The philosophical problem we address here is the closing provocation of the previous essay in this series— that artificial intelligence may be changing not just how we learn but what it means to actually know a thing. Before we can assess what AI does to knowing, we need to be clear on what knowing actually is.

—

Consider a thought experiment put forth in a children’s book we all probably read in the sixth grade.

In Lois Lowry's The Giver, society has solved the problem of knowledge management and emotional pain with certain collective knowledge efficiencies. Memory—the accumulated experiential knowledge of generations—is held by a single designated figure in society, the Receiver of Memory. Everyone else is relieved of the burden of knowing and, consequently, of the emotional experience of memory. And yet something has been lost in the community that no one can name (because the capacity to name it has also been surrendered).

What our protagonist Jonas discovers when he becomes the Receiver of Memory is not information. His training does not involve the transfer of facts. He is being given weight—the sense of what things cost, what they mean, what it is to have been cold and hungry and loved and afraid. The knowledge the Receiver holds is not propositional. It cannot be looked up. It is constitutive—it makes the knower into a different kind of being than the one who merely has access to the same content secondhand through a knowledge base, or the one who does not have access to it at all.

Lowry wrote this as a children's novel. It is also, as it turns out, a surprising analogy for the epistemological questions that our current implementations of artificial intelligence are forcing us to confront.

—

Our philosophical teachings have long recognized that not all knowledge is equal. Aristotle distinguished between episteme (systematic, demonstrable knowledge of universals) and phronesis (practical wisdom—the judgment that comes from having navigated real situations over time and developed cultivated intuition for both the environment and the circumstance). You can teach someone the principles of ethical reasoning. You cannot give them phronesis—ethical and moral discernment. It forms through encounter, through error, through the slow accumulation of judgment that only experience produces.

A third category is relevant here: techne, the knowledge embedded in craft. The carpenter who has spent twenty years working wood knows things about grain and tension and the behavior of material under stress that cannot be fully articulated in a manual. This knowledge lives in the hands as much as the mind. It is equally as tangible as it is notional. It is, in the philosopher Michael Polanyi's formulation, tacit — “we know more than we can tell.”

These distinctions have always mattered in theory. They matter urgently now in practice, because AI systems are extraordinarily good at one kind of knowledge and structurally incapable of the others. A large language model (LLM) can reproduce episteme with impressive fidelity. It has no phronesis. It has no techne. It has encountered nothing, navigated nothing, developed no judgment through the friction of real consequence. What it offers us is access—extraordinarily broad, mind-expanding access—to the surface of what humans have managed to collectively articulate on the internet. What it cannot offer is the weight that Jonas receives as the Receiver of Memory.

This is not a criticism of AI systems. It is an acknowledgment of what they are (and are not). The confusion arises when we treat access as equivalent to knowledge, when we mistake the map for the territory, the summary for understanding, fluency for discernment. This problem is particularly acute for novice thinkers, new practitioners, and students.

—

There is a concept in the German educational tradition—Bildung—that has no precise English equivalent. Wilhelm von Humboldt, the philosopher and educational reformer who shaped the modern research university, described it as "the free, harmonious development of a human being's powers into a complete and consistent whole, through encounter with the world in its variety and resistance." The key word is resistance. Bildung is not the accumulation of information. It is the transformation of a person through sustained encounter with something that pushes back.

“A tree does not exist in order to produce lumber. Something is growing there under its own power, toward its own form, and the growing is not a means to some further end.” Brendan McCord’s analogy clarifies what is at stake: a person is not an input-output machine to be optimized for productivity. People are beings in the process of becoming.

What happens to Bildung when the resistance is removed? When every question finds an immediate answer, every gap is filled before it can create the productive discomfort that drives inquiry? McCord poses the sharpest version of this question: in a world where AI handles an increasing share of cognitive labor, are we approaching what Aristotle imagined—scholé, the freedom from necessity that allows genuine human flourishing? The risk, as McCord frames it, is not that AI will make us stupid. The risk is that we will reach for and achieve autocomplete for life—reliance on the generated response rather than our own discernment.

—

Institutional memory offers another concrete case for why this distinction matters at a societal level.

