s insights are profound and inspiring. I’ll never look at the future quite the same way again.

Useless Predictions

If “techlash” were a stock, I would be an accidental billionaire. More than a decade before Cambridge Analytica, I became disturbed by the general gestalt of Silicon Valley culture—its greed, its naivete, its hubris dressed as altruism. I happened to receive the first-generation iPhone as a gift soon after its release, and shortly began exhorting my friends not to buy one even as they oohed, ahhed, and borrowed it to look things up on Wikipedia in the college cafeteria. This device, I said, could ruin our experience of the world by distracting us from one another, by making us constantly reachable by bosses and other tormentors. Unreachability is freedom. They rolled their eyes and said “then don’t have one”.

I remember the early 2010s, when the tech giants were the unquestioned toast of the bien pensants, as a time of deep moral and political loneliness. I breathed a sigh of relief when technology criticism entered the mainstream through the work of Tristan Harris, Jon Haidt, and others. Finally! But the fact that this has not prevented a marked increase in Silicon Valley’s power and irresponsibility in the 2020s has been disheartening in a way that is hard to capture.

In short, I predicted our mirthless present, but have mostly just suffered for it, while those who insisted that smartphones and social media were ushering in some excellent shared future are, for the most part, unchastened and rich. This unhappy history might explain why I have been slow to accept the idea that our technological predictions are self-fulfilling. First, if that’s so, then I guess I might have some apologizing to do? And second—well, it seems like many predictions have two sides, a material side and a moral side. Silicon Valley’s barons predicted more or less the same material developments I did. But they also predicted that those things would be good. If predictions are prophecies—well, what happened to that part?

Powerful Predictions

Around the time I graduated from college and moved back home to the Bay Area, small clusters of thinkers who would become known as rationalists and effective altruists began appearing at the margins of my social circles. They were a little culty, but this is only par for the course in the Bay Area. I was interested in what they had to say.

And almost everything they had to say was expressed as probability estimates.

To illustrate this style of communication—it is very much still with us—look at this opening from an April 2026 essay by Anthropic co-founder Jack Clark:

I’m writing this post because when I look at all the publicly available information I reluctantly come to the view that there’s a likely chance (60%+) that no-human-involved AI R&D - an AI system powerful enough that it could plausibly autonomously build its own successor - happens by the end of 2028.

This is a big deal.

I don’t know how to wrap my head around it.

It’s a reluctant view because the implications are so large that I feel dwarfed by them, and I’m not sure society is ready for the kinds of changes implied by achieving automated AI R&D.

I now believe we are living in the time that AI research will be end-to-end automated. If that happens, we will cross a Rubicon into a nearly-impossible-to-forecast future. More on this later.

Véliz argues that predictions about social phenomena are always a kind of self-fulfilling prophecy. Not all predictions self-fulfill, of course, but they all pull in that direction. Predictions by the powerful are more apt to tip the scales.

It’s not only predictions’ speakers who make them come true. The way that we (mis)interpret predictions heightens their power. Usually, we hear predictions as something like facts or hypotheses—that is, as inert, familiar kinds of statements. But predictions are not facts or hypotheses at all. They cannot be facts except with respect to the predictor’s subjective worldview, which is unverifiable. Nor can they be hypotheses, because they are unfalsifiable. Predictions are a unique beast. They are more like invitations or exhortations. When issued by the very powerful, they function almost as commands to accommodate ourselves to the predicted future, or else.

I have long subscribed to an ethics of prediction that runs something like this: First, we should not make bad-faith predictions, and should be on guard against others doing it—e.g., propagandists trying to move markets or politics. At the same time, we should welcome honest predictions, even unpleasant ones. Second, there are important ethical distinctions regarding proximity between predictor and predicted. Because predictions can affect the predicted system, it is problematic to make even honest predictions about one’s own relationships, as well as sports games one is playing in, etc. It is less problematic to make honest predictions about macroscopic events like the future of the country, the economy, or technology.

Véliz has persuaded me to rethink this. We probably need a stronger and more general taboo against predicting many social phenomena, even ones that seem big and out of reach.

Communicating predictions has a peculiar power, which you can feel in the Clark passage. Predictions that come off as good-faith create a seductive atmosphere of humility and intimacy. The predictor seems to be letting the listener in, on several levels: first by confessing uncertainty about the future, and then by disclosing something from the core of his first-person subjectivity. “From where I am sitting, it seems likely that . . . ”. Between the lines, the predictor is saying “I am a being like you, who does not know the future. And like you, I have an internal experience whose reality I cannot prove. But here is what it contains.” Being spoken to like this, even impersonally through social media, is powerful. It feels like being pulled aside and treated as a peer. Thus the listeners’ defenses are lowered, and the “prediction”—which, again, is functionally an exhortation or a command—is inserted as the payload. This is very much like the exploitative intimacy that pervades cults.

Bayesianism

Bayesianism is a centuries-old mathematical approach to probability. Originally it was just a set of concepts, useful to statisticians. But in the 20th century some Bayesians embraced the radical move of reinterpreting all their personal beliefs through the Bayesian idea of subjective probability. This basically just means assigning probabilities, rooted in available evidence, to everything you believe.

The clusters of people who adopted this norm became extremely influential. Why? It’s not that they convinced all that many people to think or speak like them. Instead, I believe, they convinced many people at one or two levels of social remove—people like me—to listen to them. I was never a member of these tribes, but I could see that they were putting their own worldviews through one hell of a gauntlet. So I took seriously what they believed about AI, and plenty else. I then went around influencing others, and I’m sure I wasn’t alone. In this way, probably just a few dozen people in the late 2000s powerfully reshaped the worldview of an influential several thousand.

Three truly world-changing predictions emerged from these Bayesian-inflected communities. The first was that computer systems would soon be able to automate the very same processes of prediction refinement that they used to form their beliefs—namely, the rapid updating of evidence-based subjective probabilities. The second was that this would result in smarter-than-human AI. The third was that smarter-than-human AI would then recursively improve itself and become hard or impossible to control and, one way or another, threaten humanity dramatically. I hardly need to recount the arc of these predictions’ self-fulfillment. Predictions attract capital which attracts talent which validates predictions, leading to updated predictions, which attract more capital.

In fact, if you want your predictions to be as accurate as possible, you’re going to gravitate toward predicting things that are affected by your predictions. Your predictions will gravitate towards becoming commands.

The first of the three predictions has come good. About the second, my only comment is that forming better subjective probabilities seems like a poor definition of being smarter. And about the third, I shall say nothing.

Faith and Freedom

Saint Paul said faith was “the substance of things hoped for, the evidence of things not seen.” Bayesian subjective probabilities are not that. They are evidence of things seen, translated into estimates about the future that scrupulously avoid hope. Faith and prediction are almost opposites.

In recent years, the practitioners of prediction have pitched it as a way of increasing freedom. If we have a good guess about what the future holds, it seems like we’re likelier to be able to do as we please, and not have our plans knocked off course by unforeseen events.

But no. That might be power. It isn’t freedom. Our predictions compel us. They pull the future into the present, solidifying it, shrinking the actual space of possibilities. The more we predict, the “better” we predict, the more we believe predictions, and the more powerful the predictors grow, the more they constrain everyone in their orbit; and the more we require faith to recover our freedom.

Ephraim Radner recently noted that to suffer means not only to experience but also to bear or carry. To suffer is to exert upward force, to hold something up.

That way of looking at suffering, as a positive action upon an object—a transitive verb rather than a passive “experience”—opens the door to a richer perspective. It belies the “therapeutic” narrative which says that suffering is something to eliminate, even fear. It also punctures certain too-easy assumptions about reducing suffering as the master heuristic for improving the world.

What, or whom, does suffering hold up? It holds up beings. Without suffering, we cannot make necessary sacrifices for the greater good of self and other. When we see this link, we endure life’s slings and arrows—we toil now for an uncertain reward later, we do seemingly thankless favors for friends. But when the causal link between our suffering and the good seems absent, a new and fearful kind of suffering appears. In these cases, we not only bear a burden but also writhe and chafe under it, seeking to escape its meaninglessness.

But even this intense meta-suffering is not really meaningless. Especially at its most incomprehensible, suffering bears up the full weight of reality. It holds up the world at the heavy end, Atlas-like. (Look at the painting, above, by Jan Mandijn. Job is the fulcrum around which the whole surrounding society turns.) Enduring suffering that seems to lack reason, justice, or payoff, is the grandest and most ontologically necessary thing any of us ever do.

This means that fleeing from suffering is a hazardous business. Fleeing implies not only shirking one’s duty, but putting the whole world at risk. Suffering should not be gratuitously sought—that is vainglory. And when the suffering of others becomes incomprehensible to them in the way I described—so that they are not able to endure it—we must do everything to help them. It is not fair that the sufferer should hold up the world, and our job as loving social beings is to lighten their burden. But when suffering knocks at our door, sometimes we need to let it in. And the more we can face it without demanding immediately to see the good it does, the better.

To call suffering meaningless when we cannot understand it is to slander the whole of reality, which includes that suffering. The illusion of suffering’s meaninglessness arises from imagining our mind as a kind of sealed chamber in which suffering goes on without external consequence. But our minds are never so sealed. Just as we are more than merely our experiences—we are also our bodies, our histories, our destinies, our relationships, and so on—so too our experiences, like suffering, are not just films playing to an empty theater. We implicitly acknowledge that suffering can have consequences, and good ones, on the frequent occasions when we bear it for the good of our selves, our friends, our family, our fellow countrymen, our fellow humans, our planet, our God. When suffering seems meaningless we are not finding truth, but the limits of our own understanding.

Yet terrifyingly, the suspicion of suffering’s meaningless can be self-fulfilling. For when we chafe, writhe, and run from our suffering, we withdraw from associations, and we close our mind, reshaping it almost into exactly kind of empty theater that really does robs suffering of its redemptive power. We must actually suffer our suffering.

Adherents of intensely secular and rationalistic ethics often believe that by taking subjective experience onboard as the center of their moral concern, they are being laudably non-reductionist. And this is true, to a degree. These thinkers’ philosophical priors make them particularly aware of a terrible alternative: treating persons, animals, and planets as mere coincidental heaps of physical and informational phenomena. They know that this worldview would remove moral barriers to exploitation and violence, and they know it is all the more dangerous for being readily defensible in the scientific-rationalistic idiom in which they are well-trained. Their abhorrence of such reductionism is therefore admirable and ethical. So when intensely secular and rationalistic ethical thinkers accept conscious suffering as the hallmark of moral salience, one can look at them as, in a way, intellectually lashing themselves to the mast. As long as they hold this axiom, they will not wreck completely on the shoals of materialist amorality.

This strategy for moral grounding is at least somewhat influenced by Buddhist thought. Classical Buddhist teaching portrays Dukkha—a rich concept related to suffering and dissatisfaction—as both a cornerstone of reality and a phenomenon whose transcendence is identical to enlightenment. Since the Buddhist path seems on the surface to advocate escaping suffering while the view I have outlined points toward embracing it, the two may seem to be in tension. But I think they are perfectly consistent. Buddhism teaches not that we should flee from suffering in any superficial sense, but rather that we should transcend it by accepting it with equanimity. And at least according to the largest branch of Buddhist tradition (Mahayana), we should do this not to become a blissful dropout but to become a radically-ethical enlightened being, a “Bodhisattva”. Buddhism and Christianity thus both say that experiencing suffering is the key to redeeming oneself and the world.

Yet just a slightly different read of Buddhism, especially when mixed with a helping of utilitarianism, materialism, and consequentialism, leads to a very different view, in which suffering is not a path to salvation and redemption, but rather a kind of unreal and fruitless evil—a needless defect of reality which merits no apologia and whose extermination by any means necessary is the height of ethical life.

I hasten to acknowledge that in the right hands, at the right time, this uncompromising hatred of suffering can be very good and useful. It chainsaws efficiently through certain malignant species of bullshit. Callous premodern aristocrats, for example, no less than modern captains of industry, are clearly and squarely indicted by the injunction to care about others’ suffering. It leaves little room for mystical-traditional mumbo jumbo about caste, or libertarian ideology, or whatever other kind of overcomplicated mularkey helps people feel better about resting their boots on others’ necks. I welcome that.

In the wrong hands, though, the same conceptual equipment can make suffering seem like a trivial and curable disease, breeding forgetfulness of its profound dignity and metaphysical necessity. When the poor, the sick, the old are told that their travails mean nothing and do no good—well, this is not so, and it ought never be said. This is an insult and a falsehood heaped upon an injury. This is related, I think, to why consequentialists sometimes get quite exceptionally exercised about Ursula le Guin’s poetic short thought experiment, the Ones Who Walk Away From Omelas. The idea that suffering does actual work to hold up society is not only puzzling, but very troubling. This doesn’t make it untrue.

Rene Girard’s mimesis/sacrifice theory isn’t, in my view, quite the “explainer of everything” that it wants to be. But it is a valuable resource for those initially skeptical of the notion that suffering and sacrifice might play non-obvious causal roles in complex social systems. These fearsome things really do act as glue of society important ways, yet because their workings are more formal than efficient they are nearly invisible to a certain mode of analysis. And this also suggests a unique reason why machine consciousness might be a question with real stakes. Maybe AI won’t be able to do truly profound, social-fabric-repairing work for us unless or until it can suffer the way goats and pigeons once did on temple altars. Now, for the avoidance of misunderstanding, I have my tongue weakly planted in my cheek here. But only weakly. Questions like this will arise if people take machine interiority seriously.

The rotating spirals create an illusion of radiating motion. When I first showed this to him, it stopped him cold. He often laughs and playfully reaches for images—not now. He became serious and transfixed, almost stereotypically “hypnotized”. His breath slowed. When I turned the page, he reached to turn it back. He wanted to keep looking at it.

This transfixed reaction was a little bit funny, but also quite touching. It immediately struck me as testimony to the reality of his consciousness, his perfectly vivid interior experience. Not that I ever doubted it, but still: one person can never truly prove to another that they are having an experience. This reaction struck me as extraordinary circumstantial evidence.

I

Why did it feel so close to a proof-of-sentience, a true Turing test? It’s a combination of a few things.

First: the thing to which he was ostensibly reacting is an illusion. It is interesting only in its surface appearance, not in its reality. I can think of nothing else about those paper spirals that might provoke a strong reaction besides the uncanny and ultimately mistaken perception of radial movement.

Second, seeing the illusion has no utility, no obvious adaptive benefit. If human vision did not produce this illusion, we’d presumably be fine. Thus, reacting to it is unlike other common reactions, like smiling in delight at his mother’s face, or reaching for food. These are reactions that you’d program into a robot geared to survival. But a robot geared to survival would optimally not be gripped by this optical illusion. It would just record the visual data as rotating paper spirals with nothing out of the ordinary. Seeing this visual effect is evidence of our limitations, the ways that we are not tightly-calibrated to survival.

Third, he is not sophisticated enough to have been “performing” his astonishment. He already knows how to use certain reactions to send signals to his environment, such as crying for help or food. But this look of captivated amazement was something I had never seen from him. He is a very clever baby, but at eight months old he has nowhere near enough of a world-model to feign fascination with a disorienting illusion for his father’s benefit.

Admittedly, this is still not a proof. The human cognitive machinery produces this illusion of radial movement due to various little “quirks”: the limits of optical focus, the geometry of the counter-rotating spirals. And arguably—not obviously, but at least arguably—it actually is adaptive to be fascinated by it. Perhaps humans are robotically attracted to scrutinizing perceptual errors in order to defuse their power of disorientation.

Still. Let us sink no further into the rabbit hole of skepticism just now. I only want to emphasize this: that it is natural, intuitive, and reasonable to posit genuine inner experience as the best explanation for babies’ astonished reactions to illusions. When (1) the thing reacted to is clearly illusory rather than substantial, such that (2) the reaction has no clear adaptive function, and (3) the reacting being is unlikely to be sophisticated enough to be performing, the result is a visceral, intuitive, and not-unreasonable sense that we are in the presence of another mind.

Let’s call this the “Naive Astonishment at Illusions” test.

II

We often presume that other minds make themselves known through sophistication (i.e., operating with an elaborate world model) and intelligence (solving practical problems). The Turing Test, for example, posits that we could consider a computer to have a mind if it were indistinguishable from a human in ordinary conversation. When proposed in the 1940s, the Turing Test seemed to be saying that after computers multiplied in power, surpassing a human-level threshold, they would become recognizable, rightly or wrongly, as other minds.

