The Intelligence We Forgot
Why artificial intelligence is forcing us to rediscover the nature of human intelligence
by Ken Archer
Pre-reading for What Rough Beast...? On AI's Potential to Surpass Humanity --- part of BKC's 2026-27 Speaker Series, All Too Human: How AI is Shifting the Ways We Make Meaning, presented with Harvard's Public Culture Project.
The False Choice
The public debate about artificial intelligence (AI) has largely stalled between two camps. On one side are those who celebrate AI as the inevitable successor to human intelligence. On the other are those who fear that AI threatens uniquely human capacities and meaningful activities. As the impasse refuses to budge, the continued proliferation of AI increasingly carries an air of inevitability—of optimism for some, resignation for others.
Debates often stall when opponents unknowingly share the same assumptions, assumptions that are then reinforced by the debate itself. The AI debate is no exception. Beneath their disagreements, both optimists and skeptics tend to think of human intelligence and artificial intelligence as separate things in the world and then ask how they compare. Skeptics ask whether AI will displace uniquely human sources of meaning and creativity. Optimists ask whether human intelligence is merely one instance of a more general phenomenon that machines are beginning to realize in a different form.
These unspoken premises, in fact, are mistaken from the start. They obscure something that we rarely stop to appreciate: the profound dynamism of human cognition, a dynamism that, we will argue, is expressed in and through all technology, including AI itself, and to which AI points us back. Human beings do not merely possess intelligence. We participate in the intelligible disclosure of the world.
Recovering this richer conception of human intelligence requires us to understand human cognition in a more original way than we do today. Earlier philosophical traditions sometimes understood human intelligence not as a detached capacity, but as a participation in the intelligibility of the world itself. Renaissance philosophers writing about crafts, architecture, and mathematics often expressed this intuition by describing the human being as a "microcosm of the universe." The sciences, arts, and technologies that emerge from human making were therefore understood as expressions of humanity's participation in the world’s intelligibility.
This is not just an ancient idea, however. The philosophy known as phenomenology recovers this richer conception of intelligence at the heart of contemporary mathematics, science and technology, through the careful description of everyday cognitive experience. The same, we will argue here, is true of AI.
AI is not a free-floating technology confronting humanity from the outside, diminishing the significance of human intelligence. It is one of the latest expressions of human intelligence, when intelligence is understood not as a thing we possess but as the disclosure of an intelligible world. To understand what AI is, and the true opportunities and dangers it presents, we must first recover the human intelligence we have forgotten.
In Searching for AI, We Search for Ourselves
Moravec's Paradox, a famous discovery in AI research, offers an important clue to the nature of human intelligence. Tasks we associate with advanced intelligence, such as chess, formal reasoning, and fluent language use across sophisticated topics, have proven surprisingly amenable to automation, while tasks a child can master, such as perception, embodied action, and reading social cues, remain extraordinarily difficult. This paradox reveals something deep: the forms of intelligence most rooted in everyday embodied experience are precisely those most resistant to formalization. This should make us curious about what everyday experience actually involves. It turns out that what we call "simple" human activities are not simple at all.
Phenomenology begins from this everyday experience. Rather than treating intelligence as an abstract capacity hidden somewhere inside the mind, it seeks to describe how the world becomes intelligible in ordinary acts of perception, judgment, and action. Through such descriptions, phenomenology recovers the richer conception of intelligence introduced above—not as a detached capacity, but as a living participation in the intelligibility of the world.
Consider something that appears entirely ordinary: perceiving a car.
At first, perception unfolds as a continuous flow. As we move around the car, contours shift with perspective. Shades of red and variations of smoothness covary with our movement, which we experience kinesthetically. Yet despite these changing appearances, we do not experience a succession of disconnected profiles. We experience a single three-dimensional object that shows up through a structured series of appearances co-varying with our sense of movement. The hidden backside of the car is already anticipated as part of the object itself, even though it is not presently visible.