Every organization that has existed long enough develops a form of tacit knowledge that is not captured in its documents, its policies, or its training materials. It lives in the people who have been there long enough to remember why decisions were made, what was attempted and what failed. When those people leave, institutional knowledge leaves with them. This phenomenon has a clinical name in organizational theory: corporate amnesia.

AI systems are now being deployed in earnest as institutional memory solutions. The pitch is intuitive: if we ingest all of our documents, all of our communications, all of our recorded decisions, we can query that corpus and recover what we know. This can be genuinely useful. But it also compresses the tacit dimension—the judgment about which documents to trust, the context that explains why a decision was recorded one way but meant another, the sense of what actually happened versus what was documented in the official record. After all, history does not have a capital H.

The Giver, read through this lens, is not a dystopia about a society that forgets things. It is a dystopia about a society that confused access with knowing, outsourced institutional memory, and organized itself accordingly. The Receiver of Memory is not a search engine. He is the only member of the community who has been transformed by encounter with what happened—who carries it as weight rather than data. When Jonas begins to receive those memories, he does not become more informed. He becomes a different person. Which is why his journey as the Receiver of Memory is described to him as a lonely one—because no one else in society shares his knowledge and his experience of that knowledge.

—

More than having access to knowledge, knowing something means having been shaped by the process of its creation. The process of knowledge creation alters how a person views subsequent problems, what questions they think to ask, what one person notices that another might miss. Jonas knows what cold is not when he is told the word, but when the memory of snow enters him and he feels cold. The carpenter knows the wood not when he reads about grain, but when the chisel catches unexpectedly and his hands learn to feel for what will happen next.

I argue that this is not mysticism, nor is it too abstract to be useful. It is a description of how understanding actually forms—through the kind of iterative, embodied, often uncomfortable engagement with material that gradually builds a knowledge structure robust enough to support genuine judgment. Invoking this process, in many ways, was the cornerstone behind the modern research university, with its structural blending of teaching and inquiry, and its seminar culture.

The neuroscience of expertise confirms what the philosophical tradition has long held: expert knowledge is structurally different from novice knowledge, not just quantitatively greater. It is organized differently, retrieved differently, applied differently. The expert does not simply know more facts. They know differently—with a kind of pattern recognition and anticipatory judgment that takes years of formative encounter to develop.

The question that AI poses to this account of knowing is not whether access is useful. It obviously is. But whether access, provided frictionlessly and at scale, can substitute for the development of cultivated intuition and discernment. The evidence, both philosophical and empirical, suggests it cannot.

“You can live on an inheritance for a while. You cannot educate a civilization on inherited judgment forever.”

- Brendan McCord

There is something almost paradoxical about where this leaves us.

The argument for AI use in the preceding essay in this series involves minds that are already deeply formed. The expert who uses AI to extend the reach of a knowledge structure they have spent years building is doing something categorically different from the novice who uses AI to avoid building a knowledge structure altogether. For the former, AI provides access that a prepared mind can metabolize into genuine understanding. For the latter, AI provides access in place of formation.

This suggests that the value AI adds to knowing is, in some sense, proportional to what the knower already knows. Which raises a question that will occupy the next essay in this series: if knowing requires formation, and formation requires resistance, and AI removes resistance, what does the pathway from novice to expert look like in an AI-saturated world? And what kind of knower does it produce at the end?

These are not abstract questions. They are questions about what kind of citizens, professionals, and human beings we are in the process of forming. And they deserve more serious attention than the current discourse has yet given them.

The knowledge formation process cited above has a structure, and this structure is well understood.[1] How we reconcile this structure with AI and the untrained mind is the topic of the next essay.

—

This is the second essay in a series on AI, knowing, and what it means to think in an age of intelligent machines. The first essay, on AI and expert cognition, is available here. The third will examine novice learning in an AI-mediated learning environment.

[1] Ericsson, K. Anders, Ralf T. Krampe, and Clemens Tesch-Römer. "The role of deliberate practice in the acquisition of expert performance." Psychological review 100, no. 3 (1993): 363.
Bjork, Robert A. "Memory and metamemory considerations in the training of human beings." Metacognition: Knowing about knowing 185, no. 7.2 (1994): 185-205.