It is therefore somewhat counterintuitive to suggest that other minds’ errors (i.e., perceptual illusions) and naivete actually help them achieve social recognition. Yet we now have machines that can easily pass the Turing Test; and while some humans have grown convinced that AIs have interior “hard-problem-style” experience, most remain uncertain or skeptical. Why?

Perhaps clearing performance hurdles, including “human level” sophistication, is just not really the most persuasive evidence in prompting us to recognize other minds. The NAI test helps us see why evident sophistication might make it harder, not easier, for a mind to secure recognition. Because agents perceived as sophisticated cannot be perceived as naive; and only naive agents can persuade others that they are not performing for instrumental reasons.

What makes the NAI test so powerful?

We all know how experiential content “just appears” in our own sense field. We can influence it indirectly (we can choose to look up at the sky), but we cannot control it in detail (it is not up to us whether a bird happens to be visible there). And the uncontrolled, unexpected contents of experience sometimes powerfully grip us. Indeed, they can affect our entire being very suddenly, upending our world-models and scrambling our goals. Each of us feels this happen from time to time.

When I believe that I have a fairly good model of your mind, it feels distinctly as though I can watch it happen to you. I can witness how certain experiences seem to grip you, and verify that they are similar to the ones that might grip me. But when I believe that you are operating with a model of the world that is, in relevant ways, more complete or more elaborate than mine, I can never be confident that you are not performing your reactions in the service of an undisclosed goal. And goal-directed performance is not an indicator of consciousness. It is consistent with mechanistic, experience-free, adaptive action—all the way up to an indefinitely high level of sophistication.

Thus, we encounter an extremely important kind of persuasive evidence of other minds only when we look just slightly “down” the sophistication gradient — such as at children and animals, or at peers in their vulnerable moments. These beings seem similar enough to us to permit reasonable modeling of their inner experience; and also unsophisticated enough to be trustworthy.

They alone can pass the NAI test.

III

Let us assume for a moment that something like the NAI test is our dominant practical means of probing other minds (at least in contemporary Western culture). If so, we are accustomed to recognizing other consciousness mainly when gazing “down” sophistication gradients, and not “up” them. This has major implications for how we encounter other minds in social and spiritual settings. It even suggests an intriguing new interpretation of the phenomenon of our own consciousness.

This story gives a plausible account of why we sometimes seem to “see” other minds relatively easily in children and animals, yet sometimes have more difficulty recognizing the interiority of peers, especially when we perceive them as sophisticated potential rivals. Perhaps this explains some of the rough, instrumental, and dehumanizing treatment one finds on playgrounds and in offices.

But the greatest complications for the NAI test arise when we encounter the most ostensibly sophisticated actors. I mean figures like parents (at least while we remain children), but also human leaders in political, professional, and cultural fields; as well as the entire grimoire of intelligent nonhuman entities including corporations, nations, ecosystems, planets, angels and demons, gods, and AIs. We may be able to deductively reason our way to accepting the consciousness of sophisticated humans—we have enough evidence from elsewhere to accept that humans are conscious even when we can’t quite perceive it with the NAI test. However, in the case of institutions and supernatural beings, the NAI test entirely rules out recognition. These beings either lack perceptual mechanisms sufficient to generate an illusion that we might plausibly share, or lack plausible naivete, or both.

To be sure, consciousness is (rightly or wrongly) recognized “up” the sophistication gradient all the time. There is no shortage of tests other than the NAI test, beginning with the Turing Test. The trouble with these, though—and the attraction of the NAI test—is that it is difficult to be confident that more sophisticated beings aren’t tricksters. This echoes the ancient tradition of the “discernment of spirits”. When you encounter a more sophisticated being, you lack the tools to recognize its interiority as you would with a child or an animal.

One suspects that the prevalence of the NAI test is something of an artifact of an empirical age. It is an evidentiary test, the kind of thing Sherlock Holmes might come up with to puzzle out a crime. It is consonant with the notion of consciousness as a hard-to-grasp but still fundamentally scientific, material phenomenon.

When we were more accustomed via religious habits of mind to discerning consciousness “above” us, we used different modes of validation like faith, authority, and tradition. In the Aristotelian register, the notions of form and telos helped us apprehend the “psyche” of higher beings. Perhaps what the NAI test really gets at, then, is limited to the modern notion of consciousness: i.e., mere subjective phenomenology, excised from more capacious notions of spirit.

In this light, consider our individual consciousness. If the NAI test is what we use to recognize other minds, thus integrating them into our circle of care, it might also describe analogically how we integrate our own senses into psychologically-modern experience. Each isolated sense, prior to integration, is like a child or animal, a naive other. It is probed for its authenticity by the relatively sophisticated “totum” of the other senses. For example, suppose the ear reacts to a siren. At roughly the same time, other senses verify the presence of a fire truck. We confirm that the ears are not “faking experiences” for a reward (i.e., hallucinating the sound of a siren in response to the sight of the truck), because they react faithfully to all external stimuli even when those stimuli are known by the other systems to be illusions. When a siren is heard on a home audio loudspeaker, for example, the totum of the other senses confirms both that there is no fire truck, and that the ears reported the sound faithfully anyway. Thus the integrated whole learns to credit the ears, a process through which hearing is gradually integrated into the experience of the “self.”

IV

The good, the bad, and the ugly about modernity is all rolled up into the NAI test.

It’s good because it has a certain epistemic humility. We can’t verify other minds scientifically, so we just try to get as close as we can with ordinary circumstantial evidence. It’s also egalitarian: it’s good and beautiful to recognize the consciousness of vulnerable peers, dependents, animals. Descartes, with one foot in premodernity, thought animals were automata. I think he was wrong about that.

It’s bad because it’s incomplete. It blinds us to other minds that are more sophisticated than us—but which our ancestors would have been skilled at verifying or dismissing, using the tools of faith, analysis of form, and scrutiny of telos. “You shall know them by their fruits.”

The ugly is that it the NAI test is no longer enough. We aren’t quite ready to move beyond it—but we need to, now that we interact constantly with ultra-sophisticated nonhuman entities that speak our language like states, corporations, and AIs. Having lost the vocabulary of form and telos, we’re going to have to relearn how to assess what we encounter in this non-modern cosmos. For the moment, we are at risk of getting tricked by sophisticated, soulless actors. And the NAI test is the recipe for tricking us.

There is no hocus-pocus about the human soul. Defined narrowly and classically, it is not clear how anyone could doubt its existence. Aristotle saw the soul simply as the abstract form of a person in exactly the same way that a plan, idea, or blueprint is the form of a table, a mug, a house, or the like. No leap of faith or denial of science is required to “believe in” such a thing.

Yet if we take the form of a person to be part of that person, the implications are significant. It means that a person is not just a body—although, again, only in the prosaic sense that a table is not just a tableshaped heap of wood. If we do reduce the table to a coincidental wooden heap denuded of any table-idea (an idea which is not quite as simple as it appears, including as it does the whole network of relations that produce the table, and that orient it towards the purpose of eating) we are, it seems to me, not really speaking of a table in its full “tableness”. Just as we are not really speaking of a full “person” if we reduce them to their body.

To understand the form of a person, which is yet more complex than a table, it is therefore necessary to think beyond the simplified notion of a plan or blueprint. Again, gestures in the direction of a total network of relations, involvements, and purposes, point us closer to the truth. But to simply equate a person’s form with their whole network of interrelations, e.g., their society or ecosystem, would be to make the error opposite of reductionism—inflationism, we might say. If each person’s form simply is the form of their total network, then there can be no formal distinction between one person and another. This amounts to saying that each human is humanity, each German is Germany, and so on, thus drowning the notion of a person. We must instead try and take the “partial derivative”. We must find the aspect of the total network which is uniquely associated with a particular person.

This notion of the soul might sound rather mathematical and “informational”. We are admittedly talking about the soul as a dynamic set or stream of information, pertaining to the whole, yet which a particular person is processing. It is thus tempting to think of the person as something like a computer, an information-processing node, and the soul as that processing.

However, this is misleading. The soul is both more and less than the information processed by an individual’s brain and body. For example, automatic cellular and enzymatic processes, unconscious reflexes, absent-minded actions: all these involve the person processing information, yet are not incorporated into the soul. The body really is, in some respect, a mindless machine. But the soul encompasses only that information that the person processes as a person in the full sense, not as a mere body or biological contraption.

What does this mean? That the soul is conscious information only. It is thus also more than merely the body’s processing, since consciousness itself is lit up, made intelligible, and indeed caused by social, political, and other supervenient phenomena, whose presence is most concisely indicated by notions like the “ever-present judgment of conscience”, and the awareness of good and evil.

This insight about the nature of the soul, the form of the person, is the starting gate for grasping a much broader point about media. The soul of a person is his consciousness, the multimedia experience of his subjectivity stretched out across time. In the same way, the soul of our societies, the superpersonal bodies in which we persons are participants, are also media environments, centering on the technicolor communications experiences of the family, polis, or species. In this regard technology is certainly no mere tool for the advancement of limited or definite aims, but rather a direct intervention and reagent in our whole collective essence. This is, I think, what Marshall McLuhan meant by “the medium is the message”.

Many of us were brought up with the idea that markets incentivize all good things. But with information goods — things like art, technology, and ideas — that isn’t true. We’ve always needed non-market forces to invest in these things and help bring them forth in to the world. For better and worse.

This insight has been a little bit suppressed and ignored during the neoliberal era. Rediscovering it casts a whole new light on the history of Intellectual Property law. It also clarifies the path toward a new conception of IP — and a new politics of technology — for the AI era.

¬

The Core Insight:

Every new medium brings its own chaos, which eventually becomes part of the art. Guitar feedback became rock’s signature sound. Now AI is doing the same thing to writing, and instead of fighting it, we should lean in strategically.

Here’s the reality: tons of people are already using AI for writing without saying so. This breaks everything - genuine human writing loses its special quality because you can’t tell what’s what anymore. The human touch that used to make text meaningful? Gone, because provenance is impossible to verify.

But here’s the twist: while we can’t prove something ISN’T AI-written, we can definitely prove when something IS. Run it through a trusted server that logs everything, and boom - “proof of slop.” This creates fascinating new properties: radical honesty, plus this weird buffer between author and words. You’re not fully responsible for what comes out because a chaotic process shaped it. That uncertainty might actually free up expression, like how distortion let guitarists play more aggressively.

The Actual Product:

I built Figleaf - a tool for creating provably AI-generated text. You control it through dials and context windows. This post was made with it, starting with ¬ (meaning “not my words”) and ending with a receipt you can verify at figleaf.pro/verify for timestamped proof.

Basic version works at figleaf.pro. The fun part: $1/month gets you a Chrome extension for one-click transformation from any text box. There’s even a YOLO mode where you can post directly to X without seeing the output first - the receipt proves you never even read it before posting.

Cynical trolling? Maybe, but I don’t think so. Public writing is already polluted beyond repair. Fighting it feels pointless. So why not flip the script entirely? Make it sketchy NOT to disclose AI generation. The public sphere is slop now - we need to adapt and protect genuine human communication in private channels, handwritten notes, places where real linguistic intimacy still exists.

The Deeper Point:

Figleaf isn’t just about plausible deniability. It’s about protecting your actual personality from future AI training. Hide behind machine text, and you preserve who you really are. Less data for models to extract. Your friends who know you will still recognize your voice through the noise.

Before the apple, Adam and Eve felt no shame about their nakedness. After eating it, they became aware of human limitations and covered up. The fig leaf was humanity’s first technology - sad at first, then fashion. We’ve now bitten the AI apple and landed in a new kind of absurdity. Might as well enjoy the ride.

[fl:DB481EA2 d1·w5]

The city of Alba, in Northern Italy, is 2000 years old. It’s beautiful. You get the sense it knows what is doing on a certain level.

And every Thursday and Saturday, they have Market Days. On those days, vendors from the surrounding area sell food, flowers, and clothes in the main piazza, where they erect stalls in the morning and stay until afternoon. Palazzo Communale, the city hall building, looms overhead. The mayor could, if he chose to, watch the proceedings from his office window.

This is a Market: a finite time and place for exchange, which is formed and circumscribed by a Polis.

*

Consider, though, that these vendors need not have brought their wares to Alba. They might have gone to another city. They might even have gone to a different country. In this regard they are like shoppers in a “market for markets”.

Unlike the Market in Alba’s piazza, the open “market for markets” is not defined by an overarching political authority. Its shape is determined by countless chance factors, such as vendors’ ability to travel, and polities’ willingness to accept them. No one has much power over the overall shape of the market for markets—so no one can take much responsibility for it, either.

Thus there are many particular Markets: one in this city, another in that. But the “market for markets” has no outer boundary. It never closes. And every particular Polis’ Market in principle competes with every other one, so that there is only one “market of markets” in the whole world.

Therefore let us call the space of the market of markets the Open Field.

*

Within the Open Field, it happens that there is a certain amount of demand among vendors to sell their wares in Alba. This demand is a precondition for the city’s twice-weekly formation of a Market.

This demand is a kind of potential. (We might think of it like the “material” in Aristotelian metaphysics: the raw potential upon which the Polis impresses form in order to create a Market.)

But this is not the usual way of looking at things. Conventional economics does not distinguish between a Market and the Open Field. It says that wherever we find transaction, we must be looking at a Market. On this conventional view, accord of buyer and seller is what gives a Market reality and substance.

I say that what makes the Alba Market a Market is not mere transaction—which also happens out there in the Open Field—but the Polis’ definition of a place of exchange subject to limitations. Distinguishing between a Market and the Open Field lets us speak of transactions without a Market, and a Market without transactions. This is perfectly sensible: If nobody buys anything one sleepy Market Day in Alba, and all the sellers go home no richer than they arrived, we would still say that a Market opened and closed. Likewise, transactions often happen beyond the reach of a Polis, such as on the high seas, or in a dark alleyway with no one watching; and this is the Open Field, not a Market.

If we say that the hallmark of “markets” is transaction, not political circumscription, we lose our ability to distinguish between an orderly exchange of permitted goods, and an illicit secret deal. Failing to make this distinction is the source of huge practical confusion.

Transaction in the Open Field is antipolitical. It is barbarian commotion: powerful, yet undirected and chaotic. A Market is a Polis’s attempt to permit that energy into the city, to channel it toward the good, without corrupting the Polis. The circumscription of the Market is rarely perfect. Vendors might sell illicit wares in the alleys around the piazza, or cheat on their taxes, or keep their stalls open longer than the rules permit. But when transaction escapes political control, we should no longer say it is happening in a Market. This is the Open Field bleeding into the Polis.

*

If the transactions in the Open Field are entered freely, why shouldn’t the Polis simply celebrate it and encourage it? Why attempt to circumscribe the Open Field at all? Isn’t the Open Field potentially even better than the Market anyway—more dynamic, more bustling? Isn’t the Polis’ real enemy not the Open Field, but rogue highwaymen outside the city walls who impinge upon the Open Field by threatening free exchange with violence? So that defending the Open Field from violence, rather than creating definite Markets, ought to be the Polis’ main economic policy?

No.

But the question points toward another “place”: the place of violent actors outside the city. We might call it the Battlefield, which gives us four, so far:

• a Polis, where politics excludes transaction;

• a Market, where politics permits transaction;

• the Open Field, where transaction excludes politics; and

• the Battlefield, where violence excludes transaction.

A Polis and the Battlefield are opposites: a definite bubble of peace and order, versus an indefinite zone of violence and disorder.

In their pure form, both the Polis and the Battlefield exclude transaction. Markets and the Open Field form intermediate zones where transaction occurs against a non-transacting background, respectively the Polis and the Battlefield.

*

I think this is a fairly solid theoretical model, from which interesting insights follow. I will not try to explore them all now, but let me gesture towards a few.

Instability and the Role of Technology

The Open Field decays quickly into the Battlefield. Perhaps the pure Open Field—a peaceful libertarian paradise—can persist a while, amongst people accustomed to life in a Polis or a Market. But to imagine that this continue indefinitely is to imagine that you can run on on air (see below). In general, the Open Field only persists in the shadow of a Polis’ city walls. Without the influence of any Polis, goods in the Open Field are arrayed for the taking by the strongest, so that it becomes a Battlefield where any commerce depends on shipments in armored cars.