This dynamism of perception is left out of most accounts of cognition, according to which the mind perceives two-dimensional visual profiles and reconstructs a depth dimension to infer a three-dimensional object behind it, and similarly infers redness, smoothness and other nominal properties from these snapshots. But perceptual experience never presents us with isolated retinal “snapshots” or piecemeal reconstructions. It is bedrock, phenomenologically speaking, that we experience the car immediately as a three-dimensional unity with a backside, depth and persistence.
Every perception is therefore more than what is immediately given. Present appearances are intertwined with retained past appearances and anticipated future appearances. Phenomenologists call this horizonal experience. Every appearance points beyond itself toward further possible appearances. Think of listening to a melody. Each note points beyond itself toward notes expected but not yet heard. You do not hear isolated tones — you hear music, precisely because each moment reaches into past and future. Human cognition is dynamic because the world is disclosed through horizons of possibility that extend beyond what is immediately present.
This anticipatory structure is ordinarily taken for granted because it operates prior to explicit conceptualization. We do not merely register a sequence of appearances. We experience intelligible unities through structured variation. We experience identity through change, one through many. The car remains the same car even as every visible property shifts with our perspective. This anticipatory structure operates prior to any explicit judgment or reasoning. It is one of the deepest ways human beings participate in the intelligibility of the world — and it is precisely the kind of intelligence that Moravec’s Paradox tells us is hardest to formalize.
The motivation to anticipate experience, intrinsic to perceptual experience, in turn motivates a shift in attention to a higher level of intelligibility. Instead of perceiving the associated appearances with the unifying car in the background, we now identify the car itself as a subject-whole articulated through categorial parts. The car is red. The car is smooth. The car is moving quickly.
This shift is decisive. Where before there was simply the structured unfolding of appearances, there is now something that can be said. The object itself becomes available as a subject of possible predications. Through categorial determinations of the object as a whole, we can anticipate indefinitely many appearances in a new and more stable way. Language does not create this structure of subject and predicate. It expresses a structure of intelligibility already disclosed in experience.
Similarly, we can anticipate our experience of cars through scientific models expressing this same underlying intelligibility. We may model the car's velocity, the aerodynamic effects of its shape, or the forces acting upon it. In doing so, we move beyond this particular car and articulate the invariant relationships governing how its appearances vary under changing conditions. Scientific models make explicit the intelligibility already implicit in perceptual experience, allowing us to anticipate not merely this car's future appearances, but the behavior of cars generally and, ultimately, of physical objects as such.
Language and models arise from a more fundamental activity. We do not first possess words and models and then attach them to the world. Rather, the world is always already intelligibly disclosed in structured ways, and language and models give expression to that disclosure. The syntactic structures of language and the formal structures of scientific models externalize and express deeper structures within intelligible experience itself.
This brings us to a profound observation: every achievement of civilization — language, mathematics, science, technology — testifies to the intelligibility of the world, and in turn to human cognition as the living disclosure of that intelligibility. These achievements do not stand apart from consciousness. They presuppose it. This was powerfully argued recently by Adam Frank, Evan Thompson and Marcelo Gleiser in their book The Blind Spot: Why Science Cannot Ignore Human Experience.
Human creation continually extends this process. Once language and models become objects of reflection in their own right, their structures can themselves be investigated and formalized. Logic formalizes relations within language. Pure mathematics formalizes relations within models. Human creations advance ever farther from immediate experience while still pointing back to the living activity from which they arise. This is the Renaissance insight of the human being as a microcosm of the universe, recovered through phenomenological analysis.
Now we can see Artificial intelligence in its proper context.
Large Language Models (LLMs) do not begin with the world itself. They begin with language. By learning statistical regularities within vast bodies of text, they model patterns in how human beings articulate and communicate intelligible experience. They learn the structures through which people describe things, make judgments, tell stories, formulate explanations, and refine distinctions.
To appreciate LLMs therefore requires first appreciating language as an expression of intelligible experience. LLMs exhibit a remarkable sensitivity to the structures of language through which the world is disclosed. They can model how vague terms are used in combination with other vague terms—how "justice" pairs with "delayed" in legal contexts, or how "sweet" articulates "grief" in poetic expression. These patterns reflect genuine tendencies within human language to navigate the space between ambiguity and clarity.