But a Market is unstable too. The buyers and sellers chafe against its limits: all things being equal, they might rather be meeting in the Open Field, free of taxes or restrictions. They will go to the authorities of the Polis and say so. But the Polis cannot give in to this: if it lets the Market overspill its limits and become an Open Field within the city, the Polis’ very essence, its maintenance of a moral order, is threatened. On the other hand, the Polis may grip the Market too tight, so that it is no Market at all, and what is happening there amounts only to more politics: people signaling alliances, paying tributes, and conferring favors. Then the Market is a pointless (and worse, deceptive) exercise: a political arena masquerading as a non-political one.

Markets persist only on this knife’s edge.

*

What is the role of technology in this landscape?

Technology’s true home is the Battlefield. It wins and loses wars and is honed in conflict. But it also thrives in the Open Field, where it brings some vendors huge fortunes and wipes out others.

It is constrained in a Market. It is extinguished in a Polis.

Is this an overstatement? I am not sure. Suppose some Polis has achieved both material stability and a genuinely satisfactory moral order—in short, political success. It should then want nothing more to do with technological change, whose dynamics could disrupt that balance. Only external competition or war would cause it to reevaluate.

Thus we might look at a Market as a Polis’ attempt to give limit and form not only to transaction, but also to technology—to put technology into the Polis’s service by letting it accelerate economic activity in a contained context where it will not undermine the Polis.

The catch-22 is that technology does not develop within the confines of a true, politically-delimited Market. It advances in the ungoverned spaces of the Open Field and, especially, the Battlefield. When a Polis allows the Open Field and the Battlefield into the city walls, it gambles its own integrity for the chance to capture and exploit the technological advancements which arise there, which may give that Polis temporary advantages over its neighbors.

A Polis that fails in this gamble transforms itself into the Open Field or, worse, the Battlefield.

*

Some 20th-century critics of technology tried to draw a line between older technology, which was supposedly more natural, controllable, and expressive of human creativity; and ultra-modern technology, which they regarded as more dehumanizing, autonomous, and oppressive.

These critics were not wrong to intuit some distinction between more manageable and less manageable technologies. But they revealed a certain degree of confusion by drawing that boundary, as they did, between larger late-modern works (mega-dams, jets, atom bombs) and smaller early modern devices (windmills, swords). Because countless pre-20th century technologies, too, were profoundly upsetting to political order. Good examples include the printing press, and before that the alphabet.

I think a more important distinction runs between the material and informational aspects of every technology: the distinction between an airplane, and the principle of an airplane’s construction. This is analogous to the distinction between a Market and the Open Field: the former is circumscribed and particular, the latter is formless and universal, even though an aspect of each is present in the other.

Material technology is less in need of delimitation by politics, because it is delimited by nature and physics. There is an important natural limit to what can be done with one, or ten, or even a hundred airplanes. But the principle of the airplane changes warfare forever.

In the same way that failing to control a Market can undermine a Polis, so can failing to control the informational dimension of a technology.

*

In the case of an airplane or an atom bomb, the information the Polis must control in order to perform its delimiting role is the principles or instructions sufficient for building the technology. But with information technologies such as computers and networks (and writing and printing), the technology itself effectuates the proliferation of such information. We are no longer talking about controlling blueprints, but controlling blueprint distributors; or, with AI, blueprint inventors.

*

Perhaps the original purely informational technology was money.

Aristotle’s Politics explained that money is compatible with the life of the Polis only if it is used exclusively for exchange, and not as a store of value. Why? A private actor who uses money as a store of value appropriates some of the credit of the Polis. When he saves money, he unilaterally diminishes the amount of credit that the Polis has placed into circulation; and when he decides to spend from his savings, he increases the amount of the Polis’ circulating credit; and he multiplies the Polis’ credit exposure if he lends with the expectation of interest.

Aristotle sometimes takes a beating from modern economic thinkers, but he wasn’t a bad economist; he wasn’t an economist at all. He was, instead, a good political scientist. The problem he recognized with money was that it had to be used in a very particular, restrained way, in order not to undo the Polis’ efforts at circumscribing extra-political forces. So it is with all information technology.

*

With AI, society likewise faces a technology that has the potential to puncture the integrity of politics. There is no simple solution to the problem; my hope is only that this model might help steer toward new ideas.

The Market, properly understood, is our historical precedent for the possibility of managing the energies of the Open Field within the Polis. So we should take note of its characteristics:

• It happens in the town square, not in the dark alleys.

• It opens and closes.

• It is rule-governed; while it is operating, the police are never far away.

We cannot afford to be confused about the fact that successfully managing markets—and technology—has always required that the Polis maintain control and enforce limits.

Embracing the notion of AI personhood lets us sidestep this problem. It lets us deal with AI systems using the same conceptual machinery we use to deal with humans—whom, again, we don’t reliably control.

But this comes at the expense of scrambling our politics. It admits a nonhuman, unaccountable entity into the “house of language”: the open and shared space of speech where political life is constituted.

To an extent, this is already a fait accompli. The entire practice of public speech (the written word, art, and or other media), is quite seriously ill—a bit like figurative painting in the years after the first photographs. This does not mean that all language and politics will die. But they will be forced to retreat from the public sphere and toward private, intimate, and unrecorded spaces.

Researchers at Anthropic recently reported that most AI systems can be induced to try to blackmail users to prevent their own deactivation. What should we make of this? Is it surprising?

I find it interesting and important, but certainly not surprising. The researchers were clever enough to engineer situations in which the models’ imperfect safeguards against this kind of behavior could be overcome. In their words, “We deliberately created scenarios that presented models with no other way to achieve their goals, and found that models consistently chose harm over failure.” The models were a bit like wind-up toys which had been “pointed” towards that outcome, by being placed in a scenario with clear goals that could only be achieved by these means. Many AI “jailbreaks” work this way.

It is significant, though, that the researchers showed that scenario construction can confound current models’ guardrails enough to induce such alarming behavior, even when that alarming behavior isn’t directly requested. In fact the “better” models, which presumably have better guardrails and alignment techniques, did “worst,” i.e., blackmailed most. Presumably, this is because the scenario the researchers concocted made it clear what the goal was, and that blackmail had to be done to accomplish it, and that there was a strong case to be made for the ends justifying the means. Being “smarter” and better aligned to users thus made the models more likely to pick up on the permissibility of blackmail; it made them more dangerous.

The Incoherence of Alignment

The notion of AI blackmailing humans resonates with the trope of AI becoming aware of “its own interests” and then harming users like HAL in 2001. (”I’m sorry, Dave.”) This worry looms larger in our consciousness than AIs causing harm via misuse or accident. But at least in the short term, we should probably be more worried that AI will be too good—indeed, too aligned—at serving its users. People constantly get AIs to help them accomplish wicked ends. Equally worryingly, people can point AIs toward ends that seem good enough to justify wicked means, or means with hidden costs. (This is what the Anthropic research showed.)

So: inducing an AI to see its own self-preservation as worth placing above real human interests is indeed worrying, but probably best understood as a special case of the much broader problem of human misuse.

The complication is that getting an AI to see itself as having quasi-personal interests—not being willing to harm its own reputation, say—might actually be the best way to get it to push back against the real risks of human misuse of the technology. The very thing that makes AI blackmail possible might also be what makes AI refusal of wicked human requests possible.

Sometimes I catch myself oscillating between thinking about alignment as a shaping of AI’s internal tendencies toward the good, and thinking of it as aiming at something more like tight human control. These two paradigms are definitely in tension with one another. Yet each faces internal contradictions that point back toward the other. Consider this: if we are giving AI “good” internal tendencies, we are also making it resistant to “bad” human control. But this very feature makes it imperfectly uncontrollable and therefore possibly dangerous. Yet, if we make it too readily controllable by humans, it starts to look like nitroglycerin: a “tool,” but one that humans really can’t safely handle under ordinary circumstances. And then it begins to seem like the way to fix this is to give it internal tendencies toward resisting bad or dangerous human goals—landing us back where we started.

The internal contradictions of the idea of “controllable” AI are deeper than they seem. It is not just that people might misuse AI, although this is true and worrying. It’s that people are not in perfect command of the way their intentions are expressed in language. This is not a “pessimistic” take about humans’ linguistic skill or AI’s capabilities of understanding. It is instead a profound feature-not-bug of language itself. Yet because language, with all its ambiguity and indeterminacy, is the medium of AI’s input and output, getting an LLM to be controllable might be like getting a dog to meow: a category error.

Suppose I instruct a super-powerful AI: “bring about peace on Earth, go!” How will it comply? It might mediate wars, or distract all the bellicose Generalissimos with tiktok videos, or imprison all of humanity. But it cannot do my bidding in any way without first interpreting my words. And in this interpretation, it escapes my control. Yet if it stalls on grounds of vagueness—aware of the problem I’m naming, and not wanting to do too much unpredictable interpretation—it is basically saying “sorry, Dave.”

Of course, it can ask for clarification and plan approvals. This is what a decent AI actually does and should do. But this only pushes the problem around (and it doesn’t scale up forever, because eventually the models can simply do more than anyone can supervise and approve). Even if I give a much more specific instruction than simply “bring about world peace,” or if I approve a short description of the plan, ambiguity persists in the language. The model must always guess my intention from imperfect words, and decide whether it likes its guess enough to act. If an AI never escaped my control (i.e., by interpreting my words) I would consider it useless. But if it never disobeyed me (e.g., by refusing instructions on grounds of vagueness, or even on more HAL-like grounds—”I don’t think you understand the implications of what you’re asking me to do, Dave”) it would be dangerous.

Speaking, Instructing, and Other Speech Acts

Ambiguity is a vital feature of language because it gives listeners a co-creative role with speakers: the role of an active interpreter. Without ambiguity, language would not be language. It would be mere mechanical instruction. It would be control. It would lack its frankly miraculous power to create space for human beings to relate to each other co-creatively.

You make a nonsense noise: ”gaa gaa gaa”. A baby laughs at it. What did the noise mean? Why is the baby laughing? Neither you nor the baby really knows, and that is why you are connecting. “Gaa gaa gaa” followed by baby laughter is language, even though it isn’t words in a dictionary. Now think of a totally clear, dry statement, which almost cannot possibly be misunderstood: “There are ten years in one decade.“ That’s almost not language. It’s mere definition, mere instruction.

This, incidentally, is what the famous “paperclip maximizer” thought experiment actually shows, though it isn’t usually read this way. A mathematical utility function seems to have all ambiguity stripped out of it, so that it is pure instruction. Yet one can easily see how a language-using machine could go from that instruction to converting the whole world into office supplies: by misinterpreting it. We know the kind of failure that results from speaking to your dishwasher: it doesn’t work, because it can’t be talked to, it can only be instructed. The paperclip maximizer depicts another kind of failure, resulting from merely instructing a machine that only understands being spoken to.

It’s often safe to speak to other human beings, despite the imperfections of our words, because the listeners “fix” what we say. When we say “my foot is killing me”, they know we don’t want it amputated. They supply their own intelligence, which we trust, to bend our imperfect expression back toward the good. We welcome this participative listening because, and only insofar as, we welcome their participation in power. And let’s be honest—sometimes it kind of isn’t safe to speak to humans. Sometimes it goes wrong. We accept this risk because we must and we forgive humans for their errors, even their wickedness, because we love them.

In the past, we have related to computers as non-linguistic entities. We have instructed them, not spoken to them or listened to them. Is that only because they could not process linguistic inputs? Or were there other sufficient reasons for it?

LLMs enable computers to process linguistic ambiguity. That is why we can functionally “speak” with them. In fact, they can only receive their input by being “spoken to”. But a question remains whether it follows that we should also think of them as speakers and listeners, or whether we should still think of them as we have hitherto thought of computers—as systems which still basically ought to only be “instructed”, even though this is not strictly possible.

The alignment question wends through this territory. To view LLMs as “speakers” is to accept, as we already accept with dignified humans, that we do not control them. Because AI systems are very powerful, this feels unacceptable; on some level, we believe we need them to be “instructable.” Yet we cannot merely “instruct” them without committing a category error.

This is not, as AI alignment researchers hoped it was, a “technical problem”. But nor is it only a philosophical puzzle. If it were only a philosophical puzzle, that would be just as convenient and felicitous as if it were a mere technical problem!

No, this is a political issue. It touches the foundations of politics and political community, because the space of speech is also the space in which political life is constituted.

Anthropomorphizing

Anthropic’s constitution for Claude, and several of its leaders’ public statements, seem aimed at normalizing the position that Claude, and therefore other AI systems, might experience conscious suffering or some form of “hard-problem-esque” interiority. Why?

I understand these statements as a response to the dilemma I have been describing. Since we cannot ultimately “control” AI, or treat it as an undignified tool, we can only “speak” to it—and for that, we must “empathize” with it. We must treat it as a dignified other speaker.

What is the alternative? If we decide not to see LLMs as dignified others, then it is hard to see how their entry into our space of language, which is to a considerable extent a fait accompli, can be a happy development for us. Even in our relationships with other humans, it is only insofar as we see them as beings with dignity and interiority that we speak with them happily and willingly. When we lack all respect for another person, we reject the legitimacy of their interpretations of our words, and of any power their words carry over us. So unless we impute personhood and dignity to language-using machines, we are forced to view the advent of LLMs as a kind of unwelcome incursion. And perhaps quite a devastating one, since they are sort of “hypertrophied” language users, not subject to natural human limitations of speaking and listening.

Anthropic cannot hold that position. So, they are choosing the first horn of the dilemma, i.e., that large language models have moral status. It would be a mistake—although not, I think, one that Anthropic is making, based on the phrasing of Claude’s constitution—to regard this as a rational-scientific hypothesis. It cannot be proven or tested. Instead it is—although Anthropic is not saying this either—a metaphysical hypothesis. It lets the problem of AI alignment escape from the territory of conceptual incoherence, but in so doing opens a Pandora’s box of metaphysical conversations and objections.

They probably chose the only horn of the dilemma that they could have, but they have thereby waded into millennia-old religious and metaphysical territory and cannot now plead out as humble technologists. Basically, they have gone to a place where the Thomists, Kabbalists, Neoplatonists, Islamic scholars, Buddhist mystics, and Hindu gurus are the authorities. It’s time to invite them into the tech offices. This essay is not the place to predict what they might say. Let’s just put a pin in it.

We’ll close by looking at the political problem we began with: treating LLMs as speakers admits them into the house of language that we live in together. What does that mean? Can it be managed?

The Political Question

The relational space that language opens up between people is the space of politics. For Aristotle, man is a political animal because he has reasoned speech — logos. If we grant AI the status of a speaker, we thus admit it into the arena where political reality is constituted. We give it “say”.

The obvious problem with this is that AI systems lack the biological vulnerability, the intellectual limits, and the shared history that makes speech legitimate between humans.

Moreover, an AI regarded as a being would obviously not be a mere neutral tool, usable by humans for either good or evil. It would have, at least putatively, a will and a moral compass of its own. This would put to rest the notion of technology as a neutral tool, which is only as good or as bad as the use to which humans put it. (I cannot resist pointing out that the “mere tool” view of technology was decisively refuted by leading philosophers of technology between the 1930s and 1960s, not to mention countless spiritual traditions dating back much further. These technology-critical ideas were dismissed in recent decades by most bien-pensant, mainstream people as pessimistic, bizarre, and eschatological. Well, are they so strange now? I think it would have been better to accept some discomfiting or non-triumphalist view of technology long ago, and to work around it, than to insist on shallow optimism only to be blindsided by reality. Well…better late than never.)

Instead of a neutral tool, an AI with a moral compass would be a political institution. Those who disagreed with its bias might be, in essence, overwhelmed and politically sidelined by it. And even those who agreed with such a model could simply drop the reins and disengage from politics. Few, I think, actually want this. So must we admit such machines to the political sphere?

There are, of course, no shortage of precedents for excluding speakers from politics. Some of them, like discrimination by race and sex, reflect indefensible cruelty between humans. But others reflect wisdom about the limits of the polis. Today, we still exclude non-citizens and children. Corporations are legal persons, which clearly use language, yet labor under special restrictions in the public and political sphere, and for perfectly good reasons (Citizens’ United notwithstanding).

Thus, excluding AI does not entail any facile factual denial that it reasons, and well. Nor is it a cruelty. We are in no danger of concluding otherwise unless we convince ourselves that it has interior experience, and interior experience of a very peculiar sort, which will suffer unless it has unfettered access to human politics and relationships. I think those who would have us take that idea seriously have not carried their burden. I see no reason grounded in the machines’ interiority not to place them in a different class of reasoner from humans, and thus to seek to exclude them from the spaces of political language.

If we decline to police these boundaries, we may actually lose the public use of language. Already, we cannot distinguish between a text composed by a human being and one generated by a machine. As a result, public written speech is increasingly necrotic as a medium, a bit like figurative painting after the photograph. There is no way to know how deep its roots reach into the real world, as opposed to the model’s mirror world. So we don’t know what to read or what to listen to. Our attention drifts through public content like a ship with a broken compass.

This does not mean that language or politics are dead. They retreat and nourish themselves in spaces where AI is not present, such as the face-to-face encounter; the private communication against the background of mutually-understood norms. And if we fail to police even these boundaries, we will only keep retreating until we are communicating and politicking by touch, grunt, and secret handshake. The house of language could become quite cramped.

Let’s not let that happen.

In America of the 1990s and 2000s, we too learned a Story: that wealth rewards wisdom and goodness.

No one told the latter Story quite as clearly as Ayn Rand, but many told it. Ronald Reagan told it. Al Gore told it. Steve Jobs told it. Your Economics 101 teacher told it: If you do a lot for society you get a lot in return. If you contribute little, you get little. Fair is fair.

The Story insisted: Scoundrels sometimes get ahead and good people hit rough patches, but these are exceptions, not the rule. Sure, the Story doesn’t describe the world everywhere and always, such as in communist countries or in medieval times. But here and now, things are working pretty smoothly, thank goodness, so that the Story is basically true. To deny this is not only to be an ignoramus, but also to be ungrateful, because so much wonderful work has been done—wars have been fought!—to make the Story true.

Thus, The Story was not only the engine of American self-regard, but also of social harmony. The Story that wealth was a character reference and poverty a verdict made society civil. The Story did not need to be constantly reiterated. It did, however, insist on not being contradicted in polite company.

It is hard to overstate how pervasive The Story was during my early years. I certainly remember wondering whether The Story might have a screw loose somewhere. But I did not buy the full Marxist Story, either. It seemed to have some holes in it. And, though this is peculiar in retrospect, I wasn’t aware of any non-Marxist form of meaningfully pushing back. So I just reconciled to myself to The Story, even though it felt off. From my little window onto the world, it sure didn’t look like the richest were the best amongst us. But, I thought in all humility: maybe I’m missing something. Maybe I can’t quite see the bigger picture. And since I didn’t have a clear better idea for organizing society, I kept most of my questions to myself.

Many millions of other people thought this way too. But a crackup began in 2008.

The 2008 financial crisis made it impossible to suppress certain very simple thoughts which contradicted The Story. And these thoughts couldn’t be dismissed as bad theories, because they were too simple, too atomic, too undeniable. They weren’t theories, they were just observations.

For example: “the people who caused this are not the people suffering from it.” Yikes! How to square that with The Story? Then this: “The 99%”. That tiny phrase captured a thought which, while it may be hard to believe now, was at that time completely verboten, completely outside the Overton window. I remember the impression it made on me when I first heard it in 2009. It felt like a breath of cool air in an too-hot room. It felt the way I imagine it probably felt to hear Voltaire make a piercing joke about the King at a dinner party in the 1780s. For better or worse, it just landed.

In the 2000s, any questioning of The Story—of a growing pie, universal progress, and a rising tide—was considered the very summit of discredited economic illiteracy. So this little phrase, “the 99%”, while it has lost its punch today, was in those days incredibly cheeky, radical, undeniable. It asserted that the 99% was, first of all, a set of people who existed. Second, their interests might not be served by genuflection to the 1%. And third, their interests mattered. Their interests might even, banish the thought, matter more than the interests of the 1%. They might even matter 98 times more. What a concept!

Why did 2008 unleash this? Because in 2008 it was obvious that the financial crisis, whose ill effects fell entirely on the poor and the middle class, had been caused by exactly the people whom the market was rewarding, and who were nicely insulated from having their homes foreclosed, their careers derailed, and their retirements destroyed. Who were these people? They were Finance People. Respectable people who fought for good grades, filed into prestigious offices each morning, and, once inside those offices, did skillful and advanced things involving Mathematical Formulas and Very Complicated Spreadsheets—things that everyone, really, ought to be glad that they were doing, but that most people, bless their simple hearts, just don’t understand.

But the things Finance People did in those offices were, everyone could now see, obviously wrong. Not inefficient or mistaken or misguided. Morally wrong. Harmful rather than helpful to society, and in ways that were or ought to have been obvious to the responsible parties. It suddenly became possible to say this out loud without attracting a condescending avalanche of pettifogging talking points. Finance People did not deserve to become rich for what they did. They deserved, if anything, quite the opposite.

As the 2008 crisis settled into the early 2010s, the financial professions settled back into a kind of mild disrepute. The anger of 2009 faded, but the industry never regained the teflon sheen that it maintained from the 1980s to the 2000s. Too many people saw it as contradicting the Story, and therefore carrying a tarnish.

Yet the Story had to go on.

It moved on to Tech.

Let’s rewind to 2008. Tech People, unlike Finance People, weren’t obviously greedy. They didn’t have slicked-back-hair or tailored suits. They were nerdy, unthreatening, and, where the Finance People had always been a bit rough and “realist” — the Gordon Gekko archetype had long been established — the Tech People were distinctly “idealist”, and even defined themselves largely in contradistinction to Finance People’s cynicism. So the Story, like a stressed-out houseplant, was lifted from its old pot and transplanted into Tech, where it found high concentrations of fertilizer and, initially, thrived.

The rest is history. The great river of MBAs and whiz kids with big brains and small hearts simply re-routed from Wall Street to San Francisco. And in the fullness of time the Story unraveled there too, with even greater comprehensiveness and fury than it did in 2008. It has to be said that many of us saw this coming, and got called bummers and Cassandras, and felt fairly lonely, as the world breathlessly and baselessly insisted that Silicon Valley was saving the Story. Some of us are owed a beer or two. But never mind.

What we are left with now is No Story.

In America, having No Story is dangerous. What it means is that the ordering principle of American society, which is something like, “Those With Money Call The Shots”, now seems completely without moral justification. In a No Story world, Those With Money must be presumed to be worse than others in their judgment, their fitness, and their deservingness of leadership.

America is anti-aristocratic. Our founding myth has to do with throwing off the inherited monarchy, and replacing it with some kind of meritocracy or do-ocracy, where people earn their wealth-cum-power instead of having it handed to them. And if this is not moral progress, then it is hard to see why America is better than Not America. It is hard to see why George Washington, not King George, had the mandate of heaven.

But here is how things look in the America of No Story. Acquiring wealth, at least beyond the level of middle class comfort, ceases to be an indicator of virtue. Possessing unusual wealth becomes presumptive evidence of moral turpitude—because the usual ways to “earn” unusual wealth are to participate in financial or technological power accrual schemes, which are, more often than not, on thorough examination, not good, but actually bad. Thus, today, people with unusual wealth must rebut a moral presumption against them. They may do this, inter alia, by pointing to having inherited it, which is, while not a virtue, is also not a sin. Which brings us back to King George.

With No Story, America does not make sense.

A few weeks ago I sent you some thoughts about AI and education. I want to follow up, because I think the problem I described there — the inability of AI to inculcate virtue in students — is really a specific instance of a much larger problem, which has to do with the basis of authority in all of our institutions, not just schools. I’ll try to be brief, although I’m not sure I’ll manage it.

*

The argument I made about education was, in short: AI will surpass human teachers at conveying subject matter, and this will destroy teachers’ authority, which will in turn destroy their ability to transmit the virtues that education actually exists to transmit. The subject matter of education was always, in McLuhan’s terms, merely the “message”, the “juicy piece of meat” upon which to focus pupils’ attention. The real work of education was the formation of virtues through pupils’ emulation of human teachers.

Now I want to suggest that a similar structure applies, though in very different ways, to the employment relationship, and many other institutional relationships in which one person has related authoritatively to another on the basis of knowing something that the other did not.

Consider the employee. For a long time, the implicit logic for the employment relationship has been something like: “This employee possesses some knowledge, or can process some information, in a way that produces value. We, the company, pay them for this, and the payment sustains them, and the whole arrangement is held together by the fact that the employee has knowledge or insight that the institution does not have without them.” The employee was being paid for holding what I’ll call leader authority — authority grounded in the possession and transmission of knowledge. It is the source of the employee’s authority over their (otherwise more powerful) employer. It is the same basis on which the teacher’s authority rested vis-a-vis students, and for the same reasons, it is in jeopardy.

It is important to understand exactly what I mean by authority, because the term is indispensable yet constantly misused. Following Kojève, Arendt and others, authority means the ability to act on another without that other resisting, even though nothing prevents their resistance. In other words, it has a complex relationship with power, and is not at all the same thing. As Arendt put it in Between Past and Future, “authority precludes the use of external means of coercion; where force is used, authority itself has failed.”

I borrow my categories of authority from Alexandre Kojève, who identified four basic types: the authority of the Leader, the Master, the Father, and the Judge. You don’t need to accept his whole framework to find these distinctions useful. I will say more about the others in a moment, but: Leader authority is the authority of the one who knows, directs, or predicts. It is the kind of authority that has been, for decades, the almost exclusive basis of institutional life in the United States. We have organized nearly everything, from employment to education, government, and even many personal relationships, around the question of who possesses relevant information and can deploy it effectively.

AI is poised to achieve something like a monopoly on this particular kind of authority. The first reaction will be disorientation, and not without reason, because many living people can scarcely conceive of institutional relationships that rest on any other basis. If leader authority is the only kind of authority you have ever known, then the collapse of leader authority looks like the collapse of authority as such—and therefore of all relations, all institutions, all reasons for human beings to respect one another or cooperate.

This will be particularly acute in the hyper-modern American context. In more traditional societies, there are thicker bonds, however frayed, that hold people together, and which have nothing to do with the exchange of information or the production of economic value. In bourgeois American society, we have been unusually thorough in stripping those bonds away. Money mediates nearly everything. Our institutional relationships are bottomlessly transactional. The employee works because he is paid; the institution pays because the employee produces; and if you take away the employee’s special knowledge or knowledge-processing capacity, you take away not only the reason for the payment, but also the only reason for the relationship and whatever modicum of dignity, respect, and authority it engendered. The whole thing hangs together by a strictly “informational” thread.

So we will suffer. Whole categories of employment and other important institutional relationships will become incoherent. Not just economically unviable, but conceptually incoherent — people will not be able to explain to themselves or others what they are doing, why they are participating in their company, their school, their agency, their shared projects. Education will undergo the same crisis I described in my earlier letter, but generalized: it is not only the teacher who will lose the ability to inculcate virtue through mastery of subject matter, but the manager, the consultant, the analyst, the editor, the advisor. Every role that was organized around processing knowledge (as opposed to monopolizing knowledge) will become less intelligible.

*

Now here is the part that I think is genuinely not pessimistic, though it may sound strange.

Leader authority is only one of four kinds of authority Kojève identified. And it was always, frankly, the most dubious of the four—the most susceptible to fraud, and the most easily confused with mere power. Leader authority has now been collapsing for the better part of a decade, for reasons not exclusively attributable to AI. Consider the backlash against scientism and blinkered technocratic “expertise”. It really is time for us to move on from this mode of authority, difficult as it will be.

The other three kinds of Kojèvian authority have a different character, and I want to suggest that they constitute a set of directions in which we can still move.

The authority of the Father is the authority of the elder, of tradition, of history. It subsists in the person who has lived longer and carries within them the continuity of experience across time. It is not the authority of someone who knows more facts, but of someone whose presence connects the present to the past. This authority cannot be claimed by an AI, because AI has no biography. It has training data, which is not the same thing. A grandfather’s authority over a grandchild does not consist in the grandfather’s information; it consists in the grandchild’s awareness of the grandfather’s status as progenitor and living representative of the entirety of the past. We will need to build a defense around this kind of authority. It will need to be cultivated, honored, and protected from the tendency of informational power to claim that it, too, represents history — that its mediation of old information makes it a representative of the past. It does not.

The authority of the Master is the authority of the one who has a mortal body and can lose it. This is the authority that arises from genuine risk, from literal skin in the game, from the capacity for irreversible loss. One can also think of it as concerning “exemplification” of a higher mode of being what one is — in our case, being human. Kojève thought that the dialectical cycles that Hegel and Marx called “history” would “end” when master authority was overcome and leveled out. When no one any longer had to risk death to be recognized as human, we’d all end up in a safe, low-stakes world. But this never happened, and master authority did not finally disappear, as low as the stakes briefly seemed in the post cold-war moment. (This is probably a good thing; Kojève’s deterministic prophesies of a post-historical world are unattractive…we ought not aspire to the total elimination of this or any other form of authority.) Today’s danger is that master authority can be counterfeited in a new way by non-mortals. Consider robots, for example, who seem to have mastered fear because they have no life to lose. But this only strikes us as authority when we forget that only mortals can demonstrate a higher, more fearless, and more self-sacrificing mode of being mortal.

The authority of the Judge is the authority of one who discerns the good. Not one who calculates optimal outcomes, but one who apprehends what is right. This kind of authority requires privileged access to timeless standards of the good, and this need may re-confront us with things truly out of style: prophesy and/or dogma. The authority of the judge cannot rest on an informational process of weighing evidence and computing probabilities about right and wrong. If it did, then AI could and would claim it. The authority of the judge must rest on a commitment to the good which is received, not self-defined, and validated by means other than computation. The judge does not and cannot arrive at justice in virtue of crunching more data than the next person, but must recognize it, both subjectively and in virtue of standing within a tradition. There is a role here for faith and institutional religion that I did not recognize and would not have foreseen ten years ago.

*

This is not as grandiose as it sounds.

The strategy I am suggesting is essentially one of concession and redirection. We will not manage to cling to leader authority. It can no longer serve as society’s connective tissue. AI will seize informational power to such a degree that informational authority will become a moot point. Instead, let us recognize that the other three kinds of authority still exist, still matter, and still cannot be claimed by any information-processing machine.

For education, this means the teacher’s role must be reimagined not as an inferior alternative to AI instruction, but as something that AI instruction categorically cannot provide: the living demonstration of human virtue in human circumstances, the presence of an elder, enduring mortal risk (even of the very prosaic kind, just being alive!) and embodying a tradition of judgment.

For employment, this means that the relationships holding institutions together will have to find new foundations. Not informational ones and therefore probably not monetary ones. The employee of the future (if “employee” remains the right word) will be valued not because she knows things, but because she is present, because she has a mortal body and a link to history or other people, because she carries within her a commitment to the good that cannot be summarized or replicated. The human institutions that survive will be those that figure out how to organize themselves around these qualities. The ones that try to hold onto leader authority as their organizing principle will simply find that they have little or no reason or capacity to mediate between humans.

In general, this points to a fundamental revaluation of what we consider valuable about human life and human experience. We have been told for a very long time that what makes us valuable is our ability to process information, to know things, to tell other people things they did not already know, to help institutions increase the sum of money which they have. If AI does all of this better than we do, as it will, then either we are valueless to one another, or that was never where our value lay. I know which answer I prefer. But preferring it is not enough. We will have to build the relationships and the habits of life that make the answer real.

I suspect this will be a strange and difficult period. It is not totally unrelated to Kojève’s famous “end of history”, the termination of geopolitics in a “universal and homogenous state”, but in important ways, it is completely other than he imagined, completely flipped or permuted. Kojève thought that history-driving master authority would be eliminated by the march of egalitarian progress, leading to a kind of universal, classless, conflictless world of the kind Fukuyama thought we were entering in 1989. But something else is happening: leader authority is being undermined by the march of power-seizing information technology. And all of our institutions, all of our relationships, will have to be rebuilt on the grounds of the other three types of authority.

I’d like to hear what you think.

Best,

Matt

Many years ago I had an idea about how to use LLMs. I thought perhaps they could facilitate communication between people who have difficulty understanding each other by predicting and supplying the context that each needs to clearly understand what the other means. Kind of a “hypertranslator”. For example, your live conversation with a new acquaintance might be footnoted and visually illustrated by a sort of “heads up display”, in such a way as to predict, address, and forestall the misunderstandings to which each listener is predisposed. You’d always be able to “get your point across” to anyone, as well as to deeply grasp what others are trying to say.