AI is not a rival to human intelligence. It is one of the most sophisticated expressions yet created of capacities that originate within human cognitive life.
Yet something essential to human cognition remains absent from its creations, from language and science. Intrinsic within the intelligible disclosure of the world is a responsibility to the world’s disclosure. This is not an appeal to an external norm, but a responsibility internal to lived experience.
Imagine approaching a car expecting it to be uniformly smooth. As you move around it, you discover what appears to be a dent on one side. Is it damage? A trick of the light? A design feature? This discovery does not merely update your understanding according to a rule. It calls your understanding into question. You now face a choice. Do you look closer to clarify the situation? Or can you ignore the anomaly and continue as before? This choice calls you to step back from the world, to be free to deliberate.
To be human is to be answerable to the world in its intelligible disclosure, such that the experience of ambiguity opens onto a choice and calls us into question as responsible agents of disclosure. Human consciousness simply is the intelligible disclosure of the world - it is not computation accompanied by an inner glow of subjective experience. (Accordingly, consciousness, cognition, and intelligence are treated here as different aspects of a single phenomenon.) The responsibility to the world intrinsic within human consciousness is one with the responsibility to ourselves as the intelligible disclosure of the world.
This is why phenomenologist John Haugeland famously remarked that AI "doesn't give a damn." The most advanced AI systems today, those that self-improve through reinforcement learning, are massive achievements of science, and yet they still update themselves according to rules that are downstream formalizations of the world we disclose and then express in written language. This disclosure is ultimately a free choice that self-responsible humans make, when the world shows up in ways that frustrate their expectations.
Phenomenologists such as Husserl, Heidegger, and Arendt worried deeply about this, regarding responsibility to the world’s intelligible disclosure as indispensable. When we lose that answerability to lived experience and its call on us to step back and deliberate as free humans, thought becomes detached from lived experience and collapses into the repetition of inherited formulas, conventions, and public opinion. What begins as a genuine disclosure of the world can harden into thoughtlessness.
This danger becomes especially important once intelligence is expressed in language, models, institutions, and technologies. To understand that danger, we must next examine how human beings become absorbed in their own creations.
Forgetting Ourselves in our Creations
Human intelligence does not remain confined to the moment of disclosure. It overflows and emanates into its creations. Through these creations, human beings externalize structures of intelligibility, giving them a durable existence beyond the moment in which they were first disclosed.
These externalizations are among humanity's greatest achievements. They make cooperation possible across vast distances of space and time. They enable the cumulative growth of science, culture, and civilization itself.
This helps explain the widespread, enduring appeal of technology. When we think of technology, we usually think of particular technologies: tools, machines, and systems. Much of the philosophical tradition, however, focused on techne—right reasoning about things to be made. Technologies arise from techne. They fascinate us because they embody intelligibility in durable form. Through them, insights that originate in human understanding become shareable, repeatable, and capable of shaping the world in new ways.
The danger does not arise from these achievements themselves. The danger arises only when we alienate ourselves from their origin as we allow ourselves to become absorbed in our products.
Phenomenology describes this process as sedimentation. Language and models don’t just express structures of intelligibility. They sediment meaning. Each time we say something like “red cars cost more to insure” or “drafting behind a truck to reduce drag saves gas,” we leave behind a trace. Over time, these traces accumulate. Meanings stabilize, along with the practices from which meanings originally arose. Language and scientific models become habitual. What was once actively disclosed becomes passively inherited. Sedimented meanings press down onto experience itself, shaping how the world is anticipated and articulated in advance. This process is intrinsic to language, science and technology as externalizations of intelligible experience.
Sedimentation is both a strength and a danger. It allows meaning to endure. But it also introduces the possibility of thoughtlessness. Words can be used automatically, detached from the experiences that first gave them sense. We say that "the science is clear" or that "it is just statistics" without any direct grasp of the concrete realities these phrases are meant to illuminate. What is said is no longer a genuine articulation of the world. It is something we say because it is what "one" says. Such thoughtlessness, when reinforced by the authority of science, misuses the natural appeal of science to in turn diminish the validity of human cognition from which science arises.