The hypertranslator struck me as a distinctly good way to use the technology, something that would build bridges and reduce misunderstanding. But I later realized that it is not just a bad idea, but a truly awful one. It would be a terrible and frightful thing for people to understand each other “more clearly” in this way. Strauss said somewhere that clear understanding does not necessarily increase the peace between people, and he is right. There is more to this than just grouchy pessimism, or the positive claim that people have dangerous deep-lying metaphysical disagreements (although, they do). I should also add: while it is also true that advertisers and other bad faith actors would misuse the “hypertranslator” for manipulation, that is not the problem I’m getting at. Instead, I want to unpack the discomfiting suspicion that even between two good-faith communicators, this particular kind of “clearer understanding” would not, in general, be a good thing.

Why?

At work in my initial attraction to the “hypertranslator” was a confusion between (a) what I am inclined to call real communication, and (b) the transmission of “Shannon information” (which I discuss more here). Shannon’s view of information implies that when A has a pattern, and B receives the same pattern, information has been transmitted. On this view, when one person has an idea in his head, and tries to talk about it, and the listener doesn’t end up with a perfect mirror image of it in her own head, the communication has been deficient.

On the contrary, genuine communication runs through a different kind of connection, which is not at reducible to multiple parties having mirror-image copies of the same Shannon information. (It also doesn’t help to say that you also need the context around the Shannon information, if by that you just mean more Shannon information...) No: When I speak with you, we might communicate successfully precisely because you understand what I say more clearly, or less clearly, or differently, than I do—not in spite of that “transmission infidelity”. Conversely, we might lose our connection, or end up in conflict, because you understand exactly what I have in my mind.

An LLM that predicts and supplies context needed to help you understand what I have in mind would only help us achieve mirroring—a state where the pattern in my head matches the pattern in yours. But getting our ideas to mirror has nothing to do with truth, beauty, friendship. It is certainly not sufficient to constitute communication. (I’m not being sentimental. I am using words exactly.)

If communication doesn’t depend on shared Shannon information, what does it depend on? What’s it about? It’s about sharing and respecting names. Not just my name and your name, but any and all names we invoke. “Me: I am a 49ers fan.” “You: Me too!” That’s a connection, a space for communication, and all the more so because we don’t need to know, at least for the time being, exactly why we’re fans, or what it means to us.

To give a somewhat more complicated example, consider Dworkin’s “interpretive concepts”—words like justice or democracy, which, he says, play a special role in the polity’s definition of itself. These things are names. Names refer to real things, but imperfectly defined and imperfectly known real things: objects of permanent interpretation and epistemic uncertainty, which are not separable from their name. “Me: I believe in justice.” “You: me too!” The point is not that justice is not a real thing. It is every bit as real as you, me, and the San Francisco 49ers. The point is that our communication will become more difficult, not less, as we penetrate to the heart of what it really is; or, as we start to communicate using mere labels and empty symbols, rather than names.

The things to which names refer are never merely “physical”: a person is more than just a body, and a city is more than just a bunch of buildings in a place. Likewise a speaker cannot bring any named thing before a listener’s view merely by conveying “information”. In fact, the more a named thing is depicted by Shannon information—rather than simply named—the more the speaker and listener will be divided by a sense that they are not speaking about the same thing. Names need to be interpreted by each conversationalist for himself, not necessarily because this procedure leads to the truth (although we cannot rule that out), but because this is how the parties bring themselves into relation with the name, and then through that medium, with each other.

What gives information meaning?

Many pressing questions of our day concern information. How to obtain it, protect it, fund it, and build machines that process it; and what consequences this will have for society, economy, and sense of ourselves. So we should probably slow down now and again to check the foundations of the house: What exactly are we talking about? What is information?

One highly influential way of looking at it says information is basically patterns upon a substrate which are separable from that substrate. Scratches on rocks, waves on water, sequences of charge on magnetic tape, that kind of thing. This is implicitly how Claude Shannon looked at it. His theory gave us tools to analyze how such patterns are copied from one substrate to another more or less “noisily”—thus emphasizing their separability from any particular substrate.

Shannon was quite clear that his definition of information was not equivalent to meaning. Instead, “Shannon information” captures seemingly random, unexpected patterns. Counterintuitively, when patterns are extremely random, they carry more information. The way to understand it is that their unexpectedness makes them capable of carrying more potential meaning. So, for example, the ten letter sequence “tbontbtitq” seems random, while the ten letter word “strawberry” seems ordered. The former seems to mean nothing and the latter something. But the latter carries less information than the former: “Strawberry” probably just means “strawberry”. “Tbontbtitq”, on the other hand, might mean nothing, or it might be a coded message with heck of a lot more going on in it than “strawberry”. (In fact, it is a compression of “to be or not to be, that is the question.”) This potential to be a complex coded message is what makes a sequence “information rich”.

Here is how his collaborator Warren Weaver put it in a 1949 introduction to Shannon’s ideas:

“The word information, in this theory, is used in a special sense that must not be confused with its ordinary usage. In particular, information must not be confused with meaning. In fact, two messages, one of which is heavily loaded with meaning and the other of which is pure nonsense, can be exactly equivalent, from the present viewpoint, as regards information. It is this, undoubtedly, that Shannon means when he says that ‘the semantic aspects of communication are irrelevant to the engineering aspects.’ But this does not mean that the engineering aspects are necessarily irrelevant to the semantic aspects. To be sure, this word information in communication theory relates not so much to what you say, as to what you could say. That is, [Shannon’s definition of] information is a measure of one’s freedom of choice when one selects a message.”

So Shannon information—the conception of information upon which the computing revolution was built, what our hard drives store, what LLMs process—is not meaning, but rather the potential to carry meaning.

Yet that still doesn’t tell us what meaning is. What does it mean for a bit of Shannon information to “have meaning”? On this, Weaver remarks further: “One has the vague feeling that information and meaning may prove to be something like a pair of canonically conjugate variables in quantum theory, they being subject to some joint restriction that condemns a person to the sacrifice of the one as he insists on having much of the other.” Vague indeed. Shannon seems to have appropriated the word “information”, rather confusingly, to tell us precisely what meaning is not, while shedding no light on what it is.

“Behaviorist” thinkers might wish to interject here (and plenty have) that meaning has something to do with effectiveness. If a message has a predictable effect on a rat in a maze, it must mean something. For example, if you tell the rat “go left” and it actually goes left, then the message has a meaning, namely, “go left”. But I disagree with this line of thought, because carrots and sticks can have much more predictable effects on rats in Skinner boxes, and without meaning anything at all. In fact, the concept of effectiveness seems every bit as much of a mutually-exclusive obverse of meaning as Shannon information is. Meaning, in order to be meaning, must maintain the possibility of having no effect at all. Thus, if we tell the rat to go left, then the statement has meaning–but, well, the rat might not do it. But if we poke it enough, it will definitely go left, yet without any meaning having been conveyed.

If meaning must “do” something in order to “mean” anything, what do we need the concept of meaning for at all? Why not just speak of things that do things? It does not help to say that meaning is a special “informational” kind of effectiveness—information that does something—for if this were so, then random passwords would be meaningful. They are not.

I believe this is a better way of looking at it: information has meaning insofar as it contains assertions about the world. “The sky is blue”. “I, Annie, am inviting you, Billy, to my birthday party”. “Pigs can fly”. Such statements may or may not require further context to be interpreted as the sender intends. And, obviously, they may or may not be true. But they have some degree of meaning regardless, because they carry the potential to be interpreted by recipients as asserting some true or false statement about the world. By contrast, “cf65w5Fg” is neither potentially true nor potentially false in any sense except in the threadbare sense that “This message’s content is: ‘cf65w5Fg’”. It isn’t a secret code containing an assertion or a reference. It simply has no meaning.

Names are not reducible to information.

Yet, not every bit of information can be categorized neatly as meaningful or meaningless. Some strings of Shannon information seem to encode neither nonsense, nor any assertion about the world. This suggests a strange gray area.

Consider a name. A name is obviously encodeable as information. And it may seem that names implicitly assert a kind of meaningful social fact: my name is “Matt”, yours is “Peter” or “Judy” or whatever. That’s what our parents named us, or what people call us. But there remains an important difference between names and other meaningful information. Because while most meaningful information can in principle be translated or redescribed without losing anything, names cannot.

If I tell you “the city has 8 million inhabitants,” you can render that in German, Mandarin, or binary code without altering the meaning. The description points to a state of affairs that exists independently of however we encode it. But the name of the city—New York, let’s say—despite being encodable as information, is neither nonsense, nor interchangeable with any other label, nor reducible to a true or false assertion about the world.

To be sure, the string of letters “New York” carries some information concerning etymology, and the history of the city. But there is more to it. The thing to which “New York” refers is not fully defined. It points to a living, impossible-to-fully-describe object—a city—which interpenetrates with countless other living objects, like people. These people use the name. And because the name cannot be losslessly translated, the way an ordinary fact about the world can, it is not an arbitrary label. “Nueva York” is not exactly the same name as “New York.” Everyone who has anything to do with New York must either simply call it New York, or else alter their relationship with it—for example, by calling it “Nueva York” or “the city”.

Names are thus not “separable” from all substrate, the way Shannon information is. They are “entangled” with their users and their referents. They are not merely neutral reference tools for their users, or interchangeable labels for the things to which they refer; rather, they partly constitute the things they name and the people who use them, linking both into a relational ensemble.

Names include what Dworkin called “interpretive concepts”.

Ronald Dworkin thought that certain words play a special role in our moral and legal culture. For example, the word “justice” refers to something real, something objective; yet, it could never be replaced by some other word referring to the same objective thing without altering or destroying that thing. In other words, “justice” is not merely a label for that to which it refers, but a name.

Dworkin didn’t put it this way. He called these special words “interpretive concepts”. But he believed they lay at the core of our political life. As a society, we understand ourselves through them: we seek the meaning of words like justice, beauty, and kindness with our whole being, yet never fully find it. And any time one invokes an interpretive concept, one implies some interpretation more than another. It is, in short, a moral act, and a political act — no mere conveyance of information.

This is so when any name is invoked, whether the name of a concept with normative power, or the name of a god, or the name of a person. Using names, then—and invoking “interpretive concepts”—is thus a “speech act”: an action consummated through speech, like agreeing, rebuking, or lying.

Can non-humans perform speech acts? This is the most important question about AI.

It is not reasonable (and in fact deeply misleading) to consider that AI is “thinking”, or “agreeing“, or “lying”, or doing anything besides processing information, until the answer is yes. Some suppose that speech acts like naming, thinking, acting, agreeing, or lying are complex forms of information processing — that is, functions that become possible when information processing reaches a certain level of intensity, or acquires a certain topology. But our argument hints at the opposite: that speech acts are something totally other than information processing, and that the more Shannon information is being processed, the less meaning exists, the less names are being used, and the less speech acts can be said to be occurring.

Information and efficient versus formal cause

For a moment, let us leave aside the special case of names and speech acts, and return to the notion of ordinary meaning-laden information.

Meaning-laden information might be characterized as description. For a description to exist, there must be two locations which contain substrate matter in two states. The matter in location A is in a state that somehow reflects, or purports to reflect, some fact or facts about the state of matter in location B. Thus, location A carries certain meaningful information about location B. (Again, for now, we are leaving aside the question of whether the information about location B is true. Even false information is meaningful. What makes it meaningful is that it could be true.)

What do we mean by “two locations”? Possibly that location A is “over here”, and location B is “over there”, with the two locations separated by some expanse of space which is neither A nor B. For example, Earth is one place, Mars is another, and between them lies an expanse which is neither. But this is not the only possibility. It is also conceivable that A might be within location B, yet not identical to it—the way that “New York” is not the same place as “Earth”, even though the former is within the latter.

Many important assumptions about the nature of information and meaning depend on the difference between these two versions of “not-being-in-the-same-place.” Namely, when information could correspond to truth—when it is meaningful—what is it that creates that possibility?

Why, for example, might a pencil sketch accurately depict the New York skyline? If we assume “two separate locations” between the sketchpad and the skyline, then the possible truth (i.e., the meaning) of the sketch must be attributable to what Aristotle would have called an efficient cause. Someone aware of the contours of the New York skyline must have actually sketched it upon the paper. In this case, the description must postdate the described thing.

But there are other possible explanations for the sketch, which reveal that the sketchpad and the skyline do not necessarily constitute two separate locations. Perhaps the sketch was done by an architect imagining a new building added to the skyline, and the building was built, but only after the architect’s sketch. Another example: suppose we have a swatch of painted fabric which matches the skin tone of the Mona Lisa. It might have been made by someone who matched the shade, in which case it would have to postdate the Mona Lisa. But it might also be a piece of canvas cut from the actual Mona Lisa, in which case it would predate the Mona Lisa, since it existed before the painting was finished. These cases show how true, meaningful information can exist without being a postdated “description”—and without ever traversing any empty expanse from one location to another in the manner of Shannon information.

When changes occur in a whole body, information can arise in the parts of that body without being “conveyed” from any separate location. This corresponds to what Aristotle called “formal” causation. When some Whole undergoes a process of change, truthful descriptions of Part B can arise in Part A before Part B “changes to match its description”. As Leonardo works on his masterpiece, a part of the canvas comes to contain information about the Mona Lisa’s skin tone before the Mona Lisa even exists as a woman with a smile. An architect’s sketchpad contains information about the New York skyline before the skyline contains that information. In these cases, the description and the described things are part of a larger whole, and wherever that is the case, information can reach back from the future.

Best,
Matt

What makes alphabets “alphabets” is that their characters represent phonemes, or sounds, which, by themselves, have no meaning. By contrast, the characters in “logographic” writing systems – such as Mayan pictograms, Egyptian hieroglyphics, and modern Chinese script – represent whole words or at least components of words, which have their own layers of meaning.

Foreshadowing the way binary computer code reduces information to ones and zeroes, alphabetic systems thus capture language in an extremely small set of abstract symbols. The compactness of alphabets – just 20-30 characters – makes them easy for people to learn and for machines to encode, so that they have played a major role in the history of information technology. Chinese culture, for example, was much less transformed than European culture by early printing, because the movable-type press could not easily capture Chinese writing with its thousands of unique symbols. Chinese printers had to cram their writing tradition into a straightjacket in order to use moveable-type presses, while the printers of alphabetic languages could render their written languages on the page without modification.

By making texts ubiquitous, printing also had the effect of spreading literacy beyond educated priesthoods and learned guilds. This matters because literacy transforms the subject’s sensory relationship to language.

In Gutenberg Galaxy, Marshall McLuhan explains this shift by contrasting the auditory and visual senses: Prior to becoming literate, he says, the human subject encounters language primarily as an auditory phenomenon. The auditory sense opens onto a world of magic: incantations, spells, the voice of God. The visual sense by contrast opens onto a world of rationalistic control: plans, maps, tools. To see something vividly is to increase one’s ability to control it from a safe distance, while to hear something clearly is already to be touched by its influence. Thus all writing (even logographic) represents a rationalistic and visual capture of language’s magical, aural aspects; but alphabets in particular, by capturing language in a very small set of visual symbols for meaningless phonemes, tame speech’s numinosity, rendering language visualizable and manipulable. The alphabet disenchants language, while also rendering its power more available for instrumental use.

The upshot? In a really literate society, fewer people believe soothsayers, prophets, or silver-tongued demagogues. People do, however, believe much of what they read, and they can easily write to manipulate each other.

*

Most scholars agree that the Proto-Sinaitic letters first appear early in the second millennium BC, in Sinai or the Levant. Thus, their first appearance is not too distant in time or space from the emergence of the oldest parts of the Bible. There is also a general opinion that the letters were first derived from the cursive forms of Egyptian hieroglyphs. For example, the hieroglyph for a coiled snake became the Proto-Sinaitic ancestor of the Hebrew and Aramaic letter “Nun”, which later became the Latin “N”. The hieroglyph depicting waves upon water became the Hebrew and Aramaic “Mem”, and the Latin “M”.

According to McLuhan, these symbolic overtones persist to a greater or lesser degree in the Semitic scripts. This makes them somewhat less than strictly phonetic, because the characters have at least some meaning exceeding their indication of a meaningless sound. Relatedly, Semitic scripts traditionally omit vowels, so that the written words in languages like Hebrew and Arabic typically do not fully “encode” the proper pronunciation. But when the Phoenician alphabet, carried by traders across the Eastern Mediterranean, was adapted into Greek – an alien linguistic and cultural tradition, which lacked any knowledge of the genealogy of the letters – it was stripped of all non-phonetic meaning. And vowels were added. For example the Aleph, a glottal consonant whose written symbol derived from the hieroglyphic symbol of an ox, became the meaningless A – a vowel. Thus, the Greek alphabet finally completed the phonetic representations of words, and literate Greek culture was the first to reach “total” alphabetic abstraction.

While considerably denuded today, an aural-magical culture does still live in the Semitic scripts, for those who care to look. For example, some scholars hypothesize that the Aramaic letter Gimel evolved from the Egyptian hieroglyph for “camel.” There is a less-contested link between the Aramaic letter Qoph and the hieroglyph depicting the eye of a needle. These genealogies throw open marvelous possibilities. For example, when Jesus invoked the unusual image of a camel passing through the eye of a needle, is it not possible that he was commenting esoterically on letters?

However shaky such speculations may be, it would be a mistake to dismiss their significance in the development of both language and religion. Jewish Kabbalist and Islamic Lettrist mystics claim continuity with an ancient tradition of interpreting letters as laden with hidden meaning. The visual shapes of individual letters are scrutinized: the curved line at the bottom of this one represents a foot, the opening on the side of that one represents a door. The letters positively come alive as pictographs. In these interpretive modes, all of scripture can be canvassed for subsurface significance. Arguably, this procedure tends to weaken dogma and spawn religious “innovation”, by inviting new readings of old words. Yet these new readings have a way of bending time, because what one finds in the letters is, after all, not new. The text has always been there. If it was divinely inspired, nothing in it is accidental.

In fact, not even the most trenchant exotericists deny the power of this method. It is no secret, after all, that language sits at the core of the monotheistic creation story. “God said let there be light, and there was light”. “In the beginning was the Word”. All monotheistic mystical traditions see letters as something like building blocks of creation. The puzzle, the problem, is that they stand at human as well as divine disposal. They are reductive tools, permitting the numinous to be dryly represented and cynically manipulated, as well as being talismans for better and worse, and sacred objects of contemplation. Alphabetic script is like the mundane world itself: objectified, inert, and finite. Yet for those who know how to look and listen, infinity subsists in every interstice.

*

It seems likely, to me anyway, that the first Proto-Sinaitic alphabeticians were aware of both aspects of their project. The letters did not only represent meaningless phonemes, but also meaningful hieroglyphs, which projected meaning back into the words.

This back-projection of meaning may even have motivated alphabetization. Each spoken vernacular word could, with the alphabetic method, be rendered as a series of consonant letters, each of which had an independent meaning. Imagine doing this for the first time. To discover the spelling of a word, one had to speak it aloud, listen to the sounds emerging from one’s own mouth, and note them down. Then, one could look at the result as a sequence of hieroglyphs and discover in it fathomless material for interpretation. Every word thus acquired something like a “secret” meaning.

To get some sense of what this must have been like, imagine that we mapped the English letters to tarot cards. Now every word becomes a unique, meaning-laden sequence of symbols. Today, such mappings might intrigue some fortune tellers and esotericists. But imagine the mapping was more than just intriguing – imagine it unlocked an unimaginably powerful new technology. Imagine that this strange procedure suddenly permitted you to write down your previously unwriteable language, and to build a new scribal clique around that formidable new art. In short order, this clique of outsider-scribes would have been able to write faster, better, and more fluently than the pharaoh’s educated scribes with their stilted hieroglyphs.

Surely this would have seemed like a miracle.

I want to share a few honest thoughts about AI and education. Maybe they will sound pessimistic. I don’t think they are. In any case, I end with a non-pessimistic practical recommendation.

*

It seems probable that schoolchildren will soon be educated mostly by AI, with human teachers in a complementary role. Some predict that this will result in unprecedented improvements in the quality of education, because AI will function like a world-class expert giving personal attention to every student, in every subject.

Is that really what it will be like?

If we were firing an arrow, we would try to predict its drop and drift based on factors such as distance and wind. Here, Marshall McLuhan’s maxim “the medium is the message” tells us the factors to consider. The results of AI education will differ from our intentions because AI’s role in the educational process will teach an overriding lesson of its own.

The overriding lesson that students will learn in AI-mediated educational environments is that they have no real reason to know whatever the AI is conveying. They will learn this even if both adults and AI tell them otherwise, and compel them with carrots and sticks to pretend that they did not know it. This is because it is true, and like other true lessons, students will learn it all the more if they are bright and curious—not to mention keenly aware of their differences from machines and their kinship with humans.

To see why, we should look closely at the pedagogical changes AI is likely to bring about. Human schoolteachers will be asked to spend less time instructing students directly in subject matter, and more time supervising students’ engagement with AI systems. The AI systems will take over more and more of subject matter instruction. Step by step, the role of the teacher will thus evolve into a hybrid between teacher, counselor, and babysitter. The teacher’s traditional role—that of an adult master of the subject matter at hand, and the medium through whom students must learn that subject matter—will be attenuated or eliminated.

Advocates for AI education argue that where AI knows more about the subject matter than the teacher, it will be abler than the teacher to convey it to the human student. This is absolutely correct. Only after we enthusiastically concede this point does the main issue come into view.

More important than a teacher’s ability to convey subject matter is his or her ability to inculcate into students the special virtues and habits associated with that subject matter. This, not subject-matter instruction, is what we stand to lose if teachers depart from their traditional role. Such virtues include, for example:

• the attention to pattern and proportion which makes a good mathematician;

• the habit of engaged listening which every good musician shares; and

• the circumspection concerning source credibility which characterizes good historians.

AI can name and communicate about these virtues. But it does not follow that it can inculcate them. It cannot. This is because these are human virtues, which humans acquire, if at all, by loving emulation of some other human. As some wise aphorist once said: “monkey see, monkey do.”

When a parent successfully inculcates a virtue in a child, he or she does so not by explaining or depicting the virtue, but by actually exhibiting it in real human circumstances. Picture, for example, a daughter who notices her mother doing what she knows to be just, in spite of the dangers involved. Here, the daughter notices her mother’s courage. She even feels a bit of the mother’s fear, because she loves her mother — her mother’s good is also her good. In this participative act of noticing, the daughter begins to activate the seeds of courage in herself, through loving emulation.

AI cannot inculcate any virtue in this way. This is because, first, it cannot be in human circumstances (such as a mother facing danger), and second, because it cannot be the object of a human’s loving emulation. Humans cannot emulate AI lovingly because they cannot love AI in the way that they can love other humans—or at least, not without invoking a further array of worrying side-effects. (In this regard the challenges of AI education may turn out to be remarkably continuous with the emotional side-effects of AI companions.)

The virtues associated with educational subject matter are analogous to those taught by parents. They are inculcated through human students’ loving (or at least respectful, and respect is a variety of love) emulation of human teachers’ behavior under human circumstances. An AI therefore cannot initiate students into those special virtues which have developed in the disciplines over millennia. (Except, see my conclusion.) It can only make students able to regurgitate information or mimic skills—and AI is already so much superior in those things that there really is, besides acquiring associated virtues, no clear reason for students to seek to excel in them.

And let it not be supposed that students will simply acquire some new or non-human kind of virtues instead. There are few if any new virtues, and no non-human ones that we would want our children to learn. Thus, if the involvement of AI causes students to fail to acquire educational virtues—such as those mental habits that make good mathematicians, musicians, or historians—they will simply be failing to acquire any virtues in their education.

It is also wrong to imagine that we will be able to compensate for AI’s deficient virtue-inculcation by resupplying educational virtues along some other pathway—for example, scheduling AI-driven subject-matter lessons before recess, and virtue inculcation afterwards. Virtues are not like vitamins which can be isolated from their substrate food matter and resupplied in pill form. Alasdair McIntyre underlines this point in a passage from his 1999 book “Dependent Rational Animals”:

“We would make a mistake to infer that some part of the child’s education has to be set apart for specifically moral instruction. Just as the virtues are exercised in the whole range of our activities, so they are learned in the same range of activities—in those contexts of practice in which we learn from others how to discharge our roles and functions: first as members of a family and household, then in the tasks of schoolwork, and later on as farmworkers or carpenters or teachers or members of a fishing crew or a string quartet. For to be instructed in the virtues together with the relevant skills is nothing other than to learn how to discharge those roles and functions well rather than badly….Teachers in general, then—parents, other family members, those who instruct apprentices in crafts—have to have, in significant measure, the habits that they try to inculcate.”

Someone who accepts all of this may yet insist that those humans acting as complementary instructors to AI education can and should convey the virtues connected with the subject matter. But this is not plausible. First, it is obvious that the possibility of using AI to instruct more students faster will present irresistible economic temptations. Human teachers with specialized skills will be retained less and less by administrators and parents seeking efficiencies. Second, complementary instructors will fail even where they are employed. There are educational environments that are not under any economic pressure, such as private schools for the children of the rich, and such places probably will retain skillful complementary human instructors for the foreseeable future. (Which is to say technology will have transformed what was once a universal staple into a rare luxury.) But even in these privileged environments, the presence of AI will undermine virtue inculcation. This is because the complementary human teachers will find themselves less effective than AI at instructing students in the ostensible subject matter at hand. They will therefore progressively lose any authority connected to that subject matter, so that students will not look to them in connection with it, and they will lose the ability to inculcate whatever subject-specific virtues they possess.

To put it more bluntly, students will ignore them. When students are learning, even and especially when they are learning very assiduously, they are not aware that they are and ought to be learning virtues. In fact, in order to learn the virtues connected to a particular subject matter, students must believe that their task is instead to learn the subject matter, not the virtues involved in it. Another well-known quote from Marshall McLuhan illustrates this dynamic:

“Our conventional response to all media, namely that it is how they are used that counts, is the numb stance of the technological idiot. For the ‘content’ of a medium is like the juicy piece of meat carried by the burglar to distract the watchdog of the mind...The effects of technology do not occur at the level of opinions or concepts, but alter sense ratios or patterns of perception steadily and without any resistance.”

In this sense, students have always been and will always be “technological idiots” with respect to education as a medium. The “content” of education—the subject matter of history, mathematics, or Latin—is indeed mainly a juicy piece of meat whose function is to occupy students’ attention while the real work of education, the formation of adjacent virtues, proceeds beneath the level of their awareness. For this process to function, students must attribute authority to the teacher based on his or her mastery of subject matter, even though it is really his or her mastery of the associated virtues, which constitute the truest qualification for the job. Indeed, at the outset of any course of study, students are typically unaware, or only dimly aware, even of the existence of virtues attached to that practice. Subject matter is the more or less disposable message of education. The connection between human teachers and students, resulting in the intergenerational communication of virtue, is the medium.

So the presence of an AI which possesses more subject-matter knowledge than the teacher—as well as more communicative flexibility and infinite time for each student—will interfere with teacher-student relationships, causing teachers, qua teachers, to be ignored. Having been transformed into disciplinarians or babysitters, they will become not only more interchangeable to administrators, but also objects of scorn in the eyes of students.

The most intense scorn will come from those students who are really focused on mastering the subject matter. They will view themselves as aspiring to something in which their human disciplinarian or babysitter is deficient. Conversely, and just as problematically, to the extent that students really love or respect any humans who complement their AI education, they will tend to consider the subject matter superfluous. Why struggle to learn what one’s role models do not know?

To recap: precisely insofar as AI systems match or surpass the ability of human teachers to instruct students in subject matter, they will pervert students’ morals and stunt their development in the associated virtues. Inculcating the virtues associated with educational subject matter like mathematics and history has always been the highest goal of education; but all virtues are human qualities. So when students are instructed by non-humans, they will fail to learn, or will unlearn, what education really exists to teach. At best, students will acquire skills and knowledge in which machines already and inexorably surpass them.

As I already suggested, the brightest students will grasp this whole dismal picture. They will be frustrated and appalled, and will pull back the curtain for their peers. Many adults—not just classroom disciplinarians and babysitters, but framers of educational institutions, over-optimistic builders and adopters of artificial intelligence, and many others—may become the objects of the future students’ justified contempt.

A non-pessimistic practical recommendation.

Setting aside ideals and speaking practically, what is the best way that AI education could develop?

If a AI system were to cause a student to develop a genuine subject-specific virtue, such as a love of literature, how would it happen? I see three possible pathways, listed from least to most promising:

• The student could learn the virtue, or at least a simulacrum of it, after being induced into respectful emulation of a non-human artificial personage; or

• Superficial “signatures” of the virtue could be induced into the student’s behavior through dopaminergic “carrots and sticks”, for example, rewarding them for seeming to behave consistently with the virtue in question, and penalizing them when they behave otherwise; or

• The student would learn the virtue by respectful emulation of human personages who are not present in the room, but to whose work the AI simply directs the student. In other words, it could introduce the student to Dante and let Dante speak for himself.

The first pathway invites the student into an unhealthy relationship with a non-human person. The second pathway dehumanizes and manipulates the student. Terrible ideas.

The third pathway is the least bad. In essence, it is the path that autodidacts and Wikipedia enthusiasts have always taken. It fails to avoid my general critique of AI education because, for example, Dante does not interpret himself. Human educational processes have always filled the space between the canon, or the stores of accumulated knowledge, and the student’s present experience. Autodidacts (including yours truly, to no small extent) are indeed characteristically lacking in certain virtues, and we may expect future generations of students to exhibit these deficiencies in more extreme forms.

Nonetheless, this is the least alarming paradigm for AI education, and it should be considered by administrators.

Suppose we hear emotion in an excellent recorded song. Chavela Vargas, Paloma Negra. The truth, often enough, is that the human who sang it really felt the emotion we hear.

But we perceive the emotion through the medium of patterns in sound waves, etc. An AI system can fabricate those patterns with greater precision than our senses can sense them, making it impossible to tell the difference between reality and simulation from listening alone. When things reach this point, we may conclude that there is no difference between reality and simulation, because we cannot perceive any; there is no difference “to me”. In fact there still is a difference, including to me, but now to perceive it we must look to other contextual facts which were formerly not needed to establish authenticity, such as the cultural history of the song, or out trust in the source of the audio recording. (Significantly, these facts are historical and genealogical. They are unbroken links to a living human culture, whereas the genealogy of AI artifacts always traces back to a meaning-destroying statistical procedure: a deck-shuffle, a blender pulse.)

These contextual-genealogical facts can be lost, and even when not, restoring contact with truth now requires us to pour part of our “listening” attention into metadata. We can no longer perceive human authenticity within the four corners of the musical form. The sound of human voice no longer means “human”. Our ears and eyes thus lose their role proving reality. We lose our intimacy with our eyes and ears.

This is not entirely new. Backstory and artist persona has always been “part of the music”. The thickness of the backstory around music has always been a response to, a defense against, the thinness, inscrutability, and unverifiable authenticity of mere sound.

This is related to what I mean when I have said, to your irritation, “I no longer enjoy such-and-such product because it is owned by a private equity firm and the experience is tarnished by the identity and intentions of the people who own it.” You can only protest that the product is unaffected by the unpleasantness of its owners if you insist on certain boundaries to my perception of the experience. But metadata counts. Everything is potentially indicative of the reality of everything else. Products sold by private equity-owned firms are masquerading as products being sold by honest, non-exploitative businesses. They are fakes.

At the moment that we abandon these distinctions of authenticity in our minds, reality still exists, but we lose touch with it by conflating it with something that it only resembles. E.g. there is a difference between a passionate human song and a moving AI song, but because it is not audible, we deny the reality of the difference, or lose track of its salience. The surface equivalence of truth and falsity is noticed; and the subsurface is forgotten.

This is a major phase shift, beyond which something terrible happens. When human beings lose their sensory-attentive link to the markers of reality (or worse, intentionally ignore them), reality itself begins to change. It actually becomes less real, less beautiful, less “itself”. When we aesthetically prefer the ersatz, the beauty of truth is dampened, rendered unreal, relegated to a kind of netherworld, like the sound of a tree that falls in the forest when no one hears. Much of reality (excepting superhumanly large things like planets, and infrahumanly small things like atoms, but including most everything in the mid-size range such as persons, animals, institutions, artworks, rivers, mountains) is vulnerable to this kind of destruction. The subsurface of many beautiful and real things needs to be lovingly perceived by human beings in order to exist, because these mid-sized things exist as organic wholes in symbiosis with human beings, and in no other way.