Again and again, modern societies mistake the static products of intelligence for their dynamic, living source in our own consciousness. We transform judgment into laws, procedures, and institutions, then come to imagine that good judgment consists merely in following procedures. We transform speech into writing and then into information, then forget the interpersonal activity through which meaning is first established and renewed. We transform understanding into scientific models, then technical systems, and increasingly AI, then begin to imagine that these systems themselves are the source of understanding.
In each case, something genuinely valuable has been created. But in the process, the living dynamism that gave rise to it recedes into the background until its source is gradually substituted by its products. Responsibility is displaced from the living activity of disclosure onto the external structures that disclosure has produced. The result is precisely the thoughtlessness—the abdication of freedom and responsibility intrinsic to human cognition—that phenomenologists such as Husserl, Heidegger, and Arendt warned against.
AI represents the most recent example of this temptation. Writing and scientific models generate products that circulate independently of the activities that originally produced them, while preserving traces of the intelligibility disclosed through those activities. AI generates outputs that do the same. Paragraphs, arguments, explanations, analyses, and recommendations now appear without any visible connection to the acts of world disclosure underlying such systems in their architecture and training data.
The danger of AI is not that machines will become conscious or generally intelligent. The danger is that we will come to identify intelligence with the products of intelligence rather than with the living activity of disclosure from which those products arise.
Once intelligence is identified with its externalized products rather than with the living disclosure from which those products arise, responsibility becomes increasingly difficult to recognize. We begin looking for understanding in systems rather than in ourselves. We seek judgment in procedures, meaning in information, and intelligence in outputs. What is lost is not merely a philosophical distinction. What is lost is our sense of ourselves as the responsible agents of disclosure from whom all of these achievements ultimately arise.
What AI Demands of Us
This is why it is critical to understand AI, and technology more broadly, as bearing witness to the intelligibility of the world and pointing back to its creators. The dynamism intrinsic to human cognition overflows and emanates into its creations, which in turn refer us back to the active consciousness from which they arise. Science and technology are therefore not rivals to human intelligence but among its most remarkable expressions. Created consciously and responsibly, they have the potential to augment human endeavor and extend our participation in the intelligibility of the world.
The absorbed fascination with technology, however, continually tempts us to forget this origin. This temptation underlies both the celebration of AI as the successor to human intelligence and the fear that AI threatens uniquely human capacities. Despite their differences, both positions share a common premise. Both treat technology as something confronting humanity from the outside rather than as an expression of human intelligence that points back to its source.
For this reason, the humanistic resistance to technology often reinforces the very assumptions it seeks to challenge. In attempting to defend human meaning against science and technology, it frequently accepts a diminished conception of both. Human beings become identified with a realm of values, meaning, or subjective experience, while science and technology become identified with an opposing realm of formal systems, instrumental reasoning, and technical control. Humanity is defended only by separating it from its own creations.
This divide has been reinforced institutionally through the separation of the sciences and humanities into distinct intellectual cultures. Scientific knowledge comes to appear self-grounding and autonomous, while humanistic inquiry is confined to interpretation and critique. The result is a shared inability to recognize their common origin in lived cognition and intelligible disclosure. Technology appears increasingly autonomous precisely because we have forgotten the activity from which it arises.
This is why so much contemporary criticism of technology is both insufficiently optimistic and insufficiently critical. It is insufficiently optimistic because it narrows technology into a free-floating, instrumental logic of efficiency, thereby overlooking its deeper significance as an expression of human intelligence. Yet it is also insufficiently critical because it assumes that changing the social distribution of power would eliminate the fundamental danger. Even technologies designed by the most benevolent institutions would still tempt us to seek understanding in systems, judgment in procedures, and meaning in outputs rather than in ourselves. We need to be both more optimistic and more critical than the current debate allows.
The result is an impoverished understanding of both humanity and technology. Humanistic critiques struggle to account for the enduring appeal of science and technology. Technological triumphalism senses that appeal but often explains it only in terms of efficiency or scale. Both overlook the deeper appeal at work. Technology fascinates because it embodies intelligibility. It is compelling not because it transcends human intelligence but because it expresses it.