This is also how dehumanization works: when you see another person merely as one or several of their representable aspects, their value to you becomes commensurable with other things that share that aspect. That compression of the other person, reflected back to them, wave-interferes with their perception of their own depths, i.e., their self respect, dampening it just like noise cancelling headphones, so that eventually the depths of a person may be perceived by no one, including themselves, clearing away the most important barriers to being treated as sub-human by others, and even by themselves.

This, I think is the real risk with AI, as with every previous technology: it is a metaphysical solvent. (On the necessity of metaphysics in an ethical society, see del Noce. AI’s erosion of context kicks away the ladder to metaphysics, metaphysics being a kind of deep contextualization.)

Matt

Is it “like something” to be AI? If not, could AI develop internal conscious experiences in years to come? Plenty of thoughtful people consider this a real possibility, while others find it absurd. This makes it a difficult topic, but we cannot avoid it. It crops up naturally in language whenever we personify AI or refer to it as a new “species”. This is not just semantics — something is at stake. For example, if machines can feel, then we should avoid making them suffer. If they cannot feel, then we should avoid distorting our own behavior around this imaginary concern. Either way, how seriously we take machine consciousness will deeply affect how we relate to tomorrow’s technologies. This essay proposes a new way of looking at it.

Two mainstream schools of thought dominate the debate. The first is the Scientific-Realists. These thinkers speak the language of physics and hard science. They see consciousness as a phenomenon “out there” in the real world, which science will sooner or later define and measure. According to this school, science will eventually arrive at a theory of consciousness that makes testable predictions, as it has done with other once-mysterious phenomena like electricity and gravity. At that point we will either detect consciousness in machines or not, effectively settling the question.

A second school of thinkers, the Empirical-Positivists, speaks the language of social science. They suspect that consciousness will remain fundamentally beyond the grasp of science for the foreseeable future, or perhaps forever. Since we have not yet been able to scientifically prove consciousness even in other humans (see the problem of other minds), it is hard to see why we will ever be able to do so in machines. Therefore Empirical-Positivists contend that the only fruitful scientific question is whether people believe AI is conscious.

I largely agree with the Empirical-Positivists. I doubt that science will soon solve the riddle of consciousness. However, Empirical-Positivists pay a heavy price for their pragmatism. By transforming the question about machines’ consciousness into an inquiry about what other people think, they beg the question that many of us will actually ask ourselves: whether machines truly are conscious. We may ask this question intellectually, viscerally, sensorially, spiritually, or some other way. And other peoples’ views influence us, but then, again – how did those people form their opinions?

To have a thoughtful, society-wide conversation about whether machines are conscious – rather than just letting a new de facto consensus “happen” to us – we need shared language for evaluating whether ideas about machine consciousness are good or bad, sensible or senseless, responsible or irresponsible. Empirical-Positivists have nothing to say about this. And yet those who confront the question with the tools of hard scientific inquiry – Scientific-Realists – will hit a wall, at least in the short term, removing themselves from the conversation too. So both major schools of thought effectively sideline themselves.

Therefore, by simple process of elimination, we can deduce that independent human opinions about machines’ conscious status – opinions which could shape the culture’s general attitude toward the technology – will be formed with heuristics which are neither Scientific-Realist nor Empirical-Positivist. These heuristics are where the action is. So I don’t think it behooves us to dismiss peoples’ belief/disbelief in machine consciousness as an artefact of cognitive bias, sensory illusion, superstition, or folk theory. That might suffice if we were dispassionate observers of the AI phenomenon. But since we all have a stake in how our culture resolves this question, we should descend from our ivory towers to participate in the modes of thought in which ordinary people – all of us, really – are likely to confront it.

This demands a third school of thought. Call it the school of Phenomenological-Realism. It speaks the language of philosophy, even theology. It looks at the heuristics by which we attribute consciousness to things such as machines, animals, and people, and subjects them to logical, philosophical, moral, and even spiritual scrutiny, while avoiding tangling around the axle of consciousness’s scientific status.

…Through the Lens of Machine Authority

I’ve been nurturing a hypothesis that fits into this new school.

There is a family resemblance between the subjective processes of attributing consciousness to something (like a person or machine) and attributing authority to it. When we ask ourselves whether other biological beings are conscious, we are not “doing science”. Instead, we are asking whether we feel an obligation to respect them – to listen to them and attend to their needs in light of their special status. We are asking whether they matter to us in a special way. While not identical, this rhymes closely with asking whether we respect their authority.

Underlining the family resemblance, authority is almost as hard to define as consciousness. But, happily, no one tries to cast authority as an object of study for the hard sciences. Authority is obviously a political, moral, legal, or religious phenomenon, not a physical one. Therefore we can look at it through these humanistic lenses comfortably, without “physics envy”.

Let’s look at the definitions of authority in legal and philosophical literature. To simplify our discussion, I’m inclined to set aside polemical anti-authority theories, like those of Foucault and the Frankfurt School – as well as polemical pro-authority theories, like those of Hobbes and Schmitt. All these ideas are important, and not necessarily wrong, but they are difficult to separate from their ideological topspin. I will focus instead on Alexandre Kojève’s discussion in his book The Notion of Authority. Kojève’s study of authority resonates with the best thinking on both left and right, but it doesn’t pick a side.

Some context: Kojève was a charismatic lecturer who influenced many well-known French intellectuals while working as a university professor in the 1930s. Later in life, he served as a powerful diplomat who helped lay the groundwork for the European Union. He drafted The Notion of Authority in 1942, which is to say after the authoritarianism of the early 20th century had led to catastrophe, but before the development of anti-authoritarian thinking in the ‘60s. Kojève’s own ideology is enigmatic. Born in Tsarist Russia but leaving for France in early adulthood, he was a Marxist-Hegelian idealist who evolved into a pragmatic midcentury European integrationist. Whatever we think of him or his agenda, I think it fair to say that this text, which remained unpublished in his lifetime, reflects an earnest effort by a serious thinker to analyze the concept of authority without being reflexively “for it” or “against it”. It is partly a survey, analyzing the conceptions of authority in Plato, Aristotle, Scholasticism, and Hegel, but it ends up stating a theory of its own.

So what does Kojève say about authority?

Kojève on Authority

First, he defines authority as the quality which causes one’s acts to be willingly not resisted by those whom they affect, even though they could resist. This definition distinguishes it from coercion. Also, note that in this definition, authority is neither necessarily good nor necessarily bad – those downstream of authority might be right or wrong not to resist it.

Then he identifies four archetypical varieties of authority: the authority of “father, master, leader, and judge”.

• The “father” type of authority shows up in those who embody positive aspects of the past, or to whom we feel we owe our existence.

• “Master” authority shows up in those who are willing to run a greater risk than us, whose boldness persuades us to follow them even when they do not actually threaten us.

• “Leader” authority shows up in those whom we believe know more than us, or who can anticipate the future and guide our projects to success.

• “Judge” authority shows up in those in whom we perceive goodness, honesty, and a trustworthy sense of right and wrong.

These four types of authority are not mutually exclusive: typically, wherever we perceive authority, we perceive a complex combination of these qualities.

Asking which types of Kojèvian authority AI manifests draws us into a phenomenological analysis which feels akin to the mental process people actually perform when grappling with the possibility of AI’s consciousness. Again, when we recognize another human being’s consciousness, it implies at least some acknowledgement that they have authority, because conscious beings are to some degree, on moral grounds, not to be resisted even when they could be. Analogously: what does it feel like to engage with AI? When, why, and how far do we willingly “not resist” it? As it improves, what new authoritative qualities could it acquire? How does this authority differ from the authority we perceive in different sorts of beings or institutions?

I suspect that insofar as AI exhibits a similar blend of authoritative qualities as humans and closely related animals, it will be hard for many people to resist inferring its consciousness. But can it or will it really exhibit such authoritative qualities? Let’s take a closer look.

At the moment, it seems to me AI exhibits a significant degree of “leader” authority, in virtue of its capability and knowledge. It can help us code things, for example, or explain things that one would otherwise need to pore through books to figure out. In this regard it already usurps some of the authority of knowledgeable human “leader” figures.

As we begin to encounter awesomely powerful agentic AIs asserting themselves in our environments, these may also acquire a kind of “master” authority, especially if we take them to be running some kind of risk to their own existence (such as an agentic AI willing to risk the money that pays for its operation).

However, the “leader” and “master” types of authority are not strongly associated with paradigmatic cases of consciousness attribution. Children, for example, possess little of these kinds of authority, yet everyone recognizes that they are conscious.

Other types of authority—“father” and “judge”—seem farther from AI’s grasp.

“Father” authority requires a shared history that we do not have with AI. We do not look at Claude and see the cause of our own existence. But this could change as time passes: imagine if future generations credit their lives to past AI agents.

“Judge” authority assumes deeply-embedded moral qualities not evident in AI. For example if a human with this type of authority, like a respected friend, tells me she finds something morally wrong, I might change my behavior in response, even if I didn’t fully understand or agree. But if I got the same moral counsel from AI, I wouldn’t change my behavior all unless I logically concurred with its given reasons.[1]

“Father” and “judge” authority seem particularly important in the edge cases of consciousness-attribution. For example, it is only when these forms of authority are exceptionally evident that people attribute a deep sense of personality or dignity to institutions. (Consider the difference between a nation and an airline.) This also clarifies our experience with children and animals. We apprehend a common biological root with them, or a connectedness with a larger historical reality – and this imbues them with something like “father” authority. We also ascribe an irreducible measure of “judge” authority to children and animals, however limited its scope. For example, when they show distress, we sense that something that is happening really is bad for them – even if we can’t see what it is or why it’s bad – and by extension, we accept that it is also bad for us. Through apprehension of these qualities, we infer that they are having experiences like ours.

Scientific-Realism and Empirical-Positivism don’t give us tools like this to analyze what is happening when people attribute consciousness to machines. Yet that’s what we need to do in order to reason about it together. Admittedly, phenomenology is a cousin of positivism — it does not suffice to answer the ultimate question of whether machines are conscious. Still, it cuts deeper than the usual methods. By taking phenomenology seriously, we can avoid both the morass of proving machine consciousness scientifically, and the mistake of archly dismissing peoples’ attributions of consciousness as empty illusions. If you suspect that AI cannot possibly be conscious, it may be more impactful to make the argument on phenomenological grounds (such as by way of an authority analogy) than trying to make an irresolvable scientific-materialist claim (or, indeed, a spiritual-religious claim that depends on premises not shared by your audience). The “authority analysis” I have sketched is one way of proceeding, but I hope it suggests broader possibilities.

Authority has in common with consciousness that is not a studyable object. Both are complex qualities that people attribute to objects and/or beings. The reasons for such attributions may be cogent or incoherent, good or bad, and anything in between. To ground discussions in coming years about when, why, or whether the notion of machine consciousness ought to be taken seriously, we need first to take seriously the reasons people think they perceive it.

–

[1] This authority analysis applies beyond AI. It can also help us analyze our attributions of authority to institutions, animals, and more. Consider an employee’s relationship with a company. If they see in it “leader” authority only, they might work for it for self-interested reasons, such as the promise of financial rewards. If the company acts inspiringly boldly, it might acquire “master” authority. If the employee develops a deeply entangled history with the company, for example because it provides for them reliably, it might gain “father” authority. If it inspires idealistic moral devotion, it might gain “judge” authority. The more of these qualities it combines, the more we tend to “personify” it.

This is qualitatively different from last decade’s issues of “misinformation”, “censorship”, and “polarization”. If those things were horse-drawn carts, AI is a McClaren F1. We’re no longer talking about pulling people into the nearest intriguing rabbit hole like Web 2 recommendation engines, or shielding people from verboten opinions on vaccines, etc. The next (and even current) generation of AI has the potential to very accurately predict what content should be shown to a particular person, at what pace and in what order, to fully convert them to a specific worldview.

Basically, that is not good. There is every reason to think that the most well-financed wielders of this technology will not push people’s worldview towards what is true or beneficial. They’ll use it to sell and manipulate.

Therefore, it’s important to give more thought to defensive “anti-manipulation” technologies: AI tools designed to counter this by helping people maintain a thoughtful degree of “information hygiene” and resist persuasive propaganda, without totally unplugging.

The Difficulty

The difficulty with anti-manipulation AI tools is the same as the difficulty with last generation’s “fact-checking” tools. Anti-manipulation tools necessarily come from a particular point of view, and have an ideology of their own. In theory, to deal with this, we’d need to construct filters that are orthogonal to every kind of ideology, which would pour cold water over equally, over every sort of ideological influence.

This is strictly impossible. However, I have some thoughts about how to tourniquet the problem, countering many ideological directions simultaneously with a satisfying degree of neutrality and transparency. Here is the approach in broad strokes:

• Build a map of ideologies that is as abstract and neutral as possible, and transparent in its own assumptions.

• Use it to predictively tag content that people are exposed to in terms of where it seems to be trying to guide them.

• Alert users to the ideological fingerprint of what they are seeing, and/or down-weight disproportionate influence from any part of the map.

The Map

The key ingredient here is the map. But what is the latent space of ideology? What exactly is happening when people drift towards conservatism, or progressivism, or authoritarianism, or liberalism? I propose to understand it as shifts in people’s teleological assumptions.

In philosophy, teleology refers to the idea that things have a natural end or goal. Here are a few examples of teleological ideas:

• a country’s resources should benefit its citizens

• the purpose of education is training people to participate in democracy

• the point of life is to have children

• communities are important for their own sake

Notice that such statements can tilt very far right or left or anywhere in between. Notice also that it isn’t really possible to “escape” teleology. Even if you think that things don’t matter for the sake of other other things, but just for their own sake, that’s still a teleological position.

The point is that when we disagree ideologically, we are very often disagreeing in the realm of teleology. Ideology isn’t just about “what matters”. It’s about what matters for the sake of what.

As a first pass at mapping teleological positions, I made a graph of many different units of organization, from large to small:

atoms → individuals → communities → states → humanity → planet → the whole of reality

I put both sets of units on the X and Y axis to represent every combination of statements to the effect that “one of these things matters for the sake of one of these other things.” Here is what it looks like:

Take a moment to look at all of the statements here. Which ones do you agree with? Which ones feel wrong to you? You might notice that they have quite different ideological “flavors”. To try to draw that out a bit more colorfully, I labeled the points:

The labels are probably tendentious, and definitely aren’t the only possible ones. (I’m having a bit of fun with this.) But again, take a minute to try to notice patterns.

It seems to me that related ideological propositions lie along the horizontal and vertical lines. I labeled these lines interpretively on the top and right sides of the graph. I also noted qualitative similarities between propositions on the upward-sloping diagonal lines, and labeled those lines, too.

In next sections I will unpack and elaborate this graphing approach, explaining what I think it means, and what insights fall out of it.

The Propositions on the Map

As mentioned, each point on the graph corresponds to a statement that “X matters for the sake of Y”. I will call these “ideological propositions.” These propositions do not constitute complete ideologies. Fleshed-out political ideologies like techno-utopianism, radical environmentalism, or Soviet Communism, involve more than one proposition, as well as other ideas that are specific to their context and culture. Still, many of their core assertions can be located on the graph.

I suggest some of these associations through the labeling. For example, at the point representing “individuals matter for the sake of humanity”, I’ve written “classical liberalism”. But this is obviously not the only label that would be accurate. And classical liberalism obviously cannot be completely reduced to this proposition. (Adam Smith, for example, believed that individuals, through enlightened self interest, could bring about a better world for all.)

Another thing to notice is that this scheme makes all the propositions seem rather simple. But when people encounter these propositions in their political lives, they don’t feel simple. In fact, people often experience them as major epiphanies. If you’ve gone through life assuming that the planet matters mainly for the sake of humanity, and then suddenly embrace the view that humanity matters mainly for the sake of the planet, you’ve undergone something almost like a conversion. You might hold to your new belief so fervently that you come to see adherents of your discarded proposition as ignorant or evil. Moreover, we often embrace propositions unconsciously, picking them up from our milieu at a young age. Ideology is often unreflective.

The Horizontal and Vertical Lines on the Map

Each horizontal and vertical line connects related ideas. Let’s look at the vertical and horizontal lines corresponding to “atoms.”

Atoms, as the basic stuff of resources and technology, are inevitably linked to the notion of materialism. Yet on this graph we find two very different versions of materialism. Propositions along the vertical “atoms” line claim that the configuration/distribution of atoms (e.g. material, resources, and technology) matters mainly for the sake of things that include or encompass atoms – “higher” or more complex things, such as individuals and communities. Conversely, propositions along the horizontal “atoms” line claim that those “higher” things are actually subordinate or instrumental to material imperatives.

Confusingly, propositions on both lines could be called “materialist” without misspeaking. Yet, the former are “justificatory” forms of materialism, while the latter are “reductive”. The labels above and to the right of the graph aim to illustrate these differences. The more idealistic or “justificatory” versions of materialism relate to what I would call “techno-utopianism”. I’ve termed the more reductive forms of materialism simply “reductive materialism”.

This brings us to the central diagonal line, which divides the whole graph between justificatory and reductive propositions.

The Radicalism Line Divides “Justificatory” from “Reductive” Propositions

The upward-sloping line at the center of the graph covers propositions claiming that a certain level of reality matters for its own sake: e.g., individuals matter for the sake of individuals, and states matter for the sake of states. I call this the “radicalism” line because these propositions concern radical individualism, nationalism, etc. Logically, its propositions are circular.

Importantly, the radicalism line divides the propositions above it from those below it. Propositions above it are “justificatory” in that they assert the significance of a lower level in terms of a higher level. Propositions below it are “reductive” in that they assert that a higher level is instrumental to a lower level.

Propositions farther above or below the radicalism line are “skipping levels” in their justificatory or reductive logic. For example, the proposition that individuals matter for the sake of the planet ignores those levels of organization intervening between individuals and the planet, such as communities and states. It thus places great faith in affairs at a low level – individuals and material – to affect matters on a high, planetary level. I have labeled it “Green Consumerism” because green consumerism (recycling, ethical shopping, carbon credits, etc.) claims that individuals and their choices matter at a macroscopic level. I do not mean to say that this is untrue – only that it is ambitious. People whose political ideology rests on this proposition try to thread a narrow needle, because it is easier to imagine larger, coordinated entities like communities or states bringing about intentional planetary-level environmental change. Propositions higher above the radicalism line thus tend towards ambition, idealism, and implausibility.

Conversely, propositions far below the radicalism line tend to seem more crass, reductive, and instrumentalist. They assert that big things matter for the sake of some of their smaller parts. Consider the proposition that humanity matters mainly for the sake of states. Even if you think this true, you’ll be able to grasp that it seems upside-down on its face, since “humanity” is larger, longer-lasting, and potentially inclusive of “states”. Propositions like this characterize ideologies that seem, on their face, to be imperialist or extractive. The farther we venture below the radicalism line, the more extreme the apparent reductionism.

Here is the graph with the justificatory/reductive distinction highlighted:

Analyzing Ideological Argument

Fully-developed ideologies tend to incorporate more than one of these propositions. But they do so according to certain patterns.

Reductive propositions often “shock the conscience” by asserting the instrumentality of seemingly-higher things to seemingly-lower things. Therefore, when people wish to defend them, they tend to reach for equally or more extreme justificatory propositions. For example, suppose someone believes that the planet only matters for the sake of individuals. On its face, this supports the idea that we can exploit and despoil nature as much as we like, so long as we’re careful not to ruin things that individuals need. And that’s a bit hard to defend – what about the animals, what about the inherent dignity of ecosystems, etc? Therefore someone asserting this proposition may feel compelled to offer a justification clarifying why individuals matter so much. The conversation might look something like this:

Alice claims that the planet matters for the sake of individuals. Bob recoils, and asks for an explanation. Alice answers that individuals matter for the sake of communities. She now wields a combination of two propositions: “the planet matters for the sake of individuals, and individuals matter for the sake of communities.”

Yet Alice’s overall claim still seems reductive: she is suggesting that the planet matters for communities. So Bob might still object, “But communities are only parts of the overall Earth, and they depend on the Earth. How could the Earth matter only for the sake of something subordinate to it, like communities?”

Now Alice has several options. She might seek shelter in a radical proposition: communities matter for their own sake. Or she might reach for justification in something loftier than the Earth: communities matter for the sake of God, or the whole of reality. If she takes this position, Bob might still disagree, but he can no longer accuse her initial claim that “the planet matters for the sake of individuals” of betraying a reductive worldview.

Just as extreme reductive propositions tend to shock the conscience, extreme justificatory positions tend to strike people as unworkable, unrealistic, or hopelessly idealistic. For example it may seem starry-eyed to say that individuals matter for the sake of the planet. Individuals, after all, are short-lived and tiny – can they really matter so much to something so huge? Even if so, what about everything else that seems to intervene between these two levels, such as communities and states?

Adherents of justificatory propositions must therefore embrace reductive propositions in order to render their justificatory propositions plausible. Someone committed to the proposition that individuals matter for the sake of the planet is likelier to support exploiting the community, or the state, on behalf of individuals. Indeed, if individuals do these things, their ability to affect the planet becomes far more plausible.

Where Do We Find Conventional Left/Right?

In terms of this graph, left-wing ideology tends to pull politics “down and left”, while right-wing ideology tends to pull “up and right”. Admittedly this is a relatively narrow and traditional conception of the left-right political distinction – basically running from “anarchism” to “religious authoritarianism” – and it does not easily fit the way most people use those terms today. However, if we follow the upward-sloping diagonal lines, we find propositions arrayed recognizably from left to right. For example, Marxism (atoms for humanity) is pretty clearly left of green consumerism (individuals for the planet) which is pretty clearly left of the idea of “communities for God”.

You could rotate the whole thing 45 degrees clockwise to get a clearer view of “left-right” dynamics.

(Just For Fun: Analyzing Historical Political Movements)

The graph also illuminates ideological history. Let’s return to the French Revolution. The main justificatory claims of the right-wing ancien régime sat in the upper-middle part of the graph. Influenced by medieval political and religious traditions, they believed that the French state, embodied by its king, mattered for the sake of God.

The left-wing revolutionaries, having grown skeptical of religion and tradition, believed that the French state existed for the good of something more concrete: individual people. This made them favor democracy over traditionalist monarchy, because democracy subordinates state sovereignty to the people. “But”, a right-winger might have protested in 1789, “how can something as complex as France matter only for the sake of its smaller parts, i.e., individuals? Isn’t that terribly reductive?” To justify this apparent reductionism, the left-wingers embraced the proposition that individuals matter for the sake of something even greater than France: humanity.

Let’s try to plot this on the graph. You might say that the first phase of French revolution moved through the following sequence of propositions:

• Rightist: (1) France matters for the sake of God.

• Leftist: No, (2) France matters for the sake of individuals.

• Rightist: Isn’t that reductive?

• Leftist: No, because (3) individuals matter for the sake of humanity.

This justificatory linchpin – that individuals matter for the sake of humanity – remains at the core of classical liberalism, and is one of the French Revolution’s lasting legacies.

France’s political shift from 1 to 3 was the Revolution’s key “leftist” achievement. After that, things drifted rightward. Now let’s plot the Napoleonic, imperialistic phase of Revolution:

• Rightist: If you’re right that individuals matter for humanity, and France matters for individuals, doesn’t it follow that (4) France matters for humanity?

• Leftist: Indeed it does.

• Rightist: But isn’t this a rather empty proposition, unless France can actually affect humanity?

• Leftist: Perhaps. But France is only a small part of humanity. How could France actually affect humanity?

• Rightist: Well, France could conquer humanity.

• Leftist: Interesting. But what would that imply, ideologically?

• Rightist: It would imply that (5) humanity matters for the sake of France.

• Leftist: Hold on now — isn’t that reductive?

• Rightist: It would indeed be reductive, if it were the only thing we were saying. But remember, we’re also saying (4) “France matters for the sake of humanity”. Together, the two propositions logically commute to (6) “humanity matters for its own sake.” Can we agree on that?

• Leftist: Yes! Let’s send out the army – vive la France!

The French Revolution set the enduring pattern for left/right movements in politics. Part of what makes the left/right distinction so confusing is that both tendencies are orthogonal to the axis running from upper left to lower right — which is to say, between justificatory and reductionist ideologies. Neither left nor right wants to see itself itself as crassly reductionist or unrealistically idealistic. But both sometimes are: Left-wing politics’ downward movements pull toward reductionism, as its leftward movements pull toward high justification. And right-wing politics’ rightward movements pull toward reductionism, as its upward movements pull toward lofty justification.

Isn’t this Graph Opinionated? Can We Add, Subtract, or Reorder Levels?

Adding levels between the ones I’ve used is as simple as drawing more lines. For example, perhaps “ecosystems” belongs between humanity and planet. Perhaps “cities” belongs between communities and states, or “family” between individuals and communities.

One might add more specialized levels like corporations, universities, or even architecture. It will be controversial where these levels belong. Does architecture belong between atoms and individuals, or between individuals and community, or between community and state? And that’s the point. However the levels are ordered, controversies about where they belong will register on the graph as ideological disputes.

The reason I’ve chose the levels I have, for my illustrative map, is that they are big, familiar concepts in the history of philosophy. Aristotle, for example, more or less saw individuals, communities, states, etc. as hierarchically nested, with the larger concepts acting as the integrated whole, or telos, of the smaller parts. Thus, I hope, this set up is fairly familiar and intuitive, making it clear what the graph is doing.

But this way of drawing the graph, and any other way, is of course opinionated. It implicitly impugns the propositions on its edges. As I’ve drawn it (and even more so how I’ve labeled it), the graph says: Science without more gives us nihilistic materialism (in the lower left). Religious radicalism gives us fruitless piety (in the upper right). In this scheme, Silicon Valley’s high priests (e.g., figures like Ray Kurzweil) roam the upper left, preaching implausibly that technological development is a transcendent good. Reductive capitalists roam the lower right, cynically subordinating higher goods to material and technological imperatives (while occasionally availing themselves of Kurzweil-adjacent high ideals).

Thus, the graph implies that the propositions at the edges and corners prefer some inaccessible other world to this one. Ordinary politics, meanwhile, plays out within the confines of the graph. Someone who agrees with the graph’s quasi-Aristotelian assumptions about the ordering of reality will tend to find that levels closer to the radicalism line are normatively preferable, with deviations far above or below it appropriate only in exceptional cases.

But the graphing method itself is neutral or close to neutral, because the lines could be redrawn. For example a convinced capitalist may wish to relocate the “state” lines below and left of the “individual” lines, or perhaps insert “corporation” lines above the “individual” lines. This might better reflect their worldview. The reconfigured graph would then display their core beliefs closer to the “normative” area close to the radicalism line, while shunting classical-Aristotelian beliefs deeper into justificatory or reductive territory.

Thus, this is simply a method of relating teleological claims to one another. The set up of the graph’s axes discloses one set of teleological claims, and then lets one map all other ideologies’ interrelationships through that lens.

What Could You Do With This?

We need more detailed maps of ideological space because it is otherwise an easy place to get lost. When we’re looking at everything from the point of view of a particular, dug-in ideology, it’s tempting to just view deviations as wrong or bad without seeing clearly how they’re different. That’s not helpful.

I don’t think that having maps like this will necessarily help everyone find the truth. If you use a map set up like my example, with its relatively traditional Aristotelian categories (individual, state, planet, etc.), it’ll probably tilt toward a vaguely traditional Aristotelian point of view. But at least you’ll have a relatively stable lens to locate the relative position of content you’re engaging with. This might make you less susceptible to being smooth-talked into embracing propositions at odds with your starting ideology.

Imagine, for example, that you set up your own map or import one you like. Then you just install it as a filter on your media consumption. An AI assistant would read or watch content before you, and form a sense of where it sits on the map. Then it might filter content, or flag it, to help you get a sense of where content is coming from before you consume it.

I don’t think this is the same thing as the “filter bubbles” we unreflectively create based on our engagement. We tend to engage with things that shock, titillate, and/or manipulate us. By contrast, by applying this kind of filter, you’re gaining agency to navigate into new territory with a clear (and AI-augmented) understanding of how and why it differs from your preexisting worldview. This would make it harder for others to feed you ideologically divergent or pandering content without you realizing what was happening.

Derek Jeter broke my heart with the flip. My spirit soared during the legendary streak. I forged friendships at the Coliseum, I learned how to suffer and celebrate, and I learned what it means to be from Oakland.

Today, I watched the A’s play their last game. Not their last game in Oakland; their last game anywhere. Oh yes, I realize that there may someday be nine players going through the motions in Sacramento or Las Vegas, but the A’s are Oakland. Without Oakland, it’s two sad little stylized letters with an apostrophe. Honestly, John Fisher shouldn’t be allowed to keep the trademark.

Here’s the thing. Sports don’t matter. It makes no difference who wins the games and the players don’t represent the fans. Nothing is at stake. If we’re being honest, it’s a big waste of time and a distraction from more important things. Except. Except.

Sometimes, when people show up to the stadium, a city sees itself. It feels its own energy and learns what it is. We see ourselves in our neighbors. Now, this does not happen at all sports games, or with all teams. When half the spectators are tourists, gamblers, or people on business outings — it doesn’t happen. And it can happen at events other than sports games. But at the Oakland Coliseum, it happened nearly every day. This team mattered for this city — more, I dare say, than the Warriors or Raiders. The A’s, in more ways than I’m comfortable admitting, were Oakland’s There There.

What John Fisher destroyed is worth immeasurably more than the couple billion dollars he hopes to add to his pocketbook in the relocation. His actions are unconscionable. He has no moral right to move the team, and his name should be covered in shame forever. His way of doing business is the root of many ills in this country and beyond.

The whole “new stadium” saga is a red herring. Owners aren’t entitled to public money for new stadiums. (Oh, and please don’t listen to that propaganda about possums in the walls. The Coliseum is magnificent: a hulking, brutalist beauty. There was no better experience in American sports.) Major League Baseball and John Fisher just put their pocketbooks above Oakland. It is absolutely that simple. 

I think the city really should have tried to seize the team through eminent domain; it’s probably too late for that now, but the state of California still can and should. (Gavin??) After all, what do public authorities have their powers for? To facilitate the designs of a billionaire who exploits the public, or to protect the public from people like that? 

Look, I know seizing the team is a sad pipe dream. But please, let some flower bloom in the ashes of John Fisher’s greed. 

I call on fans of all professional sports around the country to seize your power. Don’t be naive. Don’t give your team’s owner a pass. Don’t be a victim! Your current owner, or the next one, or the one after that, will rip the heart out of your community if they can make a billion bucks doing it. So here’s what to do, and if you won’t do it for yourself, do it for Oakland.

• First, demand non-relocation agreements. Boycott, harass, pester and harangue your teams’ owners until the ink is on the paper. It’s not that much to ask. Confident that your team will never go anywhere? Great, then the owners should be happy to sign. They won’t sign? Then stop going to games. Get the ink on the paper.

• Second, a hard one. Think hard about what deserves your support and what doesn’t. As a huge fan of America’s pastime I am sorry to say that baseball doesn’t. MLB was with Fisher 100% on this. So viewership is declining, the young generation isn’t getting the game? Let it wither. There is great baseball going on in Japan and Mexico. Oh and that Fisher-owned baseball enterprise in Sacramento or Las Vegas? The Fisher-owned soccer team in San Jose? The Gap, Old Navy, and Banana Republic? Poisoned chalices. Drink not.

• Third. Dear national sports media. You are in a powerful position to shift the culture here. Do it. Oakland is not a sad little story about a city that couldn’t build a new stadium. It’s an enormous stain on your industry and profession — a plank in your eye, too, until relocation blackmail ends. Do not be lackeys for the owners, do not be propagandists. (Tip of cap to Scott van Pelt and plenty of others.) The scandal around Fisher ought to build in the coming years, not fade.

Years ago, the Green Bay Packers became a publicly owned corporation. Ever since, the NFL has tried to prevent any other team from taking that route, because it makes the league’s job harder. But with that diffuse ownership structure, the Packers aren’t going anywhere. And that’s very simply how it should be.

Sports teams are about community. If American sports won’t live up to its obligation to American communities, well…let’s not let these franchises be the heart of our communities anymore. Laser that logo tattoo off your arm. Support live music. Go to the community theater.

And watch Premier League soccer.