It is therefore not enough merely to observe that AI is a product of human ingenuity. One must understand how AI, like technology more broadly, arises from the same cognitive structures that underlie language, creative arts, human relationships, and so forth. Only then can AI be understood as a human achievement, and only then can it be responsibly directed toward human ends.
This brings us to the central practical implication of the argument developed here. If AI is an expression of human intelligence, responsibility for its effects cannot be displaced onto, or even shared with, the model itself. The temptation to do so is precisely the form of thoughtlessness that phenomenologists warned against. It is the temptation to transfer responsibility from the living activity of disclosure onto the externalized products of that activity.
If we are builders of AI, we are responsible for the systems we create and the consequences they produce. We cannot evade that responsibility by appealing to the autonomy, complexity, or opacity of the models themselves. If we are users of AI, we are responsible for the interpretation and application of its outputs. We cannot delegate judgment to systems whose significance ultimately depends upon our own acts of understanding.
The challenge posed by AI is therefore not primarily technological. It is anthropological - a personal challenge to each of us. AI confronts us with a question about ourselves, one that both triumphalism and resistance - as Arendt would say, the twin bannisters we lean on when we talk about technology and about AI - obscure or even evade. Will we interpret AI over against us, as either surpassing or threatening human intelligence? Or will we recognize in AI a formalization of structures that originate in our participation in the world's intelligible disclosure, and thereby recover our responsibility for what we create and for ourselves as the intelligible disclosure of the world?
The answer matters because technology can be developed in very different ways. Historians such as Harry Braverman and David Noble have shown that some technologies diminish human agency by transferring skill, judgment, and responsibility into systems that deskill workers and users. Yet they also describe technological paths that expand human autonomy and responsibility, enabling people to exercise broader forms of judgment, creativity, and self-direction. Technology can deskill, but it can also upskill. It can narrow responsibility, or it can enlarge it. Each path is motivated by anthropological ideas of what intelligence and skill really are. The deeper source of the non-neutrality of technology thus lies in the conception of intelligence embodied in a technology, not merely in the concentration of power of those who build it.
The ubiquity of computers and of the Internet in daily work has made it less plausible to argue that industrial technology proceeds deterministically in the direction of deskilling and degradation of work. Several inventors of the first personal computer, especially Lee Felsenstein, were inspired by Ivan Illich’s description of convivial tools, tools which are not specialized to serve a proprietary machine or process, but which “can be easily used, by anybody, as often or as seldom as desired, for the accomplishment of a purpose chosen by the user.” The question raised by AI is whether it will function more like the convivial tools envisioned by Illich—tools that enlarge human capacities and initiative—or whether it will transfer those capacities into increasingly autonomous systems.
Writing provides a familiar example. Writing is thinking because writing is one of the ways the world comes to clearer disclosure. We often begin writing without fully knowing what we think. Through the activity of writing itself, possibilities are explored, distinctions emerge, and what was initially vague becomes clarified. Writing is not merely a means of expressing thought. It is one of the principal ways thought comes to clarity. The act of writing is an act of disclosure.
AI can be used to expand this human capacity by helping a writer explore alternatives, discover relevant sources, or refine a draft, all of which remain subject to the writer's judgment regarding what genuinely clarifies the matter at hand. Used in this way, the technology supports the practice of writing as thinking while leaving responsibility for thinking with the writer. Yet AI can also be used in ways that diminish those same capacities, when writing becomes disconnected from the world-disclosive activity from which thought arises.
The difference lies in whether we remain conscious of ourselves as the responsible agents of disclosure from whom technology arises. A society that forgets this will continually seek salvation in procedures, systems, and machines. A society that remembers it will treat technology as an expression of human intelligence rather than as its replacement.
The deepest lesson of AI is therefore not that machines are becoming more like human beings. It is that human beings must become more conscious of themselves. AI has not revealed the obsolescence of human intelligence. It has revealed how much of that intelligence we had forgotten. The future of AI therefore depends less on what machines become than on whether we remember ourselves as the responsible agents of disclosure from whom they arise